Effects of Online Social Netwo

Effects of Online Social Networking Use on Inter-personal and Psychological
Well-being among Students in China: A Longitudinal Study
LI, Jibin
A Thesis Submitted in Partial Fulfilment
of the Requirements for the Degree of
Doctor of Philosophy
in
Public Health
The Chinese University of Hong Kong
May 2016
ProQuest Number: 10300438
All rights reserved
INFORMATION TO ALL USERS

The quality of this reproduction is dependent upon the quality of the copy submitted.
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.
ProQuest 10300438
Published by ProQuest LLC ( 2016 ). Copyright of the Dissertation is held by the Author.
All rights reserved.

This work is protected against unauthorized copying under Title 17, United States Code
Microform Edition © ProQuest LLC.
ProQuest LLC.
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, MI 48106 - 1346
Thesis Assessment Committee
Professor WONG Chi Sang Martin (Chair)
Professor SU Xuefen (Thesis Supervisor)
Professor MO Kit Han Phoenix (Thesis Co-supervisor)
Professor LAU Tak Fai Joseph (Thesis Co-supervisor)
Professor MAK Wing Sze Winnie (Committee Member)
Dr. IP Patrick (External Examiner)

Abstract (English)
Effects of Online Social Networking Use on Inter-personal and Psychological
Well-being among Students in China: A Longitudinal Study
LI, Jibin
Doctor of Philosophy in Public Health
The Chinese University of Hong Kong (May 2016)
Introduction
Depression is the most widely reported psychiatric disorder among adolescents
worldwide. Social networking use as a new phenomenon has gained exponential
popularity, along with increased levels of high Internet penetration. Internet addiction
(IA) overall has been linked with higher level of depressive symptoms in many
cross-sectional studies. However, the causal direction is still under-researched due to
limitations of the cross-sectional study design, especially in the area of social
networking addiction (‘SNA’). Moreover, the impacts of social networking use
intensity on psychosocial and mental health have been raised as a public health
concern, but remain under-researched particularly among adolescents. Current
cross-sectional studies reveal a mixed association due to the study design and lack of
validated instruments used. The mechanisms underlying the relationship between
social networking use and depression have been minimally explored. Therefore,
conducting longitudinal studies among adolescents is highly warranted.
Objectives
This study consisted of two phases. The phase I cross-sectional study aimed to
i
develop and validate an instrument measuring social networking use intensity among
adolescents in Guangzhou, China. The phase II longitudinal study aimed to (1)
estimate the prevalence and incidence of depression, ‘SNA’, and IA among
adolescents in China; (2) explore the causal directions between Internet addictive
behaviors (‘SNA’ in specific and IA in general) and depression; (3) investigate the
predictive effects of intensity and emotional connection related to social networking
use on depression, by using the scales validated in the phase I study; (4) investigate
the impacts of social networking use intensity on psychosocial outcomes, including
friendship quality, perceived social support, self-esteem, adolescent-parents conflict
and social non-confidence; (5) explore the underlying mechanisms of psychosocial
factors and Internet addictive behaviors for the association between social networking
use intensity and depression among adolescents.
Methods
In the phase I validation study, 910 participants who were social networking
users were investigated in junior middle schools throughout Guangzhou, China. In the
phase II nine-month longitudinal study, 4,237 students who were social networking
users were investigated both at baseline and follow-up. Statistical methods including
univariate analyses, multilevel regression models, and mediation and moderation
analyses were used in this study.
Results
Among the longitudinal sample, the prevalence of depression, ‘SNA’, and IA
were respectively 24.57%, 13.69% and 6.56% at baseline, and the corresponding
incidence were 23.37, 12.80 and 6.77 per 100-person-years respectively during the
nine-month period. Longitudinal analyses showed bi-directional relationships between
ii
‘SNA’/IA and depression among adolescents.
The developed instrument, SNAIS, showed acceptable psychometric properties.
Two subscales were identified: SFUI and EFUI. The adjusted results showed that
SFUI and EFUI scores at baseline significantly and positively predicted incident
depression (AOR=1.017 for SFUI score and AOR=1.046 for EFUI score), while the
prediction of emotional connection on incident depression was not statistically
significant.
By using changes in scores during the longitudinal period as variables, increased
social networking use intensity was positively associated with increased perceived
social support (both SFUI and EFUI) and friendship quality (EFUI only). Additionally,
increased social networking use intensity was also positively associated with
increased social non-confidence (both SFUI and EFUI) and adolescent-mother
conflict (SFUI only).
Mediational analyses indicated that changes in addiction (i.e. ‘SNA’ and IA)
scores and changes in psychosocial factors (i.e. increased social non-confidence and
adolescent-mother conflict) altogether mediated 78.77% of the effect of change in
SFUI score on change in depressive symptoms. The effect of change in EFUI on
change in depressive symptoms was mediated completely through the paths of change
in addiction (i.e. ‘SNA’ and IA) scores and change in social non-confidence score.
Furthermore, the results of the mediational analyses revealed an interesting finding
that change in friendship quality and change in perceived social support negatively
suppressed the impact of change in social networking use intensity on change in
depressive symptoms, but the effect size was relatively small.
iii
Conclusion
Depression and Internet addictive behaviors are present in epidemic proportions
among adolescents, and the two issues are highly co-occurred. Social networking use
is harmful for adolescents’ psychosocial and mental well-being. The results of this
study identified underlying mechanisms and provided robust insight for designing
effective intervention programs to prevent depression among adolescents. Future
longitudinal studies with three or more waves of observations are needed to further
confirm these findings. Translational research is warranted to use this information to
form effective intervention programs.
iv
Abstract (Chinese)
社交網絡使用強度對青少年人際和社會心理健康影響的前瞻性隊列
研究
李濟賓
公共衛生哲學博士
香港中文大學二零一六年五月
背景
青少年人群中，抑鬱症的流行現況非常嚴峻。隨著互聯網技術的發展和普及，
社交網絡作為一種新興的網絡活動受到青少年的廣泛使用。多數橫斷面研究報道
認為網絡成癮可能會導致抑鬱症。但是，網絡成癮（包括社交網絡成癮）與抑鬱
症之間的因果關聯仍不清。此外，對於社交網絡使用強度對青少年社會心理健康
的影響的研究甚少。由於缺少有效的測量工具，目前為數不多的幾篇橫斷面研究
給出了矛盾的結果。社交網絡使用強度對抑鬱症影響的機制亦不清楚。尚無在該
人群中開展的前瞻性隊列研究。
目的
橫斷面研究的主要目的是研製適用於中國初中生的社交網絡使用強度測量
量表，并考核其信度和效度。前瞻性隊列研究的目的包括：1）估計抑鬱症，一
般網絡成癮和社交網絡成癮的患病率和發病率；2）探討網絡成癮與抑鬱症的因
果關聯；3）採用橫斷面研製的量表，調查社交網絡使用強度和社交網絡情感依
v
賴對抑鬱症的前瞻性影響；4）探討社交網絡使用強度對青少年社會心理健康的
影響，包括朋友質量，感知的社會支持，自尊，親子衝突和社交不自信；4）探
討社會心理因素和網絡成癮在青少年社交網絡使用強度和抑鬱症關聯中的機製
作用。
方法
本研究由橫斷面調查和前瞻性隊列研究兩部分組成。橫斷面研究調查了 910
名廣州初中生社交網絡使用者。前瞻性隊列研究調查了 4237 名廣州初中生社交
網絡使用者。統計分析方法包括单因素分析、多水平模型、中介效應和交互效應
分析等。
結果
抑鬱症、社交網絡成癮和一般網絡成癮的基線患病率分別為 24.57%、13.69%
和 6.56%。抑鬱症、社交網絡成癮和一般網絡成癮的發病率分別為 23.37、12.80
和 7.7 每 100 人年。多水平模型顯示一般網絡成癮以及社交網絡成癮與抑鬱症的
因果關聯是雙向的。
橫斷面研究中研製的初中生社交網絡使用強度測量量表具有較好的信度和
效度，因子分析顯示有兩個分量表：社交功能使用強度和娛樂功能使用強度。分
析顯示，社交功能使用強度和娛樂功能使用強度基線得分是抑鬱症發病率的危險
因素（社交網絡使用強度的 AOR=1.017，娛樂功能使用強度的 AOR=1.046）

，而
社交網絡情感依賴得分與抑鬱症發病率的關聯無統計學意義。
採用基線調查與隨訪調查間量表得分的改變量作為分析變量顯示，社交網絡
使用強度得分的改變量與感知的社會支持（包括社交功能使用強度和娛樂功能使
vi
用強度）和朋友質量（只有娛樂功能使用強度）的得分改變量成顯著正相關。同
時，社交網絡使用強度與社交不自信（包括社交功能使用強度和娛樂功能使用強
度）和親自衝突-母親（只有社交功能使用強度）的得分改變量成顯著正相關。
中介效應分析顯示，在社交功能使用強度得分改變量對抑鬱症得分改變量的
影響中，一般網絡成癮改變量、社交網絡成癮改變量、社交不自信得分改變量和
親自衝突-母親得分改變量四個變量對其的中介效應達 78.77%。同時，一般網絡
成癮改變量、社交網絡成癮改變量和社交不自信得分改變量三個變量完全中介了
娛樂功能使用強度得分改變量對鬱症得分改變量的影響。此外，朋友質量和感知
的社會支持得分改變量在社交網絡使用強度改變量和抑鬱症得分改變量的關聯
中表現出抑製作用。
結論
抑鬱、網絡成癮行為等心理健康問題和危險行為在中國青少年中的流行非常
嚴重。抑鬱症和網絡成癮行為之間的影響表現為雙向的。社交網絡使用不利於青
少年社會心理健康的發展。本研究的結果（如中介變量）將有助於制定高效的干
預研究策略以便有效預防和降低青少年人群中抑鬱症的水平。未來，需要更多具
有三次及以上觀察次數的前瞻性隊列研究來進一步研究社交網絡使用對青少年
健康的影響。同時，利用本研究的結論開展相應的轉化性的干預研究。
vii
Acknowledgement
My Ph.D. study in the Chinese University of Hong Kong has been like a
marathon full of challenges, frustrations, expectations, and excitement. I will carry
these invaluable experiences with me throughout my whole life and research career in
the future. As my study in CUHK is coming to a close, I would like to take this
opportunity to greatly thank all of those people who unconditionally provided help
and support in various ways during my academic journey.
Firstly, I would like to give my deepest thanks to my supervisors, Prof. Xuefen
Su, Prof. Phoenix Kit Han Mo and Prof. Joseph Tak Fai Lau. They encourage, support,
and push me to achieve different works, which led to the completion of this final
research thesis. They always walked me through all of the stages of thesis research,
from literature reviews, to extracting research topic, writing a proposal, contacting
study sites, assisting with field work, data analysis, to the final draft. Without their
consistent support and illuminating instructions, my Ph.D. study could not have been
achieved well. Their patience and support have made it possible for me to challenge
difficulties with strong confidence. Their hands-on supervision made me a much
stronger and motivated critical thinker and a better writer in scientific research.
During the study period, I am impressed and motivated by Prof. Lau’s wisdom, vision,
and passion for scientific research. I will always be grateful for their time and
guidance provided by my mentors.
Secondly, I would like to give my sincere gratitude to Director Jie Tang and Mr.
Zuguo Qin from the Institute of Health Education in Guangdong province, as well as
Director Jincheng Mai and Ms. Yuxia Chen from the Center for Health Promotion of
viii
primary and secondary schools in Guangzhou, for their strong support and
coordination for school contact and fieldwork of this research. I would also like to
thank all teachers of the junior middle schools and all student helpers for their
assistance and help in the field survey and data collection. I would also like to thank
all participants for their voluntary participation and cooperation. This study could not
have been done smoothly and successfully without their generous cooperation.
Thirdly, I deeply appreciate the help and support from my team members, Dr.
Zixin Wang, Dr. Jinhua Li, Dr. Chunrong Li, Ms. Wangnan Cao, Ms. Xueying Yang,
Mr. Mason Lau and Ms. Danielle Walden. I would also like to give my thanks to my
classmates. Their support and companionship made my three-year study enjoyable
and fruitful.
Fourthly, I would like to thank Ms. Wayser, Graceanne R for her help in editing
language of this thesis.
In addition, I would like to give my hearty thanks to professors, colleagues,
schoolmates and all teaching staff at the Jockey Club School of Public Health and
Primary Care, the Chinese University of Hong Kong, for their great support and help.
Lastly, but most importantly, my greatest gratitude goes to my beloved family
for their loving consideration and great confidence in me throughout the years. I thank
my beloved wife, Lifen Feng, for her love and support. My deepest gratitude also goes
to my parents and parents-in-law for loving and caring for me. My beloved daughter,
Yuxi Li, who came into my life during my Ph.D. study period, is the sunshine in my
life. We will always grow together.
ix
Table of Contents
Abstract (English) ......................................................................................................... i

Abstract (Chinese) ....................................................................................................... v
Acknowledgement .....................................................................................................viii
Table of Contents ......................................................................................................... x
List of Tables .............................................................................................................xvii
List of Figures ............................................................................................................. xx
Abbreviations ..........................................................................................................xxiii
Chapter 1 Introduction ................................................................................................ 1
1.1 Background of the study ...................................................................................... 1
1.1.1 Prevalence and adverse consequences of depression among adolescents .... 1
1.1.2 Increased high penetration of Internet and social networking use ................ 2
1.1.3 Prevalent Internet addictive behaviors .......................................................... 2
1.1.4 Online social networking use as a new threat for adolescents ...................... 4
1.1.5 Knowledge gaps related to Internet and social networking use among
adolescents .................................................................................................... 5
1.2 Overall study design .......................................................................................... 10
1.3 Aims ................................................................................................................... 10
1.4 Structure of the thesis......................................................................................... 12
Chapter 2 Literature review ...................................................................................... 16
2.1 Depression epidemic among adolescents ........................................................... 16
2.1.1 High disease burden caused by depression ................................................. 16
2.1.2 Depression epidemic among adolescents worldwide ................................. 17
2.1.3 Depression epidemic among adolescents in China ..................................... 18
2.2 Internet use worldwide and in China ................................................................. 18
2.3 Online social networking use among adolescents ............................................. 20
2.3.1 Online social networking definition............................................................ 20
2.3.2 Online social networking functions/activities ............................................. 21
2.3.3 High prevalent online social networking use among adolescents .............. 23
2.4 Internet addiction, social networking addiction and depression ........................ 24
2.4.1 Concept and classification of Internet addictive behaviors ........................ 24
x
2.4.2 Internet addiction epidemic......................................................................... 25
2.4.3 Social networking addiction epidemic ........................................................ 25
2.4.4 Relationship between IA and depression .................................................... 26
2.4.5 Relationship between ‘SNA’ and depression .............................................. 29
2.5 Impacts of social networking use on adolescents’ psychosocial status and
depression ........................................................................................................... 30
2.5.1 Assessment for social networking use intensity.......................................... 30
2.5.2 Association between social networking use and depression ....................... 35
2.5.3 Associations between social networking use and psychosocial status........ 38
2.6 Associations between social networking use and Internet addictive behaviors . 41
2.7 A summary of research gaps found in the literatures ......................................... 42
2.8 Statistical methodological considerations for longitudinal study ...................... 43
2.8.1 Causal direction testing in the longitudinal data ......................................... 43
2.8.2 Mediational effects and modeling specification ......................................... 46
2.8.3 Moderation effects and statistical strategies ............................................... 52
Chapter 3 Participants and methods ........................................................................ 53
3.1 Study design ....................................................................................................... 53
3.2 Study site ............................................................................................................ 54
3.3 A summary of rationale to traget junior middle school students........................ 56
3.4 Phase I: Validation study .................................................................................... 57
3.4.1 Participants and sampling ........................................................................... 57
3.4.2 Data collection procedure ........................................................................... 58
3.4.3 Measurement ............................................................................................... 60
3.4.4 Statistical Analysis ...................................................................................... 63
3.5 Phase II: A nine-month longitudinal study......................................................... 64
3.5.1 Participants and Sampling........................................................................... 64
3.5.2 Data collection ............................................................................................ 66
3.5.3 Measurements ............................................................................................. 68
3.5.4 Quality control ............................................................................................ 72
3.5.5 Statistical strategies ..................................................................................... 73
Chapter 4 Description of the independent and dependent variables in the
longitudinal sample .................................................................................................... 75
4.1 Objectives .......................................................................................................... 75
xi
4.2 Statistical methods ............................................................................................. 75
4.3 Results ................................................................................................................ 76
4.3.1 Background characteristics of the entire sample at baseline assessment .... 76
4.3.2 Attrition analysis for loss to follow-up ....................................................... 77
4.3.3 The longitudinal sample (n=4237) .............................................................. 77
4.4 Discussions ........................................................................................................ 85
4.4.1 Internet addiction and online social networking among adolescents.......... 85
4.4.2 High prevalence and incidence of depression among adolescents ............. 88
4.4.3 Natural remission from depression, ‘SNA’, and IA status .......................... 89
Chapter 5 Directionality for the causal relationships between Internet/social
networking addiction and depressive symptoms................................................... 107
5.1 Background ...................................................................................................... 107
5.1.1 Relationship between IA and depression .................................................. 107
5.1.2 Relationship between ‘SNA’ and depression ............................................ 107
5.1.3 Knowledge gaps ........................................................................................ 108
5.2 Objectives ........................................................................................................ 109
5.3 Statistical analysis ............................................................................................ 111
5.3.1 Subsamples ............................................................................................... 111
5.3.2 Modeling methods .................................................................................... 112
5.4 Results .............................................................................................................. 115
5.4.1 Causal relationship between ‘SNA’ and depression.................................. 115
5.4.2 Causal relationship between IA and depression........................................ 118
5.5 Summary .......................................................................................................... 120
5.5.1 Causal relationship between ‘SNA’ and depression.................................. 120
5.5.2 Causal relationship between IA and depression........................................ 121
5.6 Discussion ........................................................................................................ 122
5.6.1 Depression as predictor of incident ‘SNA’/IA .......................................... 122
5.6.2 ‘SNA’/IA as predictors of incident depression.......................................... 124
5.6.3 The bidirectional relationship ................................................................... 125
5.6.4 Implications on prevention ....................................................................... 125
5.6.5 Implications for future research ................................................................ 127
5.6.6 Strength and limitation.............................................................................. 128
Chapter 6 Validation of the Social Networking Activity Intensity Scale............. 140
xii
6.1 Background ...................................................................................................... 140
6.2 Objectives ........................................................................................................ 142
6.3 Sample and Statistical method ......................................................................... 142
6.3.1 Sample used in this chapter....................................................................... 142
6.3.2 Statistical method ...................................................................................... 143
6.4 Results .............................................................................................................. 144
6.4.1 Social-demographic characteristics .......................................................... 144
6.4.2 Factor structures ........................................................................................ 144
6.4.3 Ceiling and floor effects of SNAIS and its two subscales ........................ 145
6.4.4 Item analysis ............................................................................................. 145
6.4.5 Reliability.................................................................................................. 145
6.4.6 Correlation between SNAIS, EC, ‘SNA’, and IA ..................................... 145
6.4.7 Relationship between SNAIS and participants’ characteristics of social
networking use.......................................................................................... 146
6.5 Discussion ........................................................................................................ 146
Chapter 7 Intensity and emotional connection of social networking use as
predictors of incident depression ............................................................................ 153
7.1 Background ...................................................................................................... 153
7.2 Objectives ........................................................................................................ 155
7.3 Statistical methods ........................................................................................... 156
7.4 Results .............................................................................................................. 156
7.4.1 Background factors associated with incident depression.......................... 156
7.4.2 Prediction of baseline SFUI, EFUI, and EC score on incident depression
.................................................................................................................. 157
7.4.3 Moderation effects of EC on the relationships between SFUI, EFUI and
incident depression ................................................................................... 157
7.5 Discussion ........................................................................................................ 158
Chapter 8 Mechanisms underlying the relationship between social networking
use intensity and depressive symptoms: the roles of psychosocial factors.......... 164
8.1 Background ...................................................................................................... 164
8.2 Objectives ........................................................................................................ 165
8.3 Part I: Testing mediation and suppression effects ............................................ 166
8.3.1 Some examples of mediation and suppression effect for the relationships
between Internet use and mental health outcomes ................................... 166
xiii
8.3.2 Mediation and suppression hypotheses for the associations between change
in social networking use intensity and change in depressive symptoms .. 167
8.3.3 Implications of testing the mediation and suppression hypotheses .......... 178
8.3.4 Statistical analysis for mediation and suppression effects ........................ 179
8.3.5 Results of mediation and suppression effects ........................................... 182
8.3.6 Summary for the mediation and suppression effects ................................ 189
8.4 Part II: Testing moderation effects ................................................................... 190
8.4.1 Moderation effects of psychosocial factors on the association between social
networking use intensity and depressive symptoms ................................. 190
8.4.2 Statistical methods for testing moderation effects .................................... 199
8.4.3 Results ....................................................................................................... 200
8.5 Discussion ........................................................................................................ 202
8.5.1 Associations of social networking use intensity on psychosocial factors . 202
8.5.2 Mediation effects of change in social non-confidence and adolescent-mother
conflict ...................................................................................................... 205
8.5.3 Suppression effect of change in perceived social support and friendship
quality ....................................................................................................... 206
8.5.4 Implications............................................................................................... 207
8.5.5 Moderation effects .................................................................................... 208
8.5.6 Limitations ................................................................................................ 208
Chapter 9 Relationship between social networking use intensity, addiction,
psychosocial factors and depression ....................................................................... 222
9.1 Background ...................................................................................................... 222
9.2 Objectives ........................................................................................................ 225
9.3 Statistical methods ........................................................................................... 227
9.3.1 Linear regression analysis ......................................................................... 227
9.3.2 Path analysis for indirect effect model ...................................................... 227
9.4 Results .............................................................................................................. 228
9.4.1 Associations between ΔSFUI/ΔEFUI and ΔIA/ΔSNA ............................. 228
9.4.2 Associations between ΔIA, ΔSNA and ΔCES-D....................................... 229
9.4.3 Mediation effects via ΔIA and ΔSNA for the association between ΔSFUI and
ΔCES-D .................................................................................................... 229
9.4.4 Mediation effects via ΔIA and ΔSNA for the association between ΔEFUI and
ΔCES-D .................................................................................................... 231
xiv
9.4.5 Path analysis for indirect effect model via ΔIA and ΔSNA ...................... 234
9.4.6 Combined mediation effects of change in addiction and psychosocial factors
altogether on the associations between ΔSFUI (or ΔEFUI) and ΔCES-D 235
9.4.7 Summary for mediation effects of all mediation models .......................... 238
9.5 Discussion ........................................................................................................ 239
9.5.1 Associations between change in social networking use intensity and change
in Internet addictive behaviors ................................................................. 239
9.5.2 Associations between change in ‘SNA’, change in IA and change in
depressive symptoms ................................................................................ 241
9.5.3 Mediation effects for the association between change in social networking
use intensity and change in depressive symptoms via change in ‘SNA’ and
change in IA ............................................................................................. 242
9.5.4 Path model for indirect effects via ‘SNA’ and IA ..................................... 243
9.5.5 Combined mediation effects via Internet addictive behaviors and
psychosocial factors altogether................................................................. 244
9.5.6 Limitations ................................................................................................ 245
9.5.7 Conclusion ................................................................................................ 246
Chapter 10 Conclusions ........................................................................................... 256
10.1 Summary and interpretation of key findings ................................................. 256
10.1.1 Depression and Internet addictive behaviors epidemic among adolescents
.................................................................................................................. 256
10.1.2 Causal directions between Internet addictive behaviors and depression 257
10.1.3 A validated scale for assessing social networking use intensity ............. 258
10.1.4 The impact of social networking use intensity on depression ................ 258
10.1.5 Impact of social networking use intensity on psychosocial status .......... 259
10.1.6 Underlying mechanisms for the relationship between social networking use
intensity and depression ........................................................................... 259
10.2 Implications of the study ................................................................................ 261
10.2.1 Intervention programs for reducing depressive symptoms ..................... 261
10.2.2 Intervention programs for preventing Internet addictive behaviors ....... 262
10.2.3 Collaboration and integration of Internet addictive behaviors and
depression intervention ............................................................................. 263
10.2.4 Assessment for social networking use intensity...................................... 264
10.2.5 Methodological considerations in longitudinal causal-effect models ..... 265
10.3 Limitations of the study ................................................................................. 266
xv
10.3.1 Study design ............................................................................................ 266
10.3.2 Sampling ................................................................................................. 267
10.3.3 Causal inference for the relationships between ’SNA’/IA and depression
.................................................................................................................. 267
10.3.4 Measures and reporting bias ................................................................... 268
10.4 Recommendations for future research ........................................................... 270
10.4.1 Longitudinal study with three or more observations .............................. 270
10.4.2 Mining online social networking content ............................................... 271
10.4.3 Research on benefits of proper online social networking use................. 272
10.4.4 Considerations for other measures in future direction ............................ 272
10.4.5 Moderation effects of gender and personality for the effects of online social
networking use intensity on psychosocial and mental health among
adolescents ................................................................................................ 273
10.4.6 Intervention studies using social networking as an innovative avenue .. 273
10.5 Overall remarks .............................................................................................. 274
Reference .................................................................................................................. 276
Appendix I ................................................................................................................ 296
Appendix II ............................................................................................................... 302
xvi
List of Tables
Table 2-1 Instruments measuring online social networking use .................................. 34

Table 2-2 Summary of the literature review for the association between social
networking use and depression .................................................................... 37
Table 3-1 Psychometric properties of the scales used in the longitudinal study.......... 74
Table 4-1 Background characteristics of the participants of the entire sample at baseline
(n=5365) ....................................................................................................... 91
Table 4-2 Attrition analysis for loss to follow-up ........................................................ 92
Table 4-3 Background characteristics of participants among the longitudinal sample
and the subsamples ...................................................................................... 93
Table 4-4 Comparision of background characteristics between those with SNA and
those without SNA at baseline (n=4237) ..................................................... 94
Table 4-5 Comparision of background characteristics between those with IA and those
without IA at baseline (n=4237) .................................................................. 95
Table 4-6 Comparision for the mean differences of baseline IA and SNA scores between
four groups (n=4237) ................................................................................... 96
Table 4-7 Prevalence of depression, ‘SNA’, and IA at baseline and follow-up by gender
in the longitudinal sample (n=4237) ............................................................ 97
Table 4-8 Age- and gender-specific incidence of depression among participants who
were non-depressed at baseline (n=3196) .................................................... 98
Table 4-9 Age- and gender-specific incidence of ‘SNA’ among participants who were
classified as non-‘SNA’ at baseline (n=3657) .............................................. 99
Table 4-10 Age- and gender-specific incidence of IA among participants who were
classified as non-IA at baseline (n=3959) .................................................. 100
Table 4-11 Age- and gender-specific remission from depression among participants
who were depressed at baseline (n=1041) ................................................. 101
Table 4-12 Age- and gender-specific remission from ‘SNA’ among participants who
were classified as ‘SNA’ at baseline (n=580)............................................. 102
Table 4-13 Age- and gender-specific remission from IA among participants who were
classified as IA at baseline (n=248) ........................................................... 103
Table 4-14 Frequency distribution of change in status of ‘SNA’, IA, and depression in
subsample Ia~IIIa ....................................................................................... 104
Table 4-15 Score changes of social networking related scales among longitudinal
sample (n=4237) ........................................................................................ 105
Table 4-16 Score difference of interpersonal and psychosocial scales between baseline
and follow-up among longitudinal sample (n=4237)................................. 106
Table 5-1 Background factors as predictors of new incident depression among
participants who were non-depressed at baseline (n=3196) ...................... 131
Table 5-2 Background factors as predictors of new incident ‘SNA’ among participants
who were non-‘SNA’ at baseline (n=3657) ................................................ 132
xvii
Table 5-3 Predictive effect of baseline ‘SNA’ on incident depression among participants
who were non-depressed at baseline (n=3196) .......................................... 133
Table 5-4 Predictive effect of baseline depression on incident ‘SNA’ among participants
who were non-‘SNA’ at baseline (n=3657) ................................................ 133
Table 5-5 Predictive effect of change in ‘SNA’ status over time on new incident
depression among participants who were non-depressed at baseline (n=3196)
.................................................................................................................... 134
Table 5-6 Predictive effect of change in depression status over time on new incident
‘SNA’ among participants who were non-SNA at baseline (n=3657) ....... 134
Table 5-7 Background factors as predictors of new incidence IA among participants
who were non-IA at baseline (n=3959) ...................................................... 135
Table 5-8 predictive effect of baseline IA status on new incident depression among
participants who were non-depressed at baseline (n=3196) ...................... 136
Table 5-9 predictive effect of baseline depression on incident IA among participants
who were non-IA at baseline (n=3959) ...................................................... 136
Table 5-10 Predictive effect of change in IA status over time on incident depression
among participants who were non-depressed at baseline (n=3196) .......... 137
Table 5-11 Predictive effect of change in depression over time on incident IA among
participants who were non-IA at baseline (n=3959) .................................. 137
Table 5-12 Summary of the causal relationship between ‘SNA’ and depression....... 138
Table 5-13 Summary of the causal direction between IA and depression ................. 139
Table 6-1 Factor loadings and path loading estimated by exploratory and confirmatory
factor analyses for the Social Networking Activity Intensity Scale (SNAIS)
.................................................................................................................... 149
Table 6-2 Item analysis of the SNAIS (n=910).......................................................... 150
Table 6-3 Correlations among SNAIS, emotional connection, social networking
addiction and Internet addiction (n=910) ................................................... 151
Table 6-4 Mean score differences of SNAIS, SFUI, and EFUI by groups (n=910) .. 152
Table 7-1 Predictive effects of baseline SFUI, EFUI, and EC scores on incident
depression by multilevel logistic regression (n=3196) .............................. 162
Table 7-2 Testing the moderation effects of emotional connection on the relationships
between SFUI, EFUI, and incident depression (n=3196) .......................... 163
Table 8-1 Summary of the mediation/suppression hypotheses to be tested and the
rationales based on the literature................................................................ 168
Table 8-2 Background variables associated with change in depressive symptoms in
longitudinal sample (n=4237) .................................................................... 210
Table 8-3 Associations of change in SFUI and EFUI on change in depressive symptoms
by multilevel linear regression models (n=4237) ...................................... 211
Table 8-4 Associations between change in SFUI, EFUI and change in psychosocial
variables by multilevel linear regression (n=4237) ................................... 212
Table 8-5 Association of change in psychosocial scale scores on change in depressive
symptoms by multilevel linear regression (n=4237) ................................. 213
Table 8-6 Testing the mediation effects of change in psychosocial scale scores on the
xviii
association between △SFUI and △CES-D by multilevel linear regression
(n=4237) ..................................................................................................... 214
Table 8-7 Summary for the mediation effects of change in psychosocial scale scores on
the association between △EFUI and △CES-D by multilevel linear regression
(n=4237) ..................................................................................................... 215
Table 8-8 Testing the suppression effects of change in psychosocial variables on the
association between △SFUI and △CES-D (n=4237) ................................. 216
Table 8-9 Testing the suppression effects of change in psychosocial scale scores on the
association between △EFUI and △CES-D by multilevel linear regression
(n=4237) ..................................................................................................... 217
Table 8-10 Summary of interaction effects of change in psychosocial variables on the
association between △SFUI and △CES-D (n=4237) ................................. 218
Table 8-11 Summary of interaction effects of change in psychosocial variables on the
association between △EFUI and △CES-D (n=4237)................................. 220
Table 9-1 Associations of ΔSFUI and ΔEFUI on ΔIA/ΔSNA among longitudinal
sample (n=4237) ........................................................................................ 248
Table 9-2 Associations between change in ‘SNA’, change in IA and change in CES-D
among longitudinal sample (n=4237) ........................................................ 249
Table 9-3 Hierarchical regression models for testing mediation effects of ΔIA and
ΔSNA on the association between ΔSFUI and ΔCES-D among longitudinal
sample (n=4237) ........................................................................................ 250
Table 9-4 Hierarchical regression models for testing mediation effects of ΔIA and
ΔSNA on the association between ΔEFUI and ΔCES-D among longitudinal
sample (n=4237) ........................................................................................ 251
Table 9-5 Summary for all significant mediators in single mediator models for the
association between change in SFUI (or change in EFUI) and change in
CES-D ........................................................................................................ 252
Table 9-6 Multivariate linear regression models for testing the joint mediation effects of
addiction and psychosocial factors on the associations between
ΔSFUI/ΔEFUI and ΔCES-D among longitudinal sample (n=4237).......... 253
Table 9-7 Summary of total, direct, and combined mediation effects of multiple
mediator models on change in depressive symptoms (n=4237) ................ 254
Table 9-8 Summary for mediation effects (% mediated) of all mediation models in
Chapter 8 and Chapter 9 ............................................................................ 255
xix
List of Figures
Figure 2-1 Top 10 causes of DALYs lost among adolescents by sex .......................... 17
Figure 2-2 Growth in Internet usage from 2005 to 2014 worldwide ........................... 19
Figure 2-3 Trend of Internet penetration among adolescents in China ........................ 19
Figure 2-4 Main history of online social networking development ............................. 21
Figure 2-5 Illustration of mediation/suppression effects ............................................. 48
Figure 2-6 Statistical strategy of mediation/suppression hypothesis testing based on
Baron and Kenny’s method .......................................................................... 50
Figure 2-7 Illustration of moderating effect ................................................................. 52
Figure 3-1 Study design flowchart ............................................................................... 54
Figure 3-2 Internet penetration rate in Guangzhou, Guangdong and national level .... 55
Figure 3-3 Location of the study site ........................................................................... 55
Figure 3-4 Flowchart of phase I data collection .......................................................... 59
Figure 3-5 Flowchart of participants in the longitudinal study.................................... 66
Figure 4-1 Illustration of the subsamples..................................................................... 78
Figure 5-1 Causal direction between ‘SNA’ and depression using approach I .......... 113
Figure 5-2 Causal direction between IA and depression using approach I ................ 113
Figure 5-3 Causal direction between ‘SNA’ and depression using approach II......... 114
Figure 5-4 Causal direction between IA and depression using approach II .............. 115
Figure 8-1 Hypothesis illustration for suppression effect of change in friendship quality
on the association between change in SFUI/EFUI and change in CES-D 169
Figure 8-2 Hypothesis illustration for the suppression of change in perceived social
support on the association between change in SFUI/EFUI and change in
CES-D ........................................................................................................ 171
support on the association between change in SFUI/EFUI and change in
CES-D ........................................................................................................ 171
Figure 8-4 Hypothesis illustration for the mediation of change in adolescent-father
conflict on the association between change in SFUI/EFUI and change in
CES-D ........................................................................................................ 173
Figure 8-5 Hypothesis illustration for the mediation of change in adolescent-mother
conflict on the association between change in SFUI/EFUI and change in
CES-D ........................................................................................................ 173
Figure 8-6 Hypothesis illustration for the mediation of change in social non-confidence
on the association between change in SFUI/EFUI and change in CES-D . 175
Figure 8-7 Hypothesis illustration for the suppression of change in self-esteem on the
association between change in SFUI/EFUI and change in CES-D............ 177
Figure 8-8 Hypothesis illustration for the mediation of change in self-esteem on the
association between change in SFUI/EFUI and change in CES-D............ 177
Figure 8-9 Illustrations for conceptual framework of change in psychosocial factors for
xx
the associations between △SFUI/△EFUI and △CES-D ............................ 181
Figure 8-10 Mediation effect of change in adolescent-mother conflict on the association
between change in SFUI and change in CES-D ........................................ 183
Figure 8-11 Mediation effect of change in social non-confidence on the association
Figure 8-12 Illustration for the mediation effect of △Social non-confidence on the
association between △EFUI and △CES-D ................................................ 185
Figure 8-13 Suppression effect of change in friendship quality on the association
Figure 8-14 Suppression effect of change in perceived family support on the association
Figure 8-15 Suppression effect of change in perceived friend support on the association
Figure 8-16 Suppression effect of change in perceived family support on the association
between change in EFUI and change in CES-D ........................................ 189
Figure 8-17 Suppression effect of change in perceived friend support on the association
Figure 8-18 Illustration of the poor-get-poorer hypothesis ........................................ 193
Figure 8-19 Illustration for modified poorer-get-poorer and buffering effect hypothesis
.................................................................................................................... 195
Figure 8-20 Hierarchical regression models to test the interaction effects between
change in SFUI (or change in EFUI) and change in psychosocial variables on
change in depressive symptoms ................................................................. 200
Figure 8-21 Graphical illustration for the interactions between change in SFUI/EFUI
and change in conflict with father on change in depressive symptoms3 ... 202
Figure 9-1 The hypothesized path model for indirect effects between △SFUI, △EFUI
and △CES-D via △SNA and △IA. △SFUI: Change in Social Function Use
Intensity; △EFUI: Change in Entertainment Function Use Intensity; △SNA:
Change in Social Networking Addiction; △IA: Change in Internet Addiction;
△CES-D: Change in depressive symptoms measured by Center for
Epidemiological Study Scale-Depression. “+” indicates a positive
association between two variables ............................................................. 226
Figure 9-2 Illustration for mediation effect of change in IA on the association between
change in SFUI and change in CES-D ....................................................... 230
Figure 9-3 Illustration for mediation effect of change in ‘SNA’ on the association
Figure 9-4 Illustration for mediation effect of change in ‘SNA’ and IA altogether on the
association between change in SFUI and change in CES-D...................... 231
change in EFUI and change in CES-D ...................................................... 232
Figure 9-7 Illustration for mediation effect of change in ‘SNA’ and IA altogether on the
xxi
association between change in EFUI and change in CES-D ..................... 234
Figure 9-8 Path model for indirect effects between △SFUI, △EFUI and ΔCES-D via
△SNA and △IA. Paths shown in dashed lines represent non-significant
pathway at the 0.05 level. Standardized path coefficients (β) were presented
in the model. Model fit indexes: χ2 (2) =2.595 with p=0.273, CFI=1.000,
IFI=1.000 and RMSEA=0.008................................................................... 235
Figure 9-9 Illustration of combined mediation effect via change in IA, SNA, social
non-confidence and adolescent-mother conflict on the association between
change in SFUI and change in CES-D ....................................................... 237
Figure 9-10 Illustration of joint mediation effect via change in IA, SNA and social
non-confidence on the association between change in SFUI and change in
CES-D ........................................................................................................ 238
xxii
Abbreviations
CES-D Center for Epidemiological Studies-Depression scale

CDI Children Depression Inventory
IA Internet Addiction
SNA Social Networking Addiction
SNUI Social Networking Use Intensity
SNAIS Social Networking Activity Intensity Scale
SFUI Social Function Use Intensity subscale
EFUI Entertainment Function Use Intensity subscale
EC Emotional Connection (to social networking)
DALYs Disability Adjusted Life Years
MDD Major Depressive Disorder
BDI-II Beck Depression Inventory-version II
GPIUS2 Generalized Problematic Internet Use Scale 2
SEM Structural Equation Model
ANOVA Analysis of variance
ANCOVA Analysis of Covariate
FIS Facebook Intensity Scale
SMUIS Social Media Use Integration Scale
EFA Exploratory Factor Analysis
CFA Confirmatory Factor Analysis
OR Odds Ratio
95% CI 95% Confidence Interval
SE Standardized Error
-2LL -2*Log (Likelihood)
df Degree of Freedom
NNFI Non-Normed Fit Index
CFI Comparative Fit Index
RMSEA Root Mean Square Error of Approximation
xxiii
Chapter 1 Introduction
1.1 Background of the study
1.1.1 Prevalence and adverse consequences of depression among adolescents
As a key developmental phase transitioning from childhood to adulthood, the
rapid changes in biological (e.g. puberty and sexual development), psychological (e.g.
personal identity development), and social development (e.g. intimate relationship
development and social autonomy) affect every aspect of adolescents’ development [1].
During this period, adolescents are at high-risk of suffering from diverse
psychological problems, especially along with increased penetration of Internet use [2].
Mental illness, depression in particular, is a worldwide challenge in public health
among adolescents. Due to the onset of puberty, depression is the most widely
reported psychiatric disorder among adolescents [3, 4], and it has been observed that the
risk of depression substantially increased upon entering into adolescence [5].
Depression, a major public health concern globally, is the leading cause of illness and
disability. Depression is associated with a variety of physical and psychological
problems [6], e.g. substance use, disordered eating, and suicidal ideation and attempts.
A high prevalence of depression has been reported among adolescents. Over 9% of
adolescents reported moderate/severe depression (CES-D≥24), and a one-year
incident rate of moderate/severe depression was estimated at 3% in U.S. [7] . The
prevalence of mild, moderate and severe depression (measured by the Children’s
Depression Inventory, CDI) was 14.6%, 15.3% and 12.4% respectively among middle
school students in Mainland China [8].
1
1.1.2 Increased high penetration of Internet and social networking use
Personal Internet use has increased dramatically over the past two decades.
Globally, the Internet penetration exponentially grew seven-fold from 6.5% to 43%
between 2000 and 2015 [9]. Internet penetration in China also significantly increased
from 6.5% in 2002 to 48.8% in 2015 [10, 11]. Adolescents have become one of the
fastest growing segments of Internet users. In China, Internet penetration among
adolescents sharply increased to 79.6% in 2014 [12].
Online social networking as a relatively new phenomenon has gained substantial
popularity among adolescents in recent years. This trend is mainly due to the rapid
advances in information and communication technology, which contribute to easy
accessibility of the Internet. Online social networking is defined as web-based
platforms that allow individuals to create their own personal profile and build a
network of connections with other users [13]. Eight in ten of U.S. teenage Internet users
were also social networking users in 2011 [14]. In Europe, it is estimated that 49% of
adolescents aged 11 to 12 years old, 73% of adolescents aged 13 to 14 years old, and
82% of adolescents aged 15 to 16 years old had their own online social networking
accounts in 2011 [15]. Prevalence data relating to social networking use among
adolescents is limited in China. One small survey reported that almost all sampled
college students (98%) are social networking users [16]. The high popularity of social
networking use among adolescents signifies that they have integrated such use into
their daily activities.
1.1.3 Prevalent Internet addictive behaviors
Along with high penetration and an increasing trend of excessive Internet use, IA
has raised as an important and noticeable public health challenge among
2
adolescents [17]. IA is defined as excessive Internet use that interferes with users’ daily
life, such as preoccupation and feeling restless or irritable when attempting to cut
back [18, 19]. As a condition of inappropriate and excessive Internet use, IA has
considerable negative influence on adolescents’ physical and psychological
development, leading to negative consequences (e.g. severe distress, depression) and
functional impairment [17, 20]. Davis categorizes IA into generalized and a more
specific forms [21], and such classification has been supported by other researchers [22,
23]
. It has highlighted the significance and necessity for distinguishing addiction to
specific Internet activities from IA in general [24]. ‘Social Networking Addiction
(SNA)’ in specific has been recently raised as a public health concern along with
excessive social networking involvement among adolescents. Research has revealed
that those people who are addicted to social networking experience symptoms similar
to those who suffer from addictions to substances and other behaviors [25]. ‘Social
Networking Addiction (SNA)’ is defined as “being overly concerned about social
networking use, to be driven by a strong motivation to log on to or use social
networking, and to devote much time and effort to social networking that impairs
other social activities, studies/jobs, interpersonal relationships, and/or psychological
health and well-being” [26]. Give that ‘SNA’ is a relatively new concern, and its
assessment and diagnosis is still in the way of development, we therefore put it in
quotation mark as an emphasis (i.e. ‘SNA’).
In recent years, a high prevalence of generalized IA has been reported globally.
The IA prevalence among adolescents ranged widely from 0.8% in Italy to 26.7% in
Hong Kong, due to different instruments and cut-off values used [27, 28]. The IA
prevalence among Chinese adolescents was up to 26.0% in 2011, while the proportion
3
of IA among adolescents aged 13-17 was even greater at 30.5%. Current studies
related to ‘SNA’ prevalence in specific usually involved small and non-representative
student samples by using various screening methods. Facebook studies reported an
‘SNA’ (Facebook addiction) prevalence of 1.6% in Nigerian undergraduate
students [29] and 8.6% in Peruvian undergraduate students [30]. The high prevalence of
‘SNA’ was also observed in Chinese college students by modified Young’s Internet
Addiction Test (24.0%~34.3%) [31, 32].
Internet addictive behaviors, IA in general and ‘SNA’ in specific, are strongly
associated with multiple physical, psychosocial and mental health problems among
adolescents, such as obesity, anxiety, substance use and depression [33, 34]. Therefore,
Internet addictive behaviors should be given priority in public health prevention
among adolescents.
1.1.4 Online social networking use as a new threat for adolescents
Online social networking use has profoundly changed the ways that people
communicate and interact with each other. Furthermore, online social networking
applications have brought about tremendous changes in terms of connectivity,
self-expression and daily activities in adolescents’ daily life. Some researchers have
associated online social networking use with several psychiatric disorders, such as
depressive symptoms, anxiety, and low self-esteem [35-37]. However, since online
social networking is a relatively new phenomenon, many questions regarding their
potential impact on psychosocial and mental health remain unanswered. Any future
research that confirms the connection between social networking use and psychiatric
problems would illustrate a serious public health concern.
4
1.1.5 Knowledge gaps related to Internet and social networking use among
adolescents
1.1.5.1 Unclear causal directions between Internet addictive behaviors (IA, SNA)
and depression
It would be beneficial for intervention programs if the causal directions and
magnitude of the relationships between Internet addictive behaviors and depression
among adolescents were clearly defined. However, most of the existing studies that
have investigated the association between IA and depression are cross-sectional in
nature. Due to limitations of the cross-sectional study design, the results cannot
provide definitive evidence for the causal inference between IA and depression. In
contrast, longitudinal studies are advantageous for detecting development or changes
in the characteristics of the target population over time.
Our literature review only found eight longitudinal studies that investigated the
longitudinal relationship between IA and depression among adolescents. Four of the
studies were conducted in Taiwan [38-41], two in Mainland China [42, 43], one in South
Korea [44] and one in Spain [45]. Of these eight longitudinal studies, four studies tested
the prediction of depression on IA, three studies tested the prediction of IA on
depression, and only one study tested the bidirectional relationship between IA and
depression. Unfortunately, these longitudinal studies yielded mixed results, and the
causal direction between IA and depression is still inconclusive. It may be
hypothetical that IA and depression among adolescents is likely to be bi-directional:
depressed adolescents might indulge in excessive Internet use as a means of
alleviating their negative mood, while the negative consequences of such excessive
Internet use (e.g. social isolation) would further exacerbate their level of depressive
5
symptoms [28, 46].
There are fewer studies that examined the causal relationship between ‘SNA’ in
specific and depression than there are for IA in general. Research related to ‘SNA’ and
depression is very scarce. Our literature review found only a few of cross-sectional
surveys that investigated the association between ‘SNA’ and depression among
adolescents [47, 48]. No longitudinal studies were found that explored the relationship
between ‘SNA’ and depression among adolescents.
Therefore, there is a dearth of longitudinal studies that test bi-directional
relationships between ‘SNA’/IA and depression among adolescents in the same study.
It is still unclear whether ‘SNA’/IA are causes or outcomes of depression, and
large-scale longitudinal studies are warranted for clarifying longitudinal relationships
between ‘SNA’/IA and depression, especially in Mainland China.
1.1.5.2 Lack of longitudinal studies related to social networking use
(1) Validated instrument for assessing social networking use intensity
In order to sufficiently assess levels of social networking use intensity and its
impact on adolescents’ psychosocial and mental health outcomes (i.e. depression),
fully validated instruments for such assessment are a prerequisite. To date, there have
been no satisfactory measures to estimate social networking use intensity. Most of the
existing studies measured social networking behaviors only by simple items, such as
weekly frequency of use and daily time spent on social networking use [36, 49, 50]. These
simple questions usually perform poorly, are unreliable, and also cannot adequately
cover the full spectrum of complex activities related to social networking use.
Actually, adolescents might conduct diverse activities for different purposes on online
6
social networking platforms.
There are a few existing scales (mainly focusing on Facebook, e.g. Facebook
Intensity Scale [51]) aiming to assess the social networking use intensity, but they are
not fully validated and psychometrically weak in measuring social networking use
intensity. The existing scales are also potentially not suitable in Chinese cultural
context. Therefore, there is a lack of acceptable instruments to assess social
networking use intensity that take diverse social networking activities and functions
into account.
(2) Impacts of social networking use intensity on depression
Currently, the health impact of social networking use among adolescents is still
controversial. Some studies have associated social networking use with a high level of
depressive symptoms [35, 37], while other studies found no association between social
networking use and depression [52, 53]. All of these studies are based on a
cross-sectional design and were conducted in western countries (e.g. the United
States). Again, longitudinal studies are highly warranted. The present study is
expected to fill these knowledge gaps.
(3) Mechanisms underlying the relationship between social networking use
intensity and depressive symptoms
It is also important to understand the mechanisms underlying the relationship
between social networking use intensity and depression (e.g. potential mediation,
moderation, or suppression effect). Knowing this information would be helpful for
designing effective intervention programs to reduce the impact of social networking
use on depressive symptoms among adolescents. However, there is still a lack of
7
clarity on how and why social networking use affects the levels of depressive
symptoms among adolescents.
Psychosocial factors
Given that adolescents use online social networking mainly for social interaction
and communication purposes (e.g. seeking friends and social support, maintaining and
establishing friendship) [54, 55], it is possible that social networking use intensity may
have significant impacts on adolescents’ psychosocial status. Friends-related status
(friendship quality, perceived social support), family-related status (adolescent-parents
conflict), and intrapersonal-related status (social non-confidence and self-esteem) may
all be impacted by social networking use intensity. Some cross-sectional studies have
investigated the associations between social networking use and some of these
psychosocial factors among adolescents [36, 56-58]. First, it is established that
adolescents’ offline friends largely overlap with their online social networking friends,
and they mainly use social networking to connect with people they known from an
offline context [59]. It is generally believed that social networking use is potentially
beneficial for a sense of friendship quality and perception of social support [16, 60].
Second, adolescents’ preoccupation in social networking use might interfere with their
daily family activities, and causes less time interacting with family members and
greater conflict with parents [58, 61]. However, some parents may use social networking
to maintain closer contact with their children. The impact of social networking use on
adolescent-parents relationship remains unclear. Third, intensive online social
networking use inevitably decreases offline social activities. Due to insufficient
offline interactions, confidence in social occasions may be weakened. Although social
confidence is important for adolescents’ development, there are presently no studies
8
that have investigated the influence of social networking use on social confidence as a
construct among adolescents. Fourth, self-esteem as an important protective factor for
adolescents’ health development has also been researched in a social networking
context among adolescents. However, the results are mixed, with some studies
reporting a positive association between social networking use and self-esteem [62, 63]
and other studies reporting a negative associations [36, 64]. Therefore, the impact of
social networking use intensity on psychosocial factors among adolescents needs to be
further investigated, especially in longitudinal studies.
Moreover, the protective effects of high level of friendship quality, high level of
perceived social support, and high level of self-esteem on depression, in addition to
the harmful effects of social non-confidence and adolescent-parents conflicts on
depression, have been well researched and documented in previous studies among
adolescents. Therefore, the aforementioned psychosocial factors may have the ability
to partially explain how and why social networking use intensity impacts depression
in adolescents. However, there is a lack of longitudinal studies to explore the roles of
these psychosocial factors on the relationship between social networking use intensity
and depressive symptoms among adolescents, i.e. mediation, suppression or
moderation effects.
Internet addictive behaviors
Higher intensity of social networking use has significantly associated with
increased addictive tendency to Internet/social networking use [48, 65], and such
addictive use, in turn, causes adolescents’ mental health problems (e.g. depression).
However, no studies have estimated the extent to which the impact of social
networking use intensity on depressive symptoms is attributable to addictive
9
Internet/social networking use (i.e. mediation effects). Such pathways via Internet
addictive behaviors (i.e. IA, SNA) are likely to be present, but have not been
investigated in previous studies.
In summary, there is a dearth of longitudinal studies that examine psychosocial
factors and Internet addictive behaviors (i.e. ‘SNA’ and IA) altogether to explain the
mechanisms underlying the relationship between social networking use intensity and
depressive symptoms.
1.2 Overall study design
This thesis consists of two phases. The phase I study is a validation study, in
which an instrument measuring social networking use intensity was developed
according to a standard scale development process, and then formally evaluated its
psychometric properties. Using the validated instrument, a nine-month longitudinal
study was conducted to explore the longitudinal relationships between social
networking use, ‘SNA’, IA and depression, as well as the potential underlying
mechanisms for the relationship between social networking use intensity and
depression among adolescents in China.
1.3 Aims
There are several objectives in this study, including:
(1). This study aims to estimate the depression, ‘SNA’ and IA prevalence, as well as
respective incidence in a longitudinal study sample of adolescents who are social
networking users in China. Such information is important to expanding our
understanding of the depression, ‘SNA’ and IA epidemic among adolescents over
10
time in China.
(2). This study aims to explore the causal directions between ‘SNA’/IA and
depression among adolescents who are social networking users. In specific, two
causal directions were investigated: (a) For the causal directions from ‘SNA’/IA
to incident depression, the predictions of baseline ‘SNA’/IA status on incident
depression were firstly tested (modeling approach I), and then the predictions of
changes in ‘SNA’/IA status over time (i.e. persistent non-‘SNA’/IA, remission
from ‘SNA’/IA, persistent ‘SNA’/IA, incident ‘SNA’/IA) on incident depression
were conducted (modeling approach II); (b) vice versa, for the causal directions
from depression to incident ‘SNA’/IA, the similar two modeling approaches were
conducted.
(3). This study aims to develop and validate an instrument for robustly assessing
social networking use intensity among adolescents in China. This validation study
was applied as a cross-sectional sample that differed from the longitudinal sample.
The psychometric properties and factor structure of the Social Networking
Activity Intensity Scale (SNAIS) were thoroughly examined.
(4). Using the scale developed in the validation study, this study aims to investigate
the prediction of social networking use intensity on incident depression among
adolescents in China. In addition, given that emotional connection is an important
construct related to social networking use and has not been investigated in
previous studies, its prediction and interaction effect with social networking use
intensity on depression were also investigated.
(5). Last but not least, this study aims to explore the mechanisms underlying the
11
relationship between change in social networking use intensity and change in
depressive symptoms. Specifically, it explores (a) the potential mediation,
suppression and moderation effects of change in psychosocial factors in the
relationship; (b) the potential mediation effects of change in Internet addictive
behaviors (‘SNA’ and IA) in the relationship; (c) the magnitude of combined
mediation effects of significant psychosocial mediators and Internet addictive
behaviors in the relationships. No such studies have previously been conducted
for these research purposes in China or worldwide.
1.4 Structure of the thesis
Chapter 1 is a brief introduction to the background, rationale, overall study design,
objectives, and thesis structure of the study.
Chapter 2 provides a thorough literature review on relevant topics of the thesis. It
includes the following topics: (1) The depression epidemic and disease burden. (2)
Increased high Internet penetration rates worldwide and in China. (3) The definition
of online social networking, related functions/activities, as well as the high prevalence
among adolescents. It was found that there is a dearth of data related to social
networking use among adolescents in China. (4) Research related to IA and ‘SNA’
concepts and their corresponding epidemics, as well as a summary of the
cross-sectional and longitudinal relationships between ‘SNA’/IA and depression. (5)
The literature review also summarized the measurements for assessing social
networking use. (6) The association between social networking use and depression, as
well as the associations between social networking use and psychosocial factors,
including friendship quality, perceived social support, self-esteem, adolescent-parents
conflict, and social non-confidence, are summarized. These psychosocial variables
12
served as underlying mechanisms for exploring for the relationship between social
networking use intensity and depressive symptoms (i.e. potential mediation,
moderation or suppression). (7) The associations between social networking use and
Internet addictive behaviors (‘SNA’ and IA) are summarized. Lastly, topics related to
longitudinal data analysis methods are summarized.
Chapter 3 describes the participants and methodology of the thesis in details. The
setting of study site, study design, participants, sampling method, data collection and
procedure, quality control, measurements of the phase I validation study and phase II
longitudinal study are described. The psychometric properties (i.e. Cronbach’s α) of
the scales that were used in this study are also reported in this chapter.
Chapter 4 presents the age-and gender-specific prevalence and incidence of
depression, ‘SNA’ and IA in the longitudinal sample. Meanwhile, natural remission
rates from depression, ‘SNA’ and IA over time were also estimated in the longitudinal
sample. In addition, descriptive statistics (i.e. gender-specific central tendency and
variance) of social networking-related scales and psychosocial scales are also
presented.
Chapter 5 thoroughly explores the causal directions between IA in general, ‘SNA’ in
specific and depression in the longitudinal sample. Two modeling approaches were
applied: the first modeling approach uses baseline measures to predict incidence, and
the second modeling approach uses changes in status of independent variables
between baseline and follow-up to predict incidence.
Chapter 6 presents the validation results of the developed instrument (SNAIS) in the
phase I study, which was used to assess the level of social networking use intensity.
13
There were no similar instruments available in China. The SNAIS was a short and
easy tool to implement. Two subscales, Social Function Use Intensity (SFUI) and
Entertainment Function Use Intensity (EFUI), with satisfactory psychometric
properties were identified.
Chapter 7 investigates the longitudinal predictive effects of baseline social
networking indicators (SFUI, EFUI, and EC) on incident depression, and explores the
moderation effects of EC on the associations between baseline SFUI, EFUI and
incident depression in the longitudinal sample.
Chapter 8 explores the mediation, suppression and moderation effects of change in
psychosocial factors, including friendship quality, perceived social support,
self-esteem, adolescent-parents conflict, and social non-confidence, on the
associations between change in social networking use intensity (SFUI and EFUI) and
change in depressive symptoms based on the longitudinal sample.
Chapter 9 explores the relationship between change in social networking use intensity,
change in Internet addictive behaviors (i.e. ‘SNA’ and IA), and change in depressive
symptoms among adolescents by the longitudinal sample. The mediation effects of
change in ‘SNA’ and IA on the association between change in social networking use
intensity and change in depressive symptoms are also assessed. In addition, the
magnitude of joint mediation effects of significant psychosocial factors and Internet
addictive behaviors (i.e. ‘SNA’ and IA) for the association between change in social
networking use intensity and change in depressive symptoms are explored and
summarized.
Chapter 10 is the conclusion section. Key findings of the thesis are firstly summarized
14
and interpreted. The implications of the findings, limitations of the study and
recommendations for future research are discussed in details.
15
Chapter 2 Literature review
2.1 Depression epidemic among adolescents
2.1.1 High disease burden caused by depression
Depression is a common mental disorder that is characterized by depressed mood,
loss of interest or pleasure, decreased energy, feelings of guilt or low-worth, disturbed
sleep or appetite, feelings of tiredness, and poor concentration [66]. Depression often
occurs during adolescence and young adulthood and substantially impairs an
individual’s ability to function at work or school or cope with daily life. Especially
when long-lasting and with moderate or severe intensity, depression may become a
serious public health problem.
It is estimated that depressive disorder accounted for 40.5% (95%CI:
31.7%~49.2%) of the Disability Adjusted Life Years (DALYs) caused by mental and
substance use disorders, and the highest proportion of total DALYs occurred in people
aged 10 to 29 years [67]. It is the first cause of DALYs among adolescents (Figure
2-1) [68]. The effects of depression are considerable and associated with a range of
negative outcomes (i.e. suicide, substance use) [69].
16
Figure 2-1 Top 10 causes of DALYs lost among adolescents by sex
2.1.2 Depression epidemic among adolescents worldwide
A community-based epidemiological survey in 10 countries reported that lifetime
prevalence of depression ranged from approximately 3% in Japan to 16.9% in the
United States, with most countries falling in the range of 8% to 12% [70]. As one of the
most frequently diagnosed psychiatric disorders among adolescents globally, it is
estimated that depression affects 1% to 6% of adolescents each year worldwide [4].
One recent systematic review summarized that the lifetime prevalence of major
depressive disorder (MDD) ranged from 1.1% to 14.6%, with one-year prevalence of
MDD ranging from 0.8% to 13.0% [71]. In Iranian adolescents, the estimated
prevalence of depression ranged from 13.1% to 55.4% based on different instruments
used for measurement [72]. In Chilean adolescents, the prevalence rate of clinical
depression is a markedly high 35% for girls and 28% for boys [73].
17
2.1.3 Depression epidemic among adolescents in China
Since the early 1990’s, the level of depression has significantly increased across
birth cohorts among Chinese adolescents [74]. China has been on the road towards
rapid modernization and social environmental change in recent years. Surrounded by
unprecedented competitive social and academic environments, adolescents have been
exposed to increasing amounts of stress (such as school examinations, family conflicts,
poor peer relationships, etc.), and reported a high prevalence rate of depression. For
instance, a cross-sectional study in twelve middle school students reported a high
depression prevalence of 42.3%, in which the prevalence of mild, moderate and
severe depression was 14.6%, 15.3% and 12.4% respectively [8]. The depression
prevalence in Shandong province was 16.9% [75] and 22.9% in Hunan province [76].
The depression prevalence among adolescents in Yunnan was 22.81%, in which the
prevalence among adolescents aged 11-13, 14-15, and 16-19 years were 20.17%,
24.71%, and 24.92% respectively [77].
2.2 Internet use worldwide and in China
Internet use has become an indispensable part of adolescents’ daily lives. It
provides an easy and immediate way for them to explore information and
communicate with other people across the world. Globally, Internet use has
substantially grown in the past two decades, with the world penetration of Internet use
sharply increasing from 0.4% in 1993 to 40.4% in 2014 [78] (Figure 2-2).
18
100 Developed
Individuals using the Internet (%)

90 World
Developing 82.2
80
70
60
50 43.4
40
30 35.3
20
10
0
2001 2003 2005 2007 2009 2011 2013 2015*
∗ 2015data are estimated. Data taken from International
Source: ITU World Telecommunication /ICT Indicators database
Figure 2-2 Growth in Internet usage from 2005 to 2014 worldwide
The Internet penetration in Mainland China has also dramatically increased in
recent years. It is estimated that there are 668 million Internet users in China by the
[11]
end of June 2015, accounting for nearly 49% of total population . Moreover, the
Internet penetration among adolescents in China sharply increased from 54.4% in
2009 to 79.6% in 2014 according to a recent official report from China Internet
Network Information Center (Figure 2-3) [12].
100 million
3 2.77 100%
2.56
2.32 2.35
2.12 80%
1.95 79.6%
2 71.8%
64.4% 66.4% 60%
60.1%
54.4%
40%
1
20%
0 0%
2009 2010 2011 2012 2013 2014
No. Penetration rate
Figure 2-3 Trend of Internet penetration among adolescents in China
19
2.3 Online social networking use among adolescents
2.3.1 Online social networking definition
Since the launch of the first recognizable online social networking platform
(Sixdegrees.com) in 1997, diverse online social networking services have been created
and are used by millions of people on the Internet globally (Figure 2-4). Online social
networking has rapidly become an important medium for social interaction and has
profoundly changed the ways that people communicate with each other during the
past two decades, especially among adolescent and young adults [79]. However, there
is neither one generally accepted term nor well-established definition for online social
networking. Numerous terms exist in the literatures (e.g. social networking sites,
online social networking, or social networking services), however, these terms are
often used synonymously. Boyd and Ellison [13] used the term “social network sites”
and defined them as Internet-based services that allow individuals to 1) construct a
public or semi-public profile; 2) identify a list of other users with whom a connection
is shared; 3) view and traverse their list of connections as well as those made by
others within the system. However, the definition provided by Boyd and Ellison is too
broad. Schneider et al used the term of online social networking and defined it as
web-based online communities among people with common interests, activities,
backgrounds, and/or friendships that allow users to upload profiles (text, images, and
videos) and interact with others in numerous ways [80]. Online social networking is
more egocentrically structured in that the users are the center of his/her own
community. In the present study, the term online social networking is used.
20
In China
Figure 2-4 Main history of online social networking development
2.3.2 Online social networking functions/activities
Although the nature and nomenclature of online social networking varies around
different platforms/services, most of their core functions are fairly consistent. Along
with technological development of the Internet, online social networking has
combined different types of Internet functions. The core of online social networking
consists of personalized user-oriented profiles, messaging/comments, sharing, and
game functions. A profile is a personal webpage that contains an individual’s
identifying information (i.e. name, photo, date of birth, school, hometown, and
contact information) and interests. A profile provides users a place to express their
individual personalities and hobbies, display online friend lists, and messages from
friends. They can share their lives intimately with their social network online through
their personal profiles. Messaging or commenting (e.g. private/public instant
21
messaging and chatting) is another core function provided by online social networking
services, which can facilitate social interaction with other people. On the message
board (or wall), users can send messages, update their self-status, and post photo or
links. Moreover, users can comment on such messages and respond to comments from
their friends. Messaging and commenting functions are popular features of online
social networking. Additionally, individuals can actively share content (i.e. photos,
videos, music, and news feed) among their friend network through online social
networking. In addition, many online social networking services have implanted
entertainment functions and activities, such as online games and movies.
In summary, individuals can use online social networking for social interaction,
for sharing information about oneself, or for entertainment purposes. Online social
networking users may use personal websites for self-presentation, instant messaging
for communication, message boards for discussion and photo sharing, and gaming for
entertainment. Online social networking has transformed the traditional face-to-face
communication to a new, one-to-many or even many-to-many interactions in a virtual
community.
Adolescents use social networking for different purposes. It has been reported
that adolescents mainly use online social networking for the purposes of social
connection, relationship maintenance, shared identities, photographs, content sharing,
social network surfing and status updates [81-84]. The most common social networking
activities used by adolescents include sending messages/chatting with friends,
reading/responding to notes/messages, reading comments/posts on their profiles,
browsing friends’ pages/profiles/walls, writing comments on friends’ pages/posting
on other people’s walls, and status updating [14, 54, 85]. Moreover, around one fourth of
22
sampled adolescents indicate that they use online social networking mainly for
entertainment purposes (i.e. playing online games, watching videos, listening to
music) [55].
2.3.3 High prevalent online social networking use among adolescents
2.3.3.1 Prevalent online social networking use worldwide
Popularity of online social networking has exponentially increased among
adolescents in recent years. As the most popular social networking service [86],
Facebook has more than one billion active users worldwide and it has been estimated
that this number would significantly increase in the future. In the United States, a
national survey among youth in 2007 by the Pew Internet and American Life Project
reported that around 41% of adolescents aged 12 to 13 years old and 61% of
adolescents aged 14 to 17 years old were online social networking users [55]. In 2012,
more than 70% of U.S. teens were social media users [87]. In Europe, reports from
Livingstone et al in 2011 showed that 49% of adolescents aged 11 to 12 years old,
73% of adolescents aged 13 to 14 years old, and 82% of adolescents aged 15 to 16
years old had their own online social networking accounts [15]. In Canada, an online
survey reported that 85% of college students possessed a Facebook account, and
moreover, 79% of them spent 10~60 minutes on Facebook daily [85]. Besides, it is
reported that more than half of adolescents (51%) checked their online social
networking profiles more than once a day, and more than one fifth (22%) of
adolescents checked their social networking profiles even more than 10 times a
day [88].
23
2.3.3.2 Prevalent online social networking use in China
The China Internet Network Information Center reported in 2012 that 48.8% of
Internet users also used online social networking in Mainland China, while 28.8% of
these social networking users were between 10 and 19 years old [89, 90]. One recent
small survey in Chinese college students (n=337) reported that almost all sampled
college students (98%) used Qzone, 18% used Renren.com, and 7% used both [16].
Qzone is one of the most popular types of online social networking services. It was
launched by Tencent company in 2005 and allows users to write blogs, keep diaries,
send out photos, listen to music, and watch videos. Meanwhile, Renren.com is called
the "Chinese Facebook" as most of its functions are just the same as those on
Facebook. A survey among Chinese college students reported that more than
three-fifths of participants spent 15~60 minutes per day on using social networking,
and over 12% of participants spent more than one hour on social networking use
daily [91]. No data was found pertaining to middle school students.
2.4 Internet addiction, social networking addiction and depression
2.4.1 Concept and classification of Internet addictive behaviors
Based on Davis classification for IA, there are two distinct forms of IA, namely
generalized IA and specific IA [22, 92]. Generalized IA refers that individual use of
Internet as a whole in a compulsive manner, and is defined as excessive Internet use
that interferes with daily life such as preoccupation and feeling restless or irritable
when attempting to cut back [19]. In contrast, specific IA refers to problematic use of
distinct online activities (e.g. online gaming addiction, online gambling addiction).
Previous studies have highlighted the importance of distinguishing between
24
generalized and specific IA [22].
There is an increased trend to distinguish specific IA from generalized IA in
recent research. Excessive use of online social networking raises concerns related to
‘social networking addiction’ (SNA) [93], which is considered a subtype of IA in
general [94]. Previous studies indicate that ‘SNA’ incorporates the experiences of the
classic addictive symptoms like other behavioral addiction [25, 95-97], including mood
modification (engagement in online social networking leads to a favorable change in
emotional states), salience (behavioral, cognitive and emotional preoccupation with
online social networking use), tolerance (increasing use of online social networking
over time), withdrawal symptoms, conflict (interpersonal and intra-psychic problems
because of online social networking use), and relapse.
2.4.2 Internet addiction epidemic
Internet addiction (IA) in general has drawn important attention and has become
a serious public health issue among adolescents. A meta-analysis of 31 nations across
seven world regions estimated a global IA prevalence of 6% [95% CI: 5.1%~6.9%],
the Middle East with the highest prevalence of 10.9%, and Asia (including Mainland
China, Hong Kong, India and South Korea) at 7.1% [98]. In China, the IA prevalence
among adolescents dramatically increased to 26.0% in 2011 from 14.1% in 2009,
while proportion of IA among adolescents aged 13~17 was markedly high at 30.5% in
2011. A representative sample among college students in Taiwan also reported a high
prevalence of IA (15.3%) [99].
2.4.3 Social networking addiction epidemic
Social networking addiction (‘SNA’) in specific is a recent public health concern
25
along with the phenomenon of excessive social networking use. It has been reported
that about one tenth (9.78%) of college students self-perceived with Facebook
addiction [55]. Another study among Irish adolescents reported that about 33% of
participants felt that they overused online social networking [100]. However, one study
among Nigerian undergraduate students reported a relatively low percentage (1.6%)
of Facebook addiction [29], while a survey among undergraduate students in Peru
reported that the prevalence of Facebook dependence was around 8.6% [30]. Moreover,
three studies modified the Young’s Internet Addiction Test to measure ‘SNA’. Wu et al
reported 12% of probable problematic social networking use among Chinese
smart-phone users [101], while 24.0%~34.3% of sampled Chinese college students
were classified with ‘SNA’ [31, 32]. ‘SNA’ has therefore become a significant public
health issue, especially with an exponentially increasing availability of new
technologies (such as smartphones, laptops).
2.4.4 Relationship between IA and depression
The positive association between IA and depression has been reported in a
substantial amount of cross-sectional studies [102, 103]. Although the association
between IA and depression has been repeatedly reported and reviewed, it is still
unclear as to whether IA is a cause or consequence of depression among adolescents.
It is important to clearly understand causality to develop effective interventions and
treatments for both disorders. Our literature review found eight longitudinal studies
among adolescents that investigated the longitudinal relationship between IA and
depression. From this literature, three possible causal directions were described to
explain the relationship. First, it posits that depression as a psychiatric disorder
contributes to and/or deteriorates the symptoms and course of Internet addiction
26
development. Second, that Internet addiction leads to or contributes to the
development of depression. Third, that the relationship between IA and depression
might be bidirectional and may form a vicious circle.
For the causal direction from depression to IA, Davis’ cognitive-behavior
theoretical model proposes that pre-existing psychopathology (i.e. depression, social
anxiety, and substance dependence) is the diathesis and a distal necessary cause of
symptoms of Internet addictive behaviors. Along with this theoretical model, it has
been argued that the presence of depressive symptoms predisposes individuals to
develop Internet addictive behaviors. There are four existing longitudinal studies that
estimated the predictive effect of depression/depressive symptoms on IA, including
one in South Korea and three in Taiwan. Two studies used baseline depression to
predict incident IA, including the study of Cho SM et al in South Korea which
reported that children who had been classified with anxious/depressive symptoms
during childhood had increased risk of IA in adolescence during the 7-year follow-up
period [44], and a two-year prospective study in Taiwan adolescents showing that
pre-existing depression significantly predicted the occurrence of IA [39]. Another study
in Taiwanese high school students found that baseline depression and its increase in
CES-D score during the one-year follow-up period significantly predicted the
incidence of IA [38]. However, the Ko CH et al one-year longitudinal study in Taiwan
found that there was no significant difference in baseline depressive scores between
the incident IA group and the non-IA group [40]. Therefore, these four longitudinal
studies revealed a mixed prediction of depression on IA among adolescents.
Alternatively, according to the displacement hypothesis theory proposed by
Kraut et al [61], excessive Internet use declines with social involvement, then
27
deteriorates the symptoms of psychiatric disorders (i.e. depression, loneliness). Three
longitudinal studies tested the causal direction from IA to depression among
adolescents, including two conducted in Mainland China and one in Taiwan. For
instance, a longitudinal study among Chinese high school students reported that
participants with IA at baseline were 2.5 times more likely to develop depression
during the nine-month follow-up period [43]. Another one-year longitudinal study in
Mainland China reported that a significant increase in depressive scores was observed
between the before and after time periods of participants becoming addicted to the
Internet[42]. A prospective study in Taiwan also demonstrated that adolescents in the
incident IA group had a larger increase in CES-D score compared to the non-addiction
group at the one-year follow-up [41]. It should be noted, however, that research studies
delving into this topic area are quite limited in number. It should be noted that there is
a dearth of such studies delving this topic.
For the bi-directional relationship between IA and depression, Davis notes that
maladaptive cognitions (i.e. self-doubt, low self-efficacy, and negative self-appraisal)
and dysfunctional behaviors (i.e. compulsive Internet use that leads to negative
outcomes, denying about Internet use, and using the Internet to escape from emotional
problems) of pathological Internet use would intensify and continue to produce
negative outcomes over time, resulting in a diminished sense of self-worth and
increased social withdrawal. As symptoms of IA worsen, they would exacerbate
existing psychopathologies, resulting a vicious dysfunctional cycle [17, 21]. To our
knowledge, only one recent one-year longitudinal study among Spanish adolescents
analyzed the bi-directional relationships between depressive symptoms and four
components (i.e. preference for online social interaction, use of the internet for mood
28
regulation, deficient self-regulation and manifestation of negative outcomes) of IA.
The aforementioned components of IA were measured using Generalized Problematic
Internet Use Scale 2 (GPIUS2) in the study. Furthermore, the results from Structural
Equation Modeling revealed that depressive symptoms at baseline positively predicted
components of preference for online relationships, mood regulation, and negative
outcomes (three subscales of GPIUS2) one-year later, while, in turn, the only
component of negative outcomes (the subscale of negative outcomes) at baseline
positively predicted depressive symptoms at follow-up. Thus, the authors concluded
that depressive symptoms is a risk factor of the components of preference for online
social interaction and mood regulation, while the relationship between the component
of negative outcomes and depressive symptoms was reciprocal [45].
The relationship between IA and depression remains controversial as researchers
continue to debate over the causal direction. In particular, there is a dearth of research
to theoretically test a bi-directional relationship between IA and depression among
adolescents. It is still unclear as to whether IA is a cause or a consequence of
depression, or even if a bi-directional relationship exists between IA and depression.
Large-scale longitudinal studies are highly warranted.
2.4.5 Relationship between ‘SNA’ and depression
‘SNA’ may have potential detrimental effects on mental health and other
psychological well-being. Some studies have identified that ‘SNA’ is associated with
several types of health outcomes, including emotional problems (i.e. depression,
anxiety, and poor self-esteem), relational problems (i.e. family conflict, loss of friends,
and relationship dissatisfaction), and other health-related problems (i.e. sleep
problems and poorer sleep quality, somatic symptoms) [26, 104].
29
Research related to ‘SNA’ and depression is very scarce. Only a few
cross-sectional surveys investigated the association between ‘SNA’ and depression
among adolescents, one of which was conducted among Turkish students and reported
that severe depression is significantly associated with ‘SNA’ (specifically Facebook
addiction) [48]. Among Taiwanese college students, it was also found that depressive
symptoms were positively associated with Facebook addiction [47]. One
cross-sectional survey among general Facebook users reported a positive association
between depression and Facebook addiction [105]. Another study reported that
compulsive social networking users scored a higher level of depressive mood than
those non-compulsive users [106]. However, research addressing influences of ‘SNA’
on mental health, especially depression, is still limited.
In summary, there are no longitudinal studies exploring the causal relationship
between ‘SNA’ in specific and depression among adolescents. A knowledge gap hence
exists.
2.5 Impacts of social networking use on adolescents’ psychosocial status and
depression
2.5.1 Assessment for social networking use intensity
2.5.1.1 Using general questions to estimate social networking use intensity
The majority of the studies assessed social networking use behaviors by using
investigator-developed simple or single-item measures. Researchers often estimated
social networking use behaviors by frequency of use [62, 107], number of Facebook
friends [60], and average daily time spent on online social networking [35, 108]. These
simple questions often fail to assess complex constructs and may involve sizable
30
measurement errors. The functions and diverse activities of online social networking
(i.e. instant messaging, comments, sharing personal photo/information, and playing
on-line games with friends), have not been taken into account, which poses an issue
since the natures of activities on online social networking, rather than the frequency of
online social networking use, may have direct effects on adolescents’ psychosocial
well-being [62].
2.5.1.2 Instruments related to social networking use intensity
There are a few studies that have developed specific scales for their research
involving social media use. However, these measures tend to focus only on a single
social networking platform (mainly Facebook), and are often not fully validated.
These scales include the Facebook Intensity Scale (FIS) [51], Facebook
questionnaire [85], Social Media Use Integration Scale (SMUIS) [109], and the Facebook
social connectedness scale [110] (See Table 2-1).
The Facebook Intensity Scale (FIS) is developed by Ellison et al [51] in 2007.
After that, it became a commonly used instrument in the literature [36, 111, 112]. The FIS
aims to measure the extent to which an individual is emotionally connected to
Facebook and integrated its use into his/her daily activities. It includes two parts: the
first part includes six items with a five-point Likert scale (5=strongly agree,
1=strongly disagree) to measure individual’s attitudes towards Facebook (a sample
item is “Facebook is part of my everyday activity”), and the second part includes two
items that are self-reported assessments for Facebook behaviors (i.e. the number of
Facebook friends and the amount of time spent on Facebook in the past week). For the
FIS, Ellison et al reported the internal reliability, but not mention other psychometric
properties. The application of the FIS is inconsistent in the literature. Some
31
researchers argue that the two items for Facebook behaviors (i.e. the number of
Facebook friends and the amount of time spent on Facebook in the past week)
estimated different constructs from emotional connections, and the six attitude-related
items were used as a separate construct to evaluate the extent to which an individual is
emotionally connected to social networking [36].
[85]
The Facebook questionnaire, developed by Ross et al , is another scale that
has been used by other studies. The Facebook questionnaire included the six
attitudinal items from FIS along with one additional item created by the authors
(“How satisfied are you with your Facebook, overall?”), and a number of items
assessing frequency of use for basic features of Facebook. The principle component
factor analysis revealed a two-factor solution, including Attitude (7 items, α=0.83)
and Online Sociability Functions (5 items, α=0.74) (see Table 2-1).
Social Media Use Integration Scale (SMUIS) was objectively developed to
measure the social networking use integration of social behaviors and daily routines,
along with the importance of and emotional connection to social networking use [109].
The 10-item SMUIS included two subscales using six Likert points ranging from
strongly disagree to strongly agree, namely Social integration and emotional
connection subscale with 6 items and Integration into social routines subscale with 4
items by exploratory factor analysis and confirmatory factor analysis (Table 2-1). This
scale was developed based on a formal scale development procedure and conducted
formal validation analysis. The SMUIS has one item related to social routines that
overlapped with two items from FIS (i.e. “Facebook has become part of my everyday
activity and Facebook is part of my everyday activity”).
[110]
The Facebook Social Connectedness Scale was adapted from the Social
32
Connectedness Scale by revising to it specifically for the Facebook environment. It
consists of 20 items (10 positive and 10 negative) to reflect an individual’s positive
and negative feelings towards their Facebook environment. Examples of items include,
“I find myself actively involved in Facebook environment” (positive perception), and
“Even among my Facebook friends, there is no sense of brother/sisterhood” (negative
perception).
In summary, all of the aforementioned scales only examine one specific site (i.e.
Facebook) and mainly emphasized the integration and emotional connection to
Facebook. Additionally, all studies were conducted among college students (see Table
2-1). Currently, the majority of adolescents use multiple types of social media [14] and
involved multiple types of social networking activities (e.g. messaging with friends,
posting comments, status updating) [113]. Adolescents have potentially become the
prominent populations that use online social networking frequently.
33
Table 2-1 Instruments measuring online social networking use
Authors-year Sample Measures Response Items Reliability Methods
Ellison et al-2007 [51] Undergraduate Facebook Intensity Scale 5-point 8 items α=0.83 EFA
students (n=286)
Ross et al-2009 [85] Undergraduate Facebook questionnaire PCF
students (n=97) - Attitudes towards Facebook Attitude: 5-point 7 items Attitude: α=0.83
- Online sociability function OSF: 9-point 5 items OSF: α=0.74
Jenkins-Guarnieri MA et Undergraduate SMUIS 6-point SMUIS: α=0.91 EFA
al-2013 [109] students (n=552) - SIEC 6 items SIEC: α=0.89 CFA
- ISR 4 items ISR: α=0.83
Test-retest=0.80
Grieve et al-2013 [110] College students Facebook social connectedness 6-point 20 itemsα=0.92 EFA
(n=344)
Note: FIS=Facebook Intensity Scale; Attitude towards Facebook subscale includes 6 items from FIS of Ellison et al study and one additional
item composed by the author (How satisfied are you with your Facebook, overall?); Online sociability function subscale was composed of 5
questions relating to the frequency with which individuals engaged in different Facebook activities.
SMUIS: Social Media Use Integration Scale, SIEC=Social Integration and Emotional Connection subscale, ISR=Integration into Social Routines
subscale
EFA=Exploratory Factor Analysis; CFA=Confirmatory Factor analysis; PCA=Principle Component Analysis
34
2.5.2 Association between social networking use and depression
Some studies have linked online social networking use with several psychiatric
problems. The most in-depth mental health measures could be classified into two
panels, including (1) negative outcomes like depression/depressive symptoms [35, 37, 52,
53, 108, 110]
, social anxiety [16, 37, 58, 110] and loneliness [57, 58, 84, 111, 114], and (2) positive
outcomes such as satisfaction with life [16, 49, 60, 62, 110, 114, 115]. Most of these studies are
based on the cross-sectional study design and were conducted in western countries (i.e.
the United States) by focusing on Facebook, which is the most popular social
networking service worldwide. However, the relationships between social networking
use and mental health outcomes remain controversial and unclear. Our thesis would
apply depression as the outcome considering its high prevalence and disease burden
among adolescents.
The literature review found seven cross-sectional studies that investigated the
associations between diverse measures of social networking use and depressive
symptoms (Table 2-2), including two in the United States [52, 53], two in the
Philippines [37, 108], and one each in Australia [110], Serbia [35] and Belgium [116]. Five
studies were conducted in college/university students; while two of the studies, one in
Serbia and another in Belgium, were conducted among high school students. No
longitudinal study was found that investigated the association between social
networking use and depression among adolescents. In the aforementioned
cross-sectional studies, the results on the association between social networking use
and depression are inconsistent. For instance, three surveys found a non-significant
association between social networking use and depression [52, 53, 108], while two studies
reported that prolonged use of social networking (measured by amount of time spent
35
on social networking) was significantly correlated with high level of depression [35, 37].
However, the study in Australia reported that Facebook social connectedness was
significantly associated with lower level of depression [110]. The American Academy
of Pediatrics (APP) revealed a new term “Facebook depression”, that describes the
potential effects on mental health problems due to excessive amount of time spent on
social media (like Facebook), in which individuals affected by the condition will
subsequently begin to exhibit classic symptoms of depression [117].
Apart from the differences in the methodology of these studies, it is possible that
social networking use may have diverse impacts on different age groups, given that
most of the previous studies were conducted among older adolescents and young
adults (i.e. college students). There are many potential reasons why social networking
users have the tendency to suffer from depressive disorder. Contrastingly, there are
numerous factors that may influence already depressed individuals to begin using
online social networking. However, this issue has not yet been interpreted and
discussed sufficiently, and factors that contributed to heterogeneity of findings have
not yet been identified thoroughly. Considering the potentially different effects of
social networking use on depression between early and late adolescence, it is
necessary to further conduct related studies among early adolescents in particular.
Moreover, longitudinal studies with large-scale samples are warranted to estimate the
predictive effects of this new phenomenon on depression among adolescents.
In summary, all of the studies that investigated the association between social
networking use and depression are cross-sectional, and the results are inconsistent.
The impact of social networking use on adolescents’ depression is still under-
researched. Longitudinal studies are needed to robustly investigate the relationship.
36
Table 2-2 Summary of the literature review for the association between social networking use and depression
Author-year-country Design Subjects Measure for SNS use Main results
Frison and CSS 910 high school Active public Facebook use, - Passive Facebook use and active public
Eggermont-2015 [116] students Active private Facebook use Facebook use is positively associated with
Belgium Passive Facebook use depressed mood
Labrague L-2014 [37] CSS 76 college - Facebook intensity scale - Time spent on Facebook positively correlated
Philippines students - Number of Facebook friends with depression (r=0.23) and anxiety (r=0.26)
- Time spent on Facebook in a
typical day
Simonic TE et al-2014 [52] CSS 237 university - Facebook activity subscale - No direct association between Facebook use
United States students and depressive symptoms
- Moderating effect was observed: more frequent
Facebook use was associated with lower
depressive symptoms only for females with
high neuroticism
[108]
Datu JA et al-2013 CSS 200 college - Number of hours per week - No significant association
Philippines students aged
17~20
[110]
Grieve R et al-2013 CSS 274 university - Facebook social - Facebook facilitated social connections and
Australia students connectedness associated with lower depression (r=-0.27) and
anxiety (r=-0.21), and greater satisfaction with
life (r=0.30)
Jelenchick LA et CSS 190 - Average time spent on social - No association between social networking use
al-2013 [53] undergraduate networking per day based on and depressive symptoms
United States students experience sample method
one week
Pantic I, et al-2012 [35] CSS 160 high school - Daily time spent on social - Positive correlation was observed between time
Serbia students networking spent on social networking and depression
CSS: Cross-sectional study; r: correlation coefficient.
37
2.5.3 Associations between social networking use and psychosocial status
2.5.3.1 Association between social networking use and friendship quality
Establishing intimate relationships and emotional connection with others is an
important developmental skill during adolescence. Researchers suggest that
adolescents online and offline worlds are indeed connected and overlap [54]. They use
social networking mainly to maintain connections with offline friends and family
members [54, 59]. A previous study showed that 43% of sampled adolescents felt that
social networking use made them feel closer with their friends [59]. Another study
among adolescents aged 9 to 13 years old also showed that social networking users
reported a strong sense of belonging to their friendship group compared to non-users,
and a positive linear relationship between social networking use intensity and feelings
of belongingness among males was found [118]. However, one survey reported that the
majority (73%) of sampled participants thought that social networking use had not
made any difference to their relationships with friends, whereas around 20% felt that
social networking use made them closer with their friends, and on the contrary, a very
small proportion (2.5%) felt that it had negatively impacted their relationships [54]. A
few cross-sectional studies reported that social networking use was significantly
associated with enhanced friendship quality and intimacy [51, 110].
2.5.3.2 Association between social networking use and perceived social support
Perceived social support has emerged as a prominent concept that characterizes
social support as the cognitive appraisal of being reliably connected to others [119],
which refers to an individual’s belief in the availability of support. One study among
U.S. undergraduate students showed that the number of Facebook friends was
38
associated with increased perceived social support [60]. Another study among South
Korea undergraduate students reported that self-disclosure through online social
networking was positively associated with social support [115]. In general, social
networking use is considered to be beneficial for enhancing adolescents’ social
connection and perceived social support.
2.5.3.3 Association between social networking use and family relationships
During adolescence, young people strive for dependency and become
autonomous to establish their own identities. On the other hand, family still exerts a
potent influence and support on their development. Harmonious family relationships,
especially between adolescents and their parents, are important for adolescents’
development by providing emotional and instrumental support for adolescents. Such
support could assist adolescents to solve certain problems. Intra-familial conflicts,
particularly adolescents’ conflicts with their parents, are common during this period.
Adolescent-parents relationships have a strong impact on adolescents’ psychological
development. Engagement in online social networking potentially interferes with
adolescents’ family activities and reduces communication with parents [61]. Social
networking use among adolescents has been linked with greater parent-youth
conflicts [54] and less time spent with parents [120]. However, some parents may utilize
social networking to maintain greater contact with their children, and require them to
be connected through online social networking [121]. Hence, the association between
social networking use and adolescent-parents relationships remains controversial.
Future research into this area is warranted.
39
2.5.3.4 Association between social networking use and social non-confidence
Social confidence represents the extent to which an individual can disclose to,
relate to, and successfully interact with others on social occasions [122]. High levels of
social confidence are generally viewed as a protective factor that is associated with
good mental health and psychosocial outcomes among adolescents. Building
confidence to interact with others is an important aspect of adolescents’ development.
Social networking interaction potentially reduces face-to-face interaction offline.
Online interaction is substantially different from face-to-face interaction in terms of a
lack of non-verbal cues (i.e. voice tone, eye contact), and a reduced degree of social
presence [123]. Due to a reduced probability in face-to-face interactions in real life,
social confidence might be diminished. Although social confidence is an important
protective psychosocial factor for adolescents’ mental health, no existing study has
tested the impact of social networking use on social confidence as a construct among
adolescents.
2.5.3.5 Association between social networking use and self-esteem
Self-esteem refers to the evaluation of individuals’ overall sense of self-worth or
personal value. Self-esteem is an important protective psychosocial factor in
developing and maintaining mental health and overall quality of life among
adolescents. A low level of self-esteem is associated with numerous mental health
problems (e.g. depression) [124, 125]. Previous studies reported conflicting results
regarding the influence of social networking use on self-esteem. In Kalpidou et al’s
study [36], spending long time on Facebook was related with lower self-esteem.
However, some researchers argue that online social networking allows teenagers to
have enough time to control information that they want others to have about them and
40
enables users to present themselves in a selective way. These selective
self-presentations of themselves on online social networking influence their self-
impressions. A cross-sectional survey among Dutch teenagers reported a positive
association between social networking use and youth self-esteem [62]. It should be
noted that there is presently a dearth of studies that examine the relationship between
social networking use and self-esteem.
2.6 Associations between social networking use and Internet addictive behaviors
As a popular Internet activity, online social networking has been frequently and
excessively used by adolescents. One recent study among Chinese college students
reported that more than three-fifths of participants spent 15 to 60 minutes per day
using online social networking, and over 12% of participants spent more than one
hour on online social networking daily [91]. Another study reported that around 6.55%
of sampled participants spent more than 16 hours on social networking per week [106].
The characteristics of online social networking (i.e. anonymous, selective
self-disclosure, and convenient communication without the limitation by geography
and distance) could facilitate online social interaction, and then function as an
attractive factor for excessive Internet/social networking use. Keeping in touch with
friends was the greatest reason for high frequencies of social networking use [126].
Some studies have investigated the association between diverse measures of
social networking use and addictive behaviors and tendency. One survey among
college students reported that Internet addicts had higher Facebook use intensity (i.e.
frequency and minutes per day) as compared to non-addicts, and moreover, a
regression model revealed that time spent online and frequency of Facebook use were
41
significant predictors of IA [127]. A one-year longitudinal study reported that high
frequency of online social networking use significantly predicted compulsive Internet
use [65].
Similarly, a study from the Belgian general population reported that compulsive
social networking users spent considerably more time on these Internet applications
than other users [106]. Another survey in Turkish college students reported that students
who frequently used Facebook for social interaction reported a higher level of
Facebook addiction [48]. One recent cross-sectional study reported that daily time
spent on Internet use was significantly associated with Facebook addiction [105].
2.7 A summary of research gaps found in the literatures
There are several knowledge gaps found in the literature, including:
(1) The causal relationships between ‘SNA’/IA and depression remain controversial
and under-researched. There is a lack of longitudinal studies to test the two
predictive directions (‘SNA’/IA predicts depression and depression predict
‘SNA’/IA) in the same longitudinal study. Clarification for the causal directions
between ‘SNA’/IA and depression would be helpful for the development of
effective interventions.
(2) There have been no fully validated and acceptable instruments for measuring
social networking use intensity robustly, especially in Chinese context.
Instruments with good psychometric properties are highly needed, which would be
important to robustly estimate the impact of social networking use on adolescents’
mental health problems (e.g. depression).
42
(3) The impact of social networking use on depression among adolescents is
under-researched and mixed.
(4) Moreover, it is still unclear as to how and why social networking use intensity
impacts adolescents’ depression, i.e. underlying mechanisms. Literature reviews
imply that psychosocial factors (i.e. friendship quality, perceived social support,
self-esteem, social non-confidence, adolescent-parents conflict) and Internet
addictive behaviors (i.e. SNA, IA) have the potential to explain the association
between social networking use intensity and depression. However, the role of
these variables on the association between social networking use intensity and
depression is not investigated in previous studies.
2.8 Statistical methodological considerations for longitudinal study
2.8.1 Causal direction testing in the longitudinal data
The statistical methodologies in longitudinal observational studies that aim to
test longitudinal causal relationships are diverse and need to be addressed. In the
literature, there are at least six modeling approaches to analyze longitudinal
observational data in studying IA/depression or other outcome variables: (1) Baseline
independent variable to predict the dependent variable (i.e. incident depression or IA),
with or without controlling for the baseline level of the corresponding dependent
variable [43, 44, 128]; (2) Change in status of the independent variable (i.e. changing from
non-depression at baseline to depression at follow-up, changing from depression at
baseline to non-depression at follow-up) to predict the dependent variable [129, 130]; (3)
Score change over time of the independent variable to predict the dependent
variable [131]; (4) Baseline independent variable to predict the score change of the
43
dependent variable [132]; (5) Change in status of the independent variable to predict
score change of the dependent variable [133]; (6) Structural Equation Modeling
(SEM) [45, 134]. Thereinto, change in score is different from change in status, change in
scale score indicates the difference in score from baseline to follow-up, which is a
continuous variable, while change in status indicates the conversion of the
psychopathological condition over time (i.e. participants transitioned from
non-depression at baseline to depression at follow-up, or recovery from depression at
baseline to non-depression at follow-up), which is a categorical variable [130]. For the
aforementioned modeling approaches, the previous longitudinal studies involving IA
applied modeling approach 1 and the SEM approach.
Examples for the first approach as mentioned above include the Lam LT et al
study which was conducted in China and used baseline IA to predict incident
depression by fitting logistic regression models [43], and the Cho SM et al study which
was conducted in South Korea and used baseline depression to predict IA at follow-up
by simple logistic regression models [44]. In these studies, IA or depression was
generally measured at one point in time with the assumption that the condition
remains unchanged over time. This modeling method has the strength of providing an
explicitly temporal sequence for the effect between the independent and dependent
variables, meaning that the independent variable is definitely prior of the outcome,
which is one of the important criteria for causal inference. Therefore, we applied this
statistical approach in the present study.
However, the aforementioned modeling method does not take into account
remission from IA to non-IA during the study period (and remission from depression
to non-depression). Previous studies have reported that about half of participants
44
(49.5% [40] and 51.5% [135]) naturally remitted from IA during the one-year follow-up
period. Considering the potential remission effects, some studies have used the change
in status of independent variables over time for modeling specification, such as
classifying participants into four categories based on depression status at baseline and
follow-up, including (1) persistent non-depression (persistent non-depression both at
baseline and follow-up), (2) incident depression (incidence from non-depression at
baseline to depression at follow-up), (3) remission from depression (remission from
depression at baseline to non-depression at follow-up), and (4) persistent depression
(persistent depression both at baseline and follow-up). However, these studies mainly
related to the relationship between physical activity and depression. No longitudinal
studies aiming to investigate the relationship between IA and depression have applied
this modeling method. Therefore, we also applied this modeling approach to take into
account the status transition of the independent variable (e.g. remission from ‘SNA’,
IA or depression) over time.
Moreover, as previously discussed, some studies used the change in score from
baseline to follow-up (i.e. score of depressive symptoms at follow-up minus the score
at baseline) for modeling specification. However, in order to definitively define the
change in status for ‘SNA’, IA, and depression over time, we applied the modeling
approach with status change rather than score change. Besides, Structural Equation
Modeling (SEM) techniques have been increasingly used in medical statistics in
recent years. SEM has the advantage of testing the theoretical models involving
complex patterns of relationships among multiple predictors and dependent variables
simultaneously [136]. As a general rule, SEM is indicated when more than one
regression equation is necessary for statistical modeling of the phenomena under
45
investigation [137]. However, SEM is mainly used for continuous variables, and is still
in its development for handling categorical variables. It has been argued that SEM
should be used only when 1) it is truly interested in exploring a wide range of
different effects and pathways (i.e. direct and indirect effects) across an entire set of
variables for several different outcomes, and 2) it is being used principally for
exploratory and hypothesis-generating purposes [138]. Therefore, in this chapter, the
SEM is not applied as a statistical method for causal direction.
In summary, this study applied the first two aforementioned modeling
approaches when analyzing our longitudinal data, aiming to investigate the causal
directions between ‘SNA’/IA and depression from two angles – Approach I which has
predictive implications by the advantage of an explicitly temporal sequence and
Approach II which takes the changes in status of the independent variables (including
remissions from pathological states) into account (Chapter 5).
2.8.2 Mediational effects and modeling specification
2.8.2.1 Concepts of mediation and suppression effects
A mediator conceptually means a variable that has a position in the causal
sequence and could account for all or part of the relationship between an independent
variable and a dependent variable (Figure 2-5a) [139]. Statistically, the mediation effect
would reduce the magnitude of the relationship between an independent variable and
a dependent variable when the effect of a mediator is controlled for in the model [139,
140]
. The effect of the independent variable on the dependent variable is shown to
include two causal paths (Figure 2-5b): the direct impact of the independent variable
(path c’) and the impact of the mediator (path b). Moreover, there is a path from the
independent variable to the mediator (path a). For the mediation effect, indirect effect
46
of the independent variable on the dependent variable via the mediator has the same
sign of the direct effect of the independent variable on the dependent variable. For
example, one study among high school students in Italy investigated the associations
between self-esteem, preference for online social interaction and IA [141], and the
results showed that the negative association between self-esteem and IA was partially
mediated by preference for online social interaction. When controlling for preference
for online social interaction in the model, the regression coefficient of self-esteem on
IA significantly reduced from -0.711 to -0.411. This means that the negative effect of
self-esteem on IA is partially through the increase of preference for online social
interaction.
A suppressor is defined as a situation in which a third variable would increase the
predictive validity of an independent variable on a dependent variable by its inclusion
in a regression model [140, 142], thus the strength of the relationship between an
independent and a dependent variable is suppressed by a third variable. In the
statistical model (Figure 2-5b), if path a and path b mentioned above are fulfilled, but
when paths a and path b are controlled for in the same model, and a previously
significant relationship between the independent variable and dependent variable
significantly increased, a suppression effect should hence be considered. A
suppression analysis proposes that the indirect effect of the independent variable on
the dependent variable through a suppressor and the direct effect of the independent
variable on the dependent variable have different signs. From this perspective,
suppression is also known as inconsistent mediation [143]. For instance, Paulhus DL
and colleagues presented the suppression effect of self-esteem on the association
between narcissism and antisocial behavior, reporting that when adding self-esteem
47
into the regression model, the positive effect of narcissism on antisocial behavior
significantly increased from 0.21 to 0.32. A suppression effect from self-esteem is
hence present [144]. In detail, results from this study showed that narcissism is
positively associated with antisocial behavior (β=0.21), self-esteem was negative
associated with antisocial behavior (β=-0.27), and self-esteem and narcissism were
positively correlated. It showed that the direct effect of narcissism on antisocial
behavior is positive, and the indirect effect of narcissism on antisocial behavior
through self-esteem is negative. When these two effects are considered together, they
cancel each other out and result in a diminished total effect. Another cross-sectional
study examines whether the Medicaid suppresses racial disparities on health care
needs (including child’s need for professional care coordination, child’s need for
mental health services, and the family’s need for mental health services), and found
that race has a negative direct effect for all three outcomes, and Medicaid coverage
suppressed the racial disparities in reporting the need for services [145]. However, the
suppression effect was small, ranging from 0.011 to 0.017 for the three outcomes.
Independent c Dependent
Independent
variable variable
variable
Mediator/
Mediator/
Suppressor
Suppressor a b
Independent c’ Dependent
Dependent variable variable
variable
Total effect=c
Direct effect=c’; Mediation/suppression effect=ab
(a) Conceptual model (b) Statistical model
Figure 2-5 Illustration of mediation/suppression effects
48
2.8.2.2 Statistical requirements for testing mediation and suppression effects
A variable functions as a mediator when it meets the following requirements
from Baron and Kenny’s method [139]: (a) variations in levels of the independent
variable significantly account for variations in the presumed mediator (i.e. path a), (b)
variations in the mediator significantly account for variations in the dependent
variable (i.e. path b), and (c) when path a and path b are controlled for, a previously
significant relationship between the independent variable and the dependent variable
is significantly reduced or becomes non-significant. If these requirements are fulfilled,
the mediation effect is hence support. Based on the Baron and Kenny’s method, in
order to test for the mediation effect, three following regression equations should be
estimated (Figure 2-6): (1) the dependent variable is regressed on the independent
variable; (2) the mediator is regressed on the independent variable; (3) the dependent
variable is regressed on both the independent variable and the mediator. Separate
regression coefficients for each model should be estimated and tested. These three
regression equations provide the tests of the linkages of the mediational model.
To establish mediation, the following conditions must be fulfilled: (1) the
independent variable must affect the dependent variable in the first regression; (2) the
independent variable must be shown to affect the mediator in the second regression;
and (3) the mediator must affect the dependent variable in the third regression; (4) if
these conditions all hold in the predicted direction, the regression coefficient of the
independent variable on the dependent variable in the third model must be
significantly less than that in the first regression model. Complete mediation holds
true if the independent variable has no effect when the mediator is controlled in the
third regression model.
49
For suppression effect analysis, the first three aforementioned conditions are held.
It is the fourth condition that is used to distinguish the suppression from mediation.
Specifically, the regression coefficients of the independent variable on the dependent
variable in the third regression must be significantly larger than that in the first
regression model.
e (1)
Independent Step 1: c Dependent

variable X variable Y
Y = b 0(1) + cX + e (1)
e (2)
Intervening
variable M
Step 2: a Step 3: b
e (3)
Step 3: c’
Independent Dependent
variable X variable Y
M = b 0(2) + aX + e (2)
Y = b 0(3) + c ' X + bM + e (3)
Figure 2-6 Statistical strategy of mediation/suppression hypothesis testing based on

Baron and Kenny’s method
2.8.2.3 Testing mediation hypotheses for longitudinal data
Compared with cross-sectional data, longitudinal data could add potential
improvement for interpretation. However, longitudinal research introduces a number
of modeling options and additional considerations when testing longitudinal
mediational relationships. One of the advantages of using longitudinal data to evaluate
causal mediational processes comparing with using cross-sectional data is that
longitudinal data could examine associations within changes across waves of data (i.e.
within subjects) [146].
50
With regard to two-wave longitudinal data, there are mainly two mediational
modeling approaches as summarized in the literature [147, 148]. The first approach is the
raw change score approach. In this modeling approach, change scores are calculated
for each of independent, mediating, and dependent variables, and then these change
scores are used for mediation modeling through the traditional mediation strategy
(Baron and Kenny [139]). This modeling approach has been widely used in previous
longitudinal studies. For instance, a 16-week interval, two-wave longitudinal study
used the variables of change in consumption of food away from home, change in diet
quality/dietary intake and change in BMI/percent body fat to test the mediation effects
of change in diet quality/dietary intake on the relationships between change in
consumption of food away from home and change in BMI/percent body fat [149].
Another longitudinal study also applied the raw change score testing the mediation of
changes in HRV on the relationships between changes in smoking frequency and
changes in erectile tumescence using the Baron and Kenny’s strategy [150].
The second approach is the Analysis of Covariance (ANCOVA) approach. This
approach uses time one measurements as the covariates in the model with either raw
scores at time two or change scores as dependent variables. However, the ANCOVA
approach is appropriate in the case of an intervention effects in randomized controlled
trial, but is not recommended for observational studies [148, 151, 152].
Considering the advantage that the raw change score approach over time could
remove the influence of cross-sectional correlations between subjects and allow the
analysis to exclusively focus on within-subject effects [148], we applied the raw change
score modeling approach for the longitudinal mediational modeling analysis in the
present study (Chapter 8 and Chapter 9).
51
2.8.3 Moderation effects and statistical strategies
A moderator is a variable that affects the direction and/or strength of the
relationship between an independent variable and a dependent variable (Figure
2-7a) [139, 153]. In statistics, a moderation effect can be tested by using an interaction
term, whereby the effect of an independent variable on a dependent variable varies by
the level of another variable. There are three causal paths that feed into the outcome
variable (Figure 2-7b): the independent variable (path a), the moderator (path b) and
the interaction of the independent variable and the moderator (path c). The moderating
hypothesis is supported if the regression coefficient of the interaction term (path c)
onto the dependent variable is statistically significant. Moderation hypothesis testing
can signify how the association between an independent variable and a dependent
variable varies in terms of the level of the third variable.
Independent
variable a
Moderator
b Dependent
Moderator
variable
Independent Dependent
variable variable c
Predictor
×
Moderator
(a) Conceptual model (b) Statistical model
Figure 2-7 Illustration of moderating effect
52
Chapter 3 Participants and methods
3.1 Study design
Data of this thesis was obtained based on a two-phase study (see Figure 3-1).
Phase I, a cross-sectional validation study, aimed to develop and validate a new
instrument (named as Social Networking Activity Intensity Scale, SNAIS) for
measuring social networking use intensity by considering diverse social networking
activities and functions among Chinese junior middle school students.
By using instruments validated in the phase I study, we conducted a nine-month
longitudinal study (Phase II), aiming to objectively investigate the causal directions
between Internet addictive behaviors (‘SNA’ and IA) and depression, prediction of
social networking use intensity and emotional connection to social networking onto
depression, and also to test the potential underlying mechanisms for the association
between social networking use intensity and depression (i.e. mediation, suppression,
and moderation effects of psychosocial status and mediation effects of Internet
addictive behaviors).
53
Panel, literature review,
focus group, compile items
Pilot survey and scale revision � Validation analysis

Phase Ⅰ: - Construct validity: EFA/CFA
Validation survey - Content validity: Spearman correlation
- Test-retest reliability: ICC
Validation survey among junior high
- Internal reliability: Cronbach alpha
school students in Guangzhou
Revise and get the validated scales
Background variables Psychosocial scales
Baseline survey for junior high

school students
Phase Ⅱ: 9-month
Longitudinal study Data analysis
interval
Follow-up survey
Figure 3-1 Study design flowchart
3.2 Study site
The present study was conducted in Guangzhou, Mainland China. Guangzhou,
the capital of the Guangdong province, is located in the Pearl River Delta Economic
Zone in China with a total permanent population of 12.8 million in 2012, in which
there were 373,481 students in 354 junior middle schools. As one of the three
megacities in Mainland China, the Internet penetration rate was a high 71.7% among
the general population in 2010, which was more than two-fold of the national level
(Figure 3-2). Weekly time spent on the Internet was around 23.9 hours in Guangzhou
in 2010.
Geographically, administrative regions in Guangzhou consist of ten districts:
54
Yuexiu, Haizhu, Liwan, Tianhe, Baiyun, Huangpu, Huadu, Panyu, Nansha and
Luogan, and two counties: Conghua and Zengcheng. Based on geography and
educational development conditions, these districts and counties were divided into
three regions: Core region (Yuexiu, Liwan, Tianhe, and Haizhu), Suburb region
(Panyu, Huangpu, Baiyun, and Luogang), and Outer suburb region (Nansha, Huadu,
Cunhua and Zengcheng) (Figure 3-3)
80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0
2009 2010
Guangzhou 65.0 71.7
Guangdong 50.9 55.2
National 28.9 34.3
Figure 3-2 Internet penetration rate in Guangzhou, Guangdong and national level
Guangdong
Guangzhou
Figure 3-3 Location of the study site
55
3.3 A summary of rationale to traget junior middle school students
We targeted junior middle school students as our study population based on the
following considerations: (1) Adolescence represents the second fast growth spurt
after infancy [154]. Junior middle school students are an important subpopulation of
adolescents. As a key stage of early adolescence, the rapid physical (e.g. early
puberty), psychological (e.g. concrete thinking but early moral concepts) and social
(e.g. emotional seperation from parents, start of strong peer identification; early
exploratory behaviours) developemtal changes produce specific disease patterns,
unusual presentations of symptoms, unique communication and management
challenges, and they gradually establish identity and autonomy in the context of the
socialcultural environment [1, 155]. During this period, adolescents are at high-risk of
suffering from many potential psychological and mental health problems. Moreover,
adolescence has been increasingly recognised as a life period that poses specific
challenges for treating disease and promoting health. (2) The high prevalence of
depression among adolescents has been reported in previous studies. The rates of
depression rise sharply after entering puberty with immediate and long term risks [4].
It has been observed that risk of depression is low in childhood but increasing
substantially with adolescence [5], and the meaningful risk begins in the early
teens [156]. (3) Online social networking use has received considerable popularity
among adolescents in recent years. A large-scale school-based survey among students
in middle and high schools reported that majority (81.5%) of students are dialy users
of online social networking [157], and proportion of students who reported daily online
social networking use of more than 2 hours significantly increased with students’
grade [158]. (4) A few of cross-sectional studies (see our literature review in chapter 2
56
in detail) have investigated the detrimental effects of online social networking use on
mental health among post-secondary (college and university) students [52, 53]. However,
little is known about the effects of online social networking use on mental health in
middle and high school students. Therefore, from the perspective of early prevention,
it is necessary and important to investigate the effects of online social networking use
on mental health (i.e. depression) among junior middle school students.
3.4 Phase I: Validation study
3.4.1 Participants and sampling
For the validation study in phase I, eligible participants were students from grade
seven to grade nine in public secondary schools in Guangzhou who were willing to
voluntarily participate in the survey. We excluded the migrant and private secondary
schools, considering the potential differences of cultural and familial environments
from public schools.
The survey was conducted in October of 2013, and a stratified cluster sampling
method was used to select participants. First, two areas were selected from
Guangzhou, including one urban and one rural country. Second, two public secondary
schools, one from the urban area and one from the rural area, were conveniently
selected. The students in the selected classes from the two schools were voluntarily
recruited to participate in the validation survey.
During this period, a total of 1088 students voluntarily completed the
questionnaire. As a measure of quality control, 73 cases with at least one scale having
more than 20% of its items being missing (SNAIS, emotional connection to social
networking, ‘SNA’, and IA) were excluded from data analysis. The remaining 1015
57
questionnaires had only 0.11% to 3.74% of item responses with missing values, which
were imputed by using the modal value of the other items of the same scale. Finally,
data from 910 (89.7%) of the 1015 students who were social networking users
(defined as those currently possessing at least one social networking account) were
used in the validation analysis.
3.4.2 Data collection procedure
Assisted by teachers of the selected schools and the Institute of Health Education
in Guangdong province, leaflets with the significance and procedures of the study
were approached to the selected school principals and students simultaneously several
days before the data collection. School consent and permission to administer the
in-school survey was approved by school principals prior to data collection.
An anonymous structured questionnaire was self-administered by the participants
in the absence of teachers in classroom settings, and supervised by a well-trained field
worker from our research team. Information on the study’s background and purpose
and confidentiality of the study was printed in the cover page of the questionnaire.
The nature of voluntary participation was also clearly announced by the field workers,
and students were informed that they have the right to terminate their participation in
the study at any time. Verbal informed consent was also obtained from participants
before they filled out the questionnaires. After distributing the questionnaires, field
workers stood by for enquiries and supervised the students to complete the
questionnaires. The questionnaires were checked for completeness and collected
immediately one-by-one by the field workers upon students’ submissions. All
questionnaires were sealed in prepared envelopes by our field workers in the presence
of students in the classroom, and names of the school and class, field workers, as well
58
as the date of data collection were wrote down on the envelopes as a means of quality
control. During the data collection period, field workers were also supervised by a
researcher in our team. Additionally, to gauge test-retest reliability, 114 students from
the urban secondary school were invited to complete the same questionnaire after two
weeks.
No incentive was given to students. The study was approved by the Survey and
Behavioral Research Ethics Committee of the Chinese University of Hong Kong. The
flowchart of phase I data collection is presented in Figure 3-4.
Eligible secondary schools
Informed consent from school principals
Two public schools selected

(one urban and one rural)
Students in the selected classes completed the

surveys
73 surveys
excluded
1015 valid surveys
Response rate=93.3%
910 online social networking users for

validation analysis
114 participants retested after 2 weeks
Figure 3-4 Flowchart of phase I data collection
59
3.4.3 Measurement
The questionnaire for the phase I validation study is included in Appendix I.
3.4.3.1 Development of online Social Networking Activity Intensity Scale (SNAIS)
A literature review on online activities and related measures on social networking
use was conducted [54, 55, 84, 114], and a pool of 38 items was extracted. A panel,
consisting of a behavioral scientist, two psychologists and an epidemiologist, held
several meetings, and eliminated some overlapping items while combined others with
similar meanings. The 14 items hence remained were written as questions: “How often
had you performed the following online social networking activities in the last month?”
The item response scale included six categories: 0 (never), 1 (few), 2 (occasional), 3
(sometimes), 4 (often) and 5 (always). The 14-item scale was then pilot tested among
85 students from a selected secondary school in the urban area (52.9% were grade
seven and 47.1% were grade eight students). A total of 77 students were identified as
social networking users. In the pilot study, the average item response rate was 98.3%
(ranging from 96.1% to 98.7%), and the item mean ranged from 1.1 to 3.8, suggesting
satisfactory acceptance and variance. All students had a sufficient understanding of
the items and found the questions not difficult to answer. Four additional students
answered an open-ended question on whether there were any other types of online
social networking activities unmentioned. Answers obtained included “chatting with
strangers”, “online shopping”, and “reading/searching materials”. These suggested
activities, however, overlapped with the current 14 items (table 6-1 in Chapter 6). In
order to further refine the existing items and response scale, individual interviews
were conducted among 20 students (10 from seventh grade and 10 from eighth grade)
by the researchers of our team to further test the feasibility and content of the scale.
60
The interviewees expressed that the 14 items captured the common functions and
activities of online social networking use and were suitable to the Chinese cultural
context. Based on their comments, we removed the item response category of
“occasional” to improve the discrimination of the response options. Thus a 5-point
scale was used with a higher score indicating higher intensity of social networking
use.
3.4.3.2 Other measures
3.4.3.2.1 Social-demographic variables
All participants were asked about their gender, grade, parental educational
attainment, and smartphone ownership.
3.4.3.2.2 Characteristics related to social networking use
Participants were asked whether they currently possess any social networking
accounts. Those providing an affirmative answer to the question were asked about
their experiences with social networking use, including the duration of social
networking use, number of days per week using social networking on average,
amount of time spent on social networking on a typical day, and number of online
social networking friends.
3.4.3.2.3 Emotional connection to social networking (EC)
Emotional connection to social networking was measured by six-attitudinal items
adapted from the Facebook Intensity Scale in the Ellison et al study [51]. The word
“Facebook” was replaced with “online social networking”. It was translated to
English and then translated back to Chinese. The original items were translated from
61
English into Chinese, and then translated back to English by two bilingual researchers.
Five-point Likert scales were used, ranging from 1 (strongly disagree) to 5 (strongly
agree). Higher scores indicated higher levels of emotional connection to social
networking. A sample item was “Online social networking is part of my everyday
activity”. In our sample, an exploratory factor analysis revealed a one-factor solution
for this scale, explaining 60.0% of the total variance, and the factor loadings of the six
items ranged from 0.73 to 0.85. The Cronbach’s alpha coefficient was 0.87, and the
test-retest intra-class correlation coefficient was 0.78.
3.4.3.2.4 Social Networking Addiction (SNA)
‘SNA’ was measured by a modified version of the Facebook Addiction Scale [48],
which included eight items describing addictive symptoms (i.e. cognitive and
behavioral salience, conflict with other activities, euphoria, loss of control,
withdrawal, and relapse and reinstatement). Similarly, the word “Facebook” was
replaced with “online social networking”, and translation and back-translation
processes were implemented by two bilingual researchers. Response categories rated
from 1 (not true) to 5 (extremely true), thus the total score of the scale would range
from 8 to 40, with a higher score indicating a higher level of addictive tendency to
social networking. A sample item was “I have difficulties in focusing on my academic
work due to my social networking use”. In this study, an exploratory factor analysis of
this scale showed a one-factor solution that explained 51.2% of the total variance,
with high factor loadings ranging from 0.64 to 0.77. The Cronbach’s alpha coefficient
was 0.86, and the test-retest intra-class correlation coefficient was 0.84.
62
3.4.3.2.5 Internet addiction (IA)
IA was measured by the 8-item Young’s Diagnostic Questionnaire (YDQ), which
involved “yes/no” response categories. Participants who provided five or more “yes”
answers were classified as cases of IA [19, 159]. The YDQ is easy to administer. The
scale has commonly been used in the Chinese student population and showed
acceptable validity and reliability [160, 161]. In our sample, the exploratory factor
analysis revealed a one-factor solution for this scale that explained 37.8% of the total
variance, and the factor loadings ranged from 0.46 to 0.69. The scale showed good
internal reliability (Cronbach’s α=0.73).
3.4.4 Statistical Analysis
Data were double-entered and the two datasets were compared using the EpiData
software (EpiData for Windows; The EpiData Association, Odense, Denmark). The
confirmatory factor analysis was conducted using LISREL 8.70, while other data
analyses were performed using Statistical Analysis System software (SAS 9.3 for
Windows; SAS Institute Inc., NC). Statistical significance was defined by a two-tailed
p value <0.05.
The detailed statistical methods used were described in the Chapter 6.
63
3.5 Phase II: A nine-month longitudinal study
3.5.1 Participants and Sampling
For the nine-month longitudinal study, participants were grade seven and grade
eight students from public junior middle schools in Guangzhou, China. We excluded:
(1) Grade nine students based on the consideration that they need to prepare for the
high school entrance examination and would leave their schools at the end of the
academic year. Therefore, there would be practical difficulties following-up with
these students later on; (2) Migrant and private junior middle schools due to potential
differences in cultural and familial environments; (3) those unwilling to participate in
the study. The participants enrolled in the longitudinal study differed from those in the
validation study in phase I.
A stratified cluster sampling method was applied to recruit participants.
1) One district (or country) was conveniently selected from three regions
(Core region, Suburb region and Outer suburb region), respectively.
2) Three junior middle schools were conveniently selected from each selected
district/county. In total, nine secondary schools were selected.
3) Students in the grades seven and eight in the selected schools were all
voluntarily recruited for baseline and subsequent follow-up surveys.
The baseline survey was conducted between March and April 2014, while the
follow-up survey was conducted during December 2014 to January 2015. After the
sampling procedures were applied, 5472 students completed the survey at baseline
(n=2645 in grade seven, 48.34%; n=2872 in grade eight, 51.66%), and 5527 students
64
were investigated at follow-up (Figure 3-5).
As a measure of quality control, we inspected the data for missing values. An
invalid response was defined as those with any scales having more than 25% missing
values. As a result of this process, 107 participants at baseline and 134 participants at
follow-up were excluded from the study. The response rates were 98.04% at baseline
and 97.58% at follow-up. For the remaining data of the 5365 participants at baseline
and 5393 participants at follow-up, the mean percentages of missing values per item
were 0.27% (range=0.00 to 1.19%) at baseline and 0.42% (range=0.00 to 3.15%) at
follow-up. The missing values were imputed by the mean values of the other items
from the same scale.
Among 5365 students investigated at baseline, 4951 participants (92.28%)
reported that they were social networking users, while 4905 out of 5393 (90.95%)
participants were social networking users at follow-up. After the matching process,
4871 participants were investigated at both baseline and follow-up. The 4237
participants who were social networking users at both baseline and follow-up
constituted the longitudinal sample (Figure 3-5).
65
Baseline Follow-up
Baseline Follow-up
5472 students were 5527 students were
investigated from 9 schools investigated from 9 schools
107 students were 134 students were
excluded due to excluded due to
incomplete data incomplete data
5365 data were valid 5393 data were valid
4951 students were 414 students did not use 4905 students were 487 students did not use
social networking users social networking social networking users social networking
4871 students were investigated

at both baseline and follow-up
4515 students were social 356 students were non-

networking users at baseline users at baseline
4237 students provided data 278 students who were users at 212 students who were 144 students who were
related to social networking at baseline did not provide data related non-users at baseline non-users at both
both baseline and follow-up to social networking at follow-up became users at follow-up baseline and follow-up
4237 in baseline follow-up

constituted longitudinal sample
Figure 3-5 Flowchart of participants in the longitudinal study
3.5.2 Data collection
The data collection procedures were similar to those used in the phase I study.
We collaborated with the Center for Health Promotion of primary and secondary
school in Guangzhou, which has a strong relationship with the secondary schools in
Guangzhou. As the collaboraters of our research, they provided strong support and
coordinations for school contact in the fieldwork. No incentives were provided to the
Center. School consent and permission for the in-school survey were approved by
school principals before the survey was administered. Information on the study’s
background and purpose and confidentiality of the study was printed on the cover
66
page of the questionnaire. The nature of voluntary participation was clearly mentioned
in an announcement and students were informed that they had a right to terminate
their participation in the study at any time. Verbal informed consent was also obtained
from participants before the survey.
Field workers were university students (upperclassmen in Year 4 and 5) majoring
public health at Sun Yat-sen University in Guangzhou. They all received
comprehensive training to conduct field work by the research team from the Chinese
University of Hong Kong (CUHK). After the training, they were also supervised in
the field by a researcher from CUHK during the data collection period.
The well-structured questionnaire used in the longitudinal study was designed by
a panel consisting of epidemiologists and psychologists. The questionnaire was pilot
tested among 46 junior middle school students in grade seven for the feasibility and
acceptability before conducting a baseline survey. In accordance with the feedback
from the pilot test, the questionnaire was then finalized by the panel for the baseline
survey and subsequent follow-up. Both surveys used similar questionnaires.
Since an anonymous approach was used to maximize protection of personal
information of the participants, students were asked to write down multiple codes on
the cover page of the questionnaire to assist with correctly matching the two surveyed
questionnaires completed by each student at baseline and follow-up. The codes
included the last four digits of participants’ home telephone number, last four digits of
both parental mobile phone numbers, last four digits of their identity card, participants’
date of birth, and the last letter of self and both parental spell-name. It was thought
that the use of such codes would make it difficult to identify the specific students, thus
maximizing the confidentiality and privacy of participants.
67
No incentives were given to students. As a feedback, a simple report of the
survey (e.g. including research background, methogology, characteristics of Internet
and online social networking use, prevalence of depression and IA) was provided to
the specific surveyed schools seperately. The study was approved by the Survey and
Behavioral Research Ethics Committee of the Chinese University of Hong Kong.
3.5.3 Measurements
The questionnaire for the longitudinal study is enclosed in Appendix II.
3.5.3.1 Background variables
All participants were asked about their gender, grade, both parental education
levels, family financial situation, living arrangement with parents, self-reported
academic performance and perceived study pressure at baseline.
3.5.3.2 Social networking-related scales
The measures for social networking use include online Social Networking
Activity Intensity Scale (two subscales: EFUI and EFUI), Emotional Connection to
social networking (EC) and ‘Social Networking Addiction (SNA)’ scale. These three
scales have been well-validated in the phase I study and were measured at both
baseline and follow-up. In the longitudinal sample, the three scales showed acceptable
reliability. For SNAIS and its two subscales (SFUI and EFUI), the Cronbach’s α were
0.87, 0.88, and 0.60 at baseline, and 0.89, 0.89, and 0.62 at follow-up respectively.
For EC, the Cronbach’s α were 0.88 at baseline and 0.89 at follow-up. For ‘SNA’, the
Cronbach’s α were 0.86 at baseline and 0.89 at follow-up. (Table 3-1)
Since there is no established cutoff value for the ‘SNA’ scale, participants scoring
68
in the top 10% of scale scores would be defined as addicted to social networking. This
classification strategy has been used in previous studies [162, 163].
3.5.3.3 Internet addiction (IA)
IA was measured using the 8-item Young’s Diagnostic Questionnaire (YDQ) at
both baseline and follow-up. The YDQ was detailed and descripted in the phase I
validation study. In the longitudinal sample, the Cronbach’s α were 0.66 at baseline
and 0.74 at follow-up, showing acceptable reliability (Table 3-1).
3.5.3.4 Psychosocial variables
3.5.3.4.1 Friendship quality
Friendship quality was measured using a peer-relationship subscale of Children
and Adolescent Quality of Life Scale [164], developed by Tong Ji Medical College of
Huazhong University of Science and Technology in China, and widely used in
Chinese adolescents[165]. The peer-relationship subscale includes 6 items with
four-point Likert scale, and a sample item is “Is your friend care for you?” The total
score ranges from 6 to 24, with a higher score indicating a higher quality of friendship.
In the longitudinal sample, the Cronbach’s α were 0.82 at baseline and 0.83 at
follow-up, showing good reliability.
3.5.3.4.2 Perceived social support
Perceived social support was measured by the 12-item Multidimensional Scale of
Perceived Social Support (MSPSS) [166]. The Chinese version of MSPSS revealed two
subscales: perceived family support (4 items) and perceived friend support (8
items) [167]. A sample item was “I get the emotional help and support I need from my
69
family”. Each item was rated on a seven-point Likert scale ranging from 1=very
strongly disagree to 7=very strongly agree, with higher scores indicating higher levels
of perceived social support. In the longitudinal sample, the Cronbach’s α were 0.93
for perceived friend support and 0.90 for perceived family support at baseline, while
they were 0.95 and 0.91 respectively at follow-up, showing good internal reliability.
3.5.3.4.3 Adolescent-parents conflict
Adolescent-parents conflict were assessed using the adolescent-parents conflict
questionnaire from the Fang XY et al study [168]. The questionnaire assessed seven
aspects of the adolescent-parents conflict, including academic, housework, spending
money, daily life arrangement, appearance, relationship between family members, and
individual privacy. For each kind of conflict, a detailed explanation was given to
respondents. Respondents were asked to rate the frequency of conflict with both father
and mother separately for each aspect of the adolescent-parents conflict in the
previous month. Each item was rated on a five-point Likert scale ranging from
1=never to 5=almost every day. Higher scores indicate higher frequency of conflict
with the father/mother. The adolescent-parents conflict questionnaire showed
acceptable reliability among Chinese adolescents [168, 169]. In our longitudinal study,
the Cronbach’s α were 0.84 for adolescent-father conflict and 0.85 for
adolescent-mother conflict at baseline, while they were 0.86 and 0.87 at follow-up
respectively.
3.5.3.4.4 Social non-confidence
The four-item social confidence subscale of the Social Skills Scale (SSS) was
used to measure level of the social non-confidence [170]. Psychometric properties of
70
the Chinese version of SSS have been previously evaluated among college
students [171]. A sample item on the social non-confidence subscale is “I am very
sensitive to criticism from others”, with a five-point Likert scale ranging from
1=definitely not like me to 5=exactly like me. Higher scores indicate higher social
non-confidence, meaning there is a great concern about what others think about
oneself as well as a strong likelihood of feeling nervous on social occasions. In the
longitudinal sample, the Cronbach’s α were 0.60 at baseline and 0.64 at follow-up.
3.5.3.3.5 Self-esteem
Self-esteem was measured using the 10-item Rosenberg Self-Esteem Scale
(RSES) [172], which is a widely used self-reported instrument to measure global
self-worth by both positive and negative feelings about oneself. A sample item is “I
feel that I have a number of good qualities.” The RSES was unidimensional and all
items were answered using a four-point Likert scale ranging from 1=strongly agree to
4=strongly disagree. The Chinese version of RSES has been reported to have good
consistency reliability among adolescents [173]. In the longitudinal sample, the
Cronbach’s α were 0.79 at both baseline and follow-up, showing good consistency
reliability.
3.5.3.5 Depression
Depressive symptomatology was assessed using the Chinese version of the
20-item Center for Epidemiological Studies-Depression Scale (CES-D). The CES-D
is one of the most commonly used self-report instruments for epidemiological studies
to screen depressive symptomatology. The CES-D was originally designed by Radloff
in 1977 to measure current level of depressive symptomatology in the general adult
71
population, with emphasis on the affective component, depressd mood [174]. In 1991,
Radloff applied the CES-D in high school and college students and supported its
acceptability and suitability in adolescents, including respondents as young as junior
high school [175]. The good psychometric properties (i.e. reliability and validity) of
CES-D among adolescents have also been reported in China (e.g. Hongkong, Taiwan
and Mainland) [176-179]. The CES-D comprises 16 negative affect and 4 positive affects
items, such as “I felt depressed,” and “I felt lonely”. Participants were asked about
rating the frequency of occurrence of the symptoms such as restless sleep and feeling
lonely during the previous week. All items were responded on a four-point Likert
scale ranging from 0=rarely or none of the time (less than 1 day) to 3=almost or all of
the time (5-7 days). The total score could range from 0 to 60, with a higher score
reflecting more depressive symptoms [174]. In the longitudinal sample, the Cronbach’s
α were 0.86 at baseline and 0.87 at follow-up, showing good internal reliability. The
respondents with scores equaling to or larger than 21 were defined as depressed [180].
3.5.4 Quality control
A research panel, consisting of one behavioral scientist, two psychologists, and
an epidemiologist, were formed for this study to conduct the study design, construct
the questionnaire, write up the Standard Operation Procedure (SOP) for data
collection, train the field workers, and to analyze the data. Moreover, a pilot study
among 46 grade seven students was conducted for testing the feasibility of the
questionnaire.
To ensure quality and integrity of the data collected in our study, quality control
was performed in all steps in the baseline and follow-up survey. All field workers
72
were recruited from senior university students and were well-trained so that they were
very familiar with the questionnaire and process of data collection.
3.5.5 Statistical strategies
The database was created with EpiData 3.0 software. All statistics are two-sided,
and the statistical significance was set as p≤0.05. The detailed statistical methods are
described in the subsequent chapters. SAS version 9.2 (SAS Institute Inc., Cary, NC,
USA) and AMOS 20.0 were used for statistical analyses.
73
Table 3-1 Psychometric properties of the scales used in the longitudinal study
Number of Baseline Follow-up
items Cronbach’s α Cronbach’s α
Social networking variables
SFUI 10 0.88 0.89
EFUI 4 0.60 0.62
Emotional connection 6 0.88 0.89
‘SNA’ 8 0.86 0.89
Internet addiction (IA) 8 0.66 0.74
Psychosocial variables
Friendship quality 6 0.82 0.83
Perceived family support 4 0.90 0.91
Perceived friend support 8 0.93 0.95
Rosenberg self-esteem 10 0.79 0.79
Adolescent-father conflict 7 0.84 0.86
Adolescent-mother conflict 7 0.85 0.87
Social non-confidence 4 0.60 0.64
Mental health
CES-D 20 0.86 0.87
SFUI: Social Function Use Intensity; EFUI: Entertainment Function Use Intensity; SNA:
Social Networking Addiction scale; CES-D: Center for Epidemiological
Studies-Depression Scale.
74
Chapter 4 Description of the independent and dependent variables
in the longitudinal sample
4.1 Objectives
In this chapter, we describe the distributions of background characteristics of
participants obtained at baseline assessment and sample/subsamples involved in the
longitudinal analysis. Of particular importance are prevalence and incidence of
depression, ‘SNA’, and IA, and the remission rate from depression, ‘SNA’, and IA
during the follow-up period. Changes in the scale scores of independent variables
(including social networking related scales, psychosocial scales) from baseline to
follow-up in the longitudinal sample are also presented.
4.2 Statistical methods
Attrition analysis was performed by comparing the differences of background
characteristics at baseline between participants who remained in the follow-up and the
ones who were lost to follow-up by Chi-square test. The students who were social
networking users and were investigated both at baseline and follow-up constituted the
longitudinal sample (n=4237). The prevalence of depression, ‘SNA’, and IA were
calculated both at baseline and follow-up in the longitudinal sample. The differences
in the prevalence between genders were evaluated using the Chi-square test, while the
differences between baseline and follow-up were assessed by the McNemar test. The
differences of background variables between SNA and non-SNA cases (or IA vs.
non-IA cases) at baseline were estimated by univariate multilevel logistic regression
models, and ORu along with 95% confidence interval (95% CI) were derived. The
75
mean score differences of IA (or SNA) at baseline between four groups were assessed
by one-way ANOVA.
The age and gender-specific incidences were calculated by the number of events
divided by person-years of exposures; exposure for the event was estimated by half of
the duration between two assessments. Normal approximation for Poisson distribution
was used to calculate the 95% confidence interval (95% CI) of the point
incidence/remission estimates.
Continuous variables (including scores of social networking related scales, and
psychosocial scales) were described using the mean and standard deviation both at
baseline and follow-up in the longitudinal sample. The mean score differences of
social networking related scales and psychosocial scales between baseline and
follow-up were compared by paired t-test.
4.3 Results
4.3.1 Background characteristics of the entire sample at baseline assessment
A total of 5365 students were investigated at baseline assessment, 52.79% were
males, and 48.3% were from the seventh grade. Over 20% of the students reported
that their parents (either father or mother) achieved college education level or above.
The majority of students (87.83%) lived with both parents. About half of the students
(46.95%) perceived that their family financial situations were good/very good.
Additionally, more than one fifth of the students (21.47%) reported lower academic
performance, and 23.84% of students perceived heavy or even very heavy study
pressure (Table 4-1).
76
4.3.2 Attrition analysis for loss to follow-up
Around 9.2% of participants (494/5365) were lost to follow-up during the
nine-month longitudinal period. The results of the attrition analysis were presented in
Table 4-2. It showed that participants who were lost to follow-up were more likely to
be social networking non-users. In addition, students who were lost to follow-up were
more likely to be males, in their junior year of school, not living with both parents,
perceived poor/very poor family financial situation, and reported lower academic
performance (Table 4-2). Moreover, participants who were depressed at baseline were
more likely to be lost to follow-up compared with those who were non-depressed
(p=0.016, Table 4-2).
4.3.3 The longitudinal sample (n=4237)
The longitudinal sample was comprised of participants who were social
networking users (see chapter 3 for definition), and were surveyed both at baseline
and follow-up (n=4237). Six subsamples were also defined for subsequent data
analysis based on the status of depression, ‘SNA’, and IA, including subsamples Ia,
IIa, IIIa and Ib, IIb, IIIb (Figure 4-1).
77
Subsample Ia (n=3196), which consisted of
participants who were non-depressed at baseline
Subsample Ib (n=1041), which consisted of

participants who were depressed at baseline
Subsample IIa (n=3657), which consisted of

participants who were non-SNA at baseline
Longitudinal sample
(n=4237)
Subsample IIb (n=580), which consisted of
participants who were SNA at baseline
Subsample IIIa (n=3959), which consisted of

participants who were non-IA at baseline
Subsample IIIb (n=278), which consisted of

participants who were IA at baseline
Figure 4-1 Illustration of the subsamples
4.3.3.1 Background characteristics of the longitudinal sample and subsamples
The distributions of background characteristics in the longitudinal sample and
three subsamples (Ia, IIa, and IIIa) were presented in Table 4-3. In the entire
longitudinal sample (n=4237), about half of the students (49.68%) was males, 47.46%
of the students were in the seventh grade, and over 96% of students perceived medium
or above family financial situation. The majority of students (88.4%) lived with both
parents. In addition, over one fifth of students (20.11%) reported lower academic
performance, and about 23.44% of students perceived heavy or even very heavy study
pressure. The three subsamples showed similar distributions to the entire longitudinal
sample in terms of background characteristics (Table 4-3).
78
4.3.3.2 Comparison of background characteristics between those with ‘SNA’ and
those without ‘SNA’ at baseline
The table 4-4 presented the differences of background characteristices between
participants with SNA and without SNA at baseline. The unvariate analysis showed
that participants who were classified as ‘SNA’ were more likely to be those who
reported poor/very poor family financial situation (ORu=1.91, 95%CI: 1.21~3.01;
reference group: perceived good/very good family financial situation), those who
reported medium (ORu=1.81. 95%CI: 1.44~2.28) or lower academic performance
(ORu=3.39, 95%CI: 2.63~4.34; reference group: upper academic performance), and
those who perceived general (ORu=1.59, 95%CI: 1.20~2.11) or heavy/very heavy
study pressure (ORu=2.97, 95%CI: 2.21~3.99; reference group: perceived nil/light
study pressure). In addition, females were less likely to have ‘SNA’ as compared to
males (ORu=0.72, 95%CI: 0.60~0.86) (Table 4-4).
4.3.3.3 Comparison of background characteristics between those with IA and
those without IA at baseline
The results from univariate multilevel logistic regression analyses showed that
participants who were classified as IA cases at baseline were more likely to be those
who reported medium (ORu=1.60, 95%CI: 1.16~2.18) or lower academic
performance (ORu=2.70, 95%CI: 1.93~3.78; reference group: upper academic
performance), and those who perceived heavy/very heavy study pressure (ORu=2.92,
95%CI: 1.95~4.37; reference group: perceived nil/light study pressure) (Table 4-5).
Besides, females were less likely to have IA as compared to males (ORu=0.59, 95%CI:
0.46~0.76) (Table 4-5).
79
4.3.3.4 Discriminality between IA and ‘SNA’
The correlation between IA and SNA scores at baseline was less than 0.5 (α=0.49,
p<0.001), indicating a low positive correlation [181]. Moreover, the data showed that
only a small group of students (3.63%) were classified with both IA and ‘SNA’ at
baseline (Table 4-6).
The one-way ANOVA further revealed that students with both IA and “SNA”
have the highest mean IA score (F=1366.43, p<0.001; Table 4-6) and ‘SNA’ score
(F=1261.94, p<0.001; Table 4-6) at baseline. These findings indicated that IA and
‘SNA’were well discriminated by the YDQ and social networking addiction scale
respectively.
4.3.3.5 Depression, ‘SNA’, and IA prevalence in the longitudinal sample (n=4237)
4.3.3.5.1 Depression prevalence
The prevalence of depression was 24.57% at baseline, and it was significantly
higher in females than that in males (22.66% for males vs. 26.45% for females,
χ2=8.226, p=0.004). At follow-up, it was 26.55%, the gender difference became
non-significant (25.32% for males vs. 27.77% for females, χ2=3.251, p=0.071). The
prevalence of depression at follow-up was significantly higher than that at baseline
(paired χ2=7.459, p=0.006) (Table 4-7).
4.3.3.5.2 ‘SNA’ prevalence
The prevalence of ‘SNA’ was 13.69% at baseline, which was significantly higher
in males than that in females (15.77% for males vs. 11.63% for females, χ2=15.363,
p<0.001); while the prevalence at follow-up was 13.55%, the gender difference of
80
which was not statistically significant (14.25% for male vs. 12.85% for female,
χ2=1.773, p=0.183). Moreover, there was no significant difference for ‘SNA’
prevalence between baseline and follow-up (paired χ2=0.053, p=0.818) (Table 4-7).
4.3.3.5.3 IA prevalence
The prevalence of IA was 6.56% at baseline and 7.34% at follow-up, there was a
slight increase in IA prevalence during the follow-up period, but the difference was
non-significant (paired χ2=3.033, p=0.082). The prevalence was significantly higher in
males than that in females both at baseline and follow-up (8.17% vs. 4.97%,
χ2=17.682, p<0.001 at baseline, and 8.31% vs. 6.38%, χ2=5.828, p=0.016) (Table 4-7).
4.3.3.6 Depression, ‘SNA’, and IA incidence based on the subsamples Ia~IIIa
4.3.3.6.1 Depression incidence in the subsample Ia (n of subsample=3196)
In the longitudinal sample (n=4237), there were 3196 students who were
non-depressed at baseline (subsample Ia). Among these non-depressed cases at
baseline, there were 16.11% (515/3196) new incident depression cases during the
nine-month period, with 16.34% for males and 15.88% for females (χ2=0.125,
p=0.724).
The overall incidence of depression was 23.37 (95% CI: 21.43~25.48) per
100-person-years. The incidence among males was 23.72 (95% CI: 21.04~26.75) and
among females was 23.00 (95% CI: 20.32~26.04); However, gender difference was
non-significant. There was non-significant trend in age (Table 4-8).
4.3.3.6.2 ‘SNA’ incidence in the subsample IIa (n of subsample=3657)
Among participants who were non- ‘SNA’ at baseline (subsample IIa), around
81
9.16% (335/3657) of them transformed from non- ‘SNA’ to ‘SNA’ cases during the
follow-up period, with 9.42% for males and 8.92% for females (χ2=0.277, p=0.599).
The overall incidence of ‘SNA’ was 12.80 (95% CI: 11.50~14.25) per
100-person-years, with 13.18 (95% CI: 11.32~15.34) for males and 12.44 (95% CI:
10.70~14.48) for females. There were no significant differences in incidence in terms
of age and gender (Table 4-9).
4.3.3.6.3 IA incidence in the subsample IIIa (n of subsample=3959)
Among participants who were non-IA at baseline (subsample IIIa), about 4.95%
(196/3959) of them transformed from non-IA to IA cases during the follow-up period,
with 5.38% for males and 4.54% for females (χ2=1.481, p=0.224).
The overall incidence of IA was 6.77 (95% CI: 5.88~7.79) per 100-person-years,
with 7.37 (95% CI: 6.08~8.93) for males and 6.20 (95% CI: 5.05~7.60) for females.
The incidence of IA significantly decreased with age among females (from 8.23 in the
13 or younger age group to 4.08 in the 15 or older age group, χ2=5.15, p=0.023)
(Table 4-10). No gender difference was observed for incidence of IA.
4.3.3.7 Incidence of remission from depression, ‘SNA’, or IA based on the
subsamples Ib~IIIb
4.3.3.7.1 Remission from depression in subsample Ib (n of subsample=1041)
Among participants who were depressed at baseline (subsample Ib), around
41.40% of them were in remission from depression during the follow-up period. The
percentage of remitted cases in males and females were 44.03% and 39.18%
respectively (χ2=2.496, p=0.114).
82
The overall incidence of remission from depression was 69.61 (95% CI: 63.34,
76.51) per 100-person-years, with 75.27 (95% CI: 65.57, 86.17) per 100-person-years
in males and 64.98 (95% CI: 59.95, 74.13) per 100-person-years in females (Table
4-11). There were no significant differences for incidence of remission from
depression in terms of age and gender.
4.3.3.7.2 Remission from ‘SNA’ in subsample IIb (n of subsample=580)
Among participants who were ‘SNA’ at baseline (subsample IIb), around 58.79%
of ‘SNA’ cases spontaneously remitted during the follow-up period, with 59.94% for
males, and 57.26% for females (χ2=0.421, p=0.516).
The overall incidence of remission from ‘SNA’ was 111 (95% CI: 100, 123) per
100-person-years, with 114 (95% CI: 99,131) per 100-person-years in males and 107
(95% CI: 91,126) per 100-person-years in females (Table 4-12). There were no
significant differences for incidence of remission from ‘SNA’ in terms of age and
gender.
4.3.3.7.3 Remission from IA in subsample IIIb (n of subsample=278)
Among participants who were IA at baseline (subsample IIIb), about 58.63% of
IA cases remitted during the follow-up period, 58.72% in males and 58.49% in
females (χ2=0.001, p=0.970) (Table 4-13).
The overall incidence of remission from IA was 111(95% CI: 95, 129) per
100-person-years, with 111 (95% CI: 91, 135) per 100-person-years in males and 110
(95% CI: 86, 141) per 100-person-years in females (Table 4-13). There were no
significant differences for the incidence of remission from IA in terms of age and
83
gender.
4.3.3.8 Change in status of depression, ‘SNA’, and IA in subsample Ia~IIIa
4.3.3.8.1 Changes in status of ‘SNA’ and IA in subsample Ia (n=3196)
Among participants who were non-depressed at baseline (subsample Ia), the
frequency distribution of change in ‘SNA’ status was 84.29% for persistent non-
‘SNA’, 5.60% for remission from ‘SNA’, 2.97% for persistent ‘SNA’, and 7.13% for
incident ‘SNA’, while the frequency distribution of change in IA status was 94.87%
for persistent non-IA, 2.60% for remission from IA, 1.16% for persistent IA, and
3.38% for incident IA (Table 4-14).
4.3.3.8.2 Change in status of depression in subsample IIa (n=3657)
Among participants without ‘SNA’ at baseline (subsample IIa), the frequency
distribution of change in depression status over time was 67.27% for persistent
non-depression, 8.61% for remission from depression, 11.48% for persistent
depression, and 12.33% for incident depression (Table 4-14).
4.3.3.8.3 Change in status of depression in subsample IIIa (n=3959)
Among participants without IA at baseline (subsample IIIa), the frequency
distribution of change in depression status over time was 65.32% for persistent
non-depression, 9.47% for remission from depression, 12.83% for persistent
depression, and 12.38% for incident depression (Table 4-14).
84
4.3.3.9 Score changes of independent variables in the longitudinal sample
(n=4237)
4.3.3.9.1 Social networking-related scales
Except for the EC scale score, the scores of SFUI, EFUI, and ‘SNA’ had a slight
but significant decrease from baseline to follow-up (mean difference=-0.31 for SFUI,
paired t=-2.58, p=0.010; mean difference=-0.46 for EFUI, paired t=-8.86, p<0.001;
mean difference=-0.37 for ‘SNA’, paired t=-4.01, p<0.001) (Table 4-15).
Additionally, mean scores of SFUI and EC in females were significantly higher
than those in males (all p<0.001) (Table 4-12), whereas mean scores of EFUI were
larger in males than in females both at baseline and follow-up (p<0.001) (Table 4-15).
However, the gender differences for scores of ‘SNA’ were non-significant both at
baseline and follow-up.
4.3.3.9.2 Psychosocial scales
The descriptions of central tendency and variance for the scores of psychosocial
scales were presented in Table 4-16. The scores of friendship quality, perceived social
support, adolescent-father conflict, and social non-confidence slightly decreased from
baseline to follow-up (Table 4-16). However, there were no significant differences for
the scores of adolescent-mother conflict and self-esteem between baseline and
follow-up (Table 4-16).
4.4 Discussions
4.4.1 Internet addiction and online social networking among adolescents
IA in general has drawn increased attention in recent years. About 6.6% of IA
85
cases were identified at baseline in the present study. This finding is comparable to
previous studies in Mainland China (6.3%) [161] and Hong Kong (6.7%) [160] based on
the same assessment tool, but lower than those reported in south China (12.2%,
measured by 20-item Young-IAT) [182] and other countries (i.e. 17.6% by Young-IAT
in Qatar [183] and 21.2% by CIAS in the Philippines [184]) using different instruments.
A recent meta-analysis also estimated a similar global prevalence of IA (6.0%) [98]. As
such, in the absence of formal diagnostic criteria, diverse instruments, and cut-off
values used in the studies, prevalence estimates are highly tool-dependent and vary
widely worldwide. Caution should be taken when comparing the IA prevalence across
studies and different samples. Our results are consistent with several previous studies
that reported a male preponderance for IA [99, 161, 185]. It is likely that males are more
interested in games, pornography, and gambling activities that were associated with
IA[17] and have a lower level of self-control over Internet use [186].
It is worth noting that our longitudinal study observed a mild increase in IA
prevalence over a nine-month interval. It signifies that the IA condition is
continuously growing and is serious among this population, especially along with
increased availability of Internet access and portable devices (i.e. Wi-Fi access,
smartphone). A recent study reported that over 40% of Chinese adolescents had a
smartphone in 2014 [184]. It potentially increases the possibility of adolescents using
the Internet anywhere and anytime away from monitoring by their family or school.
Increased availability of Internet is associated with IA [17, 187]. Moreover, we observed
around 5.0% of new incident IA cases during the nine-month period, which was lower
than those reported from three one-year longitudinal studies conducted in Taiwan
(7.5%, 8.3%, and 15.8%, measured by CIAS) [38, 40, 41]. It appears that the potential
86
danger of IA among Chinese adolescents is increasingly serious. As a major public
health issue among adolescents, significant associations of IA with diverse psychiatric
and behavioral disorders (i.e. substance use, attention-deficit/hyperactivity disorder,
aggressive behaviors, social anxiety, and depression) have been documented in
previous studies [28, 188]. Therefore, more attention must be paid to this serious public
health issue, and effective IA prevention must be viewed as a necessity among
adolescents in China.
Online social networking as a new phenomenon has exponentially increased in
popularity among adolescents in the last decade. The majority of our sampled students
(92.3%) have their own social networking accounts. The extremely high popularity of
online social networking among adolescents has also been reported in the previous
studies, ranging from 85% to 94% [85, 189]. Adolescents are a subpopulation that are
disproportionately affected by online social networking [190]. Several studies have
indicated that social networking use is potentially associated with psychological
distress [50], lower self-esteem [36], loneliness [58, 111], and depression [35] among
adolescents. Online social networking has become the most popular form of online
activities globally. Health education should pay more attention to this increased
phenomenon.
Gender differences in social networking use should be emphasized. Our findings
revealed that females reported higher use intensity for social function of online social
networking, while males reported higher use intensity for entertainment function of
online social networking. These findings are consistent with previous studies,
reporting that males use the Internet mainly for entertainment and leisure purposes,
whereas females use it primarily for interpersonal communication [191, 192]. It suggests
87
that males and females use online social networking for different purposes and
reasons. These findings have a significant implication that strategies of health
education and intervention should be gender-specific.
‘Social networking addiction (SNA)’ in specific, similar to generalized IA, is
another public health concern that raises along with excessive engagement in such
Internet applications. We found a relatively lower prevalence of ‘SNA’ (13.7% at
baseline and 13.6% at follow-up) than previous data (24.0% and 34.3%) estimated by
modified Young’s Internet Addiction Test among Chinese college students [31, 32].
However, the prevalence is drastically higher than those reported in Nigerian (1.6%)
and Peruvian (8.6%) undergraduate students [29, 30]. One of the possible reasons is that
China has higher Internet penetration than these countries. Another possible reason
may be the lack of a universal instrument for ‘SNA’ assessment. Furthermore, around
9.2% of non-‘SNA’ participants converted into ‘SNA’ cases over the nine-month study
period. To our knowledge, this is the first longitudinal study documenting the
incidence of ‘SNA’ among early adolescents, specifically in Mainland China.
4.4.2 High prevalence and incidence of depression among adolescents
In the present longitudinal study, a high prevalence of depression was observed
at baseline (24.6%), which was considerably higher than that reported in the previous
study in Mainland China (14.8% measured by Children’s Depression Inventory) [193].
Moreover, it identified a significant increase in the prevalence of depression over a
nine-month interval, which has also been observed in a previous study [194]. This
signifies an increased threat of adolescents’ mental health problem.
Around 16.1% of non-depressed participants at baseline developed depression
88
during the follow-up period, which was almost double than that reported in a previous
nine-month longitudinal study conducted among high school students in Guangzhou
(8.4% new incident cases, measured by Zung Self-rating Depression Scale) [43], and
also higher than that in a seven-year longitudinal study in the United States (7.4% new
incident cases, measured by short version of CES-D). Depressive symptoms defined
by an epidemiological screening scale is a significant predictor of clinical depression
diagnosis [195] and suicidal ideation [196] among adolescents. Considering the high
prevalence, incidence, and serious consequences of depression among adolescents,
more attention should be paid to adolescents’ mental health (especially depression),
and effective depression intervention and prevention are highly warranted among this
population in Mainland China.
4.4.3 Natural remission from depression, ‘SNA’, and IA status
Another interesting finding in the present study is the large proportional
remission and recovery cases from depression, ‘SNA’, or IA status during the study
period without beforehand intervention, including 41.4% remission cases from
depression, 58.8% remission cases from ‘SNA’, and 58.6% remission cases from IA.
The remission rate from IA observed in this study are slightly higher than those
reported in previous longitudinal studies (one-year remission rate from IA: 49.5% and
51.5%, measured by CIAS) [40, 135]. As online social networking services are a very
recently developed phenomenon, the remission rate for ‘SNA’ has not yet been
reported. High recovery proportions from existing psychiatric and behavioral
disorders have several implications and concerns. Firstly, it suggests that depression,
‘SNA’, and IA status dynamically fluctuated over time rather than chance. The time
interval for follow-up should be cautiously defined in future longitudinal research, as
89
these conditions would be conversion between two assessment points. Prospective
studies involving a long-term period with little measurement waves may not reflect
the true status of depression, ‘SNA’, or IA. For example, although some participants
would be classified as non-IA both at baseline and follow-up, there is still a large
possibility that these participants may suffer from IA at some time in between baseline
and follow-up. Hence, in order to truly reflect the natural course of these negative
conditions, it is necessary to conduct more assessments within short time interval in
future longitudinal studies. Secondly, regression to the mean should be considered for
the observed remissions. Those with higher scores in depression, ‘SNA’, or IA scales
at baseline may have more likely tendency to move down in their scores at follow-up.
Special analysis strategies should be applied to adjust for regression to the mean [152].
Thirdly, it raises a concern whether these behavioral disorders are long-lasting or
short-term. Further investigation is warranted on the nature of the course of behavioral
disorders from a new perspective. Moreover, the status (i.e. depression, ‘SNA’, or IA)
observed in the initial assessment cannot be treated as an unchangeable condition over
time. It may be better to measure these predictors at each assessment instead of just
measuring at the initial assessment. The prospective studies and analysis strategies
should cautiously consider the change in such predictors during the course of the
study; for example, change in IA status: persistent non-IA, remission from IA,
persistent IA, and incident IA.
90
Table 4-1 Background characteristics of the participants of the entire sample at baseline (n=5365)
n %
Social-demographic variables
Gender
Male 2832 52.79
Female 2533 47.21
Grade
Seven 2592 48.31
Eight 2773 51.69
Father’s education level
Primary school or below 356 6.64
Junior secondary school 1816 33.85
Senior secondary school 1646 30.68
College or above 1317 24.55
Don't know 230 4.29
Mother’s education level
Primary school or below 588 10.96
Junior secondary school 1909 35.58
Senior secondary school 1497 27.90
College or above 1143 21.30
Don't know 228 4.25
Family financial situation
Very good/Good 2519 46.95
Average 2664 49.66
Poor /Very poor 182 3.39
Lives with both parents
Yes 4712 87.83
No 653 12.17
School-related variables
Academic performance
Upper 1817 33.87
Medium 2396 44.66
Lower 1152 21.47
Perceived study pressure
Nil/light 1034 19.27
General 3052 56.89
Heavy /Very heavy 1279 23.84
91
Table 4-2 Attrition analysis for loss to follow-up
Remained in study Lost to follow-up
p-value
(n=4871) (n=494)
Gender
Male 51.71 63.36 <0.001
Female 48.29 36.64
Grade
Seven 47.85 52.83 0.035
Eight 52.15 47.17
Primary school or below 6.69 6.07 0.975
Junior secondary school 33.75 34.82
Senior secondary school 30.69 30.57
College or above 24.55 24.49
Don't know 4.31 4.05
Primary school or below 10.92 11.34 0.950
Junior secondary school 35.72 34.21
Senior secondary school 27.92 27.73
College or above 21.19 22.47
Don't know 4.25 4.25
Very good/Good 47.38 42.71 0.005
Average 49.46 51.62
Poor /Very poor 3.16 5.67
Yes 88.28 83.40 0.002
No 11.72 16.60
Upper 34.22 30.36 <0.001
Medium 45.12 40.08
Lower 20.65 29.55
Nil/light 19.48 17.21 0.149
General 57.01 55.67
Heavy /Very heavy 23.51 27.13
Social networking users

Yes 92.69 88.26 <0.001
No 7.31 11.74
Depression at baseline
No 75.36 70.45 0.016
Yes 24.46 29.55
p values were obtained by Chi-square test.
92
Table 4-3 Background characteristics of participants among the longitudinal sample and the
subsamples
Longitudinal Subsample Ia~IIIa
sample Subsample Ia Subsample IIa Subsample IIIa
(n=4237) (n=3196) (n=3657) (n=3959)
n % n % n % n %
Gender
Male 2105 49.68 1628 50.94 1773 48.48 1933 48.83
Female 2132 50.32 1568 49.06 1884 51.52 2026 51.17
Grade
Seven 2011 47.46 1549 48.47 1755 47.99 1887 47.66
Eight 2226 52.46 1647 51.53 1902 52.01 2072 52.34
Primary school or below 273 6.44 186 5.82 226 6.18 250 6.31
Junior middle school 1425 33.63 1066 33.35 1217 33.28 1338 33.80
High middle school 1312 30.97 990 30.98 1141 31.20 1222 30.87
University or above 1053 24.85 817 25.56 922 25.21 985 24.88
Don't know 174 4.11 137 4.29 151 4.13 164 4.14
Primary school or below 445 10.50 302 9.45 370 10.12 413 10.43
Junior middle school 1507 35.57 1138 35.61 1304 35.66 1417 35.79
High middle school 1199 28.30 931 29.13 1040 28.44 1116 28.19
University or above 913 21.55 684 21.40 790 21.60 845 21.34
Don't know 123 4.08 141 4.41 153 4.18 168 4.24
Very good/good 2047 48.31 1618 50.63 1795 49.08 1928 48.70
Medium 2072 48.60 1509 47.22 1771 48.43 1922 48.55
Poor/very poor 118 2.78 69 2.16 91 2.49 109 2.75
Yes 3747 88.44 2854 89.30 3238 88.54 3512 88.71
No 490 11.56 342 10.70 419 11.46 447 11.29
Upper 1465 34.58 1193 37.33 1347 36.83 1403 35.44
Medium 1920 45.32 1440 45.06 1653 45.20 1794 45.31
Lower 852 20.11 563 17.62 657 17.97 762 19.25
Nil/light 811 19.14 698 21.84 745 20.37 778 19.65
Average 2433 57.42 1941 60.73 2128 58.19 2297 58.02
Heavy/very heavy 993 23.44 557 17.43 784 21.44 884 22.33
Subsample Ia was comprised of participants who were non-depressed at baseline.
Subsample IIa was comprised of participants who were non-SNA at baseline.
Subsample IIIa was comprised of participants who were non-IA at baseline.
93
Table 4-4 Comparision of background characteristics between those with SNA and those without
SNA at baseline (n=4237)
Non-SNA SNA
ORu (95% CI)
n % n %
Gender
Male 1773 84.23 332 15.77 1
Female 1884 88.37 248 11.63 0.72(0.60, 0.86)***
Grade
Seven 1755 87.27 256 12.73 1
Eight 1902 85.44 324 14.56 1.16(0.97, 1.38)
Primary school or below 226 82.78 47 17.22 1
Junior secondary school 1217 85.40 208 14.60 0.86(0.61, 1.23)
Senior secondary school 1141 86.97 171 13.03 0.76(0.53, 1.10)
College or above 922 87.56 131 12.44 0.78(0.53, 1.13)
Don't know 151 86.78 23 13.22 0.79(0.45, 1.36)
College or above 790 86.53 123 13.47 0.87(0.63, 1.21)
Don't know 153 88.44 20 11.56 0.69(0.41, 1.19)
Very good/Good 1795 87.69 252 12.31 1
Average 1771 85.47 301 14.53 1.12(0.93, 1.35)
Poor /Very poor 91 77.12 27 22.88 1.91(1.21, 3.01)**
Yes 3238 86.42 509 13.58 1
No 419 85.51 71 14.49 1.02(0.78, 1.34)
Upper 1347 91.95 118 8.05 1
Medium 1653 86.09 267 13.91 1.81(1.44, 2.28)***
Lower 657 77.11 195 22.89 3.39(2.63, 4.34)***
Nil/light 745 91.86 66 8.14 1
General 2128 87.46 305 12.54 1.59(1.20, 2.11)**
Heavy /Very heavy 784 78.95 209 21.05 2.97(2.21, 3.99)***
SNA: Social Networking Addiction.
ORu: Univariate odds ratio, obtained by univariate multilevel logistic regression models.
* p<0.05; ** p<0.01; *** p<0.001
94
Table 4-5 Comparision of background characteristics between those with IA and those without IA
at baseline (n=4237)
Non-IA IA
ORu (95% CI)
n % n %
Gender
Male 1933 91.83 172 8.17 1
Female 2026 95.03 106 4.97 0.59(0.46, 0.76)***
Grade
Seven 1887 93.83 124 6.17 1
Eight 2072 93.08 154 6.92 1.13(0.88, 1.44)
College or above 985 93.54 68 6.46 0.73(0.44, 1.21)
Don't know 164 94.25 10 5.75 0.63(0.29, 1.36)
College or above 845 92.55 68 7.45 1.02(0.65, 1.59)
Don't know 168 97.11 5 2.89 0.37(0.14, 0.96)*
Very good/Good 1928 94.19 119 5.81 1
Average 1922 92.76 150 7.24 1.25(0.97, 1.60)
Poor /Very poor 109 92.37 9 7.63 2.19(0.64, 2.61)
Yes 3512 93.73 235 6.27 1
No 447 91.22 43 8.78 1.40(0.99, 1.97)†
Upper 1403 95.77 62 4.23 1
Medium 1794 93.44 126 6.56 1.60(1.16, 2.18)**
Lower 762 89.44 90 6.56 2.70(1.93, 3.78)***
Nil/light 778 95.93 33 4.07 1
General 2297 94.41 136 5.59 1.41(0.95, 2.08)†
Heavy /Very heavy 884 89.02 109 10.98 2.92(1.95, 4.37)***
IA: Internet Addiction.
ORu: Univariate odds ratio, obtained by univariate multilevel logistic regression models.
† p<0.10; * p<0.05; ** p<0.01; *** p<0.001
95
Table 4-6 Comparision for the mean differences of baseline IA and SNA scores between four groups (n=4237)
Groups IA SNA
n (%)
Baseline IA Baseline SNA Mean score F p Mean score F p
No No 3533(83.38) 1.16±1.17 1366.43 <0.001 15.27±4.34 1261.94 <0.001
No Yes 426(10.05) 2.11±1.34 26.04±2.96
Yes No 124(2.93) 5.46±0.75 18.23±4.40
Yes Yes 154(3.63) 6.03±1.14 28.96±4.39
IA: Internet Addiction; SNA: Social Networking Addiction.
p values were obtained by one-way ANOVA.
96
Table 4-7 Prevalence of depression, ‘SNA’, and IA at baseline and follow-up by gender in the
longitudinal sample (n=4237)
Total prevalence Gender
(%) Male Female p for gender
Depression
Baseline 24.57 22.66 26.45 0.004
Follow-up 26.55 25.32 27.77 0.071
‘SNA’
Baseline 13.69 15.77 11.63 <0.001
Follow-up 13.55 14.25 12.85 0.183
IA
Baseline 6.56 8.17 4.97 <0.001
Follow-up 7.34 8.31 6.38 0.016
SNA: Social Networking Addiction; IA: Internet Addiction.
p for gender were obtained by Chi-square test.
97
Table 4-8 Age- and gender-specific incidence of depression among participants who were non-depressed at baseline (n=3196)
% of Incident cases Incidence per 100-person-years (95% CI)
Subsample Ia
Age Male Female Total Male Female Total
(n=3196) $
(n=1628) (n=1568) (n=3196) (n=1628) (n=1568) (n=3196)
≤13 991 15.98 16.10 16.04 23.16(18.55, 28.92) 23.35(18.78, 29.03) 23.26(19.91, 27.17)
14 1582 16.98 15.88 16.43 24.74(20.90, 29.29) 23.00(19.31, 27.41) 23.88(21.14, 26.96)
≥15 623 15.36 15.47 15.41 22.19(16.95, 29.04) 22.35(16.58, 20.14) 22.26(18.22, 27.19)
Total 3196 16.34 15.88 16.11 23.72(21.04, 26.75) 23.00(20.31, 26.04) 23.37(21.43, 25.48)
$
Participants without depression at baseline.
98
Table 4-9 Age- and gender-specific incidence of ‘SNA’ among participants who were classified as non-‘SNA’ at baseline (n=3657)
Subsample IIa
(n=3657) $
(n=1773) (n=1884) (n=3657) (n=1773) (n=1884) (n=3657)
≤13 1129 8.16 10.51 9.39 11.35(8.44, 15.25) 14.79(11.53, 18.97) 13.14(10.86, 15.89)
14 1822 9.43 8.62 9.00 13.19(10.61, 16.41) 12.01(9.68, 14.89) 12.57(10.78, 14.65)
≥15 706 11.20 6.95 9.21 15.82(11.69, 21.41) 9.69(6.38, 14.44) 12.87(10.09, 16.41)
Total 3657 9.42 8.92 9.16 13.18(11.32, 15.34) 12.44(10.70, 14.48) 12.80(11.50, 14.25)
$
Participants who were classified as non-SNA at baseline.
99
Table 4-10 Age- and gender-specific incidence of IA among participants who were classified as non-IA at baseline (n=3959)
Subsample IIIa
(n=3959) $
(n=1933) (n=2026) (n=3959) (n=1933) (n=2026) (n=3959)
≤13 1217 5.50 5.98 5.75 7.54(5.33, 10.66) 8.23(5.98, 11.30) 7.90(6.25, 9.98)
14 1970 5.08 4.19 4.62 6.96(5.24, 9.23) 5.71(4.23, 7.70) 6.30(5.13, 7.74)
≥15 772 5.90 3.01 4.53 8.10(5.43, 12.09) 4.08(2.26, 7.37) 6.19(4.44, 8.61)
Total 3959 5.38 4.54 4.95 7.37(6.08, 8.93) 6.20(5.05, 7.60) 6.77(5.88, 7.79)
$
Participants who were classified as non-IA at baseline.
100
Table 4-11 Age- and gender-specific remission from depression among participants who were depressed at baseline (n=1041)
% of remitted cases Remission per 100-person-years (95% CI)
Subsample Ib
(n=1041) $
(n=477) (n=564) (n=1041) (n=477) (n=564) (n=1041)
≤13 307 45.58 38.13 41.69 78.71(61.95, 1.00) 62.81(48.87, 80.72) 70.23(89.06, 83.52)
14 516 45.98 41.44 43.41 79.61(65.63, 96.57) 69.69(58.32, 83.28) 73.93(64.85, 84.27)
≥15 218 37.74 24.82 36.24 62.02(45.49, 84.54) 56.22(41.07, 76.94) 59.01(47.33, 73.57)
Total 1041 44.03 39.18 41.40 75.27(65.57, 86.17) 64.98(56.95, 74.13) 69.61(63.34, 76.51)
$
Participants with depression at baseline.
101
Table 4-12 Age- and gender-specific remission from ‘SNA’ among participants who were classified as ‘SNA’ at baseline (n=580)
Subsample IIb
(n=580) $
(n=332) (n=248) (n=580) (n=332) (n=248) (n=580)
≤13 169 60.42 54.79 57.99 115(89, 149) 101(74, 137) 109(89, 133)
14 276 62.50 59.48 61.23 121(100, 147) 113(89, 143) 118(101, 137)
≥15 135 53.95 55.93 54.81 99(73, 134) 104(74, 146) 101(80, 126)
Total 580 59.94 57.26 58.79 114(99, 131) 107(91, 126) 111(100, 123)
$
Participants who were classified as SNA at baseline.
102
Table 4-13 Age- and gender-specific remission from IA among participants who were classified as IA at baseline (n=248)
% of Incident cases Remission per 100-person-years (95% CI)
Subsample IIIb
(n=278) $
(n=172) (n=106) (n=278) (n=172) (n=106) (n=278)
≤13 81 60.38 64.29 61.73 115(82, 163) 126(80, 200) 119(90, 157)
14 128 61.33 60.38 60.94 118(88, 157) 115(82, 163) 117(94, 146)
≥15 69 52.27 48.00 50.72 94(63, 142) 84(48, 148) 91(65, 126)
Total 278 58.72 58.49 58.63 111(91, 135) 110(86, 141) 111(95, 129)
$
Participants who were classified as IA at baseline.
103
Table 4-14 Frequency distribution of change in status of ‘SNA’, IA, and depression in subsample
Ia~IIIa
Subsample Ia Subsample IIa Subsample IIIa
(n=3196) (n=3657) (n=3959)
n (%) n (%) n (%)
Change in ‘SNA’ status
Persistent non-‘SNA’ 2694 (84.29) -- --
Remitted ‘SNA’ 179 (5.60)
Persistent ‘SNA’ 95(2.97)
Incident ‘SNA’ 228 (7.13)
Change in IA status
Persistent non-IA 2968(94.87) -- --
Remitted IA 83(2.60)
Persistent IA 37(1.16)
Incident IA 108(3.38)
Change in depression status

Persistent non-depression -- 2471(67.27) 2586(65.32)
Remitted depression 315(8.61) 375(9.47)
Persistent depression 420(11.48) 508(12.83)
Incident depression 451(12.33) 490(12.38)
SNA: Social Networking Addiction; IA: Internet Addiction.
Subsample Ia consisted of participants who were non-depressed at baseline.
Subsample IIa consisted of participants who were non-SNA at baseline.
Subsample IIIa consisted of participants who were non-IA at baseline.
--: Not applicable.
104
Table 4-15 Score changes of social networking related scales among longitudinal sample (n=4237)
Mean difference p for
Range Baseline Follow-up
(95% CI) time
SFUI
Male 0-40 16.91±8.52 16.82±8.80 -0.10(-0.47, 0.27) 0.608
Female 0-40 19.82±7.83 19.29±7.69 -0.53(-0.83, -0.23) <0.001
Total 0-40 18.38±8.31 18.06±8.35 -0.31(-0.55, -0.08) 0.010
p for gender <0.001 <0.001
EFUI
Male 0-16 8.53±3.37 8.05±3.46 -0.48(-0..64, -0.32) <0.001
Female 0-16 7.89±2.85 7.45±2.78 -0.45(-0.57, -0.32) <0.001
Total 0-16 8.21±3.14 7.75±3.15 -0.46(-0.56, -0.36) <0.001
p for gender <0.001 <0.001
EC
Male 6-30 14.77±5.55 14.83±5.67 0.06(-0.18, 0.31) 0.614
Female 6-30 15.74±5.25 15.94±5.15 0.20(-0.02, 0.41) 0.072
Total 6-30 15.26±5.42 15.39±5.44 0.13(-0.03, 0.29) 0.118
p for gender <0.001 <0.001
‘SNA’
Male 8-40 17.10±6.13 16.54±6.36 -0.56(-0.84, -0.28) <0.001
Female 8-40 16.78±5.49 16.60±5.83 -0.18(-0.42, 0.05) 0.124
Total 8-40 16.94±5.81 16.57±6.10 -0.37(-0.55, -0.19) <0.001
p for gender 0.073 0.765
SFUI: Social Function Use Intensity; EFUI: Entertainment Function Use Intensity; EC: Emotional
Connection to social networking; SNA: Social Networking Addiction.
Mean difference: Scores at follow-up minus corresponding scores at baseline.
95% CI: 95% confidence interval.
p for time was obtained by paired t test for mean difference between baseline and follow-up.
p for gender was obtained by two sample t-test.
105
Table 4-16 Score difference of interpersonal and psychosocial scales between baseline and
follow-up among longitudinal sample (n=4237)
Range Baseline Follow-up Mean difference (95%CI) p
Friendship quality
Male 6-24 18.91±3.51 18.79±3.62 -0.12(-0.27, 0.03) 0.105
Female 6-24 19.01±3.23 18.78±3.19 -0.22(-0.35, -0.10) <0.001
Total 6-24 18.96±3.37 18.79±3.41 -0.17(-0.27, -0.08) 0.001
Perceived family support

Male 4-28 20.28±5.51 19.72±5.48 -0.56(-0.80, -0.32) <0.001
Female 4-28 20.45±5.62 20.58±5.30 0.13(-0.08, 0.34) 0.230
Total 4-28 20.36±5.57 20.15±5.41 -0.21(-0.37, -0.05) 0.009
Perceived friend support

Male 8-56 40.76±10.24 39.58±10.71 -1.19(-1.65, -0.72) <0.001
Female 8-56 43.21±9.78 41.77±10.16 -1.44(-1.86, -1.02) <0.001
Total 8-56 42.00±10.08 40.68±10.49 -1.31(-1.63, -1.00) <0.001
Adolescent-father conflict
Male 7-35 13.58±6.13 13.17±6.15 -0.41(-0.73, -0.10) 0.010
Female 7-35 12.24±5.35 12.15±5.38 -0.09(-0.35, 0.17) 0.484
Total 7-35 12.91±5.79 12.66±5.80 -0.25(-0.45, -0.05) 0.016
Adolescent-mother conflict
Male 7-35 14.96±6.59 14.89±7.02 -0.07(-0.41, 0.28) 0.698
Female 7-35 13.94±5.92 14.30±6.35 0.36(0.06, 0.65) 0.018
Total 7-35 14.45±6.28 14.59±6.69 0.15(-0.08, 0.37) 0.207
Social non-confidence
Male 4-20 12.57±3.20 12.26±3.19 -0.31(-0.47, -0.15) <0.001
Female 4-20 13.04±2.97 12.90±2.92 -0.14(-0.27, -0.004) 0.044
Total 4-20 12.81±3.09 12.58±3.07 -0.22(-0.33, -0.12) <0.001
Rosenberg self-esteem
Male 10-40 27.83±4.38 27.77±4.43 -0.07(-0.25, 0.12) 0.482
Female 10-40 27.52±4.38 27.41±4.40 -0.10(-0.27, 0.06) 0.214
Total 10-40 27.67±4.38 27.59±4.42 -0.08(-0.21, 0.04) 0.175
Mean difference: scores at follow-up minus corresponding scores at baseline.
p values were obtained by paired t test for mean difference between baseline and follow-up.
106
Chapter 5 Directionality for the causal relationships between
Internet/social networking addiction and depressive symptoms
5.1 Background
5.1.1 Relationship between IA and depression
Substantial cross-sectional studies have reported a positive association between
IA and depression among adolescents [197-201]. However, the results from these studies
are mixed [202, 203]. Due to limitations of the cross-sectional design, the causal
relationship between IA and depression could not be identified, although it is essential
to examine the causal relationship and magnitude effects for these two public health
issues. Our literature review found that only eight longitudinal studies investigated the
longitudinal relationship between IA and depression among adolescents. However, the
results are mixed. The causal direction between IA and depression remains inclusive.
There is a dearth of the research to test two causal directions in the same study.
Further studies on the direction(s) underneath the relationship between IA and
depression are therefore warranted. This chapter tackles such conceptual issues. It is
possible to hypothesize a bi-directional relationship existing for the relationship
between IA and depression.
5.1.2 Relationship between ‘SNA’ and depression
The aforementioned research questions on causality persist for the study of the
relationship between ‘SNA’ and depression, but are even less well studied as
compared to those between IA and depression. Online social networking could
provide adolescents with social convenience, social capital, selective self-disclosure,
107
and potential social support [51, 112]. Individuals who suffer from psychiatric disorders
(i.e. depressed, anxious) might view online social networking as safe and important
virtual communities where they can escape from emotional difficulty in the real
world [26, 97]. On the other hand, adolescents who have poorly adjusted to their
depression may suffer more detrimental effects from excessive social networking
use [204]. Therefore, a vicious circle between this behavioral addictive disorder and
psychopathology (i.e. depression) is also theoretically reasonable.
Our literature reviews did not find any longitudinal studies that explore the
causal relationship between ‘SNA’ and depression among adolescents and other
populations. However, due to limitations of the cross-sectional study design, it is
impossible to determine whether ‘SNA’ causes depressive psychopathology, or vice
versa. This chapter is also an attempt to fill up this knowledge gap.
5.1.3 Knowledge gaps
Conceptually, there are no longitudinal studies exploring the causal relationship
between ‘SNA’ and depression among adolescents. The existing longitudinal studies
reveal mixed results for the causal relationship between IA and depression.
Researchers continue to debate the causal directions of the relationships. Research
that theoretically tests bi-directional relationships between ‘SNA’/IA and depression
in a single longitudinal study is particularly scarce. Therefore, it is still unclear
whether ‘SNA’/IA are causes or outcomes of depression, and large-scale longitudinal
studies are warranted for clarifying such causal issues. Furthermore, most of the
studies that have been conducted on the psychological associations of IA fail to
differentiate between generalized IA and specific IA (i.e. ‘SNA’) simultaneously.
108
Moreover, methodologically, all of the current longitudinal studies involving
addictive behaviors and depression do not take into account the fluctuating change in
‘SNA’/IA status and/or depression status over time. Considering the dynamic changes
in ‘SNA’, IA, and depression status over time would provide more convincing insight
into the longitudinal nature of ‘SNA’, IA, and depression development. This study
therefore took into account change in ‘SNA’, IA or depression status over time (i.e.
remission) in the statistical analyses.
5.2 Objectives
This chapter aims to explore the causal directions between ‘SNA’/IA and
depression among adolescents. Specifically, for the causal relationship between ‘SNA’
and depression, two modeling approaches were used: Approach I: using baseline
‘SNA’ to predict incident depression among participants without depression at
baseline, and vice versa, using baseline depression to predict incident ‘SNA’ among
participants without ‘SNA’ at baseline; and Approach II: using change in SNA status
over time (persistent non-‘SNA’, remission from ‘SNA’, persistent ‘SNA’, and
incident ‘SNA’) to predict incident depression among participants without depression,
and vice versa, using change in depression status over time (persistent non-depression,
remission from depression, persistent depression, and incident depression) to predict
incident ‘SNA’ among participants without ‘SNA’ at baseline. These two types of
analyses were repeated for studying the causal relationship between IA and
depression.
The following hypotheses have been drawn:
• For the prediction of ‘SNA’/IA on incident depression
109
- Hypothesis 1: Baseline ‘SNA’ status would predict a higher probability of
incident depression during the follow-up period as compared to non-‘SNA’ cases
at baseline after adjusting for significant background variables and baseline
CES-D score among participants without depression at baseline.
- Hypothesis 2: Baseline IA status would predict a higher probability of incident
depression during the follow-up period as compared to non-IA cases at baseline
after adjusting for significant background variables and baseline CES-D score
among participants without depression at baseline.
- Hypothesis 3: Persistent ‘SNA’ and incident ‘SNA’ of change in ‘SNA’ status
from baseline to follow-up (defined as persistent non-‘SNA’, remission from
‘SNA’, persistent ‘SNA’, and incident ‘SNA’) would predict a higher probability
of incident depression as compared to persistent non-‘SNA’ cases after adjusting
for significant background variables and baseline CES-D score among
participants without depression at baseline.
- Hypothesis 4: Persistent IA and incident IA of change in IA status from baseline
to follow-up (defined as persistent non-IA, remission from IA, persistent IA, and
incident IA) would predict a higher probability of incident depression as
compared to persistent non-IA cases after adjusting for significant background
variables and baseline CES-D score among participants without depression at
baseline.
• For the prediction of depression on incident ‘SNA’/IA
- Hypothesis 5: Baseline depression status would predict a higher likelihood of
incident ‘SNA’ as compared to non-depressed cases at baseline after adjusting for
110
significant background variables and baseline ‘SNA’ score among participants
without ‘SNA’ at baseline.
- Hypothesis 6: Baseline depression status would predict a higher likelihood of
incident IA as compared to non-depressed cases at baseline after adjusting for
significant background variables and baseline IA score among participants
without IA at baseline.
- Hypothesis 7: Persistent depression and incident depression of change in
depression status over time (defined as persistent non-depression, remission from
depression, persistent depression, and incident depression) would predict incident
‘SNA’ as compared to persistent non-depressed cases after adjusting for
significant background variables and baseline ‘SNA’ score among participants
without ‘SNA’ at baseline.
- Hypothesis 8: Persistent depression and incident depression of change in
depression status over time (defined as persistent non-depression, remission from
depression, persistent depression, and incident depression) would predict incident
IA as compared to persistent non-depressed cases after adjusting for significant
background variables and baseline IA score among participants without IA at
baseline.
5.3 Statistical analysis
5.3.1 Subsamples
In this chapter, three subsamples were used for longitudinal data analyses
including: 1) subsample Ia (n of subsample=3196) which consisted of participants
111
who were non-depressed at baseline, 2) subsample IIa (n of subsample=3657) which
consisted of participants who were classified as non- ‘SNA’ at baseline, 3) subsample
IIIa (n of subsample=3959) which consisted of participants who were classified as
non-IA at baseline.
5.3.2 Modeling methods
Since the participants in the present study were selected using a stratified cluster
sampling method at the school level, two-level logistic regression models (level 1:
student, level 2: school) were fit (SAS GLIMMIX procedure) to investigate
longitudinal prediction effects by accounting for the clustering effect from the school
level, with the specification of random effects for the intercept in the regression
models. Ignoring the clustering effect would lead to an underestimation of standard
errors [205].
Background factors of incident depression, ‘SNA’ or IA (including gender, grade,
parental education level, family financial situation, living with both parents or not,
academic performance, and perceived study pressure) were selected by univariate
multilevel logistic regression analyses. Univariate odds ratios (ORu) with
corresponding 95% confidence intervals (95% CI) were derived. The significant
background factors hence found were adjusted for in the subsequent longitudinal
analyses.
5.3.2.1 Approach I: baseline status of measurement as a predictor
We firstly used baseline status of measurements to predict new incident
outcomes, including two models with incident depression as the dependent variable
and baseline ‘SNA’ or IA as predictors (see Figure 5-1a and Figure 5-2a), and two
112
models with incident ‘SNA’ or IA as dependent variables and baseline depression as
the predictor (see Figure 5-1b and Figure 5-2b). The model adjusted for significant
background variables (Model 1a), and the model adjusted for significant background
variables along with baseline scores of corresponding outcomes (Model 1b) were
constructed. A similar approach has been used in a previous study, which stated that
the difference in associations between models while controlling and not controlling
for baseline level of corresponding outcome measures enables identification of causal
pathways [206].
Univariate Model 1a Model 1b Univariate Model 1a Model 1b
Baseline Baseline Baseline Baseline Baseline Baseline

SNA SNA SNA depression depression depression
＋＋＋＋
Background Background Background Background
variables variables variables variables
＋＋
Baseline Baseline
CES-D score SNA score
(a) Baseline SNA to predict incident depression (b) Baseline depression to predict incident SNA
Figure 5-1 Causal direction between ‘SNA’ and depression using approach I
Baseline Baseline Baseline

Baseline IA Baseline IA Baseline IA
depression depression depression
＋＋＋＋
＋＋
Baseline Baseline IA
CES-D score score
(a) Baseline IA to predict incident depression (b) Baseline depression to predict incident IA
Figure 5-2 Causal direction between IA and depression using approach I
113
5.3.2.2 Approach II: change in status from baseline to follow-up as a predictor
Changes in status of ‘SNA’/IA from baseline to follow-up were categorized as
persistent non-‘SNA’/IA (participants without ‘SNA’/IA both at baseline and
follow-up), remission from ‘SNA’/IA (participants with ‘SNA’/IA at baseline but
transitioning without ‘SNA’/IA at follow-up), persistent ‘SNA’/IA (participants with
‘SNA’/IA both at baseline and follow-up) and incident ‘SNA’/IA (participants
without ‘SNA’/IA at baseline but transitioning with ‘SNA’/IA at follow-up); Similarly,
change in depression status from baseline to follow-up also included four categories:
persistent non-depression, remission from depression, persistent depression and
incident depression, the definition is the same as those for ‘SNA’/IA.
Similarly, there were also two sets of prediction models: two models with change
in ‘SNA’ or IA status as the predictors and incident depression as the dependent
variable (Figure 5-3a and Figure 5-4a), and two models with change in depression
status as the predictor and incident ‘SNA’ or IA as the dependent variables (Figure
5-3b and Figure 5-4b). The two adjusted multilevel logistic regression models were fit:
one adjusted only for significant background variables (Model 2a) and the other
additionally adjusted for baseline scores of corresponding outcomes (Model 2b).
Change in Change in Change in Change in Change in Change in

SNA status SNA status SNA status depression status depression status depression status
＋＋＋＋
＋＋
Baseline Baseline
CES-D score SNA score
(a) Change in SNA status to predict incident depression (b) Change in depression status to predict incident SNA
Figure 5-3 Causal direction between ‘SNA’ and depression using approach II
114
Change in Change in Change in Change in Change in Change in

IA status IA status IA status depression status depression status depression status
＋＋＋＋
＋＋
Baseline Baseline IA
CES-D score score
(a) Change in IA status to predict incident depression (b) Change in depression status to predict incident IA
Figure 5-4 Causal direction between IA and depression using approach II
5.4 Results
5.4.1 Causal relationship between ‘SNA’ and depression
5.4.1.1 Background factors associated with incident depression for adjustment
The results from univariate multilevel logistic regression analyses showed that
three background variables significantly and positively predicted incident depression
during the follow-up period, including perceived average (ORu=1.32, 95% CI:
1.08~1.60, p=0.005) or poor/very poor family financial situation (ORu=1.98, 95% CI:
1.12~3.49, p=0.018; reference group: perceived good/very good family financial
situation), self-reported lower academic performance (ORu=1.66, 95% CI: 1.28~2.16,
p<0.001, reference group: upper academic performance), and perceived heavy/very
heavy study pressure (ORu=1.63, 95% CI: 1.20~2.20, p=0.002, reference group:
perceived nil/light study pressure) (Table 5-1).
5.4.1.2 Background factors associated with incident ‘SNA’ for adjustment
The results showed that participants who perceived poor/very poor family
115
financial situation (ORu=2.76, 95% CI: 1.60~4.76, p<0.001, reference group:
perceived good/very good family financial situation), those who did not live with both
parents (ORu=1.53, 95% CI: 1.12~2.09, p=0.005, reference group: living with both
parents), those who reported lower academic performance (ORu=1.61, 95% CI:
1.19~2.19, p=0.003, reference group: upper academic performance), and those who
perceived heavy/very heavy study pressure (ORu=1.65, 95% CI: 1.17~2.32, p=0.003,
reference group: nil/light study pressure) had a higher likelihood of suffering from
‘SNA’ during the follow-up period (Table 5-2).
5.4.1.3 Approach I: Using baseline ‘SNA’ to predict incident depression and using
baseline depression to predict incident ‘SNA’
Among participants who were non-depressed at baseline, results from univariate
analysis showed that baseline ‘SNA’ significantly predicted incident depression during
the follow-up period (ORu=1.65, 95% CI: 1.22~2.22, see Table 5-3). After adjusting
for significant background factors that included family financial situation, academic
performance and perceived study pressure, the predictive effect of baseline ‘SNA’ on
incident depression remained significant (AOR=1.48, 95% CI: 1.09~2.01, p=0.009;
see Model 1a in Table 5-3). However, when further adjusting for baseline CES-D
score on the basis of model 1a, this predictive effect become non-significant
(AOR=1.17, 95% CI: 0.86~1.60; see Model 1b in Table 5-3). Hypothesis 1 was hence
not supported by the results.
In reverse, in the univariate multilevel logistic analysis, participants with
depression at baseline were twice more as likely to develop ‘SNA’ during the
follow-up period compared to those without depression at baseline (ORu=2.02, 95%
CI: 1.58~2.58; see Table 5-4). After adjusting for significant background factors
116
(family financial situation, living with both parents or not, academic performance, and
perceived study pressure), the predictive effect of baseline depression on incident
‘SNA’ slightly attenuated but remained significant (AOR=1.78, 95% CI: 1.37~2.30,
p<0.001; see Model 1a in Table 5-4). This prediction was still statistically significant
even when further adjusting for baseline ‘SNA’ score in the model (AOR=1.47, 95%
CI: 1.13~1.92, p<0.001; see Model 1b in Table 5-4). Hypothesis 5 was therefore
supported by the current results.
5.4.1.4 Approach II: Using change in ‘SNA’ status over time to predict incident
depression and using change in depression status over time to predict incident
‘SNA’
A significant univariate prediction was observed between change in ‘SNA’ status
and incident depression during follow-up period (ORu=1.77, 95% CI: 1.21~2.58 for
remission from ‘SNA’; ORu=2.46, 95% CI: 1.54~3.93 for persistent ‘SNA’;
ORu=4.89, 95% CI: 3.67~6.52 for incident ‘SNA’; reference group: persistent
non-‘SNA’; see Table 5-5). After adjusting for significant background factors
(including family financial situation, academic performance, and perceived study
pressure), change in ‘SNA’ status still significantly predicted incident depression
(AOR=1.61, 95%CI: 1.10~2.36 for remission from ‘SNA’; AOR=2.23, 95%CI:
1.39~3.57 for persistent ‘SNA’; AOR=4.67, 95%CI: 3.49~6.24 for incident ‘SNA’,
reference group: persistent non- ‘SNA’; see Model 2a in Table 5-5). After additional
adjustment for baseline CES-D score on the basis of model 2a, this prediction
remained statistically significant (AOR=1.66, 95% CI: 1.02~2.70 for persistent ‘SNA’,
p=0.038; AOR=4.30, 95% CI: 3.17~5.82 for incident ‘SNA’, reference group:
persistent non- ‘SNA’; see Model 2b in Table 5-5). Therefore, hypothesis 3 was
117
supported by the adjusted results.
Reversely, univariate multilevel logistic regression showed that change in
depression over time significantly and positively predicted incident ‘SNA’ among
participants who were non- ‘SNA’ at baseline (ORu=4.62, 95% CI: 3.43~6.21 for
persistent depression; ORu=4.88, 95% CI: 3.67~6.50 for incident depression;
reference group: persistent non-depression; see Table 5-6). The predictive effect of
change in depression status over time on incident ‘SNA’ did not change after
adjustment for significant background factors and baseline ‘SNA’ score (AOR=3.40,
95% CI: 2.48~4.68 for persistent depression; AOR=4.47, 95% CI: 3.35~5.99 for
incident depression; reference group: persistent non-depression; see Model 2b in
Table 5-6), which supported hypothesis 7.
5.4.2 Causal relationship between IA and depression
5.4.2.1 Background factors associated with incident IA
Univariate multilevel logistic regression analyses found that only perceived
poor/very poor family financial situation (ORu=3.28, 95%CI: 1.82~5.91, p<0.001,
reference group: perceived good/very good family financial situation) was the
significant background risk factor of incident IA during the follow-up period (Table
5-7).
5.4.2.2 Approach I: Using baseline IA to predict incident depression and using
baseline depression to predict incident IA
Both univariate and adjusted multilevel logistic regression analyses (Model 1a
and Mode 1b in Table 5-8) showed that the prediction of baseline IA on incident
118
depression was not statistically significant. Hence, hypothesis 2 was not supported.
On the contrary, univariate multilevel logistic regression analysis showed that
baseline depression positively predicted incident IA during the follow-up period
(ORu=3.03, 95% CI: 2.26~4.06; see Table 5-9). The predictive effect of baseline
depression on incident IA remained significant after adjusting for significant
background variables (AOR=2.94, 95%CI: 2.19~3.96; see Model 1a in Table 5-9).
This predictive effect slightly attenuated, but was still significant when further
adjusting for baseline IA score (AOR=1.92, 95% CI: 1.41~2.62; see Model 1b in
Table 5-9). Hypothesis 6 was hence supported by the adjusted results.
5.4.2.3 Approach II: Using change in IA status over time to predict incident
depression and using change in depression status to predict incident IA
A significant univariate prediction was observed between change in IA status and
incident depression at follow-up (ORu=2.79, 95%CI: 1.39~5.60 for persistent IA;
ORu=4.62, 95%CI: 3.12~6.85 for incident IA; reference group: persistent non-IA; see
Table 5-10). This prediction remained significant after adjusting for significant
background variables, including family financial situation, academic performance and
perceived study pressure (AOR=2.61, 95%CI: 1.29~5.28 for persistent IA; AOR=4.59,
95%CI: 3.08~6.84 for incident IA, reference group: persistent non-IA; see Model 2a
in Table 5-10). However, when further adjusting for baseline CES-D score, only
incident IA significantly predicted incident depression (AOR=4.37, 95%CI: 2.87~6.67,
p<0.001, reference group: persistent non-IA; see Model 2b in Table 5-10). Hence,
hypothesis 4 was supported by the results.
Reversely, after adjusting for the significant background factor of family
119
financial situation, change in depression status significantly predicted incident IA
(AOR=2.53, 95% CI: 1.52~4.22 for remission depression; AOR=6.44, 95% CI:
4.47~9.28 for persistent depression; AOR=4.65, 95% CI: 3.14~6.89 for incident
depression; reference group: persistent non-depression; see Model 2a in Table 5-11).
After additional adjustment for baseline IA score on the basis of the model 2a, the
predictive effect of change in depression status on incident IA was still statistically
significant (AOR=3.91, 95% CI: 2.66~5.74 for persistent depression; AOR=4.10, 95%
CI: 2.73~6.17 for incident depression; reference group: persistent non-depression; see
Model 2b in Table 5-11). Therefore, hypothesis 8 was supported by the results.
5.5 Summary
5.5.1 Causal relationship between ‘SNA’ and depression
The results of the causal relationship between ‘SNA’ and depression by two
approach methods are summarized in Table 5-12. Based on the approach I, adjustment
for background variables and the baseline score of the corresponding outcome,
baseline ‘SNA’ did not predicted incident depression. Reversely baseline depression
significantly predicted incident ‘SNA’. The results indicated that the causal
relationship between ‘SNA’ and depression was unidirectional from earlier depression
to later ‘SNA’.
Results from approach II, adjusting for background variables and the baseline
score of the corresponding outcome, showed that change in ‘SNA’ status significantly
predicted incident depression, and reversely, change in depression status also
significantly predicted incident ‘SNA’ (Table 5-12). Specifically, for the prediction of
change in depression status on incident ‘SNA’, those who were persistently depressed
120
over time and those who became depressed during the follow-up period showed an
increased likelihood of developing ‘SNA’ compared with those who were free of
depression both at baseline and follow-up. On the other hand, for the prediction of
change in ‘SNA’ status on incident depression, persistent ‘SNA’ cases and new
incident ‘SNA’ cases showed an increased risk of developing depression compared
with persistent non- ‘SNA’ cases over time. Hence, a bi-directional relationship was
supported between change in ‘SNA’ and change in depression.
5.5.2 Causal relationship between IA and depression
The results for the causal relationship between IA and depression by two
approach methods are summarized in Table 5-13. After adjusting for background
variables and baseline score of the corresponding outcome, baseline depression
significantly predicted incident IA, but the predictive effect of baseline IA on incident
depression was not statistically significant. It is hence indicated a unidirectional
causal relationship from earlier depression to incident IA by modeling approach I.
When considering the change in status, the results from adjustment of
background variables and baseline score of the corresponding outcome showed that
change in IA status significantly predicted incident depression. Reversely, change in
depression status also significantly predicted incident IA (Table 5-13). Furthermore,
for the prediction of change in depression on incident IA, participants with persistent
depression and those with incident depression showed higher risk of developing IA
compared with those who were not depressed both at baseline and follow-up. In a
reverse trend, for the prediction of change in IA status on incident depression, incident
IA cases showed increased odds of developing depression when compared with those
without IA both at baseline and follow-up. Therefore, a bi-directional association was
121
supported between change in IA and change in depression.
5.6 Discussion
In this large-scale and school-based longitudinal study, we thoroughly examined
the directionality of the causal relationships between ‘SNA’/IA and depression using
two modeling approaches. In the first modeling approach, we used baseline ‘SNA’/IA
status to predict incident depression and vice versa, using baseline depression to
predict incident ‘SNA’/IA (approach I). In the second modeling approach, we used
change in ‘SNA’/IA status from baseline to follow-up (i.e. persistent non- ‘SNA’/IA,
remission from ‘SNA’/IA, persistent ‘SNA’/IA and incident ‘SNA’/IA) to predict
incident depression, and vice versa, using change in depression status (i.e. persistent
non-depression, remission from depression, persistent depression and incident
depression) to predict incident ‘SNA’/IA (approach II).
5.6.1 Depression as predictor of incident ‘SNA’/IA
Using the first approach, we found that baseline depression was a significant
predictor of incident ‘SNA’ (AOR=1.47) and incident IA (AOR=1.92) (Approach I).
These results are consistent with previous longitudinal studies among adolescents in
South Korea [44] and in Taiwan [39] that used the same statistical methods as our
approach I, which also reported that depressed participants are more likely of
developing IA. Consistently, the second approach showed that persistent depression
and incident depression over time significantly predicted incident ‘SNA’ (AOR=3.40
for persistent depression and AOR=4.47 for incident depression), and incident IA
(AOR=3.91 for persistent depression and AOR=4.10 for incident depression;
reference group: persistent non-depression). Therefore, change in depression status, as
122
compared to baseline depression as a predictor, was more predictive of incident ‘SNA’
and IA. There are no longitudinal studies that used this approach to predict new cases
of ‘SNA’/IA; thus no direct comparison can be made.
The two modeling approaches consistently revealed that depression is an
important determinant in the development of Internet addictive behaviors among
adolescents. These findings support the cognitive-behavioral theory [21], positing that
pre-existing psychopathology (i.e. depression, social anxiety, or substance dependence)
renders an individual vulnerable to Internet addictive behaviors. The
cognitive-behavioral theory proposes that maladaptive cognitions (i.e. rumination,
self-doubt, low self-efficacy, and negative self-appraisal) and dysfunctional behaviors
(i.e. using the Internet to escape from personal problems) are critical to the
development of Internet addictive behaviors. Depressed individuals always present
cognitive symptoms and usually possess positive expectancies for their Internet use
that the Internet could help them distracting from negative and personal problems (i.e.
depression, loneliness, etc.) [101, 207]. Online interaction (i.e. online social networking)
is attractive because of anonymity and absence of social cues (i.e. facial expression,
voice inflection, and eye contact), which make Internet-based interaction less
threatening in comparison to face-to-face communications [208, 209]. These features lead
depressed individuals to prefer for online social interactions as a more secure and less
threatening way and feel more comfortable in interaction online to escape from
emotional difficulty suffered in the real world, and the positive feelings and pleasures
from Internet-based interactions use cause excessive and addictive Internet/social
networking use.
123
5.6.2 ‘SNA’/IA as predictors of incident depression
Contrarily, the first approach did not support the prediction of baseline ‘SNA’
and IA onto incident depression, which is inconsistent with a previous longitudinal
study in China from Lam TL et al [43] that reported that participants with IA at
baseline were more likely to develop depression at follow-up. Such a difference may
be attributed to the fact that the results from Lam TL et al only controlled for the
background confounding variables, but the baseline level of depressive symptoms was
not controlled for even though the baseline level of depressive symptoms has a strong
association with future depression [210]. However, using the second modeling approach,
when change in ‘SNA’ and IA status over time (i.e. remission from ‘SNA’/IA at
baseline to non- ‘SNA’/IA at follow-up) were considered in the analysis, the results
showed a strongly positive prediction of persistent ‘SNA’/IA (AOR=1.66~1.88) and
incident ‘SNA’/IA (AOR=4.30~4.37) onto incident depression.
The difference in results obtained by approach I and II again suggests that the
latter approach is a potential methodological improvement, as it is more predictive in
both the case of depression predicting incident ‘SNA’/IA and the reverse.
Improvement may come from the fact that a high natural recovery rate from IA during
the follow-up period has been observed [40], and hence ignoring the remission effect
potentially underestimates the predictive effects of ‘SNA’/IA onto depression, such as
our non-significant findings for prediction of ‘SNA’ and IA at baseline on incident
depression. The modeling approach considering the dynamic change in ‘SNA’ and IA
status hence provides more convincing and robust estimation as such analysis
methodology rules out the potential effects from those recovery cases. One of the
explanations is that addictive Internet/social networking use would cause withdrawal
124
from interpersonal offline activities and displace the time that was originally spent
with family and peers in the real world, which then causes and increases mental health
problems (i.e. depression, loneliness) [61].
5.6.3 The bidirectional relationship
We hence have found unidirectional relationships from baseline depression to
incident ‘SNA’ and IA in approach I and bidirectional relationships between ‘SNA’/IA
and depression in approach II using change in status over time. The bi-directionality
suggests that a vicious circle between Internet addictive behaviors and depression
among adolescents may exist. It indicates that psychiatric individuals (i.e. depressed
individuals) are more likely to be prone to use the Internet and social networking
excessively to regulate their negative mood (i.e. alleviate negative emotions, anxiety
and personal problems) and prefer for online social interactions [45, 211-214]. These poor
cognitive and coping strategies accelerate the development and maintenance of
Internet addictive behaviors [207]. And as a feedback, increased addictive behaviors
further intensify and continue to produce negative outcomes (i.e. depressive mood),
producing diminished sense of self-worth and increased social withdrawal. As Internet
addictive behaviors worsen, they exacerbate the psychopathologies, creating a vicious
dysfunctional circle [17, 21]. Moreover, the strongest relationships were observed
between incident ‘SNA’/IA and incident depression in modeling approach II. This
suggests that the development of depression is highly likely to be accompanied by and
co-occurred with the development of Internet addictive behaviors.
5.6.4 Implications on prevention
The findings of this chapter entail several practical implications for designing
125
prevention and intervention programs. First, for the significant prediction of baseline
depression on incident Internet addictive behaviors, it is implied that adolescents with
depression are at high-risk group for increased probability of suffering from Internet
addictive behaviors later. Therefore, the prevention of Internet addictive behaviors
(‘SNA’ and IA) should be emphasized in students with depressive symptoms.
Intervention programs for Internet addictive behaviors should consider involving
specific skill training and components of reducing depressive mood (i.e. reducing the
maladaptive belief of positive outcome expectancies from the Internet, training social
skills, and planning leisure activities beyond the Internet use), which would
effectively prevent Internet addictive behaviors [215]. Second, epidemiological experts
and clinicians may suggest considering an assessment of depressive symptoms as a
marker of vulnerability of Internet addictive behaviors. Effective screening and
intervention for these psychiatric symptoms would be essential and effective for
preventing development of Internet addictive behaviors among adolescents. When
depression is screened, further interventions should be implemented to treat
depression. If depression is not well-treated, adolescents may spend more and more
time on the Internet to alleviate their depressive mood and then progress to addiction.
Therefore, interventions and preventions that target at high-risk group with identified
depressive symptoms would effectively prevent Internet addictive behaviors among
school adolescents.
For the strong prediction of change in ‘SNA’/IA status (i.e. persistent ‘SNA’/IA
and incident ‘SNA’/IA) on incident depression and the prediction of change in
depression status (i.e. persistent depression and incident depression) on incident
‘SNA’/IA, it is implied that Internet addictive behaviors are the concurrent predictors
126
of depression, and vice versa. Until now, no comprehensive therapeutic strategy for
Internet addictive behaviors has been established. Well-established therapy for
depression [216] could be considered as a prevention for Internet addictive behaviors
among adolescents. Accordingly, professionals working with adolescents who present
symptoms of Internet addiction also need to evaluate the potential presence of
depressive symptoms and prevent it simultaneously. Second, the vicious circle
relationship between depression and Internet addictive behaviors indicates a negative
reinforcement mechanism. Intervention campaigns should make efforts to block such
vicious circle.
5.6.5 Implications for future research
First, it should be cautioned that the status of Internet addictive behaviors and
depression might change during the longitudinal research period. As such, the time
interval should be carefully defined in the prospective study in order to truly reflect
the natural course of such pathological status. For example, although participants are
classified as IA both at baseline and follow-up, the IA status might be highly variable
between these two time points. Therefore, future longitudinal research is appropriate
to conduct the study with short-term interval assessments (e.g. three-month or
six-month interval). Additionally, statistical methodology should also consider such
status change in modeling specifications, such as using change in pathological status
over time rather than baseline status as predictors for mental health outcomes. Another
modeling method which could be alternatively considered in future similar research is
the cross-lagged panel model. However, this method has the limitation of controlling
diverse potential confounding factors in the model. Therefore, we did not apply this
analysis method in the present study as depression and internet addictive behaviors
127
are the consequences of multi-cause actions. Confounding factors are one of the
important issues that need to be carefully involved in the analysis. Second, ‘SNA’, one
subtypes of IA [17, 21, 217], showed a strong longitudinal relationship with depression in
the present study. This further highlights the necessity for investigating other specific
Internet addictive behaviors (e.g. gambling addiction) to explain the relationships
between Internet addictive behaviors and depression in future research. An increased
number of studies have emphasized the necessity for distinguishing specific Internet
addiction (e.g. social networking addiction) from generalized Internet addiction [22, 23].
Thirdly, the measurement and diagnostic criteria for ‘SNA’ is still in the way of its
development. It is necessary for future research to estimate an appropriate cut-off
value for the ‘SNA’ instrument with acceptable fitness (high sensitivity and specificity)
for community screening or diagnosis of ‘SNA’ among adolescents. Golden standard
criterion (e.g. systematic diagnostic interview by psychiatrists) is still needed for
cut-off value estimation of the ‘SNA’ instrument.
5.6.6 Strength and limitation
The main strength of the present study is the prospective study design with the
repeated measures of both ‘SNA’/IA and depression assessed by the same validated
survey instruments, as well as the large representative sample from the stratified
clustering sampling method that was used. Another major advantage is that two
directional relationships, including the longitudinal relationships from ‘SNA’/IA to
incident depression and the longitudinal relationships from depression to incident
‘SNA’/IA, were tested in the same study.
However, there are still some limitations that need to be emphasized in this
chapter specifically. Firstly, the two longitudinal directions between ‘SNA’/IA and
128
depression were tested separately using two sets of logistic regression models in the
present study. However, all models in the present study controlled for the baseline
level of corresponding outcomes and other potential confounding factors, the results
are therefore robust. Logistic regression using pathological status rather than
continuous scores could provide more meaningful explanations in epidemiological
research. Secondly, for the ‘social networking addiction’, we used the top 10% of the
scores to define social networking addictive status considering that there is no
available golden standard instrument and diagnostic criteria for ‘SNA’. The sensitivity
and specificity of such criterion for social networking addictive status in our study is
unclear and needs to be estimated in future research. However, the ‘SNA’ scale
showed acceptable psychometric properties in our study, and this strategy for
classification has been widely used in previous high-quality studies (e.g. using the top
10% of scores to create the group for psychological problems as no cut-off has been
established for South African child assessment schedule (SACAS) in Lancet
Psychiatry Journal). Thirdly, the generalizability of the study is limited to nonclinical
samples of school-based adolescents. Studies in diverse demographic populations (i.e.
psychiatric clinical populations) are warranted to further confirm such causal
relationships found in the present study. Other general limitations are mentioned in
Chapter 10.
In conclusion, the present study observed a vicious cycle between ‘SNA’/IA and
depression among adolescents, meaning that psychiatric disorder (i.e. depression)
significantly contributes to the course of social networking/Internet addictive
behaviors development, and in turn depressed individuals suffer more deleterious
effects from addictive Internet/social networking use. Future longitudinal studies with
129
three or more time points are warranted for further confirmation of these findings.
130
Table 5-1 Background factors as predictors of new incident depression among participants who
were non-depressed at baseline (n=3196)
Total Row %
ORu(95% CI)
(n=3196) (n=515)
Gender
Male 50.9 16.3 1
Female 49.1 15.9 0.96(0.79,1.16)
Grade
Seven 48.5 16.1 1
Eight 51.5 16.1 1.00(0.83,1.21)
Primary school or below 5.8 17.2 1
Secondary middle school 33.4 17.8 1.04(0.69,1.59)
High middle school 31.0 14.0 0.80(0.52,1.23)
University or above 25.5 15.8 0.92(0.60,1.42)
Don’t know 4.3 18.3 1.14(0.63,2.04)
Don’t know 4.4 18.4 1.32(0.77,2.25)
Very good/good 50.6 14.2 1
Average 47.2 17.8 1.32(1.08,1.60)**
Poor/very poor 2.2 24.6 1.98(1.12,3.49)*
Yes 89.3 15.8 1
No 10.7 18.7 1.25(0.93,1.67)
Upper 37.3 14.2 1
Medium 45.1 15.7 1.13(0.91,1.41)
Lower 17.6 21.3 1.66(1.28,2.16)***
Nil/light 21.8 13.8 1
Average 60.7 15.7 1.16(0.90,1.48)
Heavy/very heavy 17.4 20.5 1.63(1.20,2.20)**
ORu: univariate odds ratios, obtained by univariate multilevel logistic regression models.
† p<0.10; * p<0.05; ** p<0.01; *** p<0.001
131
Table 5-2 Background factors as predictors of new incident ‘SNA’ among participants who were
non-‘SNA’ at baseline (n=3657)
Total Row %
ORu (95% CI)
(n=3657) (n=335)
Gender
Male 48.5 9.4 1
Female 51.5 8.9 0.94(0.75,1.17)
Grade
Seven 48.0 9.1 1
Eight 52.0 9.2 1.00(0.80,1.26)
Don’t know 4.1 9.3 0.79(0.40,1.59)
Don’t know 4.2 8.5 1.03(0.52,2.03)
Average 48.4 9.7 1.21(0.96,1.53)
Poor/very poor 2.5 19.8 2.76(1.60,4.76)***
Yes 88.5 8.7 1
No 11.5 12.9 1.53(1.12,2.09)**
Upper 36.8 8.1 1
Medium 45.2 8.8 1.10(0.85,1.42)
Lower 18.0 12.3 1.61(1.19,2.19)**
Nil/light 20.4 7.9 1
Average 58.2 8.4 1.05(0.77,1.44)
Heavy/very heavy 21.4 12.5 1.65(1.17,2.32)**
† p<0.10; * p<0.05; ** p<0.01; *** p<0.001
132
Table 5-3 Predictive effect of baseline ‘SNA’ on incident depression among participants who were
non-depressed at baseline (n=3196)
Model 1a Model 1b
Row % ORu (95% CI)
AOR (95% CI) AOR (95% CI)
Baseline ‘SNA’
No 15.43 1 1 1
** *
Yes 23.36 1.65(1.22,2.22) 1.48(1.09,2.01) 1.17(0.86,1.60)
ORu: univariate odds ratios, obtained by univariate multilevel logistic regression.
AOR: Adjusted odds ratios.
Model 1a was adjusted for family financial situation, academic performance, and perceived study
pressure.
Model 1b was additionally adjusted for baseline CES-D score on the basis of model 1a.
† p<0.10; * p<0.05; ** p<0.01; *** p<0.001
Table 5-4 Predictive effect of baseline depression on incident ‘SNA’ among participants who were
non-‘SNA’ at baseline (n=3657)
Model 1a Model 1b
Row % ORu (95%CI)
Baseline depression
No 7.80 1 1 1
*** ***
Yes 14.56 2.02(1.58,2.58) 1.78(1.37,2.30) 1.47(1.13,1.92)***
Model 1a was adjusted for family financial situation, living with both parents or not, academic
performance, and perceived study pressure.
Model 1b was additionally adjusted for baseline SNA score on the basis of model 1a.
† p<0.10; * p<0.05; ** p<0.01; *** p<0.001
133
Table 5-5 Predictive effect of change in ‘SNA’ status over time on new incident depression among
participants who were non-depressed at baseline (n=3196)
Row Model 2a Model 2b
ORu (95% CI)
% AOR (95% CI) AOR (95% CI)
Change in ‘SNA’ status
Persistent non-‘SNA’ 13.14 1 1 1
** *
Remission from ‘SNA’ 21.23 1.77(1.21,2.58) 1.61(1.10,2.36) 1.30(0.87,1.92)
Persistent ‘SNA’ 27.37 2.46(1.54,3.93)*** 2.23(1.39,3.57)*** 1.66(1.02,2.70)*
Incident ‘SNA’ 42.54 4.89(3.67,6.52)*** 4.67(3.49,6.24)*** 4.30(3.17,5.82)***
pressure.
† p<0.10; * p<0.05; ** p<0.01; *** p<0.001
Table 5-6 Predictive effect of change in depression status over time on new incident ‘SNA’ among
participants who were non-SNA at baseline (n=3657)
ORu (95%CI)
% AOR (95% CI) AOR (95% CI)
Change in depression status
Persistent non-depression 5.30 1 1 1
Remission from depression 6.67 1.28(0.80,2.07) 1.18(0.73, 1.92) 0.97(0.60,1.58)
Persistent depression 20.48 4.62(3.43,6.21)*** 4.12(3.02, 5.62)*** 3.40(2.48,4.68)***
Incident depression 21.51 4.88(3.67,6.50)*** 4.70(3.52, 6.27)*** 4.47(3.33,5.99)***
Model 2a was adjusted for family financial situation, living with both parents or not, academic
performance, and perceived study pressure.
Model 2b was additionally adjusted for baseline SNA score on the basis of model 2a.
† p<0.10; * p<0.05; ** p<0.01; *** p<0.001
134
Table 5-7 Background factors as predictors of new incidence IA among participants who were
non-IA at baseline (n=3959)
Total Row %
ORu (95% CI)
(n=3959) (n=196)
Gender
Male 48.83 5.38 1
Female 51.17 4.54 0.83(0.62,1.11)
Grade
Seven 47.66 5.30 1
Eight 52.34 4.63 0.87(0.65,1.16)
Don't know 4.14 4.27 0.74(0.29,1.89)
Don't know 4.24 5.36 0.93(0.42,2.05)
Average 48.55 4.79 1.03(0.76,1.39)
Poor/very poor 2.75 13.76 3.28(1.82,5.91)***
Yes 88.71 5.01 1
No 11.29 4.47 0.90(0.56,1.44)
Upper 35.44 4.21 1
Medium 45.31 5.24 1.29(0.92,1.81)
Lower 19.25 5.64 1.41(0.94,2.11)
Nil/light 19.65 5.14 1
Average 58.02 4.22 0.81(0.56,1.19)
Heavy/very heavy 22.33 6.67 1.35(0.89,2.04)
† p<0.10; * p<0.05; ** p<0.01; *** p<0.001
135
Table 5-8 predictive effect of baseline IA status on new incident depression among participants
who were non-depressed at baseline (n=3196)
Model 1a Model 1b
Row % ORu (95%CI)
Baseline IA
No 15.93 1 1 1
Yes 20.83 1.42(0.90,2.24) 1.32(0.83,2.08) 1.05(0.65,1.68)
IA: Internet addiction.
ORu: univariate odds ratio, obtained by univariate multilevel logistic regression.
AOR: Adjusted odds ratio.
Model 1a was adjusted for family financial situation, academic performance and perceived study
pressure.
† p<0.10; * p<0.05; ** p<0.01; *** p<0.001
Table 5-9 predictive effect of baseline depression on incident IA among participants who were
non-IA at baseline (n=3959)
Model 1a Model 1b
Row % ORu (95% CI)
Baseline depression
No 3.51 1 1 1
*** ***
Yes 9.97 3.03(2.26,4.06) 2.94(2.19,3.96) 1.92(1.41,2.62)***
ORu: Univariate odds ratio, obtained by univariate multilevel logistic regression.
Model 1a was adjusted for family financial situation.
Model 1b was additionally adjusted for baseline IA score on the basis of model 1a.
† p<0.10; * p<0.05; ** p<0.01; *** p<0.001
136
Table 5-10 Predictive effect of change in IA status over time on incident depression among
participants who were non-depressed at baseline (n=3196)
Model 2a Model 2b
Row % ORu (95% CI)
Changes in IA status
Persistent non-IA 14.86 1 1 1
Remission from IA 15.66 1.09(0.60,1.99) 1.00(0.55,1.83) 0.83(0.45,1.55)
Persistent IA 32.43 2.79(1.39,5.60)** 2.61(1.29,5.28)** 1.88(0.91,3.89)†
Incident IA 45.37 4.62(3.12,6.85)*** 4.59(3.08,6.84)*** 4.37(2.87,6.67)***
ORu: Univariate odds ratio, obtained by univariate multilevel logistic regression.
pressure.
† p<0.10; * p<0.05; ** p<0.01; *** p<0.001
Table 5-11 Predictive effect of change in depression over time on incident IA among participants
who were non-IA at baseline (n=3959)
ORu (95% CI)
% AOR (95% CI) AOR (95% CI)
Changes in depression status
Persistent non-depression 2.28 1 1 1
*** ***
Remission from depression 5.60 2.53(1.52,4.22) 2.52(1.52,4.22) 1.66(0.98,2.81)†
Persistent depression 13.19 6.44(4.47,9.28)*** 6.26(4.32,9.06)*** 3.91(2.66,5.74)***
*** ***
Incident depression 10.00 4.65(3.14,6.89) 4.63(3.12,6.86) 4.10(2.73,6.17)***
ORu: univariate odds ratio, obtained by univariate multilevel logistic regression.
Model 2a was adjusted for family financial situation.
Model 2b was additionally adjusted for baseline IA score on the basis of model 2a.
† p<0.10; * p<0.05; ** p<0.01; *** p<0.001
137
Table 5-12 Summary of the causal relationship between ‘SNA’ and depression
DV=Incident Depression DV=Incident ‘SNA’
Unadjusted for Adjusted for Unadjusted for Adjusted for
IV
baseline DV baseline DV baseline DV baseline DV
Approach I (Baseline)
‘SNA’ * NS -- --
Depression -- -- *** ***
Approach II (Change in status)

Change in ‘SNA’
Persistent non-‘SNA’ 1 1 -- --
Remitted ‘SNA’ * NS
Persistent ‘SNA’ *** *
Incident ‘SNA’ *** ***
Change in depression
Persistent non-depression -- -- 1 1
Remitted depression NS NS
Persistent depression *** ***
Incident depression *** ***
DV: Dependent variable; IV: Independent variable.
NS: Non-significant.
--: Not applicable.
† p<0.10; * p<0.05; ** p<0.01; *** p<0.001
138
Table 5-13 Summary of the causal direction between IA and depression
DV=Incident Depression DV=Incident IA
Unadjusted for Adjusted for Unadjusted for Adjusted for
IV
baseline DV baseline DV baseline DV baseline DV
Approach I (Baseline)
IA NS NS -- --
Depression -- -- *** ***
Approach II (change in status)

Change in IA
Persistent non-IA 1 1 -- --
Remitted IA NS NS
Persistent IA ** †
Incident IA *** ***
Change in depression
Persistent non-depression -- -- 1 1
Remitted depression *** †
Persistent depression *** ***
Incident depression *** ***
DV: Dependent variable; IV: Independent variable.
NS: Non-significant.
--: Not applicable.
† p<0.10; * p<0.05; ** p<0.01; *** p<0.001
139
Chapter 6 Validation of the Social Networking Activity Intensity
Scale
6.1 Background
In the Chapter 5, we found that ‘SNA’ and IA were predictive of depression
among adolescents. From the perspective of early prevention, it is important to
understand whether features of online social networking use, such as its use intensity
and online activities, are also predictive of depression. If it is affirmative,
interventions should efficiently target these features before problems turn into ‘SNA’
and/or IA. Some studies conducted in high school and university students have
reported that more time spent on social networking use was positively associated with
diverse negative outcomes such as depression [35] and internalizing problems [218]. In
contrast, some other studies found no significant associations between social
networking use and depression among college students [53, 108].
However, most of these studies measured online social networking use behavior
only by generic items (e.g. frequency of use per week, time spent on online social
networking in a typical day), these simple items often perform poorly and do not
adequately reflected full spectrum and complex activities related to online social
networking use. Moreover, these studies do not differentiate functions and purposes
that adolescents mainly conducted on these platforms. Some researchers have argued
that online communication use of the Internet could enhance individuals’ social
involvement and interaction with others, and then showed beneficial impacts on their
psychological well-being (i.e. reduced depression and loneliness, increased subjective
well-being), which is called the simulation hypothesis [219, 220]; while recreational use
140
of the Internet potentially influenced individuals’ psychological well-being negatively
(i.e. increased loneliness and depression) because of reduced social integration (i.e.
less time spent with friends, family and other in reality), which is consistent with the
displacement hypothesis [61]. Therefore, in order to gauge thorough impacts of online
social networking use on adolescents’ mental health outcomes, it is highly warranted
to have fully validated instruments to assess online social networking use intensity
(SNUI) by taking their functions and activities into consideration. It is a prerequisite
for studying health impact of online social networking use on mental health outcomes
among adolescents.
Several measures exist but they have serious limitations. Although some specific
scales assessed social networking use intensity [51, 81, 221] , they tend to focus only on
one single online social networking platform (especially Facebook). For instance, the
Facebook Intensity Scale focuses only on the attitudes towards Facebook while the
majority of adolescents use multiple types of social media [14]. Furthermore, the study
provided information on internal consistency without mentioning other psychometric
properties for this scale. Moreover, those measures failed to capture the intensity of
the full range of online social networking activities. For instance, a study from Yang et
al [84] only covered four specific types of Facebook activities (electronic interactions,
voyeurism, self-presentation, and gaming); questions on two of these four types of
activities only involved a single item. This is largely inadequate as adolescents usually
engage in multiple types of social networking activities [54, 84, 89, 113] (e.g. messaging
with friends, posting comments, status updating, etc.).
In summary, there is a lack of acceptable instruments to assess online social
networking use behavior. It is necessary to develop and validate a more
141
comprehensive instrument for sufficient assessment of online social networking use
intensity (SNUI), covering various types of online social networking activities and
functions without bounding into the specific platform. Our phase I study, as described
in this Chapter, thereby conducted the validation analyses for the self-developed
Social Networking Activity Intensity Scale (SNAIS) among adolescents in China.
6.2 Objectives
In the present study, SNUI was defined as self-reported frequency of using
multiple types of online social networking activities, and in multiple types of
platforms during the last month. It aimed to develop a new tool, the online Social
Networking Activity Intensity Scale (SNAIS), which can be used to assess online
social networking behavior intensity among Chinese secondary school students. We
examined its psychometric properties including internal consistency, test re-test
reliability and construct validity.
As a measure of concurrent validity, it was hypothesized that the SNAIS score
would be significantly correlated with external variables such as emotional connection
to social networking, social networking addiction, Internet addiction and general
characteristics related to social networking use (i.e. duration of social networking use,
number of days per week using social networking, time spent on social networking
per day and number of online social networking friends).
6.3 Sample and Statistical method
6.3.1 Sample used in this chapter
The sample for validation analysis in this chapter was a separate sample from the
142
longitudinal study. The sampling method and data collection procedures have been
described in Chapter 3.
6.3.2 Statistical method
Descriptive statistics (i.e. means, standard deviation, or percentage) were
presented when appropriate. Floor or ceiling effects were considered to be present if
more than 15% of respondents possessed the minimum or maximum score of the
SNAIS [222]. Internal consistency was assessed by Cronbach’s alpha coefficient, while
test-retest reliability was evaluated by Intra-class Correlation Coefficient (ICC).
Pearson correlation coefficients between the items and the overall scale, between the
items and their corresponding subscales, and between the items and the other
subscales of the SNAIS were calculated for the item analysis. Moreover, Pearson
correlation coefficients among SNAIS/subscales, emotional connection to social
networking, social networking addiction, and Internet addiction were derived.
Furthermore, relationships between the SNAIS/subscales and general characteristics
related to online social networking use were tested by both Spearman correlation
coefficients and one-way ANOVA to establish external validity.
To examine the factor structures of the SNAIS, the sample was randomly split
into two subsamples with equal sample size. In the first subsample, the factor
structure of the SNAIS was extracted by exploratory factor analysis (EFA), using
principal components extraction method and varimax rotation methods. In the second
subsample, confirmatory factor analysis (CFA) with maximum likelihood estimation
was conducted to cross-validate the factor structures which were suggested by the
EFA. The following goodness of fit statistics and cutoff criteria were used to evaluate
the factor structure model in the CFA [223-225]: χ2/df ratio<5.00, both Non-Normed Fit
143
Index (NNFI) and Comparative Fit Index (CFI) >0.95, and Root Mean Square Error
of Approximation (RMSEA) <0.08. The confirmatory factor analysis was conducted
by using Lisrel 8.70, while all other statistical analyses were performed by using SAS
version 9.2 (SAS Institute, Cary, NC, USA). The statistical significance level was
p<0.05.
6.4 Results
6.4.1 Social-demographic characteristics
Of 910 students, 44.4%, 40.2% and 15.4% were students of the seventh, eighth
and ninth grades respectively. Besides, around 59.2% were males and 64.1%
possessed a smartphone.
6.4.2 Factor structures
The EFA revealed two factors with eigenvalue larger than one. The first 10-item
factor explained 42.2% of the total variance, and was named the “Social Function Use
Intensity (SFUI) (via social networking platform)” subscale. The second 4-item factor
explained an additional 8.7% of the total variance, and was named the “Entertainment
Function Use Intensity (EFUI) (via social networking platform)” subscale. All factor
loadings exceeded 0.50 (Table 6-1), indicating that all items should be included in the
confirmatory factor analysis [226]. The results of the CFA confirmed the two-factor
solution, showing acceptable goodness of fit to the data: χ2/df=4.05 (Chi-square
=307.68, df=76, p<0.001), NNFI=0.96, CFI=0.96, RMSEA=0.082 (90%CI:
0.073~0.092). The standardized path estimates were above 0.48 (ranged from
0.48~0.76) (Table 6-1).
144
6.4.3 Ceiling and floor effects of SNAIS and its two subscales
The percentage of respondents reporting maximum scores of the SFUI, EFUI,
and SNAIS were 0.88%, 0.66%, and 0.55% respectively, while the percentage of
participants reporting minimum scores of the SFUI, EFUI, and SNAIS were 0.88%,
1.76% and 0.55% respectively. Therefore, no significant ceiling or floor effects were
observed for the SNAIS and its two subscales.
6.4.4 Item analysis
Except for Item 8 and Item 14, item means were close to the middle of the range.
The correlation coefficients between each item and the overall scale ranged from 0.43
to 0.72 (all p<0.001), while all correlation coefficients between the items and their
corresponding subscale ranged from 0.64 to 0.77 (all p<0.001). All correlation
coefficients between individual items and their respective subscale were higher than
those between the same items and the other subscale (Table 6-2).
6.4.5 Reliability
Internal reliability based on the entire sample was acceptable for the SNAIS
(Cronbach’s alpha=0.90, 0.60 and 0.89 for the SFUI, EFUI, and SNAIS respectively).
The test-retest intra-class correlation coefficients were 0.87, 0.67 and 0.85 for two
subscales and the overall scale respectively.
6.4.6 Correlation between SNAIS, EC, ‘SNA’, and IA
The scores of the SNAIS and its two subscales were positively correlated with
the scales of emotional connection to social networking and ‘SNA’. The SFUI showed
slightly stronger correlations with emotional connection (Pearson r=0.49, p<0.001)
145
and ‘SNA’ (Pearson r=0.32, p<0.001), as compared to those of the EFUI (Pearson
r=0.40 and 0.25, respectively, p<0.001). In contrast, the EFUI presented slightly
stronger correlations with IA (Pearson r=0.23, p<0.001) as compared to that of the
SFUI (Pearson r=0.18, p<0.001) (Table 6-3).
6.4.7 Relationship between SNAIS and participants’ characteristics of social
networking use
Duration of social networking use, number of days per week on average using
social networking, amount of time per day on average spent on social networking, and
number of social networking friends were all positively correlated with the SFUI
(Spearman r=0.26 to 0.38, all p<0.001), EFUI (Spearman r=0.27 to 0.30, all p<0.001),
and SNAIS (Spearman r= 0.30 to 0.40, all p<0.001).
The results of the one-way ANOVA analysis showed that those who reported
higher levels of duration of social networking use, number of days per week on
average using social networking, amount of time spent on social networking in a
typical day, and number of social networking friends scored significantly higher in the
SNAIS and its two subscales (all p<0.001, Table 6-4).
6.5 Discussion
In this chapter, we have developed and validated the Social Networking Activity
Intensity Scale (SNAIS) among secondary school students in China. To our
knowledge, this is the first study of developing a tool to assess online social
networking behavior intensity among adolescents based on diverse social networking
activities. The results showed that the SNAIS had satisfactory psychometric properties.
Considering the dramatic increasing prevalence and popularity of social networking
146
use among adolescents (90% prevalence in our sample), the development of the
SNAIS would facilitate the progression of related studies investigating the effects of
online social networking on psychosocial well-being among adolescents.
The developed SNAIS has some strength. It is a comprehensive inventory that
includes various social networking activities instead of basing on only one specific
platform (e.g. Facebook). We found high internal and test-retest reliability, good
content validity supported by the item analysis, and strong correlations with the scale
and criterion variables. Furthermore, there are no noticeable ceiling and floor effects.
Construct validity was also clearly established, as two subscales, namely the SFUI
and EFUI, were extracted by exploratory factor analysis and were cross-validated by
confirmatory factor analysis.
The two subscales that emerged from the SNAIS have practical implications.
Previous research has indicated that the effects of Internet use on psychosocial
well-being was partially dependent on the nature of Internet use [227]. A one-year
longitudinal study revealed that Internet use for communication purposes was
protective of depression, while the use for non-communicative purposes predicted
higher levels of depression and social anxiety [204]. The two constructs of the SNAIS
would hence facilitate future investigations of variations between effects of the two
main types of social networking use onto mental health outcomes among adolescents.
The EFUI subscale gave acceptable but relatively low internal (α=0.60) and
test-retested reliability (ICC=0.67), as compared to that of the SFUI subscale. Adding
additional items to the EFUI subscale may improve the scale’s reliability.
Furthermore, the specific entertainment functions included in the subscale (e.g.
information-seeking, games, shopping) did not necessarily correlate strongly with
147
each other. Further elaborations and refinements are warranted. We believe that the
EFUI subscale is still useful, considering its acceptable psychometric properties. We
suggest both subscales to be used together in the future research, as nowadays social
interactions and entertainments are not mutually exclusive. Furthermore, the results
showed that the EFUI had a slightly higher correlation with Internet addiction as
compared with that of the SFUI. Such an observation need to be confirmed by the
future research.
This study has a few limitations. First, participants were recruited from only two
secondary schools in Guangzhou, and the psychometric properties of SNAIS should
be further evaluated in a large representative sample in China, as Internet penetration
and pattern of use may vary across geographic regions in China. Besides, nature of
social networking use has been evolving rapidly; new activities and functions of
online social networking need to be considered in the future.
In summary, the present study developed and evaluated an instrument for
assessing extensive functions of social networking use intensity. It has satisfactory
psychometric properties and can easily to be self-administered by secondary school
students. Its application may catalyze the development of relevant research on the
impact of social networking use among adolescents in China and beyond.
148
Table 6-1 Factor loadings and path loading estimated by exploratory and confirmatory factor
analyses for the Social Networking Activity Intensity Scale (SNAIS)
How often had you performed the following online social Factor loading(EFA) Path
networking activities in the last month? SFUI EFUI estimate(CFA)
1 Sent messages to friends on message board 0.71 0.15 0.63
2 Chatted with friends via instant messaging function 0.54 0.38 0.67
3 Replied to comments made by social networking
0.70 0.21 0.74
friends
4 Commented on friends’ status, logs, and photos 0.80 0.08 0.76
5 Shared/Forwarded contents 0.72 0.12 0.73
6 Browsed others’ logs/photos/status/album 0.74 0.20 0.72
7 Updated self-status 0.66 0.40 0.70
8 Posted photos/videos on personal web profile 0.56 0.26 0.59
9 Wrote logs/weibo 0.66 0.29 0.65
Decorated personal web profile(changing
10 0.55 0.46 0.71
image/contact information/privacy setting)
11 Surfed entertainment/current news 0.20 0.61 0.49
12 Watched video/listened to music 0.34 0.53 0.63
13 Played games/applications -0.02 0.80 0.49
14 Bought/gave virtual Goods (e.g. birthday gifts) 0.24 0.57 0.48
Eigen value 5.91 1.22 --
Cumulative % of Variance explained 42.21 50.93 --
Note: SFUI: Social Function Use Intensity subscale; EFUI: Entertainment Function Use Intensity
subscale; EFA: Exploratory Factor Analysis; CFA: Confirmatory Factor Analysis.
149
Table 6-2 Item analysis of the SNAIS (n=910)
Cronbach's alpha if
Item-total Item-subscale Item -another
Item X ± SD item is deleted
correlation correlation subscale correlation
Subscale Total
1 1.34±1.02 0.89 0.88 0.66*** 0.69*** 0.34***
*** ***
2 2.24±1.11 0.89 0.88 0.68 0.68 0.45***
3 1.59±1.23 0.88 0.88 0.72*** 0.75*** 0.37***
*** ***
4 1.50±1.14 0.88 0.88 0.73 0.77 0.35***
5 1.76±1.25 0.89 0.88 0.70*** 0.74*** 0.36***
*** ***
6 1.87±1.15 0.88 0.88 0.72 0.75 0.40***
7 1.78±1.13 0.88 0.88 0.74*** 0.75*** 0.46***
*** ***
8 0.88±1.02 0.89 0.88 0.61 0.64 0.32***
9 1.29±1.19 0.89 0.88 0.70*** 0.71*** 0.40***
*** ***
10 1.61±1.09 0.89 0.88 0.72 0.72 0.46***
11 1.72±1.10 0.56 0.89 0.49*** 0.64*** 0.38***
*** ***
12 2.69±1.09 0.52 0.89 0.57 0.68 0.45***
13 2.22±1.25 0.49 0.89 0.43*** 0.74*** 0.26***
*** ***
14 0.87±1.07 0.56 0.89 0.51 0.64 0.39***
Note: Item-total correlation: Spearman correlation coefficient between each item and the Overall
scale. Item-subscale correlation: Spearman correlation coefficient between each item and its
corresponding subscale. Item-another subscale correlation: Spearman correlation coefficient
between each item and the other subscale.
***
p<0.001
150
Table 6-3 Correlations among SNAIS, emotional connection, social networking addiction and
Internet addiction (n=910)
SNAIS Emotional
‘SNA’ score
Overall SFUI EFUI connection
SNAIS --
SFUI 0.97*** --
***
EFUI 0.74 0.54*** --
Emotional connection 0.52*** 0.49*** 0.40*** --
*** *** ***
‘SNA’ score 0.34 0.32 0.25 0.56*** --
IA score 0.22*** 0.18*** 0.23*** 0.38*** 0.49***
SNAIS: Social Networking Activity Intensity Scale; SFUI: Social Function Use Intensity subscale;
EFUI: Entertainment Function Use Intensity subscale; SNA: Social Networking Addiction; IA:
Internet Addiction.
***
p<0.001
151
Table 6-4 Mean score differences of SNAIS, SFUI, and EFUI by groups (n=910)
SNAIS SFUI EFUI
%
X ± SD p X ± SD p X ± SD p
Duration of social networking use
<3 months 19.1 17.8±9.6 11.6±7.7 6.2±2.9
<0.001 <0.001 <0.001
3-6 months 6.9 21.6±7.9 14.7±6.7 7.0±2.5
7-12 months 8.0 21.8±9.2 15.0±7.3 6.8±2.8
1-2 years 26.6 23.7±9.4 16.3±7.7 7.4±2.6
>2 years 39.3 26.4±10.3 18.0±8.4 8.4±3.3
Number of days/week
≤1 day 29.0 17.8±9.4 11.5±7.7 6.3±2.9
<0.001 <0.001
2-3 days 46.8 24.6±8.8 17.0±7.1 7.6±2.8 <0.001
4-5 days 10.7 25.0±9.2 17.2±7.7 7.8±2.8
≥6 days 13.5 29.6±11.4 22.2±9.3 9.4±3.3
Amount of time/day
<10 mins 20.0 17.3± 9.2 10.9±7.3 6.4±3.0
<0.001 <0.001 <0.001
11-30 mins 30.3 21.7± 8.2 14.7±6.8 7.0±2.6
31-60 mins 28.1 25.2± 9.2 17.5±7.4 7.7±2.9
>60 mins 21.6 28.9±11.3 20.0±9.0 9.0±3.3
Number of social networking friends
<50 45.9 19.5± 9.1 12.9±7.4 6.6±2.8
<0.001 <0.001 <0.001
51-100 25.3 23.9± 8.5 16.3±6.9 7.7±2.8
101-200 17.8 27.0± 9.6 18.7±7.9 8.3±2.8
201-400 7.3 31.3±10.0 21.8±8.1 9.5±3.3
>400 3.7 33.8±13.2 24.1±10.3 9.7±4.3
SNAIS: Social Networking Activity Intensity Scale; SFUI: Social Function Use Intensity;
EFUI: Entertainment Function Use Intensity
X ± SD : Mean ± Standardized Deviation
p values were obtained by one-way ANOVA
152
Chapter 7 Intensity and emotional connection of social networking
use as predictors of incident depression
7.1 Background
We have validated the instrument to assess social networking behavior intensity,
and two behavioral patterns related to social networking activities (SFUI and EFUI)
are identified with acceptable psychometric properties as described in the Chapter 6,
which reflect two different but important functions involving online social networking
behaviors. Moreover, another dimension, such as emotional connection, is also
potentially important and should be considered. As mentioned before, the majority of
previous research operationalized online social networking use only through simple
behavioral items (i.e. frequency of use, number of friends, time spent on a typical day),
very few studies have involved psychological dimension related to online social
networking.
In the present study, we additionally involved the measure of emotional
connection as an index to reflect the psychological aspect related to online social
networking use. Emotional connection taps the degree to which a participant is
emotionally attached to online social networking and the extent to which online social
networking is integrated into his/her daily activities [36, 51, 111]. Previous surveys have
reported that adolescents mainly used social networking to establish and maintain
relationship. This dependency on online social networking as an alternative form for
communication and as an information-seeking channel may result in emotional
attachment to social networking use. Online social networking potentially changes the
traditional information sharing and relationship establishment environment, and may
153
transfer the traditional emotional ties between individuals to between individuals and
online social networking. From this perspective, it is important to assess the level of
emotional connection to social networking, which could provide additional insight for
clarifying the impacts of online social networking use on depressive symptoms among
adolescents. There is no study that has estimated the psychological impact of
emotional connection to online social networking. A knowledge gap hence exists.
Moreover, emotional connection has significant influence on social networking

[228]
use intensity . Participants who feel emotionally attached to social networking
would perceive social networking to be useful and possess positive feelings towards
[228]
the use of social networking . It implies that behavioral intensity and emotional
connection may interrelate with each other and ultimately affect mental health
outcomes (i.e. depression). Therefore, it is plausible to hypothesize that the
relationship between social networking behavior intensity and depression varies on
the level of an individual’s emotional connection to online social networking, such
that emotional connection might strengthen or weaken the impacts of social
networking use intensity on depression. The integrated effects of behavior intensity
and emotional connection to social networking on depression have not yet been
researched. We therefore objectively investigated the moderation effects of emotional
connection on the relationship between social networking use intensity and depression
in this chapter.
The current study is expected to fill these knowledge gaps. If the predictive
effects of social networking use on depression are affirmative in the longitudinal data,
further effective intervention programs targeting improvements in psychological
well-being should emphasize the controllability for social networking use intensity
154
among adolescents before such behaviors turn into addictive tendency.
7.2 Objectives
This chapter firstly aimed to explore the longitudinal predictive effects of
baseline SFUI, EFUI, and EC score on incident depression during the nine-month
period among those non-depressed participants at baseline.
It is hypothesized that:
1) Higher score in SFUI at baseline would predict higher likelihood of
developing depression during the follow-up period after adjusting for
significant background factors and baseline level of depressive symptoms.
2) Higher score in EFUI at baseline would predict higher likelihood of incident
depression during the follow-up period after adjusting for significant
background factors and baseline level of depressive symptoms.
3) Higher level of emotional connection towards social networking at baseline
would predict higher likelihood of incident depression during the follow-up
period after adjusting for significant background factors and baseline level of
The second objective of this chapter was to test whether emotional connection
moderates the effects of social networking use intensity (SFUI and EFUI) at baseline
on incident depression. It is hypothesized that the predictive effects of SFUI and/or
EFUI on incident depression would be stronger among participants with higher level
of emotional connection to social networking compared to those with lower level of
emotional connection to social networking.
155
The subsample Ia was used in this chapter, which consisted of participants who
were non-depressed at baseline (n of subsample=3196). The outcome variable was
incident depression during the follow-up period, while the independent variables were
SFUI, EFUI, and EC scores at baseline. The analysis methods were the same as those
used in the Chapter 5. The univariate predictive effects of baseline SFUI, EFUI, and
EC scores on incident depression were firstly assessed. And then, two sets of adjusted
models were constructed, one with adjusting for significant background factors and
one with additional adjustment for baseline level of depressive symptoms. The Odds
Ratios (ORs) and corresponding 95% CI were reported.
The moderation effects of EC on the prediction of SFUI and EFUI on incident
depression were tested by the models including the main effects (that a particular
independent variable and EC) and their interaction term (that EC times a particular
independent variable), and statistical significance of interaction terms were examined.
7.4 Results
7.4.1 Background factors associated with incident depression
As described in Table 5-1 in Chapter 5 that used the same subsample as used in
this Chapter, the results showed that participants with poor family financial situation,
those who reported lower academic performance, and those who perceived heavy
study pressure were more likely than others to develop new incidence of depression
during the nine-month follow-up period.
156
7.4.2 Prediction of baseline SFUI, EFUI, and EC score on incident depression
Univariate analyses showed that baseline SFUI, EFUI and EC scores all
significantly predicted increased odds of incident depression, with ORu=1.015 (95%
CI: 1.003~1.027) for SFUI, ORu= 1.050 (95% CI: 1.019~1.083) for EFUI and
ORu=1.021 (95% CI: 1.002~1.039) for EC respectively (Table 7-1).
When significant background factors (including family financial situation,
academic performance, and perceived study pressure) were adjusted for, the predictive
effect of baseline EC score on incident depression became non-significant
(AOR=1.018, 95%CI: 1.000~1.037, p=0.054; see Table 7-1), while the prediction of
baseline SFUI and EFUI scores on incident depression remained significant:
AOR=1.016 (95% CI: 1.004~1.028) for SFUI and AOR=1.047 (95% CI: 1.015~1.080)
for EFUI. The prediction of baseline SFUI and EFUI scores were still significant even
after further accounting for the effect of baseline level of depressive symptoms:
AOR=1.017 (95% CI: 1.004~1.029) for SFUI and AOR=1.046 (95% CI: 1.012~1.080)
for EFUI (Table 7-1).
7.4.3 Moderation effects of EC on the relationships between SFUI, EFUI and
incident depression
The interaction effects between baseline EC and baseline SFUI, and between
baseline EC and baseline EFUI on incident depression were shown in Table 7-2. None
of the interaction terms were significantly associated with incident depression (all
p>0.05 in Table 7-2). Therefore, the moderation hypothesis of baseline EC on the
relationship between SFUI, EFUI at baseline and incident depression was not
supported.
157
7.5 Discussion
The present study found that higher level of social networking use intensity (both
social function and entertainment function use) at baseline predicted increased odds of
incident depression among adolescents. However, the hypotheses about prediction of
emotional connection to social networking and its interaction with social networking
use intensity on depression are not supported by the results.
Social networking use can involve various functions; social function and
entertainment function use are two major ones. Regarding the prediction of
entertainment function use intensity on depression, similar findings were also reported
in a one-year longitudinal study among Dutch adolescents, showing that more time
per week spent on Internet surfing use significantly predicted a greater likelihood of
depression [204]. A similar association was also observed between weekly hours spent
on Internet use for non-communication (i.e. playing games, shopping) and depressive
symptoms among college students in a previous cross-sectional study [229]. The
finding that social function use intensity predicted increased odds of depression also
concurred with that of a previous two-wave longitudinal study conducted among
middle school students, which reported that high intensity of online communication
use (i.e. instant messaging use, measured by two five-point Likert items: how often
one uses such Internet function and the importance of such function of
communication for keeping in contact with friends and peers) positively predicted
depressive symptoms six months later [230].
This study is the first to involve the measure of emotional connection to social
networking as a predictor of depression. Contrary to our hypotheses, it found no
prediction of emotional connection to social networking and its interaction with social
158
networking use intensity on depression among adolescents. We only found one
cross-sectional study among college students that used the same measure as the one
used in our study and evaluated the association between emotional connection to
Facebook and well-being outcomes, which also found a non-significant correlation of
emotional connection with well-being outcomes (i.e. emotional adjustment and social
adjustment) [36]. Emotional connection to social networking measured in the present
study was modified from six attitudinal items in the Facebook Intensity scale [51]. This
emotional connection measure related to social networking use (e.g. Facebook) has
been used in previous studies [36, 231, 232], and a high Cronbach’s α is reported (e.g.
α=0.89). In our validation study, this measure revealed a unidimensional factor
solution which explained 60% of the total variance and showed acceptable validity
(Cronbach’s α=0.87, in Chapter 3). However, when further inspecting the items of this
measure, two items “Social networking is part of my everyday activity” and “Social
networking has become part of my daily routine” might mainly reflect the integration
of social networking use into one’s daily activities rather than the emotional
connection construct. One recent study deleted these two items from the emotional
connection to Facebook measure by exploratory factor analysis [233]. More studies are
needed to further investigate whether emotional connection to social networking truly
has no impact on depression or if the current non-significant findings may be
attributed to measurement features for emotional connection. Besides, refinement and
improvement for this measure might be necessary to precisely estimate the emotional
connection construct and its effect on mental health outcomes among adolescents.
Overall, the present study preliminarily explored the impacts of social
networking use on depression among adolescents. It is found that both types of social
159
networking use were predictive of depression among adolescents. The magnitude of
the prediction of EFUI on depression was comparable with that of SFUI on depression
(AOR=1.046 for EFUI score, AOR=1.017 for SFUI score). These findings implicate
that early intervention should be made to reduce both functions of social networking
use intensity as a means of preventing depression rather than waiting for social
networking addiction to occur, given that Internet addictive behaviors (IA and ‘SNA’)
strongly associated with incident depression as reported in Chapter 5.
However, there are still many research questions that need to be addressed for the
above prediction of social networking use intensity on depression in future research,
such as how and why higher social networking use intensity affects the level of
depression among adolescents. It is important to further understand the mechanisms
underlying such relationships (e.g. potential mediators), as this information would be
useful for improving intervention research. One of the plausible explanations is that
preoccupation in online social networking activities substitutes for the time that they
originally spent on offline social activities (e.g. with friends and family), and then
causes social withdrawal and decline in their social involvement and psychological
well-being (i.e. increased depression) [61]. Previous cross-sectional studies showed that
social networking use is associated with individuals’ psychosocial relationships (e.g.
friendship quality, perceived social support, self-esteem, and adolescent-parents
relationship) [58, 59, 234, 235]. Therefore, one set of potential underlying mechanisms
would be intervening/moderation effects related to aforementioned psychosocial
factors. A number of mediation/moderation hypotheses involving these psychosocial
factors (i.e. friendship quality, perceived social support, self-esteem, social
non-confidence and adolescent-parents conflict) are hence tested and the
160
corresponding results are presented in Chapter 8.
Another plausible explanation for the observed relationship between social
networking use intensity and depression might be that higher intensity of social
networking use would increase the addictive tendency to Internet/social networking
use, and such addictive use in turn causes adolescents’ mental health problems (e.g.
depression). The positive associations between social networking use intensity (e.g.
Facebook use for social purposes) and Internet addictive behaviors (IA or Facebook
addiction) have been reported in previous studies. It is therefore reasonable to explore
the potential mediation effects of Internet addictive behaviors (IA and ‘SNA’) on the
relationship between social networking use intensity and depression. The mediation
and magnitude of Internet/social networking addiction for the relationship are tested
in Chapter 9.
In conclusion, the present study is the first to investigate the effects of social
networking use intensity and emotional connection to social networking on depression
among adolescents. Social networking behavior intensity is a significant risk factor of
depression among adolescents. The potential underlying mechanisms for the
relationship between social networking use intensity and depression are investigated
in the next two chapters, including the mediation, suppression, and moderation effects
of psychosocial factors (Chapter 8), the mediation effects of IA and ‘SNA’, as well as
joint mediation effects of psychosocial factors and Internet addictive behaviors (‘SNA’
and IA) (Chapter 9).
161
Table 7-1 Predictive effects of baseline SFUI, EFUI, and EC scores on incident depression by
multilevel logistic regression (n=3196)
Univariate Adjusted a Adjusted b
OR (95% CI) OR (95% CI) OR (95% CI)
* **
Baseline SFUI 1.015(1.003, 1.027) 1.016(1.004, 1.028) 1.017(1.004, 1.029)*
Baseline EFUI 1.050(1.019, 1.083)** 1.047(1.015, 1.080)** 1.046(1.012, 1.080)**
* †
Baseline EC 1.021(1.002, 1.039) 1.018(1.000, 1.037) 1.009(0.990, 1.029)
SFUI: Social Function Use Intensity; EFUI: Entertainment Function Use Intensity; EC: Emotional
Connection to social networking.
OR: Odds Ratios.
a
models were adjusted by family financial situation, self-reported academic performance, and
perceived study pressure.
b
models were additionally adjusted by baseline CES-D score.
† p<0.10, * p<0.05, ** p<0.01, *** p<0.001
162
Table 7-2 Testing the moderation effects of emotional connection on the relationships between SFUI, EFUI, and incident depression (n=3196)
Model 1a Model 1b Model 2a Model 2b
OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI)
Social networking use measures
Baseline SFUI 1.017(1.003, 1.032)* 1.015(0.980, 1.052) -- --
Baseline EFUI -- -- 1.045(1.010, 1.082)* 1.041(0.949, 1.142)
Baseline EC 0.997(0.975, 1.018) 0.994(0.949, 1.042) 1.001(0.980, 1.021) 0.998(0.947, 1.052)
Interaction terms
Baseline SFUI × Baseline EC -- 1.000(0.998, 1.002) -- --
Baseline EFUI × Baseline EC -- -- -- 1.000(0.995, 1.006)
SFUI: Social Function Use Intensity; EFUI: Entertainment Function Use Intensity; EC: Emotional Connection to social networking.
All models were adjusted for family financial situation, academic performance, perceived study pressure, and baseline level of depressive symptoms.
OR: Odds Ratios.
--: Not applicable
† p<0.10, * p<0.05, ** p<0.01, *** p<0.001
163
Chapter 8 Mechanisms underlying the relationship between social
networking use intensity and depressive symptoms: the roles of
psychosocial factors
8.1 Background
In Chapter 7, we established that baseline social networking use intensity
(both SFUI and EFUI) are predictive of incident depression among adolescents. It is
important to further explore the potential underlying mechanisms whereby SFUI and
EFUI influence depression, as such information would be helpful for designing
intervention programs. Considering that adolescents use online social networking
mainly for social interaction and communication purpose [55, 236, 237], such as seeking
friends and social support, maintaining and establishing friendships, we contend that
SFUI and EFUI would significantly impact adolescents’ psychosocial status (e.g.
friendship quality, perceived social support, self-esteem, adolescent-parents
relationship, and social non-confidence), and these psychosocial factors would
partially explain (mediate or moderate) the effects of social networking use intensity
(both SFUI and EFUI) onto depression among adolescents.
There is a lack of longitudinal studies involving aforementioned psychosocial
factors to explain the effects of social networking use intensity on depressive
symptoms, and the roles of psychosocial factors on the relationship between social
networking use intensity and depressive symptoms are still under-researched and
unclear. This chapter attempts to fill these knowledge gaps.
164
8.2 Objectives
The first set of objectives in this chapter was to test the potential mediation and
suppression effects of psychosocial factors on the associations between change in
social networking use intensity (SFUI and EFUI) and change in depressive symptoms.
For this part, the mediation hypotheses and suppression hypotheses were tested (see
Table 8-1). To perform such mediation and suppression hypotheses testing, the
requirements of mediation and suppression effects were firstly investigated.
Specifically, we tested the significance of the associations (1) between changes in
SFUI/EFUI and change in CES-D, (2) between changes in potential mediators (or
suppressors) and change in CES-D, and (3) between changes in SFUI/EFUI and
change in potential mediators (or suppressors). For the variables fulfilling all these
requirements, the multivariate linear regression models were further fit to test the
strength and significance of the corresponding mediation (or suppression) effects on
the associations between changes in SFUI/EFUI and change in CES-D.
The second set of objectives in this chapter was to test the moderation effects
of change in these psychosocial factors (i.e. change in friendship quality, change in
perceived social support, change in self-esteem, change in adolescent-parents conflict
and change in social non-confidence) on the relationships between changes in
SFUI/EFUI and change in CES-D. For this part, the significance of interactive terms
was tested.
165
8.3 Part I: Testing mediation and suppression effects
8.3.1 Some examples of mediation and suppression effect for the relationships
between Internet use and mental health outcomes
There are some examples that investigated the mediating effects between Internet
use and psychological well-being, mainly from cross-sectional studies. For instance,
one study from Lee JS and Jeong B investigated a resilience model in the relationship
between worry, daily amount of Internet video game playing, daily sleep duration and
depression, and the results showed that the association between number of worries
and the severity of depression was partially mediated by daily amount of Internet
video game playing and daily sleep duration [238]. Nabi RL et al reported that the
number of Facebook friends was positively associated with perceived social support,
which in turn was associated with reduced stress and greater subjective well-being [60].
Another study in South Korean university students showed a full mediation effect of
social support for the association between self-disclosure on Facebook and life
satisfaction (self-disclosure means communication personal information, thoughts,
and feelings with others through social networking) [115]. A one-week experimental
research project revealed that the beneficial effect of induced increase in Facebook
status updating activity on reduced loneliness was fully mediated by the increased
feelings of social connectedness with friends [57].
In regards to the suppression effect, we found one cross-sectional study among
high school students in Belgium [116], and the results reported an opposite direct and
indirect effect, showing that active Facebook use was positively associated with
depressive mood. On the other hand, active Facebook use positively associated with
perceived online social support, and then increased perceived online social support
166
further negatively associated with depressive mood. Although the results were
consistent with the suppression hypothesis, the authors did not interpret the finding as
a suppression effect and suppression effect size was not reported.
Overall, there is a dearth of studies investigating such mediation and suppression
effects for the relationship between Internet use (especially social networking use) and
mental health outcomes among adolescents.
8.3.2 Mediation and suppression hypotheses for the associations between change
in social networking use intensity and change in depressive symptoms
In this section, we discussed the potential mediators and suppressors for the
association between social networking use intensity and depressive symptoms. In our
research context and based on Baron and Kenny’s statistical strategy, to be considered
a mediator, the following requirements must be fulfilled: (1) the independent variable
should be positively associated with the potential mediator, (2) the potential mediator
should be positively associated with the dependent variable; while as a suppressor, it
needs to fulfill the following requirements: (1) the independent variable should be
positively associated with the potential suppressor; (2) the potential suppressor should
be negatively associated with the dependent variable. We summarized literature on the
associations between the potential mediators/suppressors and depressive symptoms, as
well as the associations between social networking use and the potential
mediators/suppressors. Based on the literature search, a number of potential
suppressors (including friendship quality, perceived social support, and self-esteem)
and a number of potential mediators (including adolescent-parents conflict, and social
non-confidence) were considered, and corresponding mediation and suppression
hypotheses were hence tested in this study.
167
A summary for the literatures on requirements and the types of hypotheses to be
tested were presented in Table 8-1. Details were described in subsequent sections.
Table 8-1 Summary of the mediation/suppression hypotheses to be tested and the

rationales based on the literature
Requirements
Psychosocial Association between Association between Type of
factor social networking use potential hypothesis
to be tested and potential mediator/suppressor to be tested
mediator/suppressor and depression
Hs1 Friendship quality + - Suppression
Perceived family
Hs2 + - Suppression
support
Perceived friend
Hs3 + - Suppression
support
Adolescent-father
Hm1 + + Mediation
conflict
Adolescent-mother
Hm2 + + Mediation
conflict
Social
Hm3 + + Mediation
non-confidence
Hs4/ Suppression/
Self-esteem +/- -
Hm4 mediation
H: Hypothesis
“+” indicates a positive association between two variables
“-” indicates a negative association between two variables
8.3.2.1 Friendship quality as a potential suppressor for the relationship between
social networking use and depressive symptoms
(1) Association between friendship quality and depressive symptoms
Establishing and maintaining social relationships with friends and/or romantic
partners is one of the developmental tasks during adolescence. Strong relationships
with others can greatly reduce stress and improve mental health in general.
Substantive evidence has shown that greater friendship quality is protective of lower
depressive symptoms. For instance, it has been reported that depressed adolescents
had less optimal peer relationships and fewer friends compared to non-depressed
168
cases [239]. Another study among high school students in Australia also reported that
adolescents’ friendship significantly predicted depression [240].
(2) Association between social networking use and friendship quality
Adolescents mainly used online social networking to maintain and strengthen
existing offline connections. It has been reported that about half (49%) of participants’
top offline friends were also their top social networking friends in a small college
sample (n=131)[54], and around 43% of sampled high school students perceived that
online social networking use made them feel closer to their friends[59]. Generally,
social networking use appears to benefit adolescents’ sense of friendship. For instance,
self-disclosure on social networking platforms was positively related with friendship
quality [16], which is in line with the stimulation hypothesis [219], positing that online
social networking (i.e. online communication) facilitates social connectedness, and
then improve closer and higher-quality friendship among adolescents.
Based on the literature review, it is therefore hypothesized that:
Hypothesis s1 (Hs1): Change in friendship quality over time would suppress the
positive relationship between change in social networking use intensity (both change
in SFUI and EFUI) and change in depressive symptoms (Figure 8-1).
△Friendship △Friendship
quality quality
+ - + -
+ +
△SFUI △CES-D △EFUI △CES-D
(a) Suppression of △friendship quality (b) Suppression of △friendship quality

for the △SFUI-△CES-D link for the △EFUI-△CES-D link
Figure 8-1 Hypothesis illustration for suppression effect of change in friendship
quality on the association between change in SFUI/EFUI and change in CES-D
169
8.3.2.2 Perceived social support as a potential suppressor for the relationship
between social networking use and depressive symptoms
(1) Association between perceived social support and depressive symptoms
It has been well-documented that perceived social support is protective factor of
lower depressive symptoms in previous studies. Adolescents with lower levels of
social support (family or peer) are at high risk compared to other adolescents for
experiencing increases in psychosomatic symptoms [241]. A longitudinal study among
adolescents in East London reported that lower level of social support were associated
with higher level of depression [242].
(2) Association between social networking use and perceived social support
Online social networking provides a new way for adolescents to obtain social
support [236]. Pew Internet and American Life Project reported that Facebook users had
higher level of perceived social support compared with non-Facebook users[243], and
additionally, the number of hours spent on Facebook and the number of Facebook
friends were also positively associated with greater perceived social support [60, 234].
However, some other studies found a non-significant association between social
networking use (i.e. Facebook intensity or Facebook interaction use) and perceived
social support [56, 244]. We did not find the study reporting a negative association
between social networking use and perceived social support.
In consideration of the literature discussed above, we hypothesized that:
Hypothesis s2 (Hs2): Change in perceived family support over time would
suppress the positive relationship between change in social networking use intensity
170
(both change in SFUI and EFUI) and change in depressive symptoms (Figure 8-2).
Hypothesis s3 (Hs3): Change in perceived friend support over time would
suppress the positive relationship between change in social networking use intensity
△Perceived △Perceived
family support family support
+ - + -
+ +
(a) Suppression of △perceived family (b) Suppression of △perceived family

support for the △SFUI-△CES-D link support for the △EFUI-△CES-D link
Figure 8-2 Hypothesis illustration for the suppression of change in perceived
social support on the association between change in SFUI/EFUI and change in CES-D
△Perceived △Perceived
friend support friend support
+ - + -
+
+ △EFUI △CES-D
△SFUI △CES-D
(a) Suppression of △perceived friend (b) Suppression of perceived friend

support for the △SFUI-△CES-D link support for the △EFUI-△CES-D link
support on the association between change in SFUI/EFUI and change in CES-D
8.3.2.3 Adolescent-parents conflict as a potential mediator for the relationship
between social networking use and depressive symptoms
(1) Association between adolescent-parents conflict and depressive symptoms
Harmonious family relationships, especially adolescent-parents relationship, are
important for adolescents’ development by providing emotional and instrumental
171
support. On the other hand, higher frequent conflict with parents (father and/or mother)
could increase the probability of depressive symptoms among adolescents. Several
studies have documented that poorer relationships with parents were associated with
higher level of depression among adolescents [239, 245, 246].
(2) Association between social networking use and adolescent-parents conflict
Family relationship is an important concern related to Internet use. In line with
the displacement theory, adolescents’ preoccupations in online social networking
would interfere with their everyday family activities, which is linked with less time
spent with parents and greater adolescent-parents conflict [54, 61], and college students
have acknowledged that social networking use decreased their face-to-face
interaction [247]. Indeed, a previous online survey among college students found that
those students who used social networking to communicate with parents reported
higher levels of conflicts within the parental relationship [58]. Reich and colleagues
reported that about 9% of respondents’ parents (or their guardians) monitored their
social networking use [59], which may be another source that causes adolescents’
conflicts with parents. However, it is still worth noting that some parents might utilize
social networking to maintain closer contact with their teens. Thus, there are
significant implications for exploring the impact of social networking use on
adolescents’ relationships with parents, especially the potential conflict with parents
caused by social networking use.
Based on the literature, we hypothesized that:
Hypothesis m1 (Hm1): Change in adolescent-father conflict over time would
mediate the positive relationship between change in social networking use intensity
172
Hypothesis m2 (Hm2): Change in adolescent-mother conflict over time would
△Adolescent- △Adolescent-
father conflict father conflict
+ + + +
+ +
(a) Mediation of △adolescent-father (b) Mediation of △adolescent-father

conflict for the △SFUI-△CES-D link conflict for the △EFUI-△CES-D link
Figure 8-4 Hypothesis illustration for the mediation of change in adolescent-father
conflict on the association between change in SFUI/EFUI and change in CES-D
△Adolescent- △Adolescent-
mother conflict mother conflict
+ + + +
+
+ △EFUI △CES-D
△SFUI △CES-D
(a) Mediation of △adolescent-mother (b) Mediation of △adolescent-mother

conflict for the △SFUI-△CES-D conflict for the △EFUI-△CES-D
link link
Figure 8-5 Hypothesis illustration for the mediation of change in adolescent-mother
conflict on the association between change in SFUI/EFUI and change in CES-D
8.3.2.4 Social non-confidence as a potential mediator for the relationship between
social networking use and depressive symptoms
(1) Association between social non-confidence and depressive symptoms
Adolescents with low social confidence and who experience social anxiety tend
to appear nervous during social interactions and are prone to be rejected by peers [248],
and consequently, those with low social confidence would foster social isolation and
173
other mental health problems. Social confidence as an important psychosocial factor
has been widely used in previous studies, e.g. the role of psychosocial factors on
neck-shoulder and low-back pain [249], among stroke patients [250], its relationships
with alcohol use and antisocial behaviors among middle school students [251], and its
association with school connectedness among children [122]. Socially non-confident
individuals demonstrate insufficient confidence and may feel highly anxious in social
occasions. One study from Zhuang MK et al reported that social non-confidence was
positively correlated with depressive symptoms among Chinese university
students [171].
(2) Association between online social networking use and social non-confidence
We did not find any studies that tested the association between online social
networking use and social non-confidence. Social networking use would potentially
reduce face-to-face interaction offline. Furthermore, online interaction is substantially
different from face-to-face interaction in terms of a lack of non-verbal cues (i.e. voice
tone, eye contact) and reduced degree of social presence [123]. Due to the reduced
probability of face-to-face interaction in real life, social confidence is likely to be
diminished. We therefore involved social non-confidence as an important
psychosocial factor in the present study and tested its mechanism for the association
between social networking use intensity and depressive symptoms.
Given the information presented above, we hypothesized that:
Hypothesis m3 (Hm3): Change in social non-confidence over time would
174
△Social non- △Social non-
confidence confidence
+ + + +
+ +
(a) Mediation of △social non-confidence (b) Mediation of △social non-confidence

for the △SFUI-△CES-D link for the △EFUI-△CES-D link
Figure 8-6 Hypothesis illustration for the mediation of change in social
non-confidence on the association between change in SFUI/EFUI and change in
CES-D
8.3.2.5 Self-esteem as a potential suppressor or a potential mediator for the
relationship between social networking use and depressive symptoms
(1) Association between self-esteem and depressive symptoms
Self-esteem is defined as a person’s overall self-evaluation of his/her worth, and
adolescence is an especially important period for self-esteem formation. Self-esteem
is an important factor in developing and maintaining mental health and overall quality
of life. Substantive studies reveal that lower self-esteem is prospectively associated
with depression. For instance, one recent systematic review summarized that
self-esteem is a protective factor for lower depressive symptoms [252]. In a study from
Brage and Meredith [253], the results showed that self-esteem had a strong and negative
effect on adolescent depression. Another study among Canadian adolescents also
reported that self-esteem was a key protective factor for depressive symptoms [254]. A
23-year longitudinal study showed that low self-esteem at adolescence and decreases
in self-esteem during the follow-up period were significantly associated with
depressive symptoms after two decades as adults [124].
175
(2) Association between social networking use and self-esteem
Some studies reported that time spent on social networking was negatively
associated with self-esteem [235]. Similarly, one study from Canadian university
Facebook users reported that greater Facebook activity was related with lower
self-esteem [64]. These results are consistent with the objective self-awareness (OSA)
theory. The OSA theory claims that individuals are prone to self-evaluations based on
the broader social standards and norms, and any stimulus (i.e. looking at self in a
mirror, hearing one’s own voice, or any other situation during which individual’
attention focuses on the his/her own self) causing the self to become the object
(instead subject) of consciousness would lead to a diminished impression of
oneself [255, 256]. Social networking profiles include information about oneself similar
to the information that would be used to prompt OSA; therefore, frequent visiting to
his/her own profile may cause a reduction of self-esteem according to the OSA
theory [63].
Moreover, the hyper-personal model is another relevant theoretical approach to
understanding the effect of social networking use on self-esteem, which stresses that
selective self-presentation on online social networking positively impacts
self-evaluation and then improves self-esteem [63]. Indeed, some authors have
presented that social networking use might potentially enhance self-esteem, such as
the study from Valkenburg et al among Dutch teenagers [62] and the study from
Gonzales et al [63], both of which reported a positive effect of social networking use on
self-esteem.
Give that the literature remains inconclusive regarding the potential influence of
social networking use on self-esteem, both mediation and suppression hypotheses
176
were tested in the present study.
We hypothesized that:
Hypothesis s4 (Hs4): Change in self-esteem over time would suppress the
positive relationship between social networking use intensity (both change in SFUI
and EFUI) and depressive symptoms (Figure 8-7).
Alternatively:
Hypothesis m4 (Hm4): Change in self-esteem over time would mediate the
positive relationship between change in social networking use intensity (both change
in SFUI and EFUI) and change in depressive symptoms (Figure 8-8).
△Self-esteem △Self-esteem
+ - + -
+ +
(a) Suppression of △self-esteem for the (b) Suppression of △self-esteem for the
△SFUI-△CES-D link △EFUI-△CES-D link
Figure 8-7 Hypothesis illustration for the suppression of change in self-esteem on the
association between change in SFUI/EFUI and change in CES-D
△Self-esteem △Self-esteem
- - - -
+ +
(a) Mediation of △self-esteem for the (b) Mediation of △self-esteem for the
△SFUI-△CES-D link △EFUI-△CES-D link
Figure 8-8 Hypothesis illustration for the mediation of change in self-esteem on the
association between change in SFUI/EFUI and change in CES-D
177
8.3.3 Implications of testing the mediation and suppression hypotheses
The present study has several implications. First, the findings would determine
the impacts of social networking use intensity on adolescents’ psychosocial status.
Current studies on the impact of social networking use on psychosocial relationships
are under-researched and showed mixed results. Second, the present study is the first
to explore the underlying mechanisms of psychosocial factors using a longitudinal
sample. By investigating the roles of diverse psychosocial factors on the association
between social networking use intensity and depressive symptoms, the underlying
mechanisms of the detrimental effects of social networking use intensity on
depression could be explored. The findings could be suggestive for the development
of intervention campaigns to improve mental health outcomes among adolescents, i.e.
if the underlying mechanisms are confirmed, interventions should therefore focus on
improving their psychosocial conditions (e.g. improving self-esteem). Third, the
findings in the present study could provide robust evidence for the theoretical
understanding of social networking use and psychological well-being. Fourth, an
additional implication for future research is the use of mediation framework to
explicitly study the suppression effect. The objective of the suppression analysis is to
determine the extent to which one variable explains the variance in the outcome that is
mistakenly attributed to a second variable when that variable is uncontrolled for. This
could imply that without considering the suppressor in the regression, the observed
impact of social networking use intensity on depression would be underestimated.
178
8.3.4 Statistical analysis for mediation and suppression effects
8.3.4.1 Sample and types of variables used in this chapter
The sample used in this chapter was the entire longitudinal sample (n=4237). The
dependent variable was change in CES-D score during the follow-up period (indicated
as △CES-D), while the independent variables were change in SFUI and EFUI score
during the follow-up period (indicated as △SFUI and △EFUI). Potential suppressors
included △friendship quality, △perceived family support, △perceived friend support
and △self-esteem, while potential mediators included △adolescent-father conflict,
△adolescent-mother conflict and △social non-confidence. All change variables in
scale scores were calculated by subtracting scale scores at baseline from those at
follow-up. The mean score differences in SFUI, EFUI, psychosocial scales and
depressive symptoms were described in Table 4-12 and Table 4-13 in Chapter 4.
8.3.4.2 Mediation and suppression effects testing
Given that students were nested within the schools, the assumption of
independency between students in the same school required in traditional statistical
methods (i.e. ordinary least-squares regression) would be violated. Multilevel linear
regression models (level 1: students, level 2: schools) (SAS MIXED procedure) were
therefore applied in this chapter for the analyses so that the potential clustering effects
from schools could be removed by specification of the random intercept in the models.
Background factors that were associated with △CES-D were selected by univariate
linear regressions at the 0.05 level. The strategies from Baron and Kenny [139] were
applied to evaluate the potential mediation/suppression effects of change in
psychosocial factors on the associations between △SFUI/△EFUI and △CES-D
179
(Figure 8-9).
8.3.4.2.1 Testing requirements for mediation and suppression hypothesis
The linear regression models were fit to examine the significance of the
associations between △SFUI/△EFUI and △CES-D, associations between
△SFUI/△EFUI and potential mediators/suppressors, as well as associations between
potential mediators/suppressors and △CES-D after adjusting for significant
background factors.
8.3.4.2.2 Testing the strength and significance of the mediation and suppression
effects
For those variables fulfilling the above requirements, multivariate linear
regression models were fit to test the effects of mediator/suppressor on △CES-D when
controlling for the △SFUI/△EFUI.
The strength of indirect effects (mediation or suppression) were estimated by the
formula a*b according to the product of coefficients method, and the Sobel Z test was
used to test statistical significance of the indirect effect [143, 257]. As a rule of thumb
suggested by David A. Kenny, the indirect effect is defined as very small, small,
medium and large with cut-off values as 0.01, 0.09 and 0.25 respectively. For the
mediation effect, the proportion that the total effect was mediated by the specific
individual variable was calculated using the formula a*b/c to reflect the magnitude of
the mediation effect size [143, 146], where 𝑎 is the regression coefficient relating the
independent variable to the mediator in step 2, 𝑏 and c’ are the regression coefficients
relating the mediator to the dependent variable and independent variable to the
dependent variable in step 3, and c is the regression coefficient relating the
180
independent variable to dependent variable in step 1 (see Figure 8-9). For the
suppression effect, the total effect cannot be decomposed into indirect and direct
effect proportions because the suppression effect has an opposite sign of direct
effect [145, 258]. Therefore, the proportion of indirect effect on total effect is not
applicable and was not calculated for suppression effects. Only suppression effects
(a*b) were reported.
The mediators were firstly tested individually, and then tested as a whole to
explore the combined mediation/suppression effects. All models were adjusted by
significant background factors that were associated with △CES-D.
e (1)
Independent variable X Step 1: c Dependent variable Y

(i.e. △SFUI or △EFUI) (i.e. △CES-D)
Y = b 0(1) + cX + e (1)
e (2)
Intervening variable M/S
(i.e. Change in psychosocial
scale score)
Step 2: a Step 3: b
e (3)
Step 3: c’
Independent variable X Dependent variable Y
(i.e. △SFUI or △EFUI) (i.e. △CES-D)
M = b 0(2) + aX + e (2)
Y = b 0(3) + c ' X + bM + e (3)
Total effect=c
Direct effect=c’
Mediation/suppression effect=a*b
Figure 8-9 Illustrations for conceptual framework of change in psychosocial factors

for the associations between △SFUI/△EFUI and △CES-D
181
8.3.5 Results of mediation and suppression effects
8.3.5.1 Descriptive statistics
The description of changes in SFUI and EFUI scores over time is present in
Table 4-12 in Chapter 4. It shows that both SFUI score and EFUI score had a slight
but significant decrease from baseline to follow-up. Similarly, the results presented in
Table 4-13 in Chapter 4 show that scores in friendship quality, perceived family
support, perceived friend support, conflict with father, and social non-confidence
slightly decreased during the follow-up period. However, the mean score differences
in conflict with mother and self-esteem between baseline and follow-up were
non-significant.
8.3.5.2 Background factors associated with △CES-D
Univariate analysis showed that grade, academic performance, and perceived
study pressure were significant background factors associated with △CES-D at the 0.
05 level (Table 8-2). These significant background factors were hence adjusted for in
the following regression analyses.
8.3.5.3 Associations between △SFUI, △EFUI, and △CES-D
After adjusting for significant background factors, both △SFUI and △EFUI were
positively associated with △CES-D. The regression coefficients were 0.072 for
△SFUI and 0.139 for △EFUI (all p<0.001, see Table 8-3). These results were similar
with those reported in Chapter 7, showing that baseline SFUI and EFUI scores
significantly and positively predicted incident depression during the follow-up period.
182
8.3.5.4 Testing the mediation hypotheses
8.3.5.4.1 Mediation effects for the association between △SFUI and △CES-D
Given that △SFUI was not significantly associated with △adolescent-father
conflict and △self-esteem, hypothesis m1 and hypothesis m4 were not tested. After
adjusting for significant background factors, △SFUI was positively associated with
△adolescent-mother conflict and △social non-confidence (Table 8-4). Meanwhile,
△adolescent-mother conflict and △social non-confidence were also positively
associated with △CES-D (Table 8-5). Therefore, since hypothesis m2 and hypothesis
m3 met the requirements, mediation hypotheses were tested only for these two
potential mediators.
Hypothesis m2: When △adolescent-mother conflict was added into the model 1
in Table 8-7, the regression coefficient of △SFUI on △CES-D remained significant,
but diminished from 0.073 to 0.070 (Model 2a in Table 8-6 and Figure 8-10). It
indicated that the positive effect of △SFUI on △CES-D was partially mediated by
△adolescent-mother conflict. The mediation effect was 0.004 (95% CI: 0.001~0.007,
Sobel Z=2.434, p=0.015), which indicated a very small effect size (smaller than the
cut-off of 0.01). Furthermore, the proportion mediated by △adolescent-mother
conflict was 5.11%.
△Adolescent-
mother conflict
0.041*** 0.091***
Direct effect: 0.070***

△SFUI △CES-D
Total effect: 0.073***
Figure 8-10 Mediation effect of change in adolescent-mother conflict on the
association between change in SFUI and change in CES-D
183
Hypothesis m3: When △social non-confidence was added into the model 1 in
Table 8-7, the regression coefficient of △SFUI on △CES-D remained significant but
diminished from 0.073 to 0.058 (Model 2b in Table 8-6 and Figure 8-11). It was hence
suggested that the positive effect of △SFUI on △CES-D was partially mediated by
△social non-confidence. The mediation effect from △social non-confidence was 0.015
(95% CI: 0.009~0.021, Sobel Z=5.137, p<0.001), which indicated a small effect size
(smaller than the cut-off of 0.09). Moreover, the proportion that total effect was
mediated by △social non-confidence was 20.97%.
△Social non-
confidence
0.04S*** 0.S56***

△SFUI △CES-D
Total effect: 0.07S***
Figure 8-11 Mediation effect of change in social non-confidence on the association

between change in SFUI and change in CES-D
When these two significant mediators were tested in the same model, the
combined mediation effect via △adolescent-mother conflict and △social
non-confidence were approximately 25.33% (Model 3 in Table 8-6). These results
implied that increase in SFUI augmented probability of adolescents’ conflict with
mother and social non-confidence, which in turn further increased their level of
depressive symptoms. Moreover, the mediation effect was mainly from △social
non-confidence.
184
8.3.5.4.2 Mediation effects for the association between △EFUI and △CES-D
The adjusted results showed that △EFUI was positively associated with △social
non-confidence (Table 8-4), and △social non-confidence was also positively
associated with △CES-D (Table 8-5). Therefore, △social non-confidence (hypothesis
m3) fulfilled the requirements, and the mediation hypothesis was conducted only for
△social non-confidence. For hypothesis m1, hypothesis m2 and hypothesis m4, given
that △EFUI was not significantly associated with △adolescent-father conflict,
△adolescent-mother conflict, and △self-esteem, these three mediation hypotheses
were not tested.
Hypothesis m3: After adding the △social non-confidence into the model 1 in
Table 8-8, the regression coefficient of △EFUI on △CES-D remained significant but
diminished from 0.139 to 0.096 (Model 2 in Table 8-7 and Figure 8-12). It suggested
that △social non-confidence partially mediated the association between △EFUI and
△CES-D, and the mediation effect was 0.043 (95%CI: 0.029~0.056, Sobel Z=6.068,
p<0.001), which indicated a small effect size (smaller than the cut-off of 0.09). The
proportion of THE mediation effect to total effect was around 30.65%.
△Social non-
confidence
0.117*** 0.S58***

△EFUI △CES-D
Total effect: 0.1S9***
Figure 8-12 Illustration for the mediation effect of △Social non-confidence on the
association between △EFUI and △CES-D
185
8.3.5.5 Testing the suppression hypotheses
8.3.5.5.1 Suppression effects for the association between △SFUI and △CES-D
The previous analyses showed that △SFUI was positively associated with
△friendship quality, △perceived family support, and △perceived friend support (Table
8-4). Additionally, △friendship quality, △perceived family support, and △perceived
friend support were all negatively associated with △CES-D (Table 8-5). Therefore, the
hypotheses s1, s2 and s3 were proceeded for the suppression testing. However, given
that △self-esteem was not significantly associated with △SFUI, hypothesis s4 was not
further tested.
Hypothesis s1: When adding the △friendship quality into the model 1 in Table
8-9, the regression coefficients of △SFUI on △CES-D increased from 0.073 to 0.084
(Model 2a in Table 8-8 and Figure 8-13). The suppression effect was -0.011 for
△friendship quality (Sobel Z=-2.946, p=0.003). Hence, hypothesis s1 was supported.
The results indicated that △friendship quality suppressed the positive association
between △SFUI and △CES-D. When statistically removing the effect of △friendship
quality, the association between △SFUI and △CES-D actually became larger.
However, the suppression effect from △friendship quality was small.
△Friendship
quality
0.018*** -0.621***

△SFUI △CES-D
Figure 8-13 Suppression effect of change in friendship quality on the association

186
Hypothesis s2: When adding △perceived family support into the model 1 in
Table 8-9, the regression coefficients of △SFUI on △CES-D increased from 0.073 to
0.083 (Model 2b in Table8-9 and Figure 8-14). The suppression effect was -0.009 for
△perceived family support (Sobel Z=-3.008, p=0.003). Hypothesis s2 was supported.
It meant that △perceived family support suppressed the positive effect of △SFUI on
△CES-D. When controlling for △perceived family support in the model, the
predictive validity of △SFUI on △CES-D was enhanced. However, the suppression
effect size was very small as the absolute value was less than the suggested cut-off of
0.01.
△Perceived
family support
P.P31*** -P.298***
Direct effect: P.P83***

△SFUI △CES-D
Total effect: P.P73***
Figure 8-14 Suppression effect of change in perceived family support on the

Hypothesis s3: When adding △perceived friend support into the model 1 in table
8-9, the regression coefficients of △SFUI on △CES-D increased from 0.073 to 0.095
(Model 2c in Table 8-8 and Figure 8-15). The suppression effect was -0.022 for
△perceived friend support (Sobel Z=-6.468, p<0.001). Hypothesis s3 was therefore
supported by the results. It meant that △perceived friend support suppressed the
positive effect of △SFUI on △CES-D. However, the suppression effect size was small
as the absolute value was less than the suggested cut-off of 0.09.
187
△Perceived
friend support
P.176*** -P.124***
Direct effect: P.P95***

△SFUI △CES-D
Total effect: P.P73***
Figure 8-15 Suppression effect of change in perceived friend support on the

8.3.5.5.2 Suppression effects for the association between △EFUI and △CES-D
The adjusted analysis showed that △EFUI was positively associated with
△perceived family support and △perceived friend support (Table 8-4), and reversely,
△perceived family support and △perceived friend support were negatively associated
with △CES-D (Table 8-5). These two variables fulfilled the requirements of
suppression hypotheses s2 and s3. Therefore, the suppression hypotheses were tested
only for △perceived family support (hypothesis s2) and △perceived friend support
(hypothesis s3). For △friendship quality and △self-esteem, because of their
non-significant association with △EFUI, hypotheses s1 and s4 were not further tested.
Hypothesis s2: When adding △perceived family support into the model 1 in
table 8-10, the regression coefficients of △EFUI on △CES-D increased from 0.139 to
0.184 (Model 2a in Table 8-9 and Figure 8-16). Hypothesis s2 was hence supported.
The suppression effect was -0.036 (Sobel Z=-5.571, p<0.001). It suggested that
△perceived family support suppressed the positive association between △EFUI and
△CES-D. However, the effect size was small as the absolute value was less than the
suggested cut-off of 0.09.
188
△Perceived
family support
P.149*** -P.3P3***
Direct effect: P.184***

△EFUI △CES-D
Total effect: P.139***
Figure 8-16 Suppression effect of change in perceived family support on the

association between change in EFUI and change in CES-D
Hypothesis s3: When adding △perceived friend support into the model 1 in
Table 8-10, the regression coefficients of △EFUI on △CES-D increased from 0.139 to
0.178 (Model 2b in Table 8-9 and Figure 8-17). Hypothesis s3 was therefore
supported. The suppression effect was -0.039 (Sobel Z=-5.518, p<0.001, Table 8-10).
Hypothesis s3 was supported. It suggested that △perceived friend support acted as a
suppressor for the positive association between △EFUI and △CES-D. However, the
suppression effect was small.
△Perceived
friend support
P.322*** -P.121***
Direct effect: P.178***

△EFUI △CES-D
Total effect: P.139***
Figure 8-17 Suppression effect of change in perceived friend support on the

association between change in EFUI and change in CES-D
8.3.6 Summary for the mediation and suppression effects
8.3.6.1 Summary for the mediation effects
After adjusting for the significant background factors, change in
189
adolescent-mother conflict and change in social non-confidence partially mediated the
associations between change in SFUI and change in CES-D. The magnitudes of
mediation effect were 5.11% for change in adolescent-mother conflict and 20.97% for
change in social non-confidence respectively.
Moreover, change in social non-confidence also partially mediated the
association between change in EFUI and change in CES-D. The magnitude of
mediation effect via change in social non-confidence was 30.65%.
8.3.6.2 Summary for the suppression effects
The positive association between change in SFUI and change in CES-D was
suppressed by change in friendship quality, change in perceived family support, and
change in perceived friend support. The suppression effects were -0.011, -0.009 and
-0.022 respectively.
Change in perceived family support and change in perceived friend support
suppressed the positive association between change in EFUI and change in CES-D
after adjusting for significant background factors. The suppression effects were -0.036
for change in perceived family support and -0.039 for change in perceived friend
support.
8.4 Part II: Testing moderation effects
8.4.1 Moderation effects of psychosocial factors on the association between social
networking use intensity and depressive symptoms
Previous studies have suggested that social resources could moderate the
relationships between Internet use and psychological outcomes [219, 259]. Effect of
190
social networking use on mental health outcomes (i.e. depressive symptoms) might
vary in terms of levels of some psychosocial factors, stressing the necessity to
consider moderation effects. The psychosocial factors investigated in our study (e.g.
friendship quality, perceived family support, perceived friend support, self-esteem,
adolescent-father conflict, adolescent-mother conflict, and social non-confidence) can
be seen as participants’ social resources. In addition to being predictive factors on
depressive symptoms, they may also moderate the relationships between social
networking use intensity and depressive symptoms. Knowing such information could
clarify what people would be impacted greatly by social networking use, and could
then guide for future intervention campaigns.
The relevant and commonly mentioned hypotheses are the rich-get-richer and
poor-get-poorer effects, which forms the basis of the “Matthew effect” originated
initially by Merton RK in 1968 in sociological science [260]. Merton RK proposed the
Matthew effect originally to explain discrepancies in scientific career development,
such that the eminent scientists usually receive more credit for their contributions to
science, while relatively unknown scientists tend to get little credits for the
comparable achievement [260]. This states that a relatively favorable position acts as a
resource that produces further gains and creates inequality across any temporal
process [261]. In a general sense, the Matthew effect denotes the psychosocial process
that the “rich-get-richer and poor-get-poorer”, and its name quotes from the Bible:
“For unto every one that hath shall be given, and he shall have abundance: but from
him that hath not shall be taken away even that which he hath” (Matthew 25:29). The
Matthew effect has been introduced into other research fields, i.e. the educational field
to describe the development of individual differences in reading ability [262, 263] and in
191
economics field to explain the poverty and inequality [264]. The Matthew effect has
also been introduced into public health research. For instance, it is used to examine
whether patients with low socioeconomic status and patients with emotional problems
were more likely to become incapacitated for work after heart disease than those with
high socioeconomic status and those without emotional problems [265].
The rich-get-richer and poor-get-poorer effects had also been applied to explain
the effects of Internet use on mental health outcomes first by Kraut et al in 2002 [46].
The longitudinal results from Kraut et al showed that Internet use was associated with
better psychosocial outcomes (i.e. increased self-esteem) for extraverts and those with
more social support (rich-get-richer), while Internet use was associated with worse
psychosocial outcomes (i.e. increased loneliness) for introverts and those with less
social support (poor-get-poorer) [46]. A one-year longitudinal study among Dutch
adolescents tested the moderating effect of friendship quality on the association
between Internet use and depression and social anxiety by hierarchical regression
models with depression/social anxiety at time 2 as dependent variables and types of
Internet use and friendship quality at time 1 as independent variables. Meanwhile the
poor-get-poorer hypothesis was tested by the interaction terms of types of Internet use
and friendship quality, and the results found that the harmful effects of Internet surfing
on depression and social anxiety were significantly larger among adolescents with low
friendship quality than those with high friendship quality, which supported the
poor-get-poorer hypothesis [204]. However, a longitudinal study from Van den Eijnden
et al tested the poor-get-poorer effect based on the assumption that the positive
prediction of Internet use on depression would be larger for adolescents with high
level of loneliness than for those with low level of loneliness by multi-group
192
comparisons of SEM, but the poor-get-poorer hypothesis was not supported [230]. The
results are hence mixed.
Figure 8-21 illustrates the general concept of the poor-get-poorer hypothesis in
terms of the effect of social resources (e.g. friendship quality) on the positive
association between Internet use and depression. Based on the poor-get-poorer
hypothesis, the positive effect of Internet use on depression would be greater among
individuals with low levels of social resources as compared to those with high levels
of social resources.
Low social resource

Level of Depression
High social resource
Internet use
Figure 8-18 Illustration of the poor-get-poorer hypothesis
In this longitudinal study, score changes were used to reflect the within-subject
dynamic variations in the levels of individual’s social resources (i.e. change in
friendship quality, perceived family support, perceived friend support, self-esteem,
adolescent-father conflict, adolescent-mother conflict, and social non-confidence).
The poor-get-poorer model is thus modified to take into account longitudinal nature of
the data, which is poorer-get-poorer. We hypothesized that the positive association
between increase in social networking use intensity (SFUI and EFUI) and increase in
193
depressive symptoms would be stronger among those with ‘large deterioration’ in
social resources as compared to those with ‘small change’ in social resources during
the follow-up period (see Figure 8-19 below for the modified poorer-get-poorer
hypothesis of the ‘large deterioration’ in social resources as compared to ‘small
change’ in social resources).
Another hypothesis that is commonly applied for moderation effect testing
involving social resource variables is referred to a buffering effect model. The
buffering effect model has been elaborated in detail by Cohen sheldon [266] in 1985,
which posits that the pathogenic influence of risk events on mental health outcomes
(i.e. depression) are smaller under conditions of high perceived social support than
under conditions of low perceived social support. Studies on buffering effects (e.g.
social support) on the associations between risk behaviors and negative outcomes
have been reported in previous studies, especially those focusing on social support.
For instance, it is reported that social support could protect bullied adolescents from
poor academic achievement and depression [267]. A longitudinal study reported that
effortful control buffered the risk of adolescent family adversity on externalizing
problems (i.e. Aggressive Behavior and Rule-Breaking Behavior) [268]. However, there
are no studies that investigate the potential buffering effect of social resources on the
associations between social networking use intensity and depression among
adolescents. In the present study, for those whose personal social resources largely
improved during the follow-up period, it might be plausible that such ‘large
improvement’ in social resources would diminish the adverse effects of social
networking use on depression, suggesting that buffering effects could potentially exist.
Figure 8-19 illustrates the buffering effect hypothesis for ‘large improvement’ in
194
social resources as compared to ‘small change’ in social resource for the association
between change in social networking use intensity and change in depressive
symptoms.
'Large deterioration'
Poorer-get-poorer in social resource
Change in CES-D
Buffering effect 'Small change' in

social resource
'Large improvement'
in social resource
Change in social working use intensity
Figure 8-19 Illustration for modified poorer-get-poorer and buffering effect

hypothesis
Following the aforementioned logic and hypotheses, we developed a three-level
definition for the change scores of social resource variables (i.e. friendship quality,
perceived family support, perceived friend support, self-esteem, adolescent-father
conflict, adolescent-mother conflict, and social non-confidence) based on the strategy
of 0.5 standard deviation (SD) above and below the mean of change scores (i.e. mean
± 0.5SD). Definition for each level are as follows: ‘Large deterioration’ in social
resource (those scoring 0.5 SD below the mean of change scores), ‘Large
improvement’ in social resource (those scoring 0.5 SD above the mean of change
scores) and ‘Small change’ in social resource (those scoring between 0.5 SD below
the mean of change scores and above the mean of change scores).
195
Therefore, there are two set of hypotheses, one set for the poorer-get-poorer
hypotheses (i.e. ‘Large deterioration’ in social resources would increase the effects of
change in social networking use intensity on change in depressive symptoms as
compared to the ‘small change’ group in social resources), the other set for the
buffering effect hypotheses (i.e. ‘Large improvement’ in social resources would buffer
the effects of change in social networking use on change in depressive symptoms as
compared to the ‘small change’ group in social resources).
The specific poorer-get-poorer hypotheses for each social resource variable are
presented in detail below:
H4a: The positive effects of change in SFUI/EFUI on change in depressive
symptoms would be larger among participants with ‘large deterioration’ in friendship
quality as compared to those with ‘small change’ in friendship quality during the
follow-up period.
H4b: The positive effects of change in SFUI/EFUI on change in depressive
symptoms would be larger among participants with ‘large deterioration’ in perceived
family support as compared to those with ‘small change’ in perceived family support
during the follow-up period.
H4c: The positive effects of change in SFUI/EFUI on change in depressive
symptoms would be larger among participants with ‘large deterioration’ in perceived
friend support as compared to those with ‘small change’ in perceived friend support
H4d: The positive effects of change in SFUI/EFUI on change in depressive
symptoms would be larger among participants with ‘large deterioration’ in self-esteem
196
as compared to those with ‘small change’ in self-esteem during the follow-up period.
H4e: The positive effects of change in SFUI/EFUI on change in depressive
symptoms would be larger among participants with ‘large deterioration’ in
adolescent-father conflict (meaning a large increase in conflicts with the father) as
compared to those with ‘small change’ in adolescent-father conflict during the
follow-up period.
H4f: The positive effects of change in SFUI/EFUI on change in depressive
symptoms would be larger among participants with ‘large deterioration’ in
adolescent-mother conflict (meaning a large increase in conflicts with the mother) as
compared to those with ‘small change’ in adolescent-mother conflict during the
follow-up period.
H4g: The positive effects of change in SFUI/EFUI on change in depressive
symptoms would be larger among participants with ‘large deterioration’ in social
non-confidence (meaning a large increase in social confidence) as compared to those
with ‘small change’ in social non-confidence during the follow-up period.
The specific buffering effect hypotheses for each social resource variable are
presented in detail below:
H5a: The positive effects of change in SFUI/EFUI on change in depressive
symptoms would be smaller among participants with ‘large improvement’ in
friendship quality as compared to those with ‘small change’ in friendship quality
H5b: The positive effects of change in SFUI/EFUI on change in depressive
197
perceived family support as compared to those with ‘small change’ in perceived
family support during the follow-up period.
H5c: The positive effects of change in SFUI/EFUI on change in depressive
perceived friend support as compared to those with ‘small change’ in perceived friend
support during the follow-up period.
H5d: The positive effects of change in SFUI/EFUI on change in depressive
self-esteem as compared to those with ‘small change’ in self-esteem during the
follow-up period.
H5e: The positive effects of change in SFUI/EFUI on change in depressive
adolescent-father conflict (meaning large increases in conflicts with the father) as
compared to those with ‘small change’ in adolescent-father conflict during the
follow-up period.
H5f: The positive effects of change in SFUI/EFUI on change in depressive
adolescent-mother conflict (meaning large increases in conflicts with the mother) as
compared to those with ‘small change’ in adolescent-mother conflict during the
follow-up period.
H5g: The positive effects of change in SFUI/EFUI on change in depressive
symptoms would be smaller among participants with ‘large improvement’ in social
198
non-confidence (meaning large decreases in social non-confidence) as compared to
those with ‘small change’ in social non-confidence during the follow-up period.
8.4.2 Statistical methods for testing moderation effects
The moderating effects of △friendship quality, △perceived family support,
△perceived friend support, △self-esteem, △adolescent-father conflict,
△adolescent-mother conflict, and △social non-confidence on the relationship between
△SFUI/△EFUI and △CES-D were tested by fitting the models including one of the
independent variables (△SFUI or △EFUI), one of the psychosocial factors
(△friendship quality, △perceived family support, △perceived friend support,
△self-esteem, △adolescent-father conflict, △adolescent-mother conflict, and △social
non-confidence), and their interaction terms, after adjusting for the significant
background factors. The significance of interaction terms were evaluated by
comparing the main effect model and the model containing the main effect variables
and the interaction term by Likelihood Ratio Test [269], using the change in statistics
-2*Log Likelihood (-2LL), which followed a Chi-square distribution with degree of
freedom (df) equals to two.
When the significant interaction was shown by Likelihood Ratio Test, the
poorer-get-poorer and buffering effect hypotheses could be identified by inspecting
the sign of interaction terms. Specifically, if the sign of the interaction term between
△SFUI/EFUI and group in ‘large deterioration’ in social resources was significantly
positive, the poorer-get-poorer hypothesis would be supported (i.e. ‘large deterioration’
in social resources over time amplifies the positive influences of △SFUI/△EFUI onto
△CES-D as compared to the ‘small change’ group). For the interaction terms between
199
△SFUI/EFUI and the group in large improvements in psychosocial status, if the sign
of the interaction was negative, a buffering effect would be supported for
improvement in social resources, meaning that ‘large improvement’ in social
resources over time would weaken the positive impact of △SFUI/EFUI onto △CES-D
as compared to ‘small change’ group.
Model a Model b
Change in SFUI/EFUI Change in SFUI/EFUI
＋＋
Change in psychosocial Change in psychosocial
variables variables
＋
Change in SFUI/EFUI
×
Change in psychosocial
variables
Figure 8-20 Hierarchical regression models to test the interaction effects between
change in SFUI (or change in EFUI) and change in psychosocial variables on change
in depressive symptoms
8.4.3 Results
The moderation results are showed in Table 8-10 and Table 8-11. Models 1b~7b
were fit by adding an interaction term (e.g. △SFUI*△friendship quality, including two
dummy variables: △SFUI*large improvement and △SFUI*larger deterioration with
‘small change’ as reference group) to the corresponding main effect models (model
1a~7a). Significant interactions at the 0.05 level were detected between △SFUI and
△adolescent-father conflict (Change in -2LL between Model 5a and Model 5b =11.7,
p=0.003 in Table 8-10), and between △EFUI and △adolescent-father conflict (Change
in -2LL between Model 5a and Model 5b =7.9, p=0.019 in Table 8-11). The other
200
interactions were not statistically significant.
In particular, the interaction terms between △SFUI and the group in ‘large
deterioration’ in adolescent-conflict father as well as between △EFUI and the group in
‘large deterioration’ in adolescent-father conflict were marginally significant.
However, the direction did not support the poorer-get-poorer hypothesis as the
regression coefficients of these two interaction terms were negative, meaning that
those that deteriorated encountered smaller association between increase in social
networking use intensity and increase in depressive symptoms (see Figure 8-24).
Significant interactions between △SFUI and the group in ‘large improvement’ in
adolescent-father conflict, and between △EFUI and the group in ‘large improvement’
in adolescent-father conflict were found in the same two models. The negative sign of
the interaction terms indicated a buffering effect, meaning that the large improvement
in the relationship with the father over time (meaning a decrease in adolescent-father
conflict) buffered the positive influence of △SFUI and △EFUI onto △CES-D as
compared to the small change group. Figure 8-21 illustrated the significant
interactions found. It is seen that the positive effects of △SFUI and △EFUI on
△CES-D were smaller among those with ‘large improvement’ in adolescent-father
conflict as compared to the ‘small change’ group.
Besides, all main effect models (Model 1a~7a in Table 8-10 and Table 8-11)
showed that △SFUI and △EFUI were positively associated with △CES-D (bs ranging
from 0.063 to 0.091 for △SFUI; bs ranging from 0.108 to 0.168 for △SFUI), which
are consistent with those in part I.
201
8 Large deterioration
6 Large improvement
Small change
4
Change in CES-D
2
-2
-4
-6
-8
-40 -30 -20 -10 0 10 20 30 40
Change in SFUI
(a) Interaction between change in SFUI and change in adolescent-father conflict
8 Large deterioration
6 Large improvement
Small change
4
Change in CES-D
2
0
-2
-4
-6
-8
-20 -15 -10 -5 0 5 10 15 20
Change in EFUI
(b) Interaction between change in EFUI and change in adolescent-father conflict
Figure 8-21 Graphical illustration for the interactions between change in SFUI/EFUI
and change in conflict with father on change in depressive symptoms3
8.5 Discussion
8.5.1 Associations of social networking use intensity on psychosocial factors
In this chapter, we first described the potential influence of change in social
networking use intensity onto change in psychosocial factors over a nine-month
follow-up period, including friendship quality, perceived social support, self-esteem,
adolescent-parents conflict, and social non-confidence. First, it found that increased
202
social networking use intensity was positively associated with increased perceived
social support (both SFUI and EFUI), and friendship quality (SFUI only) during the
nine-month follow-up period. Similar results are reported in previous cross-sectional
studies among adolescents, reporting that more frequent Facebook use per day was
associated with higher perceived social support [243] and frequency of social
networking use was also associated with enhanced friendship quality [118]. The
findings imply that increase in intensity of social networking use increases
adolescents’ perceptions of increases in social support and sense of friendship quality.
Social networking use facilitates the interactions and social connections with each
other, and then fulfills adolescents’ needs for friendship maintenance during
adolescence. Engaging in diverse online social networking activities (e.g. wall posts,
instant messages, and comments) makes it easier for individuals both to broadcast
information about their activities/needs and to track the activities of their social
networking friends, which increases their feelings of closeness with friends and sense
of being supported. Another possible explanation is that social networking activities
(e.g. status updates, photo management, message posing, commenting on message
posted by their friends) encourage online self-disclosure (self-disclosure means
communicating personal information, thoughts, and feelings with others) [54, 270], and
the amount of online self-disclosure further enhances their sense of perceived social
support and friendship quality [16, 115].
Second, we also found that increase in SFUI was significantly associated with
increased adolescents’ conflict with the mother but not with conflict with the father
during the follow-up period. In general, using online social networking is a
time-demanding activity overall, and those who invest more time on social
203
networking use inevitably spend less time on shared activities with family
members [271]. Intensive social networking use and involvement in virtual
communities potentially isolates adolescents from their family, causing negative
impacts for adolescent-parents relationship. Adolescents might encounter more
conflict with their mothers as compared to that with their fathers. Previous studies
have revealed that social networking use among adolescents has been linked with
greater adolescent-parents conflict and less time spent with parents [58, 247]. We
speculate that this discrepancy is partially attributed to the differential roles of father
and mother in adolescents’ development. Mothers tend to be more likely involved in
parenting adolescents than fathers, and also interact more frequently with their
adolescents concerning personal issues (e.g. intensive social networking use, poor
time management) [272, 273], and this increases probability of conflicts with their
adolescents [274].
Third, this is the first study to estimate the influence of social networking use
intensity on the social non-confidence construct among adolescents. It is novel that
increased social networking use was significantly and positively associated with
increased social non-confidence during the follow-up period. It suggests that online
social networking activities decrease individuals’ confidence in social context. One
plausible explanation is that online social networking use reduces the opportunity of
face-to-face interaction in the real world. Adolescents with limited practice of
interactions with others in the social arena would appear to be nervous and may have
social anxiety during social interactions, manifesting as a lower level of social
confidence. Given the lack of previous data, this hypothesis needs to be tested in
future longitudinal studies among adolescents.
204
8.5.2 Mediation effects of change in social non-confidence and adolescent-mother
conflict
The mechanisms underlying the linkage between social networking use intensity
and depressive symptoms are not well understood. There is a dearth of studies to
explore the mediation effects of psychosocial factors on such associations. The
present study examined the mediation effects of change in adolescent-mother conflict
and social non-confidence on the relationship between change in social networking
use intensity and change in depressive symptoms among adolescents. The findings
revealed that the association between change in social networking use intensity and
change in depressive symptoms was partially mediated by change in social
non-confidence (both SFUI and EFUI) and change in adolescent-mother conflict
(EFUI only). Change in social non-confidence is a key mediator as compared to
change in adolescent-mother conflict. Proportions mediated by change in social
non-confidence were 20.97% for change in SFUI and 30.65% for change in EFUI
respectively.
The findings contend that the harmful impacts of increased social networking use
intensity on increased depressive symptoms are partially going through their
weakening effects on these two aspects of adolescents’ psychosocial status (i.e.
increased adolescent-mother conflict, increased social non-confidence), especially
through increased social non-confidence. It is important to strengthen social
confidence and improve adolescent-parents relationship (especially the
adolescent-mother relationship) to effectively prevent depressive symptoms among
adolescents with high intensity of social networking use.
205
8.5.3 Suppression effect of change in perceived social support and friendship
quality
It is interesting to find that change in perceived social support and friendship
quality negatively suppressed the influence of change in social networking use
intensity on change in depressive symptoms. This means that when controlling for
change in perceived social support and change in friendship quality, the magnitude of
the relationships between change in social networking use intensity and change in
depressive symptoms actually became larger. Harmful effects of increase in social
networking use intensity on depression are partially neutralized by simultaneous
improvement in perceived social support and friendship quality, as exhibited in the
first part of this chapter; such social improvements were protective against an increase
in depression. From the statistical perspective, it could be interpreted that inclusion of
change in perceived social support and friendship quality statistically removed the
outcome-irrelevant variance of change in social networking use intensity, and as a
result, the predictive validity of change in social networking use intensity on change
in depressive symptoms increased. Another interpretation is that increased perceived
social support and friendship quality partially compensate the effect of change in
social networking use intensity onto change in depressive symptoms. These
compensation effects cause researchers observing an underestimation of the harmful
effects of social networking use intensity on depressive symptoms than those that
actual existed, when not considering such opposite effects from improved perceived
social support and friendship quality. In addition, the findings implicate that harmful
effects of social networking use intensity on mental health outcomes (i.e. increased
depressive symptoms) could be suppressed by enhancing sense of perceived social
206
support and friendship quality. However, the suppression effects of change in
perceived social support and friendship quality are small. Future research is hence
needed to test such suppression effects further among adolescents.
8.5.4 Implications
The findings of this chapter have some practical implications for developing
intervention programs to improve mental health outcomes for adolescents with
intensive social networking use. Interventions that target at reducing the impact of
social networking use intensity on depressive symptoms should be effective by
improving psychosocial status (e.g. strengthening social confidence, improving
adolescent-parents relationships). Additionally, social skills training should be
involved in intervention programs for reducing depressive symptoms among
adolescents who use online social networking. Secondly, professionals and educators
should try to foster effective conflict solving strategies for adolescents to solve
conflicts with their mothers (e.g. conciliation, positive problem-copying
strategies) [275]. In addition, if possible, it would be helpful to involve parents, at least
mothers, in such training to reduce depressive symptoms of adolescents who use
social networking intensively. Thirdly, the suppression findings suggest the necessity
of measuring perceived social support and friendship quality in future research
estimating the relationship between social networking use and depression.
However, the mediation effects found in our study are relatively small. It
suggests that other potential mediators that are risk factors for depression may exist
but are not included in this Chapter. Future research is therefore warranted to explore
other potential important mediating variables that could further explain the
relationship between social networking use intensity and depressive symptoms among
207
adolescents. One set of such potentially mediating variables are Internet addictive
behaviors (i.e. ‘SNA’ and IA) As a large public concern among adolescents, intensive
social networking use would increase the tendency of Internet addictive behaviors,
and which would in turn increase the level of depressive symptoms. These hypotheses
are tested and the results are presented in Chapter 9.
8.5.5 Moderation effects
We also explored the moderation effects of these psychosocial factors on the
relationship between change in social networking use intensity and change in
depressive symptoms. A significant moderation effect on change in depressive
symptoms was found between change in social networking use intensity (both SFUI
and EFUI) and change in adolescent-father conflict. The findings reveal that the
deteriorating effect of change in social networking use intensity on change in
depression was significantly smaller among participants whose relationship with their
father largely improved during the follow-up period (as indicated by large decreases
in adolescent-father conflict) as compared to the ‘small change’ group, supporting a
buffering effect for ‘large improvement’ in adolescents’ relationship with their father.
It suggests that improvement in adolescents’ relationship with their father could
weaken the harmful effects of social networking use on mental health outcomes.
8.5.6 Limitations
There are some limitations in the present study. First, the online and offline
psychosocial relationships were not differentiated by scales in our study (e.g. online
friendship quality vs. offline friendship quality, perceived online social support vs.
perceived offline social support). There is a study in Hong Kong that reported that
208
perceived online social support could not buffer stress for adolescents [276]. Therefore,
our study cannot answer whether social networking use has different effects on online
and offline psychosocial relationships (e.g. whether social networking use mainly
increases users’ online friendship quality rather than offline friendship, or increases
both). Second, we only investigated the perceived social support. Other forms of
social support (e.g. receiving and giving social support through social networking) are
important but not involved in the present study. Social support is not unidimensional,
but a process of social exchange encompassing receiving and giving social support.
Perceived social support is one of the most effective indicators of social support in
health-related contexts. Third, although the mediation observed in the present study
explained some possible mechanisms on how social networking use affects depressive
symptoms via its negative impacts on psychosocial factors, the causal mediational
sequence cannot be derived by the restriction of the two-wave study design, and at
least three separate time points are required, one for each of the variables.
209
Table 8-2 Background variables associated with change in depressive symptoms in longitudinal
sample (n=4237)
△CES-D Univariate linear regression
Mean SD b SE p
Gender
Male 0.51 8.11 --
Female 0.45 9.00 -0.121 0.264 0.648
Grade
Seven 0.21 8.82 --
Eight 0.72 8.32 0.539 0.263 0.040
Father’s education level 0.880
Primary school or below -0.05 9.07 --
Junior middle school 0.54 8.66 0.529 0.569 0.352
High middle school 0.56 8.16 0.588 0.581 0.312
University or above 0.49 8.63 0.470 0.596 0.431
Don't know 0.14 9.49 0.273 0.837 0.744
Mother’s education level 0.053
Primary school or below -0.46 8.86 --
Junior middle school 0.81 8.47 1.237 0.465 0.008
High middle school 0.70 8.09 1.155 0.487 0.018
University or above 0.20 8.93 0.647 0.510 0.204
Don't know -0.05 9.56 0.514 0.773 0.506
Family financial situation 0.944
Very good/good 0.51 8.43 --
Average 0.44 8.49 0.023 0.270 0.933
Poor/very poor 0.67 11.66 0.274 0.810 0.734
Yes 0.57 8.46 --
No -0.19 9.29 -0.705 0.412 0.087
Academic performance 0.002
Upper 1.06 8.18 --
Medium 0.35 8.39 -0.681 0.297 0.022
Lower -0.22 9.49 -1.243 0.368 <0.001
Perceived study pressure <0.001
Nil/light 1.04 8.27 --
Average 0.68 8.15 -0.358 0.347 0.302
Heavy/very heavy -0.47 9.65 -1.464 0.404 <0.001
△CES-D: score in CES-D at follow-up minus its score at baseline.
SD: Standardized deviation.
b: Unstandardized regression coefficients from univariate multilevel linear regression model.
SE: Standard error.
--: Not applicable.
210
Table 8-3 Associations of change in SFUI and EFUI on change in depressive symptoms by
multilevel linear regression models (n=4237)
Univariate Adjusted a
b SE p β b SE p β
△SFUI 0.075 0.017 <0.001 0.069 0.073 0.017 <0.001 0.067
△EFUI 0.138 0.039 <0.001 0.055 0.139 0.039 <0.001 0.055
△SFUI: Change in Social function use intensity score; △EFUI: Change in Entertainment function
use intensity score.
a
Models were adjusted by grade, academic performance and perceived study pressure.
b: regression coefficients obtained by multilevel linear regression models.
β: Standardized regression coefficients.
SE: Standardized Error.
211
Table 8-4 Associations between change in SFUI, EFUI and change in psychosocial variables by multilevel linear regression (n=4237)
IV=△SFUI IV=△EFUI
a
DV Univariate Adjusted Univariate Adjusted a
b SE β b SE β b SE β b SE β
△Friendship quality
** **
0.017 0.006 0.136 0.018 0.006 0.141 0.018 0.014 0.061 0.018 0.015 0.060
△Perceived family support 0.031** 0.010 0.241 0.031** 0.010 0.248 0.147*** 0.024 0.499 0.149*** 0.024 0.505
△Perceived friend support 0.178*** 0.020 1.404 0.176*** 0.020 1.394 0.323*** 0.047 1.094 0.322*** 0.047 1.093
△Rosenberg self-esteem 0.011 0.008 0.088 0.012 0.008 0.096 0.017 0.018 0.057 0.020 0.018 0.067
△Adolescent-father conflict 0.023† 0.013 0.178 0.020 0.013 0.161 0.018 0.031 0.061 0.011 0.031 0.038
△Adolescent-mother conflict 0.043** 0.015 0.335 0.041** 0.015 0.320 0.063† 0.034 0.215 0.056 0.034 0.189
△Social non-confidence 0.044 ***
0.007 0.346 0.043 ***
0.007 0.340 0.121 ***
0.015 0.411 0.119 ***
0.015 0.404
IV: Independent variable; DV: Dependent variable.
SFUI: Social Function Use Intensity; EFUI: Entertainment Function Use Intensity.
△ indicates the change in interpersonal and psychosocial scale scores from baseline to follow-up.
a
†
p<0.10; * p<0.05; ** p<0.01; *** p<0.001
212
Table 8-5 Association of change in psychosocial scale scores on change in depressive symptoms
by multilevel linear regression (n=4237)
Univariate Adjusted a
b SE β b SE β
△Friendship quality -0.621*** 0.040 -0.073 -0.612*** 0.040 -0.071
△Perceived family support -0.293*** 0.024 -0.034 -0.292*** 0.024 -0.034
△Perceived friend support -0.112*** 0.013 -0.013 -0.115*** 0.013 -0.013
△Rosenberg self-esteem -0.685*** 0.031 -0.080 -0.682*** 0.031 -0.080
△Adolescent-father conflict 0.097*** 0.019 0.011 0.093*** 0.019 0.011
△Adolescent-mother conflict 0.097*** 0.017 0.011 0.094*** 0.017 0.011
△Social non-confidence 0.377*** 0.038 0.044 0.369*** 0.038 0.043
△ indicates the change of psychosocial scale scores from baseline to follow-up.
a
†
p<0.10; * p<0.05; ** p<0.01; *** p<0.001
213
Table 8-6 Testing the mediation effects of change in psychosocial scale scores on the association between △SFUI and △CES-D by multilevel linear
regression (n=4237)
Model 1 Model 2a Model 2b Model 3
△SFUI 0.073 ***
0.017 0.067 0.070 ***
0.017 0.064 0.058 ***
0.016 0.053 0.055 **
0.016 0.050
△Conflict with mother -- -- -- 0.091*** 0.017 0.011 -- -- -- 0.085*** 0.017 0.010

△Social non-confidence -- -- -- -- -- -- 0.356*** 0.038 0.042 0.349*** 0.038 0.041
Mediation effect size 5.11% 20.97% 25.33%

Sobel Z 2.434 5.137 --
p 0.015 <0.001 --
△ indicates the change scale scores using follow-up minus baseline.
All models were adjusted by grade, academic performance and perceived study pressure.
b regression coefficients obtained by multilevel linear regression models.
Mediation effect size: ratio of indirect to total effect.
†
p<0.10; * p<0.05; ** p<0.01; *** p<0.001
214
Table 8-7 Summary for the mediation effects of change in psychosocial scale scores on the
association between △EFUI and △CES-D by multilevel linear regression (n=4237)
Model 1 Model 2
b SE β b SE β
△EFUI 0.139*** 0.039 0.055 0.096*** 0.039 0.038
△Social non-confidence -- -- -- 0.358*** 0.038 0.042
Mediation effect size 30.65%

Sobel Z 6.068
p <0.001
Mediation effect size: ratio of indirect to total effect.
--: Not applicable.
†
p<0.10; * p<0.05; ** p<0.01; *** p<0.001
215
Table 8-8 Testing the suppression effects of change in psychosocial variables on the association between △SFUI and △CES-D (n=4237)
Model 1 Model 2a Model 2b Model 2c
△SFUI 0.073 ***
0.017 0.067 0.084 ***
0.016 0.078 0.083 ***
0.016 0.076 0.095 ***
0.017 0.088
△friendship quality -- -- -- -0.621*** 0.040 -0.072 -- -- -- -- -- --

△perceived family support -- -- -- -- -- -- -0.298*** 0.024 -0.035 -- -- --
△perceived friend support -- -- -- -- -- -- -- -- -- -0.124*** 0.013 -0.015
Suppression effect -0.011 -0.009 -0.022

Sobel Z -2.946 -3.008 -6.468
p 0.003 0.003 <0.001
†
p<0.10; * p<0.05; ** p<0.01; *** p<0.001
216
Table 8-9 Testing the suppression effects of change in psychosocial scale scores on the association between △EFUI and △CES-D by multilevel linear
regression (n=4237)
Model 1 Model 2a Model 2b
b SE β b SE β b SE β
△EFUI 0.139*** 0.039 -- 0.184*** 0.038 0.184 0.178*** 0.039 0.178
△perceived family support -- -- -- -0.303*** 0.024 -0.303 -- -- --

△perceived friend support -- -- -- -- -- -0.121*** 0.013 -0.121
Suppression effect -0.036 -0.039
Sobel Z -5.571 -5.518
p <0.001 <0.001
All models were adjusted by perceived family financial situation, academic performance, and perceived study pressure.
--: Not applicable.
†
p<0.10; * p<0.05; ** p<0.01; *** p<0.001
217
Table 8-10 Summary of interaction effects of change in psychosocial variables on the association
between △SFUI and △CES-D (n=4237)
Models without interaction Models with interaction
b SE p b SE p
1a 1b
△SFUI 0.088 0.016 <0.001 0.090 0.026 <0.001
Small change ref ref
Improvement -1.944 0.316 <0.001 -1.953 0.316 <0.001
Deterioration 2.781 0.305 <0.001 2,754 0.306 <0.001
Interaction term
△SFUI × Small change -- ref
△SFUI × Improvement 0.024 0.040 0.548
△SFUI × Deterioration -0.029 0.039 0.452
△-2LL (df=2) 1.6
P for △-2LL 0.449
2a 2b
△SFUI 0.077 0.016 <0.001 0.055 0.026 0.035
△Perceived family support
Improvement -1.797 0.319 <0.001 -1.788 0.319 <0.001
Deterioration 1.954 0.307 <0.001 1.974 0.307 <0.001
Interaction term
△SFUI × Deterioration 0.048 0.038 0.208
△-2LL (df=2) 1.5
P for △-2LL 0.472
3a 3b
△SFUI 0.091 0.017 <0.001 0.070 0.025 0.006
△Perceived friend support
Improvement -2.074 0.312 <0.001 -2.090 0.313 <0.001
Deterioration 1.180 0.318 <0.001 1.198 0.321 <0.001
Interaction term
△-2LL (df=2) 1.3
P for △-2LL 0.522
4a 4b
△SFUI 0.082 0.016 <0.001 0.066 0.026 0.011
△Rosenberg self-esteem
Improvement -3.512 0.291 <0.001 -3.503 0.291 <0.001
Deterioration 2.962 0.318 <0.001 2.982 0.318 <0.001
Interaction term
△-2LL (df=2) 1.3
P for △-2LL 0.522
218
(Cont’d) Table 8-10 Summary of interaction of change in psychosocial variables on the
association between △SFUI and △CES-D (n=4237)
Models without interaction Models with interaction
b SE p b SE p
5a 5b
△SFUI 0.070 0.017 <0.001 0.126 0.024 <0.001
△Adolescent-father conflict
Improvement -0.681 0.324 0.036 -0.756 0.325 0.020
Deterioration 1.213 0.330 <0.001 1.200 0.329 <0.001
Interaction term
△SFUI × Improvement -0.132 0.039 <0.001
△2LL (df=2) 11.7
P for △-2LL 0.003
6a 6b
△SFUI 0.070 0.017 <0.001 0.077 0.025 0.002
△Adolescent-mother conflict
Improvement -0.650 0.322 0.044 -0.680 0.323 0.035
Deterioration 1.352 0.318 <0.001 1.349 0.318 <0.001
Interaction term
△SFUI × Improvement -0.043 0.040 0.286
△-2LL (df=2) 1.9
P for △-2LL 0.387
7a 7b
△SFUI 0.063 0.016 <0.001 0.109 0.028 <0.001
△Social non-confidence
Improvement -1.561 0.309 <0.001 -1.574 0.310 <0.001
Deterioration 1.147 0.320 <0.001 1.159 0.320 <0.001
Interaction term
△SFUI × Improvement -0.053 0.039 0.182
△-2LL (df=2) 4.8
P for △-2LL 0.091
All models were adjusted by grade, academic performance, and perceived study pressure.
--: not applicable
219
Table 8-11 Summary of interaction effects of change in psychosocial variables on the association
between △EFUI and △CES-D (n=4237)
Models without interaction Model with interaction
b SE p b SE p
1a 1b
△EFUI 0.151 0.038 <0.001 0.119 0.058 0.041
Improvement -1.894 0.316 <0.001 -1.879 0.318 <0.001
Deterioration 2.741 0.306 <0.001 2.778 0.309 <0.001
Interaction term
△EFUI × Small change -- ref
△EFUI × Improvement 0.037 0.092 0.688
△EFUI × Deterioration 0.073 0.091 0.418
△-2LL(df=2) 0.6
P for △-2LL 0.741
2a 2b
△EFUI 0.164 0.038 <0.001 0.159 0.061 0.009
△Perceived family support
Improvement -1.819 0.319 <0.001 -1.826 0.321 <0.001
Deterioration 1.982 0.307 <0.001 2.015 0.311 <0.001
Interaction term
△EFUI × Improvement -0.046 0.095 0.628
△-2LL (df=2) 1.0
P for △-2LL 0.607
3a 3b
△EFUI 0.168 0.038 <0.001 0.167 0.060 0.006
△Perceived friend support
Improvement -2.026 0.312 <0.001 -2.028 0.313 <0.001
Deterioration 1.130 0.318 <0.001 1.159 0.325 <0.001
Interaction term
△-2LL (df=2) 0.4
P for △-2LL 0.819
4a 4b
△EFUI 0.160 0.037 <0.001 0.178 0.059 0.003
△Rosenberg self-esteem
Improvement -3.492 0.291 <0.001 -3.500 0.294 <0.001
Deterioration 2.981 0.318 <0.001 2.952 0.322 <0.001
Interaction term
△EFUI × Deterioration -0.054 0.093 0.556
△-2LL (df=2) 0.4
P for △-2LL 0.819
220
(Cont’d) Table 8-11 Summary of interaction effects of change in psychosocial variables on the
association between △EFUI and △CES-D (n=4237)
Models without interaction Model with interaction
b SE p b SE p
5a 5b
△EFUI 0.134 0.039 <0.001 0.243 0.055 <0.001
△Adolescent-father conflict
Improvement -0.671 0.324 0.039 -0.804 0.329 0.014
Deterioration 1.244 0.330 <0.001 1.176 0.332 <0.001
Interaction term
△-2LL (df=2) 7.9
P for △-2LL 0.019
6a 6b
△EFUI 0.135 0.039 <0.001 0.217 0.057 <0.001
△Adolescent-mother conflict
Improvement -0.667 0.323 0.039 -0.737 0.326 0.024
Deterioration 1.365 0.318 <0.001 1.296 0.320 <0.001
Interaction term
△-2LL (df=2) 3.7
P for △-2LL 0.157
7a 7b
△EFUI 0.108 0.039 0.005 0.073 0.065 0.259
△Social non-confidence
Improvement -1.577 0.309 <0.001 -1.478 0.315 <0.001
Deterioration 1.135 0.320 <0.001 1.149 0.321 <0.001
Interaction term
△EFUI × Improvement 0.132 0.091 0.148
△-2LL (df=2) 3.9
P for △-2LL 0.142
--: not applicable
221
Chapter 9 Relationship between social networking use intensity,
addiction, psychosocial factors and depression
9.1 Background
Chapter 7 and Chapter 8 found that social networking use intensity at baseline
and its change in scores during the follow-up period are positively associated with
depressive symptoms among adolescents. However, it is still unclear the extent to
which the detrimental effect of social networking use intensity on depressive
symptoms is due to impulsive and addictive social networking/Internet use (i.e. the
magnitude of mediation effects of ‘SNA’ in specific and IA in general on the pathway
between social networking use intensity and depressive symptoms). Such mediations
are likely to be in operation because of the significant associations between high
social networking use intensity and social networking/Internet addiction.
Some recent studies have focused on addiction to specific online activities, such
as gambling, shopping, or social networking [23, 277]. It highlights the significance of
distinguishing addiction to specific Internet activities (i.e. ‘SNA’) from IA in general.
Indeed, some online activities and behaviors have particularly strong addictive
potential, and could foster the development of addictive behaviors. Social networking
addiction (‘SNA’), which is considered as a subtype of IA [94], is a very recent concern
related with excessive social networking use (e.g. Facebook). It argues that the
capacity for online socializing carries elements that could be addictive [18, 278], such as
the characteristics of online social networking (e.g. anonymous, selective
self-disclosure and convenient communication without the limitations by geographic
distance and time), which could facilitate social interaction online and then function
222
as an attractive factor for addictive Internet use [18].
Some studies have investigated the associations between social networking use
and Internet addictive behaviors (i.e. ‘SNA’ or IA). For example, one study among
college students reported that time spent online and frequency of Facebook use were
significant predictors of IA [127], and a one-year longitudinal study still showed that
higher frequency of social networking use significantly predicted compulsive Internet
use [65]. Similarly, one survey among Turkish college students revealed that students
who frequently used Facebook for social interaction reported higher level of Facebook
addiction [48]. Considering that online social networking is an important medium for
social interaction among adolescents, intensive social networking use might be more
susceptible to ‘SNA’ compared to IA in general, i.e. the effect of social networking
use intensity on ‘SNA’ might be stronger than that on IA. However, there are no
studies that have investigated and compared such differentiating impacts of social
networking use intensity on ‘SNA’ and IA.
The positive association between IA and depressive symptoms has been widely
documented in both cross-sectional [201] and longitudinal [43] studies among
adolescents. ‘SNA’ has also been linked with depressive symptoms among adolescents,
and a significant positive association between severe depression and Facebook
addiction has been reported among college students in a previous study [48]. Although
IA and ‘SNA’ are separately documented to be positively associated with depressive
symptoms in previous studies, there is still no study that investigates and compares
their impacts on depressive symptoms among adolescents in the same study.
Given the aforementioned associations between social networking use and
Internet addictive behaviors (i.e. ‘SNA’ and IA), and the associations between Internet
223
addictive behaviors (i.e. ‘SNA’ and IA) and depressive symptoms among adolescents,
it is definitely possible that ‘SNA’ and IA would mediate the positive association
between social networking use intensity and depressive symptoms. Our literature
review found no such studies that have estimated and compared these mediation
effects. The present study attempts to fill this knowledge gap.
In the present study, to test mediation effects of change in ‘SNA’ and IA scores
on the association between change in social networking use intensity and change in
depressive symptoms, the hierarchical linear regression models were conducted
following the four-step method of Baron and Kenny’s strategy. As a supplement, path
model analysis was further applied to estimate the indirect effects of change in social
networking use intensity onto change in depressive symptoms via change in ‘SNA’
and IA scores. Path analysis can allow for the examination of a set of relationships
between two or more independent variables and one or more dependent variables [279,
280]
. The advantages of the path analysis model over the correlational and multivariate
regression analyses are that 1) it permits more than one path to predict the dependent
variable (i.e. change in depressive symptoms), and therefore can estimate the indirect
impact via other variables (i.e. ‘SNA’ and IA), and 2) the potential causal relationship
parameters can be estimated simultaneously in one single multistage regression
model [280, 281].
Moreover, some researchers argue that mediation processes may include multiple
mediators, and a multiple mediator model could likely provide a more accurate
assessment of mediation effects in the mediational research contexts [143]. Our Chapter
8 has provided the evidences that change in psychosocial factors (i.e. social
non-confidence, and adolescent-mother conflict) partially mediated the associations
224
between change in social networking use intensity and change in depressive
symptoms. Therefore, it is theoretically plausible that multiple mediational processes
might exist in the relationship between change in social networking use intensity and
change in depressive symptoms. Therefore, it is important to estimate the joint
mediation effects via multiple mediators (i.e. SNA, IA, and psychosocial factor) on
the associations between change in social networking use intensity and change in
9.2 Objectives
The first objective is to investigate the associations of changes in SFUI and EFUI
with change in ‘SNA’ (or change in IA) by several linear regression models after the
adjustment of background variables. It is hypothesized that change in SFUI and EFUI
would be positively associated with change in SNA and with change in IA.
The second objective is to investigate the associations of changes in ‘SNA’ and
IA with change in CES-D by multivariate linear regression models. It is hypothesized
that change in ‘SNA’ and IA would be positively associated with change in CES-D,
and moreover, the relative magnitude of the association between change in ‘SNA’ and
change in CES-D would be relatively stronger than that between change in IA and
change in CES-D.
The third objective is to estimate the extent to which the association between
change in SFUI (or change in EFUI) and change in CES-D is mediated by changes in
‘SNA’ and IA based on Baron and Kenny’s statistical method of fitting a set of
hierarchical linear regression models after adjusting for the significant background
factors. Besides, it further compares the magnitude of two individual mediation
225
effects in order to determine whether the mediation effect via change in ‘SNA’ is
significantly larger than that via change in IA.
Fourthly, as previously mentioned, we additionally applied path analysis to
investigate the indirect effects of change in SFUI and EFUI on change in CES-D via
two types of addiction (i.e. change in ‘SNA’ and IA score). A path model in Figure 9-1
was hence synthesized and tested.
+
ΔSFUI ΔSNA +
+
+
+
ΔCES-D
+ +
ΔEFUI ΔIA
Figure 9-1 The hypothesized path model for indirect effects between △SFUI, △EFUI
and △CES-D via △SNA and △IA. △SFUI: Change in Social Function Use Intensity;
△EFUI: Change in Entertainment Function Use Intensity; △SNA: Change in Social
Networking Addiction; △IA: Change in Internet Addiction; △CES-D: Change in
depressive symptoms measured by Center for Epidemiological Study
Scale-Depression. “+” indicates a positive association between two variables
The fifth objective is to estimate the total mediation effects of change in
addiction (‘SNA’ and IA) scores and the significant psychosocial mediators identified
in Chapter 8 altogether for the association between change in SFUI (or change in
EFUI) and change in depressive symptoms.
226
9.3.1 Linear regression analysis
The sample and statistical methods for linear regression analysis used in this
Chapter were the same as those mentioned in Chapter 8. The details have been
described in Chapter 8.
Contrasting two mediating effects in multiple mediator models
In order to quantify whether the mediation effect via △SNA was significantly
different from that via △IA for the association between ΔSFUI (or △EFUI) and
△CES-D in multiple-mediator models including ΔSNA and ΔIA as potential mediators.
The contrast was conducted to test such difference among pairs of specific mediation
effects (i.e. ΔSNA vs. ΔIA) based on the following formula [282, 283]:
a1b1 − a2b2
Z=
a1σ b21 + b1σ a21 + a2σ b22 + b2σ a22 − 2a1a2σ b1b2
This difference approximately follows a normal distribution, and the significance of
the difference with 95% CI was then established for hypotheses testing.
9.3.2 Path analysis for indirect effect model
The hypothesized path model in Figure 9-1 was fitted in our data. The
satisfactory “goodness-of-fit” for the hypothesized path model was indicated by the
Root Mean Square Error of Approximation (RMSEA) <0.08 and by the Comparative
Fit Index (CFI) >0.90 [284]. The path coefficients were estimated by the maximum
likelihood method by structure equation modeling method. The indirect effects and
227
95% confidence intervals (95% CIs) were estimated by the 1000 bootstrapping
method. Standardized path coefficients were presented in the path diagrams. The
standardized coefficients measure the relative strength and sign of the effect from an
independent variable to a dependent variable in the model, and can be used to
compare the relative magnitude of the effects of different independent variables in the
path model [285]. Path analyses were performed by AMOS 20.0 and a p<0.05 was
considered statistically significant.
9.4 Results
9.4.1 Associations between ΔSFUI/ΔEFUI and ΔIA/ΔSNA
After adjusting for the significant background factors, △SFUI (b=0.018,
SE=0.003, p<0.001) and △EFUI (b=0.054, SE=0.008, p<0.001) were all significantly
and positively associated with △IA at the 0.05 level (Adjusted models in Table 9-1).
Similar associations were also found between △SFUI and △SNA (b=0.156, SE=0.011,
p<0.001), and between △EFUI and △SNA (b=0.338, SE=0.338, p<0.001) (Adjusted
models in Table 9-1).
We also conducted multivariate linear regression models, which contained both
△SFUI and △EFUI as independent variables to predict △SNA (or △IA) after
adjustment of the significant background factors. It consistently showed that △SFUI
and △EFUI are independently and positively associated with △SNA and with △IA.
When inspecting their standardized regression coefficients within the same
multivariate model, the association between △EFUI and △IA (β=0.087, p<0.001) was
relatively stronger than that between △SFUI and △IA (β=0.044, p=0.011). Contrarily,
the association between △SFUI and △CES-D (β=0.147, p<0.001) was slightly
228
stronger than that between △EFUI and △CES-D (β=0.121, p<0.001) (Multivariate
models in Table 9-1).
9.4.2 Associations between ΔIA, ΔSNA and ΔCES-D
The adjusted models showed that △SNA was significantly associated with
△CES-D (b=0.284, SE=0.022, p<0.001, Table 9-2), and △IA was also significantly
associated with △CES-D (b=0.821, SE=0.077, p<0.001, Table 9-2).
The multivariate model, containing both △SNA and △IA as independent
variables, showed that △SNA and △IA are independently and positively associated
with △CES-D (Table 9-2). And the standardized regression coefficient between
△SNA and △CES-D (β=0.168, p<0.001) was slightly larger than that between △IA
and △CES-D (β=0.116, p<0.001).
9.4.3 Mediation effects via ΔIA and ΔSNA for the association between ΔSFUI and
ΔCES-D
The results in Chapter 8 showed that △SFUI was positively associated with
△CES-D (b=0.073, p<0.001) after adjustment for the significant background factors.
In addition, the above analyses in this chapter showed that △SFUI was also positively
associated with △IA (b=0.018, p<0.001) and with △SNA (b=0.156, p<0.001) (Table
9-3). The requirements for the mediation hypotheses of ΔIA and ΔSNA on the
association between △SFUI and △CES-D were hence fulfilled. The mediation
hypotheses were then tested by entering △IA and △SNA, both individually and
altogether, into the model that contained △SFUI as the independent variable.
When ΔIA was added into the model 1 in Table 9-3, the regression coefficient
229
between △SFUI and △CES-D was still significant, but diminished from 0.073 to
0.059 (Model 1a in Table 9-3). Given that △IA was also significantly associated with
△CES-D, △IA partially mediated the association between △SFUI and △CES-D
(Figure 9-2). The mediation effect via △IA was 0.014 (95% CI: 0.009~0.020, Sobel
Z=5.191, p<0.001). The proportion mediated via △IA was 19.65%.
△IA
0.018*** 0.797***

△SFUI △CES-D
change in SFUI and change in CES-D
Similarly, when adding △SNA into model 1 in Table 9-3, the association between
△SFUI and △CES-D became non-significant (Model 1b in Table 9-3). Given that
ΔSNA was still significantly associated with △CES-D, △SNA fully mediated the
association between △SFUI and △CES-D (Figure 9-3). The mediation effect was
0.043 (95%CI: 0.034~0.052, Sobel Z=9.397, p<0.001). The proportion mediated via
△SNA was notably high at 58.98%.
△SNA
0N156*** 0N276***
Direct effect: 0N0S0†

△SFUI △CES-D
Total effect: 0N07S***
230
When adding △IA and △SNA altogether into model 1 that contained △SFUI as
the independent variable (Model 1c in Table 9-3), the association between △SFUI and
△CES-D became non-significant. Given that △IA and △SNA were all positively
associated with △CES-D, this indicated that △IA and △SNA altogether mediated the
association between △SFUI and △CES-D (see Figure 9-4). The total mediation effect
via △IA and △SNA was 0.047 (95% CI: 0.038~0.056), which accounted for 64.38%
of total effect. The specific mediation effect via △SNA was 0.036 (95% CI:
0.027~0.045) with proportion mediated equaling to 49.31%; while the specific
mediation effect via △IA was 0.011 (95% CI: 0.006~0.015) with proportion mediated
equaling to 15.07%. The magnitude of the specific mediation effect via △SNA was
significantly larger than the specific mediation effect via △IA (Z=4.923, p<0.001,
95% CI for difference: 0.015~0.036).
△IA
0.018*** 0.586***
Direct effect: 0.026

△SFUI △CES-D
0.232***
0.156***
△3.A
Figure 9-4 Illustration for mediation effect of change in ‘SNA’ and IA altogether on
the association between change in SFUI and change in CES-D
9.4.4 Mediation effects via ΔIA and ΔSNA for the association between ΔEFUI and
ΔCES-D
The adjusted results in Chapter 8 showed that △EFUI was positively associated
231
with △CES-D (b=0.139, p<0.001), and the results in this chapter showed that △EFUI
is still positively associated with △IA (b=0.054, p<0.001) and with △SNA (b=0.338,
p<0.001). Therefore, the requirements for mediation analyses of ΔIA and ΔSNA on
the association between △EFUI and △CES-D were fulfilled. The mediation
hypotheses were then tested by entering △IA and △SNA, both individually and
altogether, into the model containing △EFUI as the independent variable.
When adding △IA into the model 2 in Table 9-4, the regression coefficient
between △EFUI and △CES-D was still significant, but diminished from 0.139 to
0.096 (Model 2a in Table 9-4). Given that △IA was significantly associated with
△CES-D, △IA partially mediated the association between △EFUI and △CES-D
(Figure 9-5). The mediation effect via △IA was 0.043 (95% CI: 0.028~0.058, Sobel
Z=5.660, p<0.001). The proportion mediated via △IA was 30.96%.
△IA
0.054*** 0.800***
Direct effect: 0.096*

△EFUI △CES-D
change in EFUI and change in CES-D
Similarly, when adding △SNA into model 2 in Table 9-4, the association between
△EFUI and △CES-D became non-significant (Model 2b in Table 9-4). Given that
△SNA was still significantly associated with △CES-D, △SNA fully mediated the
association between △EFUI and △CES-D (Figure 9-6). The mediation effect was
0.094 (95% CI: 0.073~0.115; Sobel Z=8.909, p<0.001), and the proportion mediated
232
via △SNA was 67.84%.
△SNA
0NSS8*** 0N279***
Direct effect: 0N044

△EFUI △CES-D
Total effect: 0N1S9***
between change in EFUI and change in CES-D
When adding △SNA and △IA altogether into model 2 that contained △EFUI as
the independent variable (Model 2c in Table 9-4), the association between △EFUI and
△CES-D became non-significant. Given that △IA and △SNA were still significantly
associated with △CES-D, this indicated that △IA and △SNA together mediated the
association between △EFUI and △CES-D (Figure 9-7). The total mediation effect via
ΔIA and ΔSNA was 0.111 (95%CI: 0.090~0.133; Z=10.005, p<0.001), which
accounted for 79.86% of total effect. The specific mediation effect via ΔSNA was
0.080 (95% CI: 0.060~0.099), with proportion mediated of 57.55%; while the specific
mediation effect via △IA was 0.032 (95% CI: 0.019~0.044), with proportion mediated
of 23.02%. The magnitude of the specific mediation effect via △SNA was
significantly larger than the specific mediation effect via △IA (Z=3.819, p<0.001,
95% CI for difference: 0.023~0.073).
233
△IA
0.054*** 0.587***

△EFUI △CES-D
0.338*** 0.236***
△3.A
Figure 9-7 Illustration for mediation effect of change in ‘SNA’ and IA altogether on
the association between change in EFUI and change in CES-D
9.4.5 Path analysis for indirect effect model via ΔIA and ΔSNA
The path model in Figure 9-8 showed that △SFUI and △EFUI were positively
associated with △SNA (β=0.148 for △SFUI, p<0.001; β=0.123 for △EFUI, p<0.001),
and △EFUI was positively associated with △IA (β=0.056, p<0.001). In turn, △SNA
was positively associated with △IA (β=0.268, p<0.001), while △SNA and △IA were
positively associated with ΔCES-D (β=0.172 for ΔSNA, p<0.001; β=0.120 for △IA,
p<0.001) (see Figure 9-8).
The path model suggested several indirect paths. The indirect effect of △SFUI
onto △CES-D followed two paths: (1) through △SNA to △CES-D, the indirect effect
was 0.026; and (2) through △SNA to △IA, and then to △CES-D, the indirect effect
was 0.005. The total indirect effect of △SFUI on △CES-D via two paths was 0.31
(95% CI: 0.021~0.042, p=0.002)
Regarding the indirect effect of △EFUI on △CES-D, there were three paths: (1)
through △SNA to △CES-D, the indirect effect of this path was 0.021; (2) through
△SNA to △IA, and then to △CES-D, the indirect effect of this path was 0.004; and (3)
234
through △IA to △CES-D, the indirect effect of this path was 0.007. The total indirect
effect of △EFUI on △CES-D via the three paths was 0.32 (95% CI: 0.22~0.43,
p=0.002).
ΔSFUI 0.148*** ΔSNA

0.123*** 0.172***
0.474*** 0.268*** ΔCES-D
***
0.006 0.120
ΔEFUI 0.056*** ΔIA
Figure 9-8 Path model for indirect effects between △SFUI, △EFUI and ΔCES-D via
△SNA and △IA. Paths shown in dashed lines represent non-significant pathway at the
0.05 level. Standardized path coefficients (β) were presented in the model. Model fit
indexes: χ2 (2) =2.595 with p=0.273, CFI=1.000, IFI=1.000 and RMSEA=0.008.
†
p<0.10, *p<0.05, **p<0.01, ***p<0.001
9.4.6 Combined mediation effects of change in addiction and psychosocial factors
altogether on the associations between ΔSFUI (or ΔEFUI) and ΔCES-D
9.4.6.1 Summary of significant mediators in single mediator models
A summary for the results of all significant mediators from single mediator
models in Chapter 8 (i.e. psychosocial factors) and Chapter 9 (i.e. △SNA and △IA)
was presented in Table 9-5. Change in social non-confidence and change in
adolescent-mother conflict partially mediated the association between change in SFUI
and change in CES-D, with proportion mediated of 20.97% (change in social
non-confidence) and 5.11% (change in adolescent-mother conflict) respectively.
Besides, the association between change in SFUI and change in CES-D was also
mediated by change in SNA and IA, with proportion mediated of 58.98% (change in
235
SNA) and 19.65% (change in IA) respectively.
For the relationship between change in EFUI and change in CES-D, the results in
Chapter 8 showed that this association was partially mediated by change in social
non-confidence, and the proportion mediated was 30.65%. Meanwhile, change in
SNA and IA also mediated this association, with proportion mediated of 67.84%
(change in SNA) and 30.65% (change in IA) respectively.
Given the above significant mediators in single mediator models, the multiple
mediator models were further constructed to estimate the joint mediation effects by
involving two types of addictions (i.e. △SNA and △IA) and the significant
psychosocial factors as potential mediators. The statistical analyses were separately
conducted for △SFUI and △EFUI.
9.4.6.2 Combined mediation for the association between ΔSFUI and ΔCES-D
Given that △IA, △SNA, △social non-confidence, and △adolescent-mother
conflict were individual significant mediators for the association between △SFUI and
△CES-D (Table 9-5), the multiple mediator model was hence structured by involving
these four variables as potential mediators and △SFUI as the independent variable.
The results of the multiple mediator model were shown in Table 9-6 (Model 3a). The
sizes of the total, direct, and mediation effects were summarized in Table 9-7. Figure
9-9 illustrated the multiple mediator model. All four specific mediation effects were
statistically significant. The estimate of total mediation effect via these four mediators
was 0.057 (Z=11.177, p<0.001), and the proportion of the total mediation effect was
notably high at 78.77% (Table 9-7). The proportion mediated was highest for △SNA
(44.45% of total mediation effect), followed by △social non-confidence (17.02%),
236
ΔIA (13.59%), and △adolescent-mother conflict (3.71%).
△3.A
0.156*** 0.208***
0.018*** △IA 0.551***

△SFUI △CES-D
0.043*** 0.289***
△3ocial non-
confidence
0.041*** 0.066***
△Adolescent-
mother conflict
Figure 9-9 Illustration of combined mediation effect via change in IA, SNA, social
non-confidence and adolescent-mother conflict on the association between change in
SFUI and change in CES-D
9.4.6.3 Combined mediation for the association between ΔEFUI and ΔCES-D
For the association between △EFUI and △CES-D, our findings showed that △IA,
△SNA and △social non-confidence were the individual significant mediators (Table
9-5). These three potential mediators were involved in the multiple mediator model.
The results of multiple mediator model are shown in Table 9-6 (Model 3b). The size
of the total, direct, and mediation effects are summarized in Table 9-7. Figure 9-10
illustrates the multiple mediator model via all three mediators for the association
between △EFUI and △CES-D. All three specific mediation effects were statistically
significant. The total mediation effect was 0.139 (Z=11.541, p<0.001, Table 9-7),
which completely accounted for the total effect (proportion mediated=100%). △SNA
237
again had the highest proportion mediated (52.52% of total mediation effect),
followed by △social non-confidence (25.18%), and △IA (22.30%).
△3.A
0.338*** 0.216***
0.054*** △IA 0.571***
Direct effect: -0.004

△EFUI △CES-D
0.119*** 0.296***
△3ocial non-
confidence
Figure 9-10 Illustration of joint mediation effect via change in IA, SNA and social
non-confidence on the association between change in SFUI and change in CES-D
9.4.7 Summary for mediation effects of all mediation models
Table 9-8 summarized the proportion of mediation effects (% mediated) of all
mediation models for the relationships between change in SFUI (or change in EFUI)
and change in depressive symptoms in Chapter 8 and Chapter 9. Change in SNA,
change in IA, change in social non-confidence, and change in adolescent-mother
conflict were individually significant mediators for the association between change in
SFUI and change in depressive symptoms. The proportions mediated were 58.98%,
19.65%, 20.97% and 5.11% respectively. In the multiple mediator model, these four
mediators altogether mediated 78.77% of the relationship between change in SFUI
and change in depressive symptoms (Table 9-8).
Change in SNA, change in IA and change in social non-confidence were
238
individually significant mediators for the association between change in EFUI and
change in depressive symptoms, and the proportions mediated were 67.84%, 30.96%
and 30.65% respectively. These three mediators altogether completely mediated the
association between change in EFUI and change in depressive symptoms (Table 9-8).
9.5 Discussion
9.5.1 Associations between change in social networking use intensity and change
in Internet addictive behaviors
In the present study, it was found that increased social networking use intensity
(both SFUI and EFUI) are significant risk factors of increased ‘SNA’ and IA during
the follow-up period. The positive association between change in social networking
use intensity and change IA score found in the present study is consistent with a
previous two-wave longitudinal study among adolescents from Van Rooij AJ and
Schoenmakers TM [65], which reported that change in frequency of online social
networking use was positively associated with change in IA score. However, that
study measured social networking use intensity only by one item (i.e. frequency per
week), while our study assessed such use by a multi-item validated scale (Chapter 6).
From this perspective, the results in our study are more informative than that from
Van Rooij AJ and Schoenmakers TM [65]. The findings in the present study are also
similar with the results of previous cross-sectional studies that reported that frequency
and hours per week spent on social networking use were positively associated with
SNA [48] and IA [127, 277].
Moreover, results from multivariate regression model showed that change in
SFUI (β=0.147) was slightly more predictive of ‘SNA’ than change in EFUI (β=0.121).
239
This is understandable as adolescents mainly use social networking services for social
purpose and interactions (e.g. communicating with friends, maintaining relationships,
self-presentation). The reverse magnitude was observed for prediction of IA by
change in SFUI and change in EFUI, showing that change in EFUI (β=0.087) is more
predictive of IA than change in SFUI (β=0.044). This result is similar with reports of a
previous study [65], finding that online entertainment (i.e. online gaming) has strongest
association with IA as compared to other Internet applications (i.e. chatting, blogging,
MSN use). It indicates that entertainment function use plays a significant role in the
development of IA. One possible explanation is that IA in general as a compulsive
behavior usually requires more time spent online. Entertainment function use
generally requires a large amount of commitment and time investments in order to
achieve imminent goals which may then contribute to the development of maladaptive
behaviors and addiction [277]. Besides, excessive of online gaming, information
seeking and video watching/downloading have great impact on IA in general[24].
These activities (e.g. browsing information, playing games watching video) can be
conducted on social networking platforms, and also the key elements of entertainment
function use intensity subscale that was used in the present study.
These findings highlight the importance and necessity of early intervention for
reducing social networking use intensity in order to prevent the condition progressing
to addictive status, as intensive social networking use renders adolescents vulnerable
for addiction (both ‘SNA’ and IA). Both social function and entertainment function
use could cause Internet addictive behaviors, and therefore need to be properly
controlled simultaneously for adolescents in future intervention programs.
240
9.5.2 Associations between change in ‘SNA’, change in IA and change in
depressive symptoms
The hypothesis that change in ‘SNA’ and change in IA would be positively
associated with change in depressive symptoms was supported by the multivariate
regression model. These positive findings are in accordance with those findings in the
Chapter 5 with reporting that change in ‘SNA’/IA status (categorical variables,
persistent ‘SNA’/IA and incident ‘SNA’/IA) significantly predicted incidence of
depression by modeling approach II, and are also similar with those reports of
previous studies [43, 105, 197, 286].
The results further revealed that the association between change in ‘SNA’ and
change in depressive symptoms (β=0.168) is slightly stronger than that between
change in IA and change in depressive symptoms (β=0.116) in the same model. These
findings indicate that ‘SNA’ in specific and IA in general are both significant risk
factors of depressive symptoms among adolescents. However, the ‘SNA’ in specific is
more predictive as compared to IA in general for depressive symptoms. It might be
that excessive involvement in specific online activities (i.e. social networking
activities) is more devastating for adolescents’ mental health rather than IA in general.
Therefore, it is more important to tailor and prevent ‘SNA’. It is also suggestive for
future research that measure specific Internet addiction (e.g. ‘SNA’) might be more
appropriate and informative as risk factors of depression when comparing to IA in
general. Intervention programs aiming to reduce the level of depressive symptoms
should make effort to prevent both types of Internet addiction (i.e. ‘SNA’ and IA), but
more attention should be paid to ‘SNA’ in specific.
241
9.5.3 Mediation effects for the association between change in social networking
use intensity and change in depressive symptoms via change in ‘SNA’ and change
in IA
The present chapter further estimated the mediation effects of change in
addictive scores (both ‘SNA’ and IA) on the associations between change in social
networking use intensity and change in depressive symptoms. Change in ‘SNA’ fully
mediated the association between change in SFUI and change in depressive symptoms
(proportion mediated=58.98%), and it also fully mediated the association between
change in EFUI and change in depressive symptoms (proportion mediated=67.84%).
Whereas change in IA only partially mediated these two associations (proportion
mediated=19.65% for change in SFUI and 30.96% for change in EFUI). Moreover,
multiple mediator models revealed that change in ‘SNA’ and IA altogether mediated
64.38% of the effect of change in SFUI on change in depressive symptoms, while
they altogether mediated 79.86% of the effect of change in EFUI on change in
These findings provide evidences that increased social networking use might
make adolescents at high-risk of suffering from ‘SNA’ and IA, which might in turn be
detrimental for adolescents’ depressive symptoms. However, ‘SNA’ in specific is a
more important mediator given that the higher proportion of mediation effect it
carried as compared to IA in general. This discrepancy may be accounted for by the
fact that measurement for social networking use intensity in the present study more
exclusively related to aspects of ‘SNA’ in specific as compared to IA in general. The
mediation findings of change in ‘SNA’ and IA on the associations between change in
social networking use intensity and change in depressive symptoms indicate that
242
interventions for preventing/reducing SNA and IA may alleviate the impacts of social
networking use intensity on depressive symptoms. Prevention on ‘SNA’ and IA are
hence warranted in future research.
9.5.4 Path model for indirect effects via ‘SNA’ and IA
The hypothesized path model is well supported by our data. It gives new insights
into relationships of social networking use intensity and associated addictive
behaviors on depressive symptoms, and presents more informative pathways for such
associations. First, the indirect effect of change in SFUI on change in depressive
symptoms was through two paths: via change in SNA, and via change in SNA to
change in IA. The indirect effect of change in EFUI on change depressive symptoms
was through three paths. The first two paths were the same as those between ΔSFUI
and ΔCES-D. The third path was the indirect effect justly via change in IA, which is
specified for the association between ΔEFUI and ΔCES-D.
There are some implications for these findings. First, ‘SNA’ in specific plays an
important role on the potential causal chain between social networking use intensity
and IA in general, as four of the five paths initially through ‘SNA’ (either via or not
via IA to depression), but only one path directly through IA to depression. Second,
‘SNA’ in specific might temporally precede IA in general. One explanation is that
adolescents might be attracted to specific online activities firstly (i.e. social
networking), and the potential gratification received from such use reinforces the
attachment to social networking and progresses to ‘SNA’. Addictive social networking
use then amplifies the severity of IA. Third, ‘SNA’ and IA as intervening variables
between social networking use intensity and depressive symptoms have significant
devastating effects on depressive symptoms. Future intervention programs should
243
make more efforts to block the potential causal chain from social networking use to
depression among adolescents by reducing the addictive tendency. However, more
effort should be given to ‘SNA’ prevention.
9.5.5 Combined mediation effects via Internet addictive behaviors and
psychosocial factors altogether
Results of the multiple mediator models showed that the combination of change
in addiction (i.e. ‘SNA’ and IA) and significant psychosocial mediators altogether
explained the mediating mechanisms very well for the relationship between change in
social networking use intensity and change in depressive symptoms. More than 78%
of the effect from change in SFUI to change in depressive symptoms is attributable to
the joint effects on change in addiction (i.e. ‘SNA’ and IA) and change in psychosocial
factors (i.e. increased social non-confidence and adolescent-mother conflict); while
the impact of change in EFUI on change in depressive symptoms were completely
through its joint impacts on change in addiction (‘SNA’ and IA) and change in social
non-confidence.
These findings provide strong evidence for understanding the mechanisms on
how and why social networking use intensity could increase adolescents’ level of
depressive symptoms, and gives robust informative evidences for future intervention
design. First, the results firstly evidence-based confirm that there are multiple indirect
pathways that carry the detrimental effect of social networking use onto depressive
symptoms. On one hand, higher levels of social networking use could increase
adolescents’ susceptibility to addiction (i.e. both ‘SNA’ and IA, but especially for
‘SNA’ in specific); on the other hand, higher levels of social networking use could
devastate adolescents’ psychosocial status (i.e. weakening social confidence in social
244
contexts, disharmonious adolescent-parents relationship). The pathological social
networking/Internet use and deteriorative psychosocial relationships further amplify
the level of depressive symptoms among adolescents.
Second, future intervention programs should more specifically emphasize these
intervening components in order to effectively reduce the detrimental effect of social
networking use on depressive symptoms among adolescents. For instance, on the one
hand, intervention campaigns should improve psychosocial status of adolescents who
are social networking users (i.e. strength social confidence, improve family
relationship), and on the other hand, it is also important to simultaneously encourage
adolescents lessened involvement in online social networking to prevent addictive
tendency. These intervention strategies exclusively targeting significant intervening
variables could theoretically and effectively block the negative influence of social
networking use on adolescents’ mental health.
9.5.6 Limitations
Some limitations in this Chapter need to be considered with caution. First, only
one type of specific IA (i.e. ‘SNA’) was involved. Internet addictive behaviors,
however, cover a broad-spectrum of maladaptive behaviors (e.g. cybersex, chatting,
games and gambling). These online activities also have the potential for excessive use
and may lead to symptoms similar to addiction. Future studies are warranted to
include other specific forms of Internet addiction (e.g. online gaming addiction, online
gambling addiction or other social media addiction). Second, our multiple mediator
models specified that social networking use intensity is related to multiple potential
mediators that are altogether related to the outcome, but their interrelationships
between potential mediators are not considered due to complexity and the limitations
245
by time-points observed in our study. There may have been other potential multiple
mediator models, such as social networking use increases ‘SNA’/IA, increased
‘SNA’/IA further decreases social confidence, and ultimately increases levels of
depressive symptoms, or contrastingly, social networking use decreases social
confidence, and socially non-confident adolescents further prefer interactions through
social networking/Internet rather than offline interactions, which ultimately increases
levels of depressive symptoms, etc. Future longitudinal studies with three or more
observations are highly warranted to clarify these complex interrelationships of
potential mediators on the relationship between social networking use intensity and
depressive symptoms among adolescents.
9.5.7 Conclusion
In summary, the findings of this chapter reveal that change in social function use
intensity and entertainment function use intensity were significantly associated with
change in IA and with ‘SNA’ in the multivariate analyses. change in ‘SNA’ and IA
were significant mediators for the associations between change in social networking
use intensity and change in depressive symptoms, and change in ‘SNA’ in specific
was a more important mediator as compared to change in IA in general. Moreover, the
underlying mechanisms for the relationship between change in social networking use
intensity and change in depressive symptoms were well explored by the combination
of change in addiction (‘SNA’ and IA) and significant psychosocial factors (i.e. social
non-confidence, adolescent-mother conflict), which explained the large proportion of
the effects of change in social networking use intensity on change in depressive
symptoms. These findings provide insights for developing interventions for reducing
levels of depressive symptoms. Future interventions should effort to make early
246
interventions for reducing social networking use intensity, and simultaneously
improve adolescents’ psychosocial status in order to prevent their detrimental effects
on mental health among adolescents.
247
Table 9-1 Associations of ΔSFUI and ΔEFUI on ΔIA/ΔSNA among longitudinal sample (n=4237)
DV=ΔIA DV=ΔSNA
b SE β b SE β
Univariate models
ΔSFUI 0.018*** 0.003 0.086 0.156*** 0.011 0.205
ΔEFUI 0.054 ***
0.008 0.108 0.338 ***
0.027 0.191
Adjusted models
ΔSFUI 0.018*** 0.003 0.085 0.156*** 0.011 0.204
ΔEFUI 0.054*** 0.008 0.108 0.338*** 0.027 0.191
Multivariate models
ΔSFUI 0.009* 0.004 0.044 0.112*** 0.012 0.147
ΔEFUI 0.043 ***
0.009 0.087 0.214*** 0.030 0.121
△ indicates change scale scores by subtracting baseline score from corresponding follow-up score.
SFUI: Social Function Use Intensity; EFUI: Entertainment Function Use Intensity; SNA: Social
Networking Addiction; IA: Internet Addiction.
Adjusted models: adjusted by grade, academic performance, and perceived study pressure.
Multivariate model: ΔSFUI and ΔEFUI were entered together in the model after adjustment of
background variables.
b: regression coefficients.
SE: Standardized Error of the regression coefficients.
*
p<0.05; *** p<0.001
248
Table 9-2 Associations between change in ‘SNA’, change in IA and change in CES-D among
Univariate Adjusted Multivariate model
b SE β b SE β b SE β
ΔSNA 0.290*** 0.022 0.203 0.284*** 0.022 0.199 0.239*** 0.022 0.168
ΔIA 0.841*** 0.077 0.166 0.821*** 0.077 0.162 0.590*** 0.079 0.116
SNA: Social Networking Addiction; IA: Internet Addiction; CES-D: Center for Epidemiological
Adjusted models: adjusted by grade, academic performance, and perceived study pressure.
Multivariate model: ΔSNA and ΔIA were both entered in the model after adjustment of
background variables.
SE: Standardized Error of the regression coefficient.
†
p<0.10; * p<0.05; ** p<0.01; *** p<0.001
249
Table 9-3 Hierarchical regression models for testing mediation effects of ΔIA and ΔSNA on the association between ΔSFUI and ΔCES-D among longitudinal sample
(n=4237)
ΔSFUI 0.073 ***
0.017 0.067 0.059 ***
0.016 0.054 0.030 †
0.017 0.028 0.026 0.017 0.024
ΔIA -- -- -- 0.797*** 0.077 0.157 -- -- 0.586*** 0.079 0.115
ΔSNA -- -- -- -- -- -- 0.276 ***
0.022 0.194 0.232 ***
0.023 0.163
Sobel Z -- 5.191 9.397 9.997
p -- <0.001 <0.001 <0.001
△ indicates the change in scale scores using follow-up minus baseline.
SFUI: Social Function Use Intensity, IA: Internet Addiction, SNA: Social Networking Addiction; CES-D: Center for Epidemiological Studies-Depression Scale.
--: Not applicable
†
p<0.10; * p<0.05; ** p<0.01; *** p<0.001
250
Table 9-4 Hierarchical regression models for testing mediation effects of ΔIA and ΔSNA on the association between ΔEFUI and ΔCES-D among longitudinal sample
(n=4237)
ΔEFUI 0.139 ***
0.039 0.055 0.096 *
0.038 0.038 0.044 0.039 0.018 0.027 0.039 0.011
ΔIA -- -- -- 0.800*** 0.077 0.157 -- -- -- 0.587*** 0.079 0.115
ΔSNA -- -- -- -- -- -- 0.279 ***
0.022 0.196 0.236 ***
0.023 0.166
Sobel Z -- 5.660 8.909 10.005
p -- <0.001 <0.001 <0.001
EFUI: Entertainment Function Use Intensity, IA: Internet Addiction, SNA: Social Networking Addiction; CES-D: Center for Epidemiological Studies-Depression
Scale.
--: Not applicable
*
p<0.05; ** p<0.01; *** p<0.001
251
Table 9-5 Summary for all significant mediators in single mediator models for the association
between change in SFUI (or change in EFUI) and change in CES-D
Total Mediation 95%CI for
Z % Med.$
effect effect mediation effect
ΔSFUI → ΔCES-D 0.073***
Addiction variables
Via △IA 0.014*** 0.009, 0.020 5.191 19.65%
Via △SNA 0.043*** 0.034, 0.052 9.397 58.98%
Via △social non-confidence 0.015*** 0.009, 0.021 5.137 20.97%
Via △adolescent-mother conflict 0.004* 0.001, 0.007 2.434 5.11%
ΔEFUI → ΔCES-D 0.139***

Addiction variables
Via △IA 0.043*** 0.028, 0.058 5.660 30.96%
Via △SNA 0.094*** 0.073, 0.115 8.909 67.84%
Psychosocial factor
Via △social non-confidence 0.043*** 0.029, 0.056 6.068 30.65%
SFUI: Social Function Use Intensity; EFUI: Entertainment Function Use Intensity; IA: Internet
Addiction, SNA: Social Networking Addiction; CES-D: Center for Epidemiological
Studies-Depression Scale
$
: Proportion mediated, calculated by dividing absolute value of mediation effect by absolute value
of total effect.
*
p<0.05, *** p<0.001
252
Table 9-6 Multivariate linear regression models for testing the joint mediation effects of addiction
and psychosocial factors on the associations between ΔSFUI/ΔEFUI and ΔCES-D among
Model 3a Model 3b
b SE β b SE β
Social networking use
△SFUI 0.016 0.016 0.014 -- -- --
△EFUI -- -- -- -0.004 0.038 -0.0002
Addictive variables
△IA 0.551*** 0.078 0.108 0.571*** 0.079 0.112
△SNA 0.208*** 0.023 0.146 0.216*** 0.023 0.152
△Social non-confidence 0.289*** 0.037 0.034 0.296*** 0.038 0.035
△Adolescent-mother conflict 0.066 ***
0.017 0.008 -- -- --
EFUI: Entertainment Function Use Intensity, IA: Internet Addiction, SNA: Social Networking
Addiction; CES-D: Center for Epidemiological Studies-Depression Scale.
--: Not applicable
***
p<0.001
253
Table 9-7 Summary of total, direct, and combined mediation effects of multiple mediator models
on change in depressive symptoms (n=4237)
95%CI for %
Estimates Z
estimates Med.$
ΔSFUI → ΔCES-D
Total effect 0.073*** 0.041, 0.106 -- --
Direct effect 0.016 -0.017, 0.048 -- --
Total mediated effect 0.057*** 0.047, 0.068 11.177 78.77

△IA 0.010*** 0.006, 0.014 4.573 13.59
△SNA 0.032*** 0.024, 0.041 7.625 44.45
△Social non-confidence 0.012*** 0.007, 0.017 4.829 17.02
△Adolescent-mother conflict 0.003* 0.0003, 0.005 2.235 3.71
ΔSFUI → ΔCES-D
Total effect 0.139*** 0.063, 0.215 3.590 --
Direct effect -0.004 -0.076, 0.075 -0.01 --
Total mediated effect 0.139*** 0.115, 0.163 11.541 100.00

△IA 0.031*** 0.019, 0.043 4.933 22.30
△SNA
***
0.073 0.054, 0.092 7.512 52.52
△Social non-confidence 0.035*** 0.023, 0.047 5.786 25.18
Addiction, SNA: Social Networking Addiction; CES-D: Center for Epidemiological
$
: Proportion mediated, calculated by dividing absolute values of mediation effect (specific or
total) by absolute value of total effect.
--: Not applicable
*
p<0.05; ** p<0.01; *** p<0.001
254
Table 9-8 Summary for mediation effects (% mediated) of all mediation models in Chapter 8 and
Chapter 9
Mediation models % Mediated $
ΔSFUI → ΔCES-D
Addiction variables
Via △IA 19.65
Via △SNA 58.98
Via △IA + △SNA 64.38
Via △social non-confidence 20.97
Via △adolescent-mother conflict 5.11
Via △social non-confidence + △adolescent-mother conflict 25.33
Combined addiction and psychosocial variables

Via △IA + △SNA + △social non-confidence + △adolescent-mother conflict 78.77
ΔEFUI → ΔCES-D
Addiction variables
Via △IA 30.96
Via △SNA 67.84
Via △IA + △SNA 79.86
Psychosocial factor
Via △social non-confidence 30.65
Combined addiction and psychosocial variables

Via △IA + △SNA + △social non-confidence 100.00
Addiction; SNA: Social Networking Addiction; CES-D: Center for Epidemiological
$
: Proportion mediated, calculated by dividing absolute value of mediation effect by absolute value
of total effect.
255
Chapter 10 Conclusions
10.1 Summary and interpretation of key findings
10.1.1 Depression and Internet addictive behaviors epidemic among adolescents
The high depression prevalence and incidence were observed among adolescents
in our study. The depression prevalence was respectively 24.6% at baseline and 26.7%
at follow-up, showing a significant increase over the nine-month interval. Moreover,
the depression incidence was highly 23.4 per 100-person years. These figures signify
that depression epidemic is serious and likely to worsen over time. Effective
prevention and interventions aiming to reduce depression is highly warranted among
adolescents in China. Mental healthcare prevention services may be needed to
promote adolescents’ healthy development through routine school health education.
Internet addiction (IA) in general is a real threat for adolescents’ healthy
development that has drawn extensive concern among public health researchers in the
last two decades. In this study, around 6.6% of participants were classified with IA at
baseline, and the IA prevalence mildly increased to 7.3% at follow-up. During the
nine-month period, the IA incidence was approximately 6.77 per 100-person-years. In
particular, online social networking use as a new Internet-related phenomenon has
exponentially increased among adolescents in recent years. Adolescents have become
important subpopulations that are disproportionately influenced by online social
networking use. Along with excessive social networking use, a high prevalence of
‘SNA’ was found both at baseline (13.7%) and follow-up (13.6%). The incidence of
‘SNA’ was 12.80 per 100-person-years during the nine-month interval. The findings
256
determined that Internet addictive conditions are continuously worsening and are
serious problems among this population. Similar to IA in general, ‘SNA’ as another
newly raised potential public health issue should be given significant attention in the
future research.
10.1.2 Causal directions between Internet addictive behaviors and depression
This study suggests that the longitudinal relationships between Internet addictive
behaviors (both ‘SNA’ and IA) and depression are likely to be bi-directional and
reciprocal. That is, adolescents with depression have higher risk of developing
Internet addictive behaviors, and in turn, adolescents who suffer from symptoms of
Internet addictive behaviors further aggravate their level of depressive symptoms.
These findings suggest a vicious dysfunctional circle for the relationships between
Internet addictive behaviors and depression among adolescents. In such a condition,
interventions focusing on Internet addictive behaviors without paying attention to
psychiatric disorders (i.e. depression), and vice versa, might result in inefficient or
even failure of interventions. Future intervention programs should therefore make
strong efforts to reduce both depressive disorder and Internet addictive behaviors
together. Co-treatment/prevention strategy may be appropriate for optimal effects. For
adolescents suffering from depression, alleviating emotional difficulty may be a more
important mechanism underlying Internet addictive behaviors development. And
contrastingly, for adolescents with Internet addictive behaviors, being isolated from
real-world activities and interactions caused by excessive Internet/social networking
involvement may be boosting depression development. Furthermore, adolescents
might be comorbidity of depression and Internet addictive behaviors. These
mechanisms need to be considered in future intervention studies aiming to prevent
257
depression and Internet addictive behaviors. In this sense, it could be useful to work
with different types of cognitive distortions and dysfunctional behaviors that could be
maintaining and exacerbating these problems.
10.1.3 A validated scale for assessing social networking use intensity
Social networking use is a relatively new phenomenon, but speedily attracts high
popularity among adolescents. There is, however, a scarcity of formally validated
instruments to measure social networking use intensity, which partly hinders the
research to explore the health impacts of social networking use for adolescents. The
present study is the first to formally develop and validate a comprehensive instrument
to assess social networking use intensity by considering diverse social networking
activities and functions without bonding to a specific platform. The 14-item online
Social Networking Activity Intensity Scale (SNAIS) developed in this study showed
acceptable psychometric properties. The two subscales, SFUI and EFUI, have been
definitively identified and cross-validated with internal and external validity. Such
instrument that follows the standard scale development procedure could facilitate and
serve future relevant research aiming to explore the impact of social networking use
intensity on adolescents’ mental health outcomes both in China and worldwide.
10.1.4 The impact of social networking use intensity on depression
By applying SFUI and EFUI subscales as the index of social networking use
intensity, higher levels of social networking use intensity at baseline (both SFUI and
EFUI) was a significant risk predictor of incident depression, even after accounting
for baseline level of depressive symptoms. The results suggest that the causal effect of
social networking use intensity on depression does not vary in terms of the two broad
258
types of functions that adolescents mainly conduct on such platforms. As a significant
risk factor, social networking use should be paid more attention in depression
prevention and intervention programs targeting adolescents. Early interventions
should be considered to reduce social networking use intensity as a means of
preventing depression.
10.1.5 Impact of social networking use intensity on psychosocial status
This is the first longitudinal study to investigate the effects of social networking
use intensity onto diverse psychosocial factors among adolescents. This study found
that increased social networking use intensity is positively associated with increased
sense of perceived social support and friendship quality during the follow-up period.
In addition, increased social networking use intensity is positively associated with
increased adolescent-mother conflict and social non-confidence during the follow-up
period. These results confirm the direct impacts of social networking use on
adolescents’ psychosocial statuses. These important psychosocial effects of social
networking use intensity highlight the necessity and attention for future intervention
studies.
10.1.6 Underlying mechanisms for the relationship between social networking
use intensity and depression
It is important to understand the underlying mechanisms on how and why social
networking use increases the level of depressive symptoms among adolescents. The
present study firstly demonstrates some potential mechanisms underlying the
relationship between social networking use intensity and depressive symptoms among
adolescents. The results of multiple-mediator models reveal the first evidence that
259
change in addiction (‘SNA’ and IA) and significant psychosocial factors (i.e. social
non-confidence and adolescent-mother conflict) altogether mediated a large
proportion of the effect of change in SFUI on change in depressive symptoms
(proportion mediated=78%). Meanwhile, the effect of change in EFUI on change in
depressive symptoms was completely mediated by change in addiction (i.e. ‘SNA’ and
IA) and change in social non-confidence. The mediational findings indicate that
higher intensity of social networking use could increase users’ addictive tendency to
Internet and social networking, and could also decrease their psychosocial status (i.e.
social confidence and family relationships), leading to increased level of depressive
symptoms.
Additionally, a significant moderation effect of increased adolescent-father
conflict was observed for the relationship between change in social networking use
intensity and change in depressive symptoms. These results showed that improvement
in adolescent-father relationship could significantly buffer the detrimental effect of
increased social networking use intensity on increased depressive symptoms.
Furthermore, it was interesting to find that change in friendship quality and
change in perceived social support slightly but significantly suppressed the
association between change in social networking use intensity and change in
depressive symptoms. The findings reveal that increased perceived social support and
friendship quality partially compensate the harmful effect of increased social
networking use intensity onto increased depressive symptoms. However, the
suppression effects observed in the present study are relatively small. Future research
is needed to investigate and confirm such negative suppression effects among this
population.
260
These results contribute to our understanding of the underlying mechanisms for
the relationship between social networking use intensity and depressive symptoms.
This information is important and helpful for developing effective intervention
programs in future research. These psychosocial factors and addictive behaviors
should be taken into consideration when formulating intervention programs targeting
to improve adolescents’ mental health.
10.2 Implications of the study
10.2.1 Intervention programs for reducing depressive symptoms
The positive prediction of social networking use intensity on depression and the
identified underlying mechanisms suggest some new directions for implementing
effective intervention programs to prevent or reduce depressive symptoms among
adolescents who are social networking users. This direction is particularly important
under the condition of high popularity of social networking use among adolescents.
First, intervention programs should include components to strengthen
adolescents’ confidence in social situations and to improve their relationships with
parents, such as through skills training programs that encourage positive coping
strategies. Coping effectiveness training should emphasize positive appraisal. The
optimal use of positive coping methods is effective in improving confidence in social
situations and adolescent-parents relationships.
Second, early intervention is important and necessary to reduce intensive and
excessive social networking use, as intensive social networking use renders
adolescents vulnerable for addiction. Therefore, adolescents’ social networking use in
future intervention programs should be properly restrained, rather than waiting for
261
Internet/social networking addiction to occur.
Intervention strategies in future research exclusively targeting these significant
mediation components (i.e. social non-confidence, adolescent-parents relationships,
and Internet addictive behaviors) would effectively block the negative impacts of
social networking use on adolescents’ mental health.
10.2.2 Intervention programs for preventing Internet addictive behaviors
Internet addictive behaviors are associated with wide range of adverse mental
and physical problems, and have become a noticeable public health issue. Policy
makers and health workers who work with adolescents must aid in the elimination of
these harmful behaviors. Unlike other addictive behaviors (e.g. alcohol use, gambling),
Internet and social networking use is now an integral aspect of modern life. It is not
suitable or sustainable to require students to abstain completely from Internet and
social networking use. As recommended by Young [287], the primary objective of
Internet addictive behaviors intervention should be abstinence from problematic
online applications. Moderated and controlled use of Internet/social networking is the
most appropriate to prevent Internet addictive behaviors. Moreover, school and public
health educators should encourage students to have more interactions and activities
with peers, family members and others offline. These strategies could improve social
adjustment and reduce the dependency on Internet/social networking by increasing
offline social activities.
In addition, health care providers who work with adolescents may find it useful
to evaluate social networking or other social media exposures during routine health
education sections. When large amounts of social networking use are reported, a
262
further detailed assessment of Internet addictive behaviors should be considered for
early intervention and treatment.
10.2.3 Collaboration and integration of Internet addictive behaviors and
depression intervention
The bi-directional and highly co-occurred relationships between Internet
addictive behaviors and depression highlight the necessity for integration and
collaboration of strategies to prevent these two public health challenges. Collaboration
and integration of these intervention strategies can improve efficiency,
cost-effectiveness, and health outcomes. For those students who seek help regarding
addictive Internet use, health care workers and professionals should consider whether
they have depressive psychopathology simultaneously, which may pinpoint the root of
the problem and can help to choose appropriate intervention strategies. Screening for
depression is hence recommended for these students. If depression is not found,
effective interventions should specifically focus on reducing addictive behaviors more
effectively. Meanwhile, efforts should be made to prevent the presence of depression
attributed to Internet/social networking addiction, as adolescents who use the Internet
or social networking pathologically are at high-risk group of developing depression in
the absence of effective interventions for Internet addictive behaviors. Such early
interventions and prevention by targeting at-risk groups of adolescents is effective in
reducing the burden of depression [288]. Alternatively, if comorbidities of Internet
addictive behaviors and depression are present, it may indicate that these students
treat Internet/social networking as a more comfortable and secure way to alleviate
their depressive mood as compared to seeking offline support and help. Interventions
should target at both issues to maximize the treatment outcomes. Effective
263
interventions including specific components to reduce depressed mood, such as
improving social self-efficacy, social skills, and coping strategies training, may help
these students to involve social interactions in the real-life settings and reduce the
dependency on the Internet and social networking. It is important to evaluate and treat
these psychiatric symptoms in adolescents with Internet addictive behaviors. To treat
these disorders concurrently could provide benefits to interventions for Internet
addictive behaviors among adolescents.
On the other hand, for those students with depressive symptoms, brief screenings
of Internet addictive behaviors may be appropriate as a standard battery of health
questions at care settings. Explicitly screening would aid in the early identification of
individuals at risk for potential Internet addictive behaviors, to whom timely and
appropriate counseling and interventions could be provided to reduce harms and
related complications. Furthermore, since depressive symptoms might predispose
Internet addictive behaviors, appropriate treatment for depressive symptoms might
prevent the emergence of Internet addictive behaviors. A strategy of integrated
therapy for both problems in dual diagnosis could be provide more consistent
treatment effects [289].
10.2.4 Assessment for social networking use intensity
This study provides a useful and easy-administered instrument with acceptable
psychometric properties, Social Networking Activity Intensity Scale (SNAIS), for
assessing social networking use intensity. The SNAIS takes diverse social networking
activities and functions into consideration, and generalizes the features that are
common to standard social networking use. The two subscales of SNAIS reflect the
main functions of social networking use that adolescents conduct on social
264
networking platforms. Its application could catalyze the progression of relevant
studies on the impacts of social networking use intensity on mental health outcomes
among adolescents in China and worldwide.
10.2.5 Methodological considerations in longitudinal causal-effect models
The statistical methodologies that are applied to explore causal-effects in
longitudinal observational studies are diverse and not in consensus. Our results
suggest that statistical modeling by taking change in pathological status over time into
consideration is a potential methodological improvement in exploring the
causal-effects in longitudinal studies, especially for those pathological conditions,
such as depression and Internet addictive behaviors, with significant fluctuations
during the study period. Therefore, statistical analysis should consider such
pathological status change into modeling specifications in future longitudinal research.
Moreover, it may be better for robust causal directional inference to combine results
from multiple statistical modeling methods.
In addition, some other alternative statistical modeling methods, such as
cross-lagged SEM, could be considered in future research. However, caution should
be given that this method is disadvantageous in controlling many potential
confounding factors and in handling categorical variables, which are the main reasons
why we did not apply the SEM statistical method in the present study. Future
longitudinal studies aiming to explore causal relationships between Internet addictive
behaviors and depression might properly consider such method as an alternative to
supply direct comparison with our findings to give a more convincing picture for
these research questions.
265
10.3 Limitations of the study
10.3.1 Study design
The nine-month observational time is a relatively short longitudinal period,
which is no long enough to verify the long-term effects of online social networking
use intensity on psychosocial and mental health among adolescents. Two studies by
Kraut et al have revealed that the short-term detrimental effects may dissipate over a
longer period time [46, 61]. Longitudinal studies with longer follow-up time (e.g. one or
two years) would be preferable to convincingly assess the long-term effects of online
social networking use intensity on adolescents’ psychological and mental health (e.g.
depressive symptoms) in China.
The present study explored some possible mediational mechanisms on how
social networking use intensity affects depression among adolescents. However, due
to the restriction of two-wave study design, it could not derive causal mediational
sequence. Thus, experimental studies with three or more assessments would be
needed to convincingly establish the temporal mediational evidence among
adolescents.
Since the present study focused on a very specific demographic population, i.e.
non-clinical school-based secondary students in Guangzhou of South China, the
generalizability of the results to other population and regions should be cautious.
Similar studies in other demographic populations (e.g. other age group, or psychatric
clinical population) are warranted in the future.
266
10.3.2 Sampling
The districts and schools in the present study were conveniently rather than
randomly selected in each stratification since randomized sampling is difficult and not
feasible. There may be selection bias as the sampling frame did not include all junior
middle school students in Guangzhou. Therefore, it could not fully ensure that the
reported prevalence and incidence rates (e.g. depression) represent the true situation
of adolescents in Guangzhou. Although the stratified cluster sampling method could
partially guarantee the representativeness of the sample for the targeted population,
attentions should also be made to the representativeness of the sample when
interperting the prevalence and incidence rates reported in this study.
It was a limitation that students were not screened about whether they previously
have depression (or currently receive treatment) and other mental illness by
physicians at baseline. However, to our understanding, the proportion of such students
is relately small since those with such psychiatric conditions would be refered to
receive treatment. Moreover, such students are more likely to be unwilling to
participate the survey, and then excluded from the study.
10.3.3 Causal inference for the relationships between ’SNA’/IA and depression
The present study provided the evidences of temporal sequence and strength of
the associations for causal inference between ’SNA’/IA and depression among
adolescents. However, it is still far away to establish causal relationships for these
public health issues because of the lack of evidences of other critetia for causal
inference [290], such as biological gradient, consistency, specificity etc. Moreover,
there may be other potential confounders (e.g. prolonged time of watching TV) that
267
were not included and adjusted in our analysis. From this persepctive, the findings in
this study could only imply possible causal relationships between ’SNA’/IA and
depression. More evidences from experimental or randomized controlled studies, if
possible, are highly warranted in order to establish convinced causal relationships
between these public health problems.
10.3.4 Measures and reporting bias
Depression measured by CES-D, a self-reported epidemiological screening, only
represents probable depression, which is rather than clinical depression. It should be
remembered that the CES-D was a measure for epidemiologic research and was not
intented to be used for clinical diagnosis. However, significant depressive
symptomatology is a strongly predictive of clinical depression, and CES-D has been
widely used among adolescents in China and worldwide. The depression defined by
cut-off value of 21 indicates a moderate level of depression in epidemiologic research.
Thus, we cannot completely exclude misclassification as a source of measurement
error that explains current observed associations. Future studies using clinical
diagnosis critetia for depression (or among clinical psychiatric populations) are hence
warranted to further confirm the findings in this study.
Some online social networking-related measures such as “emotional connection
to social networking” and “social networking addiction (SNA)” were firstly modified
from Facebook-related scales of previous studies. These scales were not previously
validated in Chinese cultural context. However, we strictly translated and validated
these measures, and all these measures in this study have satisfactory psychometric
properties (Cronbach’s alpha > 0.8). The internal consistency of the modified scales
should not be a problem.
268
Since there is a lack of evidence-based diagnostic criteria and cut-off value for
‘SNA’ classification, ‘SNA’ was experimentally defined by the top 10% of scores
(around mean + 1.2 Standard deviation). Thus, caution must be given when
interpreting the related results, i.e. causal direction, prevalence and incidence.
Additional refinement and improvement for these measurements are needed.
Similarly, there is currently no acceptable consensus in the diagnostic criteria for
IA in general. There are several commonly used scales for generalized Internet
addiction assessment among adolescents, e.g. 20-item Chen Internet addiction
scale [99], Young’s 20-itme Internet addiction test [185, 202], and Young’s 8-item
diagnostic questonnaire [160, 291]. The IA prevalence and incidence in this study is
assessed based on Young’s 8-item Diagnostic Questionnaire (YDQ). The YDQ is a
more stringent criterion than Chen’s Internet Addiction Scale, and therefore is less
likely to overestimate the situation. Comparison of the prevalence and incidence data
with other studies should be done carefully.
Only online social networking activities intensity were measured in the present
study. It is a limitation that the conventional offline social network activities were not
measured, which may be a major confounder in this study. Offline social network is
important and fundamental to adolescents’ living and development. For example,
adolescents with poor offline social networks (e.g. poor social skills to make friends)
may have increased use of online social networking and then increased the risk of
developing depression. It is therefore recommended to involve offline social
network-related measures in future studies.
Self-reported data collecting method was applied in this study, and reporting bias
may consequentially exist (i.e. social desirable bias, recall bias). However, the great
269
efforts were made to minimize this bias, including the anonymous and private nature
of the study, the absence of teachers during the investigation, using well-trained field
workers, and guaranteeing participants that their personal data was highly confidential
and inaccessible to third parties. The students were also told that their participations in
the study was voluntary and that they have the right to terminate their participation at
any time.
The questionnaire was relatively long and participants may have lost patience
when completing it. However, the pilot test was conducted to assess the feasibility of
the questionnaire before we put it into use in the longitudinal study, and the feedback
showed that the length was acceptable among participants. During the follow-up
period, only 9.2% of participants were lost to follow-up in the nine-month interval.
10.4 Recommendations for future research
10.4.1 Longitudinal study with three or more observations
Due to practical constraints and resource limitations, our longitudinal study only
included two time points with nine-month interval. Therefore, in order to further
confirm these causal relationships and the longitudinal mediational effects, more
longitudinal studies with at least three measurement waves are highly warranted in
different populations (e.g. other age group, or psychiatirc population) and under
different circumstances (i.e. social-cultural). Especially when testing causal
mediational effects, three or more measurement waves are essentially needed in order
to provide a clear temporal sequence for robust causal mediational relationships.
Moreover, considering the highly dynamic status changes in IA, ‘SNA’, and
depression over time, the time interval should be carefully defined in the prospective
270
study. In order to truly reflect the natural course of such pathological statuses, future
longitudinal research should appropriately conduct the follow-up within a short-term
interval (e.g. three-month or six-month interval), which could provide more pertinent
information. Additionally, these pathological conditions should be assessed at each
time point rather than just measured in the initial assessment.
10.4.2 Mining online social networking content
Most of the published studies have focused on behavioral characteristics (such as
frequency and duration of social networking use); however, the actual contents
created and exchanged through social networking by adolescents have been largely
ignored, which could be an invaluable asset to the study of adolescents. However, the
potential of social networking to provide rich and valid data about the nature of
adolescents’ computer-mediated communications remains largely under-researched.
Online social networking has been used to identify depression among college
students who display reference of depressive symptoms on Facebook [292]. Transcripts
from social networking content are potential fruitful avenues to pursue and explore in
future research. Better documentation and understanding of social networking
activities and content could be useful to identify adolescents who are at an increased
risk for a variety of psychosocial challenges, including social isolation, deviant
behavior, and psychopathology. Therefore, mining social networking content that
adolescents conduct online would be considered as a future research direction, as such
studies could facilitate the early identification of deteriorating conditions, improve
both public and professional awareness about early warning of risk behaviors
(cyber-bullying, online harassment), and enhance our understanding of the
characteristics of social networking use among adolescents [49, 62, 86, 293, 294].
271
10.4.3 Research on benefits of proper online social networking use
Our study mainly investigated the harmful effects of online social networking
use intensity on adolescents’ psychological and mental health outcomes. Some
previous studies have reported that online social networking use could increase users’
satisfaction with life [114]. It is recommended to explore and elaborate the potential
beneficial effects of proper online social networking use on psychosocial well-being
among adolescents, such as satisfaction with life, social capital, and reduced
loneliness etc.
10.4.4 Considerations for other measures in future direction
It is unknown whether (or extent to which) online social networking use
intensity has different impacts on adolescents’ online and offline psychosocial
relationships, e.g. whether online social networking use only increases users’
perceived online social support, but have no effect or even harmful impacts on
perceived offline social support. One previous study reported that Facebook use
helped college students to obtain online social support, but online social support
appears to have little direct effects on well-being [295]. Future studies are therefore
recommended to distinguish online and offline social relationships, and investigate
their potential mediating/moderating effects seperately. Such knowledge will extend
the scope and depth for understanding of underlying mechanisms.
Secure relationship with one or a few good friends is important for adolescents’
development. Apart from the measure for online social networking use intensity, it is
recommended to measure the intensity that adolescents use online social networking
to interact with particular good friends in future studies.
272
It is also recommended that other asepcts of psychosocial relationships should
further be investigated and measured apart from those measured in the present study.
For instance, quantity of friendship such as time spent with particilar good friends,
degree of social connectedness with friends and others, and other types of social
support (e.g instrumental support, or emotional social support) might be important
measures and covariates that may influence the current findings. Future research
should consider to measure these aspects of adolescents’ psychosocial relationships.
10.4.5 Moderation effects of gender and personality for the effects of online social
networking use intensity on psychosocial and mental health among adolescents
Personality and gender might interact with online social networking use intensity
to affect adolescents’ psychosocial and mental health. For instance, whether
extraverted individuals are more likely to use online social networking intensively
when compared to introverted individuals, or whether detrimental effects of online
social networking use intensity on psychosocial and mental well-being are stronger or
weaker among males than that in females, or even only existed among males/females?
Such research questions were not explored in this study, but they would contribute to
a better understanding on who would have more tendency to use online social
networking intensively and would be more influenced by online social networking
use.
10.4.6 Intervention studies using social networking as an innovative avenue
Mental health interventions through online social networking platforms can
easily reach targeted populations and could potentially be cost-effective. There is clear
evidence that online interventions using the cognitive-behavioral theory are promising
273
in reducing depression in young people. Social networking as a tool to enhance
awareness of adolescents at risk and to intervene online is an important new direction.
Further study is warranted to identify the effectiveness of online interventions by
delivering cognitive behavioral subcomponents to prevent depression among
adolescents, such as problem solving therapy or other approaches. Given the high
popularity of social networking among adolescents and adults, social networking
would be an innovative means of identifying young individuals at high risk for
depression, and can also provide a cost-effective intervention, such as disseminating
intervention-related information, or healthy-related information.
10.5 Overall remarks
This study is one of the first to thoroughly explore the causal directions between
Internet addictive behaviors (both ‘SNA’ in specific and IA in general) and depression
by a large-scale school-based sample among adolescents in China. Bidirectional
relationships between Internet addictive behaviors and depression were supported.
This study provides a useful instrument for assessing social networking use
intensity, which is a feasible, acceptable and easy-administered among adolescents. It
is firstly identified that social networking use intensity is a significant risk factor of
depression among adolescents even after controlling for the baseline level of
Additionally, this study is the first longitudinal study to provide robust evidence
on how and why social networking use intensity increases depressive symptoms
among adolescents. It has showed that the impact of social networking use intensity
on depressive symptoms are mainly indirect through its effects on increased addictive
274
behaviors and decreased psychosocial conditions (i.e. social non-confidence and
adolescent-mother conflict).
The results of this study not only enrich our understanding of the impact of social
networking use on adolescents’ psychosocial health in China, but also provide
important and useful information in designing effective intervention in preventing
depression and Internet addictive behaviors. Furthermore, the results of this study also
provide some implications for future research. More longitudinal studies with three or
more waves of observations are highly warranted in China and worldwide.
275
Reference
1. Christie D, Viner R. Adolescent development. BMJ 2005,330:301-304.

2. Merikangas KR, He JP, Burstein M, Swanson SA, Avenevoli S, Cui L, et al.
Lifetime prevalence of mental disorders in US adolescents: results from the
National Comorbidity Survey Replication–Adolescent Supplement (NCS-A).
Journal of the American Academy of Child & Adolescent Psychiatry
2010,49:980-989.
3. Knopf D, Park MJ, Mulye TP. The mental health of adolescents: A national
profile, 2008. San Francisco, CA: National Adolescent Health Information
Center 2008.
4. Thapar A, Collishaw S, Potter R, Thapar AK. Managing and preventing
depression in adolescents. BMJ 2010,340:C209.
5. Lewinsohn PM, Clarke GN, Seeley JR, Rohde P. Major depression in
community adolescents: age at onset, episode duration, and time to
recurrence. Journal of the American Academy of Child & Adolescent
Psychiatry 1994,33:809-818.
6. Birmaher B, Arbelaez C, Brent D. Course and outcome of child and adolescent
major depressive disorder. Child and Adolescent Psychiatric Clinics of North
America 2002,11:619-637.
7. Rushton JL, Forcier M, Schectman RM. Epidemiology of depressive symptoms
in the National Longitudinal Study of Adolescent Health. Journal of the
American Academy of Child & Adolescent Psychiatry 2002,41:199-205.
8. Feng ZZ, Zhang DJ. An epidemiological study on depressive symptoms in 2634
middle school students. Chinese Journal of Behavioral Medical Science
2005,14:7-9.
9. ITU ICT Facts and Figures-The world in 2015.
In: http://www.itu.int/en/ITU-D/Statistics/Documents/facts/ICTFactsFigures2
015.pdf; 2015.
10. China Internet Network Information Center (CNNIC). [The 27th statistical
report of Internet network develpoment in China].
In: http://www.cnnic.net.cn/hlwfzyj/hlwxzbg/201101/P020120709345289031
187.pdf; 2011.
report of internet network develpoment in China].
In: http://www.cnnic.net.cn/hlwfzyj/hlwxzbg/hlwtjbg/201507/P02015072354
9500667087.pdf; 2015.
12. China Internet Network Information Center (CNNIC). [Report on Internet
behavior among adloescents in China in 2014].
In: http://www.cnnic.net.cn/hlwfzyj/hlwxzbg/qsnbg/201506/P020150603434
893070975.pdf; 2015.
13. Boyd DM, Ellison NB. Social Network Sites: Definition, History, and Scholarship.
Journal of Computer-Mediated Communication 2007,13:210-230.
14. Lenhart A, Madden M, Smith A, Purcell K, Zickuhr K, Rainie L. Teens, Kindness
and Cruelty on Social Network Sites: How American Teens Navigate the New
276
World of" Digital Citizenship". Pew Internet & American Life Project 2011.
15. Livingstone S, Haddon L. EU Kids Online. Journal of Psychology
2009,217:236-239.
16. Wang JL, Jackson LA, Gaskin J, Wang HZ. The effects of Social Networking Site
(SNS) use on college students’ friendship and well-being. Computers in
Human Behavior 2014,37:229-236.
17. Shaw M, Black DW. Internet addiction. CNS drugs 2008,22:353-365.
18. Byun S, Ruffini C, Mills JE, Douglas AC, Niang M, Stepchenkova S, et al.
Internet addiction: metasynthesis of 1996-2006 quantitative research.
CyberPsychology & Behavior 2009,12:203-207.
19. Young KS. Internet addiction: The emergence of a new clinical disorder.
20. Dalbudak E, Evren C, Aldemir S, Coskun KS, Ugurlu H, Yildirim FG. Relationship
of internet addiction severity with depression, anxiety, and alexithymia,
temperament and character in university students. Cyberpsychology, Behavior,
and Social Networking 2013,16:272-278.
21. Davis RA. A cognitive-behavioral model of pathological Internet use.
Computers in Human Behavior 2001,17:187-195.
22. Montag C, Bey K, Sha P, Li M, Chen YF, Liu WY, et al. Is it meaningful to
distinguish between generalized and specific Internet addiction? Evidence
from a cross‐cultural study from Germany, Sweden, Taiwan and China.
Asia‐Pacific Psychiatry 2014,7:20-26.
23. Laconi S, Tricard N, Chabrol H. Differences between specific and generalized
problematic Internet uses according to gender, age, time spent online and
psychopathological symptoms. Computers in Human Behavior
2015,48:236-244.
24. Pontes HM, Szabo A, Griffiths MD. The impact of Internet-based specific
activities on the perceptions of Internet addiction, quality of life, and
excessive usage: A cross-sectional study. Addictive Behaviors Reports
2015,1:19-25.
25. Griffiths M. Social networking addiction: emerging themes and issues. Journal
of Addiction Research and Therapy 2013,4:e118.
26. Andreassen CS. Online Social Network Site Addiction: A Comprehensive
Review. Current Addiction Reports 2015,2:175-184.
27. Kuss D, Griffiths M, Karila L, Billieux J. Internet addiction: a systematic review
of epidemiological research for the last decade. Current Pharmaceutical
Design 2014,20:4026-4052.
28. Ko CH, Yen JY, Yen CF, Chen CS, Chen CC. The association between Internet
addiction and psychiatric disorder: a review of the literature. European
Psychiatry 2012,27:1-8.
29. Alabi OF. A survey of Facebook addiction level among selected Nigerian
University undergraduates. New Media and Mass Communication
2013,10:70-80.
30. Wolniczak I, Cáceres-DelAguila JA, Palma-Ardiles G, Arroyo KJ, Solís-Visscher R,
Paredes-Yauri S, et al. Association between Facebook dependence and poor
sleep quality: a study in a sample of undergraduate students in Peru. PLoS
277
ONE 2013,8:e59087.
31. Zhou SX. Gratifications, loneliness, leisure boredom and self-esteem as
predictors of SNS-game addiction and usage pattern among Chinese college
students. Hong Kong: The Chinese University of Hong Kong; 2010.
32. Wan C. Gratifications & loneliness as predictors of campus-SNS websites
addiction & usage pattern among Chinese college students. Hong Kong: The
Chinese University of Hong Kong; 2009:35.
33. Ko CH, Yen JY, Chen CS, Chen CC, Yen CF. Psychiatric comorbidity of internet
addiction in college students: an interview study. CNS Spectrums
2008,13:147-153.
34. Ni X, Yan H, Chen S, Liu Z. Factors influencing internet addiction in a sample of
freshmen university students in China. Cyberpsychology & Behavior
2009,12:327-330.
35. Pantic I, Damjanovic A, Todorovic J, Topalovic D, Bojovic-Jovic D, Ristic S, et al.
Association between online social networking and depression in high school
students: behavioral physiology viewpoint. Psychiatria Danubina
2012,24:90-93.
36. Kalpidou M, Costin D, Morris J. The relationship between Facebook and the
well-being of undergraduate college students. Cyberpsychology, Behavior and
Social Networking 2011,14:183-189.
37. Labrague LJ. Facebook use and adolescents’ emotional states of depression,
anxiety, and stress. Health Science Journal 2014,8:80-89.
38. Chang FC, Chiu CH, Lee CM, Chen PH, Miao NF. Predictors of the initiation and
persistence of Internet addiction among adolescents in Taiwan. Addictive
Behaviors 2014,39:1434-1440.
39. Ko CH, Yen JY, Chen CS, Yeh YC, Yen CF. Predictive values of psychiatric
symptoms for internet addiction in adolescents: a 2-year prospective study.
Archives of Pediatrics & Adolescent Medicine 2009,163:937-943.
40. Ko CH, Yen JY, Yen CF, Lin HC, Yang MJ. Factors predictive for incidence and
remission of internet addiction in young adolescents: a prospective study.
41. Ko CH, Liu TL, Wang PW, Chen CS, Yen CF, Yen JY. The exacerbation of
depression, hostility, and social anxiety in the course of Internet addiction
among adolescents: A prospective study. Comprehensive Psychiatry
2014,55:1377-1384.
42. Dong G, Lu Q, Zhou H, Zhao X. Precursor or sequela: pathological disorders in
people with Internet addiction disorder. PloS ONE 2011,6:e14703.
43. Lam LT, Peng ZW. Effect of pathological use of the internet on adolescent
mental health: a prospective study. Archives of Pediatrics & Adolescent
Medicine 2010,164:901-906.
44. Cho SM, Sung MJ, Shin KM, Lim KY, Shin YM. Does psychopathology in
childhood predict internet addiction in male adolescents? Child Psychiatry &
Human Development 2013,44:549-555.
45. Gámez-Guadix M. Depressive Symptoms and Problematic Internet Use
Among Adolescents: Analysis of the Longitudinal Relationships from the
Cognitive–Behavioral Model. Cyberpsychology, Behavior, and Social
278
Networking 2014,17:714-719.
46. Kraut R, Kiesler S, Boneva B, Cummings J, Helgeson V, Crawford A. Internet
paradox revisited. Journal of Social Issues 2002,58:49-74.
47. Hong FY, Huang DH, Lin HY, Chiu SL. Analysis of the psychological traits,
Facebook usage, and Facebook addiction model of Taiwanese university
students. Telematics and Informatics 2014,31:597-606.
48. Koc M, Gulyagci S. Facebook Addiction Among Turkish College Students: The
Role of Psychological Health, Demographic, and Usage Characteristics.
Cyberpsychology, Behavior, and Social Networking 2013,16:279-284.
49. Lee G, Lee J, Kwon S. Use of social-networking sites and subjective well-being:
a study in South Korea. Cyberpsychology, Behavior and Social Networking
2011,14:151-155.
50. Chen WH, Lee KH. Sharing, liking, commenting, and distressed? The pathway
between Facebook interaction and psychological distress. Cyberpsychology,
Behavior, and Social Networking 2013,16:728-734.
51. Ellison NB, Steinfield C, Lampe C. The benefits of Facebook “friends:” Social
capital and college students’ use of online social network sites. Journal of
Computer‐Mediated Communication 2007,12:1143-1168.
52. Simoncic TE, Kuhlman KR, Vargas I, Houchins S, Lopez-Duran NL. Facebook use
and depressive symptomatology: Investigating the role of neuroticism and
extraversion in youth. Computers in Human Behavior 2014,40:1-5.
53. Jelenchick LA, Eickhoff JC, Moreno MA. “Facebook depression?” Social
networking site use and depression in older adolescents. Journal of
Adolescent Health 2013,52:128-130.
54. Subrahmanyam K, Reich SM, Waechter N, Espinoza G. Online and offline
social networks: Use of social networking sites by emerging adults. Journal of
Applied Developmental Psychology 2008,29:420-433.
55. Pempek TA, Yermolayeva YA, Calvert SL. College students' social networking
experiences on Facebook. Journal of Applied Developmental Psychology
2009,30:227-238.
56. Li XQ, Chen WH, Popiel PW. What happens on Facebook stays on Facebook?
The implications of Facebook interaction for perceived, receiving, and giving
social support. Computers in Human Behavior 2015,51:106-113.
57. Deters FG, Mehl MR. Does Posting Facebook Status Updates Increase or
Decrease Loneliness? An Online Social Networking Experiment. Social
Psychological & Personality Science 2013,4:1-8.
58. Gentzler AL, Oberhauser AM, Westerman D, Nadorff DK. College students' use
of electronic communication with parents: links to loneliness, attachment,
and relationship quality. Cyberpsychology, Behavior, and Social Networking
2011,14:71-74.
59. Reich SM, Subrahmanyam K, Espinoza G. Friending, IMing, and hanging out
face-to-face: overlap in adolescents' online and offline social networks.
Developmental Psychology 2012,48:356-368.
60. Nabi RL, Prestin A, So J. Facebook friends with (health) benefits? Exploring
social network site use and perceptions of social support, stress, and
well-being. Cyberpsychology, Behavior, and Social Networking
279
2013,16:721-727.
61. Kraut R, Patterson M, Lundmark V, Kiesler S, Mukopadhyay T, Scherlis W.
Internet paradox. A social technology that reduces social involvement and
psychological well-being? American Psychologist 1998,53:1017-1031.
62. Valkenburg PM, Peter J, Schouten AP. Friend networking sites and their
relationship to adolescents' well-being and social self-esteem.
63. Gonzales AL, Hancock JT. Mirror, mirror on my Facebook wall: Effects of
exposure to Facebook on self-esteem. Cyberpsychology, Behavior, and Social
Networking 2011,14:79-83.
64. Mehdizadeh S. Self-presentation 2.0: Narcissism and self-esteem on Facebook.
65. Van Rooij AJ, Schoenmakers TM, Van de Eijnden RJ, Van de Mheen D.
Compulsive internet use: the role of online gaming and other internet
applications. Journal of Adolescent Health 2010,47:51-57.
66. WHO. Depression: A Global Crisis. World Federation for Mental Health.
2012:1-32.
67. Whiteford HA, Degenhardt L, Rehm J, Baxter AJ, Ferrari AJ, Erskine HE, et al.
Global burden of disease attributable to mental and substance use disorders:
findings from the Global Burden of Disease Study 2010. The Lancet
2013,382:1575-1586.
68. WHO. Health for the World’s Adolescents A second chance in the second
decade. 2014.
69. Toolan JM. Depression in children and adolescents. American Journal of
Orthopsychiatry 1962,32:404-415.
70. Andrade L, Caraveo‐anduaga JJ, Berglund P, Bijl RV, Graaf RD, Vollebergh W,
et al. The epidemiology of major depressive episodes: results from the
International Consortium of Psychiatric Epidemiology (ICPE) Surveys.
International Journal of Methods in Psychiatric Research 2003,12:3-21.
71. Wesselhoeft R, Sørensen MJ, Heiervang ER, Bilenberg N. Subthreshold
depression in children and adolescents–a systematic review. Journal of
Affective Disorders 2013,151:7-22.
72. Sajjadi H, Kamal SHM, Rafiey H, Vameghi M, Forouzan AS, Rezaei M. A
systematic review of the prevalence and risk factors of depression among
Iranian adolescents. Global Journal of Health Science 2013,5:16-27.
73. Stapinski LA, Montgomery AA, Heron J, Jerrim J, Vignoles A, Araya R.
Depression symptom trajectories and associated risk factors among
adolescents in Chile. PloS ONE 2013,8:e78323.
74. Xin Z, Niu J, Chi L. Birth cohort changes in Chinese adolescents' mental health.
International Journal of Psychology 2012,47:287-295.
75. Liu XC, Ma DD, Kurita H, Tang MQ. Self-reported depressive symptoms among
Chinese adolescents. Social Psychiatry and Psychiatric Epidemiology
1999,34:44-47.
76. Yang J, Yao S, Zhu X, Zhang C, Ling Y, Abela JR, et al. The impact of stress on
depressive symptoms is moderated by social support in Chinese adolescents
with subthreshold depression: A multi-wave longitudinal study. Journal of
280
Affective Disorders 2010,127:113-121.
77. Yang WH, Zhou T, Peng F, Liu HH. Prevalence of Depression in Adolescents in
Yueyang County of Hunan Province. Chinese Journal of Clinical Psychology
2013,6:937-941.
78. Internet Live Stats. In: http://www.internetlivestats.com/internet-users/;
2014.
79. Pantic I. Online Social Networking and Mental Health. Cyberpsychology,
80. Schneider F, Feldmann A, Krishnamurthy B, Willinger W. Understanding online
social network usage from a network perspective. In: Proceedings of the 9th
ACM SIGCOMM conference on Internet measurement conference: ACM; 2009.
pp. 35-48.
81. Alhabash S, Park H, Kononova A, Chiang YH, Wise K. Exploring the motivations
of Facebook use in Taiwan. Cyberpsychology, Behavior, and Social Networking
2012,15:304-311.
82. Spiliotopoulos T, Oakley I. Understanding motivations for Facebook use:
Usage metrics, network structure, and privacy. In: Proceedings of the SIGCHI
Conference on Human Factors in Computing Systems: ACM; 2013. pp.
3287-3296.
83. Joinson AN. Looking at, looking up or keeping up with people?: motives and
use of facebook. In: Proceedings of the SIGCHI conference on Human Factors
in Computing Systems: ACM; 2008. pp. 1027-1036.
84. Yang CC, Brown BB. Motives for using Facebook, patterns of Facebook
activities, and late adolescents' social adjustment to college. Journal of Youth
and Adolescence 2013,42:403-416.
85. Ross C, Orr ES, Sisic M, Arseneault JM, Simmering MG, Orr RR. Personality and
motivations associated with Facebook use. Computers in Human Behavior
2009,25:578-586.
86. Moreno MA, Kolb J. Social networking sites and adolescent health. Pediatric
Clinics of North America 2012,59:601-612.
87. Bull SS, Levine DK, Black SR, Schmiege SJ, Santelli J. Social media-delivered
sexual health intervention: a cluster randomized controlled trial. American
Journal of Preventive Medicine 2012,43:467-474.
88. Media CS. Is Social Networking Changing Childhood? In:
https://www.commonsensemedia.org/about-us/news/press-releases/is-social
-networking-changing-childhood. SanFrancisco,CA; 2009.
89. China Internet Network Information Center (CNNIC). [Annual report of social
networking sites use appplication among Chinese netizen in 2012].
In: http://www.cnnic.net.cn/hlwfzyj/hlwxzbg/sqbg/201302/P0201302196116
51054576.pdf; 2013.
report of internet network develpoment in China].
In: http://www.cnnic.cn/hlwfzyj/hlwxzbg/hlwtjbg/201307/P02013071750534
3100851.pdf; 2013.
91. Shi JQ, Niu QK. SNSs Usage among Chinese Internet Users: An Empirical Study.
In: Annual Review of Cybertherapy and Telemedicine 2010: Advanced
281
Technologies in Behavioral, Social and Neurosciences. Edited by Wiederhold
BK, Riva G, Kim SI. Washington, DC: ISO Press; 2010. pp. 150-154.
92. Tsitsika A, Janikian M, Schoenmakers TM, Tzavela EC, Ólafsson K, Wójcik S, et
al. Internet Addictive Behavior in Adolescence: A Cross-Sectional Study in
Seven European Countries. Cyberpsychology, Behavior, and Social Networking
2014,17:528-535.
93. Kuss D, Griffiths M. Excessive online social networking: Can adolescents
become addicted to Facebook. Education and Health 2011,29:68-71.
94. Kuss DJ, Griffiths MD. Internet gaming addiction: A systematic review of
empirical research. International Journal of Mental Health and Addiction
2012,10:278-296.
95. Griffiths M. A'components' model of addiction within a biopsychosocial
framework. Journal of Substance Use 2005,10:191-197.
96. Echeburua E, De Corral P. Addiction to new technologies and to online social
networking in young people: A new challenge. Adicciones 2010,22:91-95.
97. Griths MD, Kuss DJ, Demetrovics Z. Social Networking Addiction: An Overview
of Preliminary Findings. In: Behavioral Addictions: Criteria, Evidence and
Treatment. Edited by Rosenberg KP, Feder LC. London: Elsevier; 2014. pp.
119-141.
98. Cheng C, Li AYL. Internet addiction prevalence and quality of (real) life: a
meta-analysis of 31 nations across seven world regions. Cyberpsychology,
99. Lin MP, Ko HC, Wu JYW. Prevalence and psychosocial risk factors associated
with Internet addiction in a nationally representative sample of college
students in Taiwan. Cyberpsychology, Behavior, and Social Networking
2011,14:741-746.
100. Machold C, Judge G, Mavrinac A, Elliott J, Murphy A, Roche E. Social
networking patterns/hazards among teenagers. Irish Medical Journal
2012,105:151-152.
101. Wu AM, Cheung VI, Ku L, Hung EP. Psychological risk factors of addiction to
social networking sites among Chinese smartphone users. Journal of
Behavioral Addictions 2013,2:160-166.
102. Moreno MA, Jelenchick LA, Breland DJ. Exploring depression and problematic
internet use among college females: A multisite study. Computers in Human
Behavior 2015,49:601-607.
103. Yoo YS, Cho OH, Cha KS. Associations between overuse of the internet and
mental health in adolescents. Nursing & Health Sciences 2014,16:193-200.
104. Ryan T, Chester A, Reece J, Xenos S. The uses and abuses of Facebook: A
review of Facebook addiction. Journal of Behavioral Addictions
2014,3:133-148.
105. Błachnio A, Przepiórka A, Pantic I. Internet use, Facebook intrusion, and
depression: Results of a cross-sectional study. European Psychiatry
2015,30:681-684.
106. De Cock R, Vangeel J, Klein A, Minotte P, Rosas O, Meerkerk G-J. Compulsive
Use of Social Networking Sites in Belgium: Prevalence, Profile, and the Role of
Attitude Toward Work and School. Cyberpsychology, Behavior, and Social
282
Networking 2014,17:166-171.
107. Trepte S, Reinecke L. The reciprocal effects of social network site use and the
disposition for self-disclosure: A longitudinal study. Computers in Human
Behavior 2013,29:1102-1112.
108. Datu J, Valdez J, Datu N. Does Facebooking make us sad? Hunting relationship
between Facebook use and depression among Filipino adolescents.
International Journal of Research Studies in Educational Technology (IJRSET)
2012,1:83-91.
109. Jenkins-Guarnieri MA, Wright SL, Johnson B. Development and validation of a
social media use integration scale. Psychology of Popular Media Culture
2013,2:38-50.
110. Grieve R, Indian M, Witteveen K, Anne Tolan G, Marrington J. Face-to-face or
Facebook: Can social connectedness be derived online? Computers in Human
Behavior 2013,29:604-609.
111. Lou LL, Yan Z, Nickerson A, McMorris R. An Examination of the Reciprocal
Relationship of Loneliness and Facebook Use among First-Year College
Students. Journal of Educational Computing Research 2012,46:105-117.
112. Steinfield C, Ellison NB, Lampe C. Social capital, self-esteem, and use of online
social network sites: A longitudinal analysis. Journal of Applied Developmental
Psychology 2008,29:434-445.
113. Burke M, Marlow C, Lento T. Social network activity and social well-being. In:
Proceedings of the SIGCHI Conference on Human Factors in Computing
Systems: ACM; 2010. pp. 1909-1912.
114. Guo Y, Li Y, Ito N. Exploring the Predicted Effect of Social Networking Site Use
on Perceived Social Capital and Psychological Well-Being of Chinese
International Students in Japan. Cyberpsychology, Behavior, and Social
Networking 2014,17:52-58.
115. Lee KT, Noh MJ, Koo DM. Lonely people are no longer lonely on social
networking sites: The mediating role of self-disclosure and social support.
116. Frison E, Eggermont S. Exploring the relationships between different types of
Facebook use, perceived online social support, and adolescents’ depressed
mood. Social Science Computer Review 2015:1-19.
117. O'Keeffe GS, Clarke-Pearson K. The impact of social media on children,
adolescents, and families. Pediatrics 2011,127:800-804.
118. Quinn S, Oldmeadow JA. Is the igeneration a ‘we’generation? Social
networking use among 9‐to 13‐year‐olds and belonging. British Journal
of Developmental Psychology 2013,31:136-142.
119. Barrera Jr M. Distinctions between social support concepts, measures, and
models. American Journal of Community Psychology 1986,14:413-445.
120. Lee SJ. Online Communication and Adolescent Social Ties: Who benefits more
from Internet use? Journal of Computer‐Mediated Communication
2009,14:509-531.
121. Kanter M, Afifi T, Robbins S. The impact of parents “friending” their young
adult child on Facebook on perceptions of parental privacy invasions and
parent–child relationship quality. Journal of Communication 2012,62:900-917.
283
122. Rice M, Kang DH, Weaver M, Howell CC. Relationship of Anger, Stress, and
Coping With School Connectedness in Fourth‐Grade Children. Journal of
School Health 2008,78:149-156.
123. Chan DKS, Cheng GHL. A comparison of offline and online friendship qualities
at different stages of relationship development. Journal of Social and Personal
Relationships 2004,21:305-320.
124. Steiger AE, Allemand M, Robins RW, Fend HA. Low and decreasing self-esteem
during adolescence predict adult depression two decades later. Journal of
Personality and Social Psychology 2014,106:325-338.
125. Orth U, Robins RW, Roberts BW. Low self-esteem prospectively predicts
depression in adolescence and young adulthood. Journal of Personality and
Social Psychology 2008,95:695.
126. Cheak APC, Goh GGG, Chin TS. Online social networking addiction: exploring
its relationship with social networking dependency and mood modification
among undergraduates in Malaysia. In: Proceedings Int Conference
Management, Economics and Finance; 2012. pp. 247-262.
127. Kittinger R, Correia CJ, Irons JG. Relationship between Facebook use and
problematic Internet use among college students. Cyberpsychology, Behavior,
and Social Networking 2012,15:324-327.
128. Toseeb U, Brage S, Corder K, Dunn VJ, Jones PB, Owens M, et al. Exercise and
Depressive Symptoms in Adolescents: A Longitudinal Cohort Study. JAMA
Pediatrics 2014,168:1093-1100.
129. Van Gool CH, Kempen GI, Penninx BW, Deeg DJ, Beekman AT, Van Eijk JT.
Relationship between changes in depressive symptoms and unhealthy
lifestyles in late middle aged and older persons: results from the Longitudinal
Aging Study Amsterdam. Age and Ageing 2003,32:81-87.
130. Da Silva MA, Singh-Manoux A, Brunner EJ, Kaffashian S, Shipley MJ, Kivimäki
M, et al. Bidirectional association between physical activity and symptoms of
anxiety and depression: the Whitehall II study. European Journal of
Epidemiology 2012,27:537-546.
131. Colver A, Rapp M, Eisemann N, Ehlinger V, Thyen U, Dickinson HO, et al.
Self-reported quality of life of adolescents with cerebral palsy: a
cross-sectional and longitudinal analysis. The Lancet 2015,385:705-716.
132. Strawbridge WJ, Camacho TC, Cohen RD, Kaplan GA. Gender differences in
factors associated with change in physical functioning in old age: a 6-year
longitudinal study. The Gerontologist 1993,33:603-609.
133. Schnall PL, Schwartz JE, Landsbergis PA, Warren K, Pickering TG. A longitudinal
study of job strain and ambulatory blood pressure: results from a three-year
follow-up. Psychosomatic Medicine 1998,60:697-706.
134. Lindwall M, Larsman P, Hagger MS. The reciprocal relationship between
physical activity and depression in older European adults: a prospective
cross-lagged panel design using SHARE data. Health Psychology
2011,30:453-462.
135. Ko CH, Wang PW, Liu TL, Yen CF, Chen CS, Yen JY. Bidirectional associations
between family factors and Internet addiction among adolescents in a
prospective investigation. Psychiatry and Clinical Neurosciences
284
2015,69:192-200.
136. Chin WW. Commentary: Issues and opinion on structural equation modeling.
In: JSTOR; 1998.
137. Kupek E. Beyond logistic regression: structural equations modelling for binary
variables and its application to investigating unobserved confounders. BMC
Medical Research Methodology 2006,6:13.
138. VanderWeele TJ. Invited commentary: structural equation models and
epidemiologic analysis. American Journal of Epidemiology 2012:1-5.
139. Baron RM, Kenny DA. The moderator–mediator variable distinction in social
psychological research: Conceptual, strategic, and statistical considerations.
Journal of Personality and Social Psychology 1986,51:1173-1182.
140. Conger AJ. A revised definition for suppressor variables: A guide to their
identification and interpretation. Educational and Psychological Measurement
1974,34:35-46.
141. Fioravanti G, Dèttore D, Casale S. Adolescent Internet addiction: testing the
association between self-esteem, the perception of Internet attributes, and
preference for online social interactions. Cyberpsychology, Behavior, and
Social Networking 2012,15:318-323.
142. Cheung GW, Lau RS. Testing mediation and suppression effects of latent
variables: Bootstrapping with structural equation models. Organizational
Research Methods 2007.
143. MacKinnon DP, Fairchild AJ, Firtz MS. Mediation analysis. Annual Review of
Psychology 2008,58:593-641.
144. Paulhus DL, Robins RW, Trzesniewski KH, Tracy JL. Two replicable suppressor
situations in personality research. Multivariate Behavioral Research
2004,39:303-328.
145. Rose RA, Parish SL, Yoo J, Grady MD, Powell SE, Hicks-Sangster TK.
Suppression of racial disparities for children with special health care needs
among families receiving Medicaid. Social Science & Medicine
2010,70:1263-1270.
146. Bentley JP. An examination of statistical methods for longitudinal mediation
modeling: The University of Alabama at Birmingham; 2011.
147. MacKinnon DP. Introduction to statistical mediation analysis. New York,
London: Routledge; 2008.
148. Roth DL, MacKinnon DP. Mediation analysis with longitudinal data. In:
Longitudinal data analysis: A practical guide for researchers in aging, health,
and social sciences. Edited by Newsom JT, Jones RN, Hofer SM. London:
Routledge; 2012. pp. 181-216.
149. Altman M, Holland JC, Lundeen D, Kolko RP, Stein RI, Saelens BE, et al.
Reduction in Food Away from Home Is Associated with Improved Child
Relative Weight and Body Composition Outcomes and This Relation Is
Mediated by Changes in Diet Quality. Journal of the Academy of Nutrition and
Dietetics 2015,115:1400-1407.
150. Harte CB, Meston CM. Association between cigarette smoking and erectile
tumescence: the mediating role of heart rate variability. International Journal
of Impotence Research 2013,25:155-159.
285
151. Lydersen S. Statistical review: frequently given comments. Annals of the
Rheumatic Diseases 2015,74:323-325.
152. Twisk JW. Applied longitudinal data analysis for epidemiology: a practical
guide: Cambridge University Press; 2013.
153. Bennett JA. Mediator and moderator variables in nursing research:
Conceptual and statistical differences. Research in Nursing & Health
2000,23:415-420.
154. Rogol AD, Clark PA, Roemmich JN. Growth and pubertal development in
children and adolescents: effects of diet and physical activity. The American
journal of clinical nutrition 2000,72:521s-528s.
155. Sanders R. Adolescent psychosocial, social, and cognitive development.
Pediatrics in Review/American Academy of Pediatrics 2013,34:354-358; quiz
358-359.
156. Kessler RC, Avenevoli S, Merikangas KR. Mood disorders in children and
adolescents: an epidemiologic perspective. Biological psychiatry
2001,49:1002-1014.
157. Sampasa-Kanyinga H, Chaput J-P, Hamilton HA. Associations between the use
of social networking sites and unhealthy eating behaviours and excess body
weight in adolescents. British Journal of Nutrition 2015,114:1941-1947.
158. Sampasa-Kanyinga H, Lewis RF. Frequent use of social networking sites is
associated with poor psychological functioning among children and
adolescents. Cyberpsychology, Behavior, and Social Networking
2015,18:380-385.
159. Aboujaoude E. Problematic Internet use: an overview. World Psychiatry
2010,9:85-90.
160. Fu KW, Chan WS, Wong PW, Yip PS. Internet addiction: prevalence,
discriminant validity and correlates among adolescents in Hong Kong. British
Journal of Psychiatry 2010,196:486-492.
161. Li Y, Zhang X, Lu F, Zhang Q, Wang Y. Internet addiction among elementary
and middle school students in china: a nationally representative sample study.
162. Pereira SMP, Geoffroy MC, Power C. Depressive symptoms and physical
activity during 3 decades in adult life: bidirectional associations in a
prospective cohort study. JAMA Psychiatry 2014,71:1373-1380.
163. Verkuijl NE, Richter L, Norris SA, Stein A, Avan B, Ramchandani PG. Postnatal
depressive symptoms and child psychological development at 10 years: a
prospective study of longitudinal data from the South African Birth to Twenty
cohort. The Lancet Psychiatry 2014,1:454-460.
164. Wu HR, Liu P, Meng H. Norm, Reliability and validity if childern and
Adolescents' QOL scale. Chinese Journal of School Health 2006,27:18-21.
165. Chen JY, Fang YH, Jiang Y. Investigation of children and adolescent QoL in
Guangzhou. Chinese Journal of School Health 2009,30:647-648.
166. Zimet GD, Dahlem NW, Zimet SG, Farley GK. The multidimensional scale of
perceived social support. Journal of Personality Assessment 1988,52:30-41.
167. Chou KL. Assessing Chinese adolescents’ social support: the multidimensional
scale of perceived social support. Personality and Individual Differences
286
2000,28:299-307.
168. Fang XY, Dong Q. Parent-child conflicts during early adolescence. Psychology
Research 1998,21:122-125.
169. Fang XY, Zhang JT, Sun L, Liu Z. Parent-adolescent conflict and adolescents'
social adjustment. Chinese Journal of Applied Psychology 2003,9:14-21.
170. Baron RA, Markman GD. Beyond social capital: The role of entrepreneurs'
social competence in their financial success. Journal of Business Venturing
2003,18:41-60.
171. Zhuang MK, Gan YQ, Liu HH. The Revision and Preliminary Application of
Social Skill Scale. Chinese Mental Health Journal 2004,18:755-759.
172. Rosenberg M, Schooler C, Schoenbach C, Rosenberg F. Global self-esteem and
specific self-esteem: Different concepts, different outcomes. American
Sociological Review 1995,60:141-156.
173. Kong F, Zhao J, You X. Emotional intelligence and life satisfaction in Chinese
university students: The mediating role of self-esteem and social support.
Personality and Individual Differences 2012,53:1039-1043.
174. Radloff LS. The CES-D scale a self-report depression scale for research in the
general population. Applied Psychological Measurement 1977,1:385-401.
175. Radloff LS. The use of the Center for Epidemiologic Studies Depression Scale
in adolescents and young adults. Journal of youth and adolescence
1991,20:149-166.
176. Wang M, Armour C, Wu Y, Ren F, Zhu X, Yao S. Factor Structure of the CES‐D
and Measurement Invariance Across Gender in Mainland Chinese Adolescents.
Journal of Clinical Psychology 2013,69:966-979.
177. Lee SW, Stewart SM, Byrne BM, Wong JP, Ho S, Lee PW, et al. Factor structure
of the Center for Epidemiological Studies Depression scale in Hong Kong
adolescents. Journal of Personality Assessment 2008,90:175-184.
178. Chen ZY, Yang XD, Li XY. Psychometric feartures of CES-D in chinese
adolescents. Chinese Journal of Clinical Psychology 2009,17:443-448.
179. Cheng CP, Yen CF, Ko CH, Yen JY. Factor structure of the center for
epidemiologic studies depression scale in Taiwanese adolescents.
Comprehensive Psychiatry 2012,53:299-307.
180. Stockings E, Degenhardt L, Lee YY, Mihalopoulos C, Liu A, Hobbs M, et al.
Symptom screening scales for detecting major depressive disorder in children
and adolescents: A systematic review and meta-analysis of reliability, validity
and diagnostic utility. Journal of Affective Disorders 2015,174:447-463.
181. Mukaka M. A guide to appropriate use of Correlation coefficient in medical
research. Malawi Medical Journal 2012,24:69-71.
182. Wang H, Zhou X, Lu C, Wu J, Deng X, Hong L. Problematic Internet Use in high
school students in Guangdong Province, China. PLoS ONE 2011,6:e19660.
183. Bener A, Bhugra D. Lifestyle and depressive risk factors associated with
problematic internet use in adolescents in an Arabian Gulf culture. Journal of
Addiction Medicine 2013,7:236-242.
184. Mak KK, Lai CM, Watanabe H, Kim D-I, Bahar N, Ramos M, et al. Epidemiology
of Internet behaviors and addiction among adolescents in six Asian countries.
287
185. Cao H, Sun Y, Wan Y, Hao J, Tao F. Problematic Internet use in Chinese
adolescents and its relation to psychosomatic symptoms and life satisfaction.
BMC Public Health 2011,11:802-810.
186. Wang HJ, Lu JM. The Compilation of the Middle School Students' Self-control
Ability Questionnaire. Psychological Science 2004,27:1477-1482.
187. Mann RE. Availability as a law of addiction. Addiction 2005,100:924-925.
188. Carli V, Durkee T, Wasserman D, Hadlaczky G, Despalins R, Kramarz E, et al.
The association between pathological internet use and comorbid
psychopathology: a systematic review. Psychopathology 2013,46:1-13.
189. Thompson SH, Lougheed E. Frazzled by Facebook? An exploratory study of
gender differences in social network communication among undergraduate
men and women. College Student Journal 2012:88-98.
190. Mikami AY, Szwedo DE. Social Networking in Online and Offline Contexts. In:
Encyclopedia of Adolescence. Edited by Levesque RJR. New York: Springer;
2014. pp. 2801-2808.
191. Weiser EB. Gender differences in Internet use patterns and Internet
application preferences: A two-sample comparison. CyberPsychology and
Behavior 2000,3:167-178.
192. Joiner R, Gavin J, Duffield J, Brosnan M, Crook C, Durndell A, et al. Gender,
Internet identification, and Internet anxiety: Correlates of Internet use.
193. Wang X, Sun Y, An J, Hao J, Tao F. Gender difference on depressive symptoms
among Chinese children and adolescents. Zhong Hua Liu Xing Bing Xue Za Zhi
2013,34:893-896.
194. Fleming TM, Clark T, Denny S, Bullen P, Crengle S, Peiris-John R, et al. Stability
and change in the mental health of New Zealand secondary school students
2007–2012: Results from the national adolescent health surveys. Australian
and New Zealand Journal of Psychiatry 2014,48:472-480.
195. Pietsch K, Allgaier AK, Frühe B, Hoyler A, Rohde S, Schulte‐Körne G.
Screening for adolescent depression in paediatric care: validity of a new brief
version of the Center for Epidemiological Studies Depression Scale. Child and
Adolescent Mental Health 2013,18:76-81.
196. Kaltiala-Heino R, Rimpelä M, Marttunen M, Rimpelä A, Rantanen P. Bullying,
depression, and suicidal ideation in Finnish adolescents: school survey. BMJ
1999,319:348-351.
197. Ha JH, Kim SY, Bae SC, Bae S, Kim H, Sim M, et al. Depression and Internet
addiction in adolescents. Psychopathology 2006,40:424-430.
198. Liu TC, Desai RA, Krishnan-Sarin S, Cavallo DA, Potenza MN. Problematic
Internet use and health in adolescents: data from a high school survey in
Connecticut. The Journal of Clinical Psychiatry 2011,72:836-845.
199. Yen CF, Ko CH, Yen JY, Chang YP, Cheng CP. Multi‐dimensional discriminative
factors for Internet addiction among adolescents regarding gender and age.
Psychiatry and Clinical Neurosciences 2009,63:357-364.
200. Cheung LM, Wong WS. The effects of insomnia and internet addiction on
depression in Hong Kong Chinese adolescents: an exploratory cross‐
sectional analysis. Journal of Sleep Research 2011,20:311-317.
288
201. Şaşmaz T, Öner S, Kurt AÖ, Yapıcı G, Buğdaycı R, Şiş M. Prevalence and risk
factors of Internet addiction in high school students. The European Journal of
Public Health 2013:1-6.
202. Lam LT, Peng ZW, Mai JC, Jing J. Factors associated with Internet addiction
among adolescents. CyberPsychology & Behavior 2009,12:551-555.
203. Casale S, Fioravanti G. Psychosocial correlates of internet use among Italian
students. International Journal of Psychology 2011,46:288-298.
204. Selfhout MH, Branje SJ, Delsing M, Ter Bogt TF, Meeus WH. Different types of
Internet use, depression, and social anxiety: the role of perceived friendship
quality. Journal of Adolescence 2009,32:819-833.
205. Maes L, Lievens J. Can the school make a difference? A multilevel analysis of
adolescent risk and health behaviour. Social Science & Medicine
2003,56:517-529.
206. Hinkley T, Verbestel V, Ahrens W, Lissner L, Molnár D, Moreno LA, et al. Early
childhood electronic media use as a predictor of poorer well-being: a
prospective cohort study. JAMA Pediatrics 2014,168:485-492.
207. Brand M, Laier C, Young KS. Internet addiction: coping styles, expectancies,
and treatment implications. Frontiers in Psychology 2014,5:1256.
208. Caplan SE. Preference for online social interaction a theory of problematic
Internet use and psychosocial well-being. Communication Research
2003,30:625-648.
209. Young KS, Rogers RC. The relationship between depression and Internet
addiction. CyberPsychology & Behavior 1998,1:25-28.
210. Primack BA, Swanier B, Georgiopoulos AM, Land SR, Fine MJ. Association
between media use in adolescence and depression in young adulthood: a
longitudinal study. Archives of General Psychiatry 2009,66:181-188.
211. Gámez-Guadix M, Villa-George FI, Calvete E. Measurement and analysis of the
cognitive-behavioral model of generalized problematic Internet use among
Mexican adolescents. Journal of Adolescence 2012,35:1581-1591.
212. Yu LW. An investigation into the problematic use of Facebook: Hong Kong
Baptist University Hong Kong; 2011.
213. Caplan SE, High AC. Online Social Interaction, Psychosocial Well‐Being, and
Problematic Internet Use. In: Internet addiction: A handbook and guide to
evaluation and treatment. Edited by Young KS, De Abreu CN. Hoboken, NJ,
USA: John Wiley & Sons, Inc.; 2011. pp. 35-53.
214. Muñoz-Rivas MJ, Fernández L, Gámez-Guadix M. Analysis of the indicators of
pathological Internet use in Spanish university students. The Spanish Journal
of Psychology 2010,13:697-707.
215. Chou WP, Ko CH, Kaufman EA, Crowell SE, Hsiao RC, Wang PW, et al.
Association of stress coping strategies with Internet addiction in college
students: The moderating effect of depression. Comprehensive Psychiatry
2015,62:27-33.
216. Gelenberg AJ, Freeman MP, Markowitz JC, Rosenbaum JF, Thase ME, Trivedi
MH, et al. Practice guideline for the treatment of patients with major
depressive disorder third edition. The American Journal of Psychiatry
2010,167:1-3,9-11,13-118.
289
217. Medenica S, Račić M, Joksimović V. Internet and computer addiction:“new
age” disease of the 21st century. Биомедицинска истраживања 2015,6.
218. Tsitsika AK, Tzavela EC, Janikian M, Ólafsson K, Iordache A, Schoenmakers TM,
et al. Online Social Networking in Adolescence: Patterns of Use in Six
European Countries and Links With Psychosocial Functioning. Journal of
Adolescent Health 2014,55:141-147.
219. Valkenburg PM, Peter J. Preadolescents' and adolescents' online
communication and their closeness to friends. Developmental Psychology
2007c,43:267-277.
220. Valkenburg PM, Peter J. Online Communication and Adolescent Well‐Being:
Testing the Stimulation Versus the Displacement Hypothesis. Journal of
Computer‐Mediated Communication 2007b,12:1169-1182.
221. Chan M, Wu X, Hao Y, Xi R, Jin T. Microblogging, Online Expression, and
Political Efficacy Among Young Chinese Citizens: The Moderating Role of
Information and Entertainment Needs in the Use of Weibo. Cyberpsychology,
222. Terwee CB, Bot SD, de Boer MR, van der Windt DA, Knol DL, Dekker J, et al.
Quality criteria were proposed for measurement properties of health status
questionnaires. Journal of Clinical Epidemiology 2007,60:34-42.
223. Gil-Monte PR, Carlotto MS, Câmara SG. Validation of the Brazilian version of
the" Spanish Burnout Inventory" in teachers. Revista de Saúde Pública
2010,44:140-147.
224. Beauducel A, Wittmann WW. Simulation study on fit indexes in CFA based on
data with slightly distorted simple structure. Structural Equation Modeling
2005,12:41-75.
225. Anderson JC, Gerbing DW. The effect of sampling error on convergence,
improper solutions, and goodness-of-fit indices for maximum likelihood
confirmatory factor analysis. Psychometrika 1984,49:155-173.
226. Floyd FJ, Widaman KF. Factor analysis in the development and refinement of
clinical assessment instruments. Psychological Assessment 1995,7:286-299.
227. Vergeer M, Pelzer B. Consequences of media and Internet use for offline and
online network capital and well‐being. A causal model approach. Journal of
Computer‐Mediated Communication 2009,15:189-210.
228. Teo T. Modelling Facebook usage among university students in Thailand: the
role of emotional attachment in an extended technology acceptance model.
Interactive Learning Environments 2014:1-13.
229. Morgan C, Cotten SR. The relationship between Internet activities and
depressive symptoms in a sample of college freshmen. CyberPsychology &
Behavior 2003,6:133-142.
230. Van den Eijnden R, Meerkerk G, Vermulst AA, Spijkerman R, Engels R. Online
communication, compulsive Internet use, and psychosocial well-being among
adolescents: a longitudinal study. Developmental Psychology
2008,44:655-665.
231. Seto E. Associations between Self-Reported Narcissism, Self-Esteem, and
Social-Emotional Functions of Facebook: Baylor University; 2012.
232. Clayton R. Loneliness, Anxiousness, Alcohol, and Marijuana Use: Examining
290
the Predictors of Facebook Connections and Emotional Connectedness to
Facebook Amongst College Freshmen: Texas State University-San Marcos;
2012.
233. Wisniewski P, Xu H, Lipford H, Bello‐Ogunu E. Facebook apps and tagging:
The trade‐off between personal privacy and engaging with friends. Journal
of the Association for Information Science and Technology
2015,66:1883-1896.
234. Olson DA, Liu J, Shultz KS. The influence of Facebook usage on perceptions of
social support, personal efficacy, and life satisfaction. Journal of
Organizational Psychology 2012,12:133-144.
235. O'Dea B, Campbell A. Online social networking amongst teens: friend or foe?
Studies in Health Technology and Informatics 2011,167:133-138.
236. Kim Y, Sohn D, Choi SM. Cultural difference in motivations for using social
network sites: A comparative study of American and Korean college students.
237. Tosun LP. Motives for Facebook use and expressing “true self” on the Internet.
238. Lee JS, Jeong B. Having mentors and campus social networks moderates the
impact of worries and video gaming on depressive symptoms: a moderated
mediation analysis. BMC Public Health 2014,14:426.
239. Field T, Diego M, Sanders C. Adolescent depression and risk factors.
Adolescence 2001,36:491-498.
240. Wilkinson RB. Best friend attachment versus peer attachment in the
prediction of adolescent psychological adjustment. Journal of Adolescence
2010,33:709-717.
241. He F, Zhou Q, Li J, Cao R, Guan H. Effect of social support on depression of
internet addicts and the mediating role of loneliness. International Journal of
Mental Health Systems 2014,8:34.
242. Smith NR, Clark C, Smuk M, Cummins S, Stansfeld SA. The influence of social
support on ethnic differences in well-being and depression in adolescents:
findings from the prospective Olympic Regeneration in East London (ORiEL)
study. Social Psychiatry and Psychiatric Epidemiology 2015,50:1701-1711.
243. Hampton K, Goulet LS, Rainie L, Purcell K. Social networking sites and our lives
(Pew Internet & American Life Project). PewResearchCenter Internet, Science
& Tech 2011:1-85.
244. Vitak J, Ellison NB, Steinfield C. The ties that bond: Re-examining the
relationship between Facebook use and bonding social capital. In: System
Sciences (HICSS), 2011 44th Hawaii International Conference on: IEEE; 2011.
pp. 1-10.
245. O'Shea G, Spence SH, Donovan CL. Interpersonal factors associated with
depression in adolescents: are these consistent with theories underpinning
interpersonal psychotherapy? Clinical Psychology & Psychotherapy
2014,21:548-558.
246. Hu J, Ai H. Self-esteem mediates the effect of the parent–adolescent
relationship on depression. Journal of Health Psychology 2014:1-8.
247. Briones RL, Janoske M. How American students perceive social networking
291
sites: An application of uses and gratifications theory. In: Social Networking
and Education. Edited by Issa T, Isaias P, Kommers P: Springer; 2016. pp.
239-253.
248. Betts LR, Stiller J. Centrality in children's best friend networks: The role of
social behaviour. British Journal of Developmental Psychology 2014,32:34-49.
249. Viikari-Juntura E, Vuori J, Silverstein B, Kalimo R, Kuosma E, Videman T. A
life-long prospective study on the role of psychosocial factors in
neck-shoulder and low-back pain. Spine 1991,16:1056-1061.
250. Horne J, Lincoln NB, Preston J, Logan P. What does confidence mean to people
who have had a stroke?–A qualitative interview study. Clinical Rehabilitation
2014,28:1125-1135.
251. Griffin KW, Nichols TR, Birnbaum AS, Botvin GJ. Social competence among
urban minority youth entering middle school: Relationships with alcohol use
and antisocial behaviors. International Journal of Adolescent Medicine and
Health 2006,18:97-106.
252. Scott SM, Wallander JL, Cameron L. Protective Mechanisms for Depression
among Racial/Ethnic Minority Youth: Empirical Findings, Issues, and
Recommendations. Clinical Child and Family Psychology Review
2015,18:346-369.
253. Brage D, Meredith W. A causal model of adolescent depression. The Journal of
Psychology 1994,128:455-468.
254. Ames ME, Rawana JS, Gentile P, Morgan AS. The protective role of optimism
and self-esteem on depressive symptom pathways among Canadian
Aboriginal youth. Journal of Youth and Adolescence 2013,44:142-154.
255. Duval S, Wicklund RA. A theory of objective self awareness. Oxford, England:
Academic Press; 1972.
256. Lckes WJ, Wicklund RA, Ferris CB. Objective self awareness and self esteem.
Journal of Experimental Social Psychology 1973,9:202-219.
257. Teng MS, Hsu LA, Wu S, Sun YC, Juan SH, Ko YL. Association of CDH13
Genotypes/Haplotypes with Circulating Adiponectin Levels, Metabolic
Syndrome, and Related Metabolic Phenotypes: The Role of the Suppression
Effect. PloS ONE 2015,10.
258. Wen Z, Fan X. Monotonicity of effect sizes: Questioning kappa-squared as
mediation effect size measure. Psychological Methods 2015,20:193-203.
259. Bessiere K, Kiesler S, Kraut R, Boneva BS. Effects of Internet use and social
resources on changes in depression. Information, Community & Society
2008,11:47-70.
260. Merton RK. The Matthew effect in science. Science 1968,159:56-63.
261. DiPrete TA, Eirich GM. Cumulative advantage as a mechanism for inequality: A
review of theoretical and empirical developments. Annual Review of
Sociology 2006:271-297.
262. Bast J, Reitsma P. Analyzing the development of individual differences in
terms of Matthew effects in reading: Results from a Dutch longitudinal study.
Developmental Psychology 1998,34:1373-1399.
263. McNamara JK, Scissons M, Gutknecth N. A longitudinal study of kindergarten
children at risk for reading disabilities: The poor really are getting poorer.
292
Journal of Learning Disabilities 2011:1-10.
264. Wade RH. The causes of increasing world poverty and inequality; or, why the
Matthew effect prevails. International Journal of Health Services
2005,35:631-653.
265. Meland E, Grønhaug S, Øystese K, Mildestvedt T. Examining the Matthew
effect on the motivation and ability to stay at work after heart disease.
Scandinavian Journal of Public Health 2011:1-8.
266. Cohen S, Wills TA. Stress, social support, and the buffering hypothesis.
Psychological Bulletin 1985,98:310-357.
267. Rothon C, Head J, Klineberg E, Stansfeld S. Can social support protect bullied
adolescents from adverse outcomes? A prospective study on the effects of
bullying on the educational achievement and mental health of adolescents at
secondary schools in East London. Journal of Adolescence 2011,34:579-588.
268. Bakker MP, Ormel J, Verhulst FC, Oldehinkel AJ. Adolescent family adversity
and mental health problems: the role of adaptive self-regulation capacities.
The TRAILS study. Journal of Abnormal Child Psychology 2011,39:341-350.
269. Ene M, Smiley W, Bell BA. MIXED_FIT: A SAS® macro to assess model fit and
adequacy for two-level linear models. In: SAS Glob Forum Pap: Citeseer; 2013.
pp. 255-2013.
270. Tidwell LC, Walther JB. Computer‐mediated communication effects on
disclosure, impressions, and interpersonal evaluations: Getting to know one
another a bit at a time. Human Communication Research 2002,28:317-348.
271. Mesch GS. The Family and the Internet: The Israeli Case*. Social Science
Quarterly 2003,84:1038-1050.
272. Collins WA, Russell G. Mother-child and father-child relationships in middle
childhood and adolescence: A developmental analysis. Developmental Review
1991,11:99-136.
273. Wierson M, Armistead L, Forehand R, Thomas AM, Fauber R.
Parent-adolescent conflict and stress as a parent: Are there differences
between being a mother or a father? Journal of Family Violence
1990,5:187-197.
274. Krishnan UD. Parent-adolescent conflict and adolescent functioning in a
collectivist, ethnically heterogenous culture: Malaysia. The United States: The
Ohio State University; 2004:123.
275. Zhao HY, Xu Y, Wang F, Jiang J, Zhang XH, Wang XR. Chinese adolescents'
coping tactics in a parent-adolescent conflict and their relationships with life
satisfaction: the differences between coping with mother and father. Frontiers
in Psychology 2015,6:1572.
276. Ho TY. Functional similarities and differences between online and offline
social support on psychological well-being. Pokfulam, Hong Kong: The
University of Hong Kong; 2014.
277. Kuss DJ, van Rooij AJ, Shorter GW, Griffiths MD, van de Mheen D. Internet
addiction in adolescents: Prevalence and risk factors. Computers in Human
Behavior 2013,29:1987-1996.
278. Kormas G, Critselis E, Janikian M, Kafetzis D, Tsitsika A. Risk factors and
psychosocial characteristics of potential problematic and problematic internet
293
use among adolescents: a cross-sectional study. BMC Public Health
2011,11:595-602.
279. Yakubu A. Path coefficient and path analysis of body weight and biometric
traits in Yankasa lambs. Slovak Journal of Animal Science 2010,43:17-25.
280. Grapentine T. Path analysis vs. structural equation modeling. Marketing
Research 2000,12:12-19.
281. Chen W, Srinivasan SR, Berenson GS. Path analysis of metabolic syndrome
components in black versus white children, adolescents, and adults: the
Bogalusa Heart Study. Annals of Epidemiology 2008,18:85-91.
282. MacKinnon DP. Contrasts in multiple mediator models. In: Multivariate
applications in substance use research: New methods for new questions.
Edited by Rose JS, Chassin L, Presson CC, Sherman SJ. New York London:
Psychology Press; 2000. pp. 141-160.
283. Preacher KJ, Hayes AF. Asymptotic and resampling strategies for assessing and
comparing indirect effects in multiple mediator models. Behavior Research
Methods 2008,40:879-891.
284. Tabri N, Elliott CM. Principles and practice of structural equation modeling.
Canadian Graduate Journal of Sociology and Criminology 2012,1:59-60.
285. Grover G, Sabharwal A, Mittal J. Estimation of Survival Function and
Probability of Onset of Diabetic Nephropathy using Path Analysis and Analysis
of Covariance. International Refereed Journal of Engineering and Science
(IRJES)(India) 2014,3:50-58.
286. Kim K, Ryu E, Chon MY, Yeun EJ, Choi SY, Seo JS, et al. Internet addiction in
Korean adolescents and its relation to depression and suicidal ideation: a
questionnaire survey. International Journal of Nursing Studies
2006,43:185-192.
287. Young K. Internet addiction: diagnosis and treatment considerations. Journal
of Contemporary Psychotherapy 2009,39:241-246.
288. Bramesfeld A, Platt L, Schwartz FW. Possibilities for intervention in
adolescents’ and young adults’ depression from a public health perspective.
Health Policy 2006,79:121-131.
289. Drake RE. Dual diagnosis. Psychiatry 2007,6:381-384.
290. Rothman KJ, Greenland S. Causation and causal inference in epidemiology.
American journal of public health 2005,95:S144-S150.
291. Wang L, Luo J, Bai Y, Kong J, Luo J, Gao W, et al. Internet addiction of
adolescents in China: Prevalence, predictors, and association with well-being.
Addiction Research & Theory 2013,21:62-69.
292. Moreno MA, Jelenchick LA, Egan KG, Cox E, Young H, Gannon KE, et al. Feeling
bad on Facebook: depression disclosures by college students on a social
networking site. Depress Anxiety 2011,28:447-455.
293. Kim J, Lee JER. The Facebook paths to happiness: Effects of the number of
Facebook friends and self-presentation on subjective well-being.
294. Sampasa-Kanyinga H, Hamilton HA. Use of social networking sites and risk of
cyberbullying victimization: a population-level study of adolescents.
294
295. Liu C-Y, Yu C-P. Can Facebook use induce well-being? Cyberpsychology,
295
Appendix I
Questionnaire for phase I validation study (Chinese version)
初中生社交网络使用强度调查问卷
同学们，你们好。这是一份调查问卷，目的是了解初中生社交网络的使用情况、及其对
心理健康的影响。研究结果将有助于促进我国青少年身心的健康发展。现诚挚邀请你参加本
研究，并如实填写此调查问卷。本调查采取匿名方式填写，你无需将自己的姓名写在问卷上，
同时你们提供的宝贵信息将仅用于科学研究，个人资料绝对保密，不会告知老师、家长、同
学等人，保证不会外泄。
请注意：问卷中涉及的问题，答案并无对错之分，你只需根据个人的看法或感受在相应
选项中打“√”。在回答问卷的过程中，如有不清楚的地方，请举手询问我们的研究人员，
他们会尽力给予解释。你们的参与对我们的研究十分重要，再次感谢你们的参与！
请首先填写下列信息，以下信息将用于问卷匹配，请务必准确填写：
1.家庭住宅电话最后四位数字：X X X
2.你的身份证号最后四位数字：X X X
3.你出生于：_____年_____月_____日
4.你姓名拼音的首字母：___________ (如“张三”，请填: ZS)
第一部分、基本资料
1.1 你的性别： □1男 □2女

1.2 你现在的年级： □ 1 初一 □ 2 初二 □ 3 初三
1.3 父亲的文化程度：
□ 1 小学及以下 □ 2 初中 □ 3 高中/中专 □ 4 本科(或大专)及以上
1.4 母亲的文化程度：
□ 1 小学及以下 □ 2 初中 □ 3 高中/中专 □ 4 本科(或大专)及以上
1.5 你是否拥有自己的智能手机? □ 1 是, 用了多久：_____年____月 □2否
第二部分、网络成瘾
2. 请仔细阅读下面每一句话，根据自己的实际情况进行选择是否
1). 你是否沉溺于互联网(总想着以前上网的经历或是期待着下一次的上网) 1 2
2). 你是否需要通过增加上网时间以获得满足感 1 2
3). 你是否经常不能抵制上网的诱惑或很难下网 1 2
4). 停止使用互联网时你是否会产生消极的情绪体验和不良的生理反应 1 2
5). 每次上网实际所花的时间是否都比原定时间要长 1 2
6). 上网是否已经对你的人际关系、学习造成负面影响 1 2
7). 你是否对家人、朋友或心理咨询人员隐瞒了上网的真实时间和费用 1 2
8). 你是否将上网作为逃避问题和排遣消极情绪的一种方式 1 2
296
第三部分、社交网络使用情况
【社交网络】是指QQ空间、人人网、朋友网、微博、开心网等可以建立自己社交页面
的网络服务，能够为人们提供交流互动的平台，其中不包括QQ等即时通讯工具。请根据你
使用社交网络的实际情况回答。
3.1 你是否拥有自己的社交网络账户？ □ 1 是 □ 2 否(选“否”，请跳至“第四部分”)

a) 你主要通过什么设备登陆这些社交网络 (选最主要的一项) ？
□ 1 电脑（台式/笔记本） □ 2 平板电脑(如iPad等)
□ 3 智能手机 □ 4 其他，_________
b) 过去 1 年，你使用频率较高的社交网络有（最多选三项）
□ 1 QQ空间 □ 2 人人网(原校内网) □ 3 朋友网 □ 4 开心网
□ 5 微信 □ 6 微博(如新浪/腾讯等) □ 7 其他, 请说明___________
请根据正在使用的所有社交网络情况(包括各种社交网络的使用情况总和)回答下述问题。
3.2 你使用这些社交网络的时间有多久？（按使用时间最长的回答）
□ 1 <3 月 □ 2 3-6 月 □ 3 6-12 月 □ 4 1-2 年 □ 5 2-4 年 □ 6 >4 年
3.3 过去 1 个月，你平均每周有多少天用过社交网络？
□ 1 几乎没有用过 □2 1 天 □32 天 □43 天
□54 天 □6 5 天 □7 6 天 □87 天
3.4 过去 1 个月有使用社交网络的日子里，你平均每天花多少时间在社交网络上？
□ 1 <10 分钟 □ 2 10-30 分钟 □ 3 31-60 分钟
□ 4 61-90 分钟 □ 5 >90 分钟
3.5 在这些正在使用的社交网络上，你大约总共有多少个不同的朋友？
□ 1 ≤10 □ 2 11-50 □ 3 51-100 □ 4 101-150
□ 5 151-200 □ 6 201-250 □ 7 251-300 □ 8 301-400 □ 9 ≥400
3.6 回想你【最近 1 个月】使用社交网络（如QQ空间，人人网，朋友网等）的情况，选择

合适的选项。
最近 1 月的使用频率
【最近 1 个月】
从不较少有时经常总是
1). 在留言板上给好友留言 1 2 3 4 5
2). 通过即时通讯功能与朋友聊天 1 2 3 4 5
3). 回复朋友对自己社交网络内容的评论（如头
1 2 3 4 5
像、照片、状态等）
4). 评论朋友的动态、日志、照片等 1 2 3 4 5
5). 分享、转帖 1 2 3 4 5
6). 浏览其他人的日志、照片、动态、相册等 1 2 3 4 5
7). 浏览信息(如娱乐/时事新闻等) 1 2 3 4 5
8). 更新自己的状态 1 2 3 4 5
9). 发照片到自己的社交页面 1 2 3 4 5
10). 写日志、发微博 1 2 3 4 5
11). 编辑自己的社交页面（如更换头像、修改 1 2 3 4 5
297
联系方式、隐私设置等）
12). 看视频或听音乐等 1 2 3 4 5
13). 玩游戏或应用等 1 2 3 4 5
14). 购买/赠送虚拟物品 1 2 3 4 5
15). 其他活动，请说明：_____________ 1 2 3 4 5
3.7 请仔细阅读下面每句话，根据自己使用社交网络的经历（或感受）
，选择最能够代表你
自己想法的选项。
完全不同完全
不同意一般
同意意同意
1). 社交网络是我每天活动的一部分 1 2 3 4 5
2). 我很骄傲的告诉他人我在使用社交网络 1 2 3 4 5
3). 社交网络已经成为我每天例行项目的一部分 1 2 3 4 5
4). 当一段时间不登录社交网络，我就会感觉与外
1 2 3 4 5
界失去了联系
5). 我感觉我是社交网络群体的一部分 1 2 3 4 5
6). 如果关闭社交网络，我会感到难过 1 2 3 4 5
3.8 请选择一个合适的选项，来代表你【最近 1 个月】使用社交网络的情况。

极不不符一符非常
符合合般合符合
1). 因为使用社交网络，我很难集中精力在我的学业上 1 2 3 4 5
2). 每天起床后我脑海里的第一件事就是登陆社交网络 1 2 3 4 5
3). 我花了太多的时间在社交网络上而出现失眠 1 2 3 4 5
4). 使用社交网络妨碍了我的社交活动 1 2 3 4 5
5). 当我情绪低落的时候，登陆社交网络会让我感觉好点 1 2 3 4 5
6). 我的家人/朋友认为我在社交网络上花的时间太多 1 2 3 4 5
7). 如果我不能访问社交网络，我会感到很焦虑 1 2 3 4 5
8). 我曾经尝试少花些时间在社交网络上，但是没有成功 1 2 3 4 5
298
Appendix I
Questionnaire for phase I validation study (English version)
Questionnaire for social networking use intensity among junior middle school students
Please fill the following information accurately, which is only used for questionnaire matching.
1. Last 4 digital of home-telephone number: X X X
2. Last 4 digital of your Identity Card number: X X X
3. Your birthday: ________year______month_____day
4. Initial letter of your spellname: _______(e.g. If your names is “Zhang San”, please fill: ZS)
Part I: Social-demographic background
1.1. Your gender: □ 1 Male □ 2 Female

1.2. Your current grade: □ 1 Seven □ 2 Eight □ 3 Ninth
1.3. Father education level:
□ 1 Primary school or below □ 2 Junior high school
□ 3 Senior high school (or equal academic qualification) □ 4 University or above
1.4. Mother education level:
□ 1 Primary school or below □ 2 Junior high school
□ 3 Senior high school (or equal academic qualification) □ 4 University or above
1.5. Do you possess a smartphone?
□ 1 Yes, Duration: _____year____month □ 2 No
Part II: Internet addiction

2. Do you agree with the following statement for yourself? Yes No
1) Do you feel preoccupied with the Internet (think about previous on-line
1 2
activity or anticipate next on-line session)?
2) Do you feel the need to use the Internet with increasing amounts of time in
1 2
order to achieve satisfaction?
3) Have you repeatedly made unsuccessful efforts to control, cut back, or stop
1 2
Internet use?
4) Do you feel restless, moody, depressed, or irritable when attempting to cut
1 2
down or stop Internet use
5) Do you stay on-line longer than originally intended? 1 2
6) Have you jeopardized or risked the loss of significant relationship,
1 2
educational opportunity because of the Internet?
7) Have you lied to family member, a therapist, or others to conceal the extent of
1 2
involvement with the Internet?
8) Do you use the Internet as a way of escaping from problem or of relieving a
1 2
dysphonic mood (e.g. feelings of helplessness, guilt, anxiety, depression)?
299
Part III: Social networking use
3.1. Do you have your own social networking accounts? □ 1 Yes □ 2 No (Please jump to Part
IV)
a) Device that you most frequently used for accessing social networking platforms
□ 1 Computer (Desk/Laptop) □ 2 Tablet (e.g. Ipad)
□ 3 Smartphone □ 4 Other, _________
b) During the past year, the first three mostly frequently used social networking services
□ 1 Qzone □ 2 RenRen/Xiaonei □ 3 Pengyou □ 4 Kaixin
□ 5 WeChat □ 6 Weibo (e.g. Sina/Tencent) □ 7 Other, _______
Please answer the following questions according to the conditions that you use online social
networking (total of all online social networking use conditions)
3.2. How long have you used these online social networking (Based on the longest one)?
□ 1 <3 months □ 2 3~6 months □ 3 6~12 months
□ 4 1~2 years □ 5 2~4 years □ 6 more than 4 years
3.3. During the past one month, on average, approximately how many days per week do you
have used online social networking?
□ 1 Almost not □ 2 1 day □ 3 2 days □ 4 3 days
□ 5 4 days □ 6 5 days □ 7 6 days □ 8 7 days
3.4. During the past one month, on average, approximately how many minutes per day have
you spent on using online social networking?
□ 1 <10 mins □ 2 10~30 mins □ 3 31~60 mins □ 4 61~90 mins □ 5 >90 mins
3.5. About how many total social networking friends do you have?
□ 1 ≤10 □ 2 11-50 □ 3 51-100 □ 4 101-150
□ 5 151-200 □ 6 201-250 □ 7 251-300 □ 8 301-400 □ 9 ≥400
3.6. During the past one month, how frequent do you do the following activities on online social
networking (e.g. Qzone, Renren)?
Frequency during the past one month
During the past one month
Never Few Sometimes Often Always
1). Sent messages to friends on message board 1 2 3 4 5
2). Chatted with friends via instant messaging
1 2 3 4 5
function
3). Replied to comments made by social networking
1 2 3 4 5
friends
4). Commented on friends’ status, logs, and photos 1 2 3 4 5
5). Shared/Forwarded contents 1 2 3 4 5
6). Browsed others’ logs/photos/status/album 1 2 3 4 5
7). Updated self-status 1 2 3 4 5
8). Posted photos/videos on personal web profile 1 2 3 4 5
9). Wrote logs/Weibo 1 2 3 4 5
10). Decorated personal web profile(changing
1 2 3 4 5
11). Surfed entertainment/current news 1 2 3 4 5
300
12). Watched video/listened to music 1 2 3 4 5
13). Played games/applications 1 2 3 4 5
14). Bought/gave virtual Goods (e.g. birthday gifts) 1 2 3 4 5
15). Others, __________ 1 2 3 4 5
3.7. Please clarify the extent to which the following statement is appropriate to you?
Strongly Strongly
Disagree Neutral Agree
disagree agree
1). Social networking is part of my
1 2 3 4 5
everyday activity
2). I am proud to tell people I’m on social
1 2 3 4 5
networking
3). Social networking has become part of
1 2 3 4 5
my daily routine
4). I feel out of touch when I haven’t
logged onto social networking for a 1 2 3 4 5
while
5). I feel I am part of the social networking
1 2 3 4 5
community
6). I would be sorry if social networking
1 2 3 4 5
shut down
3.8. Please clarify the extent to which the following statement is appropriated to you during the
past one month?
Strongly Strongly
During the past one month Disagree Neutral Agree
disagree agree
1). I have difficulties for focusing on my
academic work due to mu social 1 2 3 4 5
networking use
2). The first thing on my mind when I get up
1 2 3 4 5
is to log into social networking
3). I lose sleep over spending more time on
1 2 3 4 5
social networking
4). My social networking use interferes with
1 2 3 4 5
doing social activities
5). I log into social networking to make
1 2 3 4 5
myself feel better when I am down
6). My family or friends think that I spend
1 2 3 4 5
too much time on social networking
7). I feel anxious if I cannot access to social
1 2 3 4 5
networking
8). I have attempted to spend less time on
1 2 3 4 5
social networking but have not succeeded
301
Appendix II
Questionnaire for phase II longitudinal study (Chinese version)
初中生社交网络使用及其对社会心理健康的影响调查问卷
同学们，你们好。这是一份调查问卷，目的是了解初中生社交网络的使用情况、及其对
社会心理健康的影响。研究结果将有助于促进我国青少年身心的健康发展。现诚挚邀请你填
写一份调查问卷，并于本次调查后约 10 个月再次填写一份相似的问卷。本调查以自愿参加
为原则，采取匿名方式填写，你无需将自己的姓名写在问卷上，同时你们提供的宝贵信息将
仅用于科学研究，个人资料绝对保密，不会告知老师、家长、同学等人，保证不会外泄。
请仔细阅读问卷内容，根据条目的要求认真作答，不要遗漏。问卷中涉及的问题，选项
并无对错之分。在回答问卷的过程中，如有不清楚的地方，请举手询问我们的研究人员，他
们会尽力给予解释。如果你填写了此问卷，即视为知情同意。你的参与对我们的研究十分重
要，再次感谢你的参与！
 填写方法：在相应选项的“□”处打“√”，或在“_____”处填写
请首先填写下面的信息，以下信息仅用于问卷匹配，请务必准确填写：
1.家庭住宅电话最后四位数字：X X X
2.父亲手机号码最后四位数字：X X X
3.母亲手机号码最后四位数字：X X X
4.你的身份证号最后四位数字：X X X
5.你的生日：年月日
6.你、父亲及母亲姓名最后一个字的汉语拼音的最后一个字母：
自己:_______; 父亲: _______; 母亲: _______
(例如你叫“张小天(TIAN)”,父亲叫“张海肖(XIAO)”,母亲叫“欧晓辉(HUI)”,请依次填：N O I）
第一部分、基本情况
1.1. 你的性别： □男 □女
1.2. 你现在的年级： □初一 □初二
1.3. 父亲的文化程度：
□小学及以下 □初中 □高中/中专 □本科(或大专)及以上
1.4. 母亲的文化程度：
□小学及以下 □初中 □高中/中专 □本科(或大专)及以上
1.5. 你家里的经济状况：
□非常好 □较好 □一般 □较差 □非常差
1.6. 你在班级中的学习成绩：
□中上（班中前 1/3） □中等 □中下（班中后 1/3）
1.7. 你认为自己的学习压力如何？
□没有 □较轻 □一般 □较重 □很重
302
1.8. 你是否和父母住一起？
□与父母同住 □只与父亲同住 □只与母亲同住 □父母都不与我同住
第二部分、网络成瘾
2.请仔细阅读下面每一句话，根据自己的实际情况进行选择是否
a). 你是否沉溺于互联网(总想着以前上网的经历或是期待着下一次的上网) 1 2
b). 你是否需要通过增加上网时间以获得满足感 1 2
c). 你是否经常不能抵制上网的诱惑或很难下网 1 2
d). 停止使用互联网时你是否会产生消极的情绪体验和不良的生理反应 1 2
e). 每次上网实际所花的时间是否都比原定时间要长 1 2
f). 上网是否已经对你的人际关系、学习造成负面影响 1 2
g). 你是否对家人、朋友或心理咨询人员隐瞒了上网的真实时间和费用 1 2
h). 你是否将上网作为逃避问题和排遣消极情绪的一种方式 1 2
三、社交网络使用情况
【社交网络】是指QQ空间、人人网、朋友网、微博、开心网等可以建立自己社交页面
的网络服务，能够为人们提供交流互动的平台，其中不包括QQ等即时通讯工具。请根据你
使用社交网络的实际情况回答。
3.1. 你有自己的社交网络吗？
□有 □没有（如选“没有”，请跳至“四、社会心理等健康状况”）
3.2. 回想你【过去 1 个月】使用社交网络（如QQ空间，人人网，朋友网等）的情况，你都

在社交网络上做什么？请选择合适的选项。
最近 1 个月，你在社交网络上从不较少有时经常总是
a). 在留言板上给好友留言 1 2 3 4 5
b). 通过即时通讯功能与朋友聊天 1 2 3 4 5
c). 回复朋友对自己社交网络内容的评论（如头像、
1 2 3 4 5
照片、状态等）
d). 评论朋友的动态、日志、照片等 1 2 3 4 5
e). 分享、转帖 1 2 3 4 5
f). 浏览其他人的日志、照片、动态、相册等 1 2 3 4 5
g). 更新自己的状态 1 2 3 4 5
h). 发照片到自己的社交页面 1 2 3 4 5
i). 写日志、发微博 1 2 3 4 5
j). 编辑自己的社交页面(如更换头像、修改联系方
1 2 3 4 5
式、隐私设置等)
k). 浏览信息(如娱乐/时事新闻等) 1 2 3 4 5
l). 看视频或听音乐等 1 2 3 4 5
m). 玩游戏或应用等 1 2 3 4 5
n). 购买/赠送虚拟物品 1 2 3 4 5
303
3.3. 请仔细阅读下面每句话，根据自己使用社交网络的经历（或感受）
，选择最能够代表你
想法的选项。
完全不不同完全
一般同意
同意意同意
a). 社交网络是我每天活动的一部分 1 2 3 4 5
b). 我很骄傲的告诉他人我在使用社交网络 1 2 3 4 5
c). 社交网络已经成为我每天例行项目的一部分 1 2 3 4 5
d). 当一段时间不登录社交网络，我就会感觉与外
1 2 3 4 5
界失去了联系
e). 我感觉我是社交网络群体的一部分 1 2 3 4 5
f). 如果关闭社交网络，我会感到难过 1 2 3 4 5
3.4. 请选择一个合适的选项，来代表你【最近 1 个月】使用社交网络的情况。

极不不符一符非常
符合合般合符合
a). 因为使用社交网络，我很难集中精力在我的学业上 1 2 3 4 5
b). 每天起床后我脑海里的第一件事就是登陆社交网络 1 2 3 4 5
c). 我花了太多的时间在社交网络上而出现失眠 1 2 3 4 5
d). 使用社交网络妨碍了我的社交活动 1 2 3 4 5
e). 当我情绪低落的时候，登陆社交网络会让我感觉好点 1 2 3 4 5
f). 我的家人/朋友认为我在社交网络上花的时间太多 1 2 3 4 5
g). 如果我不能访问社交网络，我会感到很焦虑 1 2 3 4 5
h). 我曾经尝试少花些时间在社交网络上,但是没有成功 1 2 3 4 5
四、社会心理等健康状况
4.1. 请阅读下面每句话，并根据你自己的实际情况，选择一个最能够代表自己的选项。
极不很不稍不中稍同很同极同
同意同意同意立意意意
a). 当我有需要的时候，总有一个好朋友在我身边 1 2 3 4 5 6 7
b). 我有一个好朋友，无论开心或者不开心，我都可
1 2 3 4 5 6 7
以同他/她分享
c). 我的家人真的十分愿意帮助我 1 2 3 4 5 6 7
d). 我的家人可以给我心理上的支持 1 2 3 4 5 6 7
e). 我有一个真的可以安慰我的朋友 1 2 3 4 5 6 7
f). 我的朋友真的愿意帮助我 1 2 3 4 5 6 7
g). 如果有什么事发生，我可以依靠我的朋友 1 2 3 4 5 6 7
h). 我可以和家人诉说我自己的问题 1 2 3 4 5 6 7
i). 我有一些朋友，无论开心或不开心，我都可以同
1 2 3 4 5 6 7
他们分享
j). 我生命中有个好朋友，他/她会关心我的感受 1 2 3 4 5 6 7
304
k). 我的家人愿意和我一起做决定 1 2 3 4 5 6 7
l). 我可以和朋友倾诉自己的问题 1 2 3 4 5 6 7
4.2. 以下是一些用来描述个人感受的句子，请选择最能够代表你的感受的选项。
很不非常
不符合符合
符合符合
a). 我感到我是一个有价值的人，至少与其他人在同一水 1 2 3 4
平上
b). 我感到我有许多好的品质 1 2 3 4
c). 归根结底，我倾向于觉得自己是一个失败者 1 2 3 4
d). 我能像大多数人一样把事情做好 1 2 3 4
e). 我感到自己值得自豪的地方不多 1 2 3 4
f). 我对自己持肯定态度 1 2 3 4
g). 总的来说，我对自己是满意的 1 2 3 4
h). 我希望我能为自己能得更多尊重 1 2 3 4
i). 我确实时常感到毫无用处 1 2 3 4
j). 我时常认为自己一无是处 1 2 3 4
4.3. 下面是对你可能存在的或最近有过的感受的描述，请按照你【最近 1 周】出现这种感

受的实际情况，选择最适合你的答案。
沒有或偶时常或一多数时间
有时
最近 1 周尔（少于 1 半时间或持续
(1-2 天)
天） (3-4 天) (5-7 天)
1). 一些通常并不困扰我的事使我心烦 1 2 3 4
2). 我不想吃东西；我胃口不好 1 2 3 4
3). 我觉得即便有家人或朋友帮助也无
1 2 3 4
法摆脱这种苦闷
4). 我感觉同别人一样好 1 2 3 4
5). 我很难集中精力做事 1 2 3 4
6). 我感到压抑 1 2 3 4
7). 我感到做什么事都很吃力 1 2 3 4
8). 我觉得未来有希望 1 2 3 4
9). 我认为我的生活一无是处 1 2 3 4
10). 我感到恐惧 1 2 3 4
11). 我睡觉不解乏 1 2 3 4
12). 我很幸福 1 2 3 4
13). 我比平时话少了 1 2 3 4
14). 我感到孤独 1 2 3 4
15). 人们对我不友好 1 2 3 4
16). 我生活快乐 1 2 3 4
17). 我曾经放声痛哭 1 2 3 4
18). 我感到忧愁 1 2 3 4
305
19). 我觉得别人厌恶我 1 2 3 4
20). 我走路很慢 1 2 3 4
4.4. 下面几个问题是想了解你与朋友的关系，请选择与你的实际情况最相符合的选项。
1). 你觉得班上的同学对你友好吗？
□不友好 □有一点友好 □比较友好 □非常友好
2). 你的朋友关心你吗？
□不关心 □有一点关心 □比较关心 □非常关心
3). 你的朋友多吗？
□有一两个 □有三四个 □有五六个 □有很多
4). 你对自己的朋友满意吗？
□非常不满意 □有一点不满意 □比较满意 □非常满意
5). 需要帮助时，你能找到可信赖的朋友吗？
□从不这样 □很少这样 □经常这样 □总是这样
6). 你能从集体活动中得到快乐吗？
□从不这样 □很少这样 □经常这样 □总是这样
4.5. 请认真阅读下面每一句话，看自己是否这样，然后选择一个合适的选项。
完全不比较不比较完全
一般
符合符合符合符合
a). 我经常会考虑别人对我的看法 1 2 3 4 5
b). 我对于别人的批评非常敏感 1 2 3 4 5
c). 在一大群人面前演讲，我感到非常困难 1 2 3 4 5
d). 一本正经地讲笑话对我来说是件很困难
1 2 3 4 5
的事情
4.6. 过去 1 个月，你与父母在下面这些方面是否经常发生冲突（如情绪、语言或身体方面）？
请根据实际情况，选择最接近实际情况的选项，并在表格内填上相应的数字。
1=从未发生 2=每月一两次 3=约一周一次 4=一周几次 5=几乎每天都发生
父亲母亲
a). 你的学业方面（如家庭作业、成绩等）
b). 家务方面（如打扫自己的房间、帮助家人干活）
c). 花钱方面（如怎样花钱、买什么等）
d). 日常生活安排方面（如睡觉时间、饮食习惯、看电视等）
e). 外表方面（如衣着、发式等）
f). 家庭成员关系方面（如尊重、礼貌、兄弟姐妹的关系）
g). 你的隐私方面（如私拆信件、偷听电话等）
306
Appendix II
Questionnaire for phase II longitudinal study (English version)
The impact of social networking use intensity on psychosocial well-being among

junior middle school students
Please fill the following information accurately, which is only used for questionnaire matching.
1. Last 4 digital of home-telephone number: X X X
2. Last 4 digital of father mobile phone number: X X X
3. Last 4 digital of mother mobile phone number: X X X
4. Last 4 digital of your Identity Card number: X X X
5.Your Birthady: year month day
6. Last letter of the last word of your, your father’s and mother’ spellname:
Self:_______; Father: _______; Mother: _______
(e.g. Your name is ”Zhang Xiaotian (TIAN)”, your father name is ”Zhang Haixiao(XIAO)”, and your mother
name is ”Ou Xiaohui (HUI)”, please successively fill in “N O I”.
Part I: Background characteristics

1.1 Your gender: □Male □Female
1.2 Your current grade: □Seven □Eight
1.3 Father education level:
□Primary school or below □Junior high school
□Senior high school (or equal academic qualification) □University or above
1.4 Mother education level:
□Primary school or below □Junior high school
□Senior high school (or equal academic qualification) □University or above
1.5 How do you think of your family financial situation?
□very good □Good □General □Poor □very poor
1.6 Your academic performance in the class:
□Above the average (The first of one third) □Average
□Below the average (The end of one third)
1.7 How do you perceive your current study pressure?
□Nil □Light □General □Heavy □Very heavy
Part II: Internet addiction

2. Do you agree with the following statement for yourself? Yes No
a). Do you feel preoccupied with the Internet (think about previous on-line activity
1 2
or anticipate next on-line session)?
b). Do you feel the need to use the Internet with increasing amounts of time in
1 2
order to achieve satisfaction?
c). Have you repeatedly made unsuccessful efforts to control, cut back, or stop
1 2
Internet use?
d). Do you feel restless, moody, depressed, or irritable when attempting to cut 1 2
307
down or stop Internet use
e). Do you stay on-line longer than originally intended? 1 2
f). Have you jeopardized or risked the loss of significant relationship, educational
1 2
opportunity because of the Internet?
g). Have you lied to family member, a therapist, or others to conceal the extent of
1 2
involvement with the Internet?
h). Do you use the Internet as a way of escaping from problem or of relieving a
1 2
dysphonic mood (e.g. feelings of helplessness, guilt, anxiety, depression)?
Part III: Social networking use

3.1 Do you have your own social networking accounts?
□Yes □No (Please jump to Part IV “psychosocial conditions”)
3.2 During the past one month, how frequent do you do the following activities on social
networking (e.g. Qzone, Renren)?
During the past one month Never Few Sometimes Often Always
a). Sent messages to friends on message board 1 2 3 4 5
b). Chatted with friends via instant messaging function 1 2 3 4 5
c). Replied to comments made by social networking
1 2 3 4 5
friends
d). Commented on friends’ status, logs, and photos 1 2 3 4 5
e). Shared/Forwarded contents 1 2 3 4 5
f). Browsed others’ logs/photos/status/album 1 2 3 4 5
g). Updated self-status 1 2 3 4 5
h). Posted photos/videos on personal web profile 1 2 3 4 5
i). Wrote logs/weibo 1 2 3 4 5
j). Decorated personal web profile(changing
1 2 3 4 5
k). Surfed entertainment/current news 1 2 3 4 5
l). Watched video/listened to music 1 2 3 4 5
m). Played games/applications 1 2 3 4 5
n). Bought/gave virtual Goods (e.g. birthday gifts) 1 2 3 4 5
3.3 Please clarify the extent to which the following statement is appropriate to you?
Strongly Strongly
Disagree Neutral Agree
disagree agree
a). Social networking is part of my
1 2 3 4 5
everyday activity
b). I am proud to tell people I’m on social
1 2 3 4 5
networking
c). Social networking has become part of
1 2 3 4 5
my daily routine
d). I feel out of touch when I haven’t
logged onto social networking for a 1 2 3 4 5
while
e). I feel I am part of the social networking
1 2 3 4 5
community
308
f). I would be sorry if social networking
1 2 3 4 5
shut down
3.4 During the past one month, please clarify the extent to which the following statement is
appropriate to you?
Strongly Strongly
During the past one month Disagree Neutral Agree
disagree agree
a). I have difficulties for focusing on my
academic work due to mu social 1 2 3 4 5
networking use
b). The first thing on my mind when I get up
1 2 3 4 5
is to log into social networking
c). I lose sleep over spending more time on
1 2 3 4 5
social networking
d). My social networking use interferes with
1 2 3 4 5
doing social activities
e). I log into social networking to make
1 2 3 4 5
myself feel better when I am down
f). My family or friends think that I spend
1 2 3 4 5
too much time on social networking
g). I feel anxious if I cannot access to social
1 2 3 4 5
networking
h). I have attempted to spend less time on
1 2 3 4 5
social networking but have not succeeded
Part IV Psychosocial conditions

4.1 We are interested in how you feel about the following statements. Read each statement
carefully. Indicate how you feel about each statement.
Very Very
Strongly Mildly Mildly Strongly
strongly Neutral strongly
disagree disagree agree agree
disagree agree
a). There is a special person who is
1 2 3 4 5 6 7
around when I am in need
b). There is a special person with
whom I can share my joys and 1 2 3 4 5 6 7
sorrows
c). My family really tries to help me 1 2 3 4 5 6 7
d). I got the emotional help and
1 2 3 4 5 6 7
support I need from my family
e). I have a special person who is a
1 2 3 4 5 6 7
real source of comfort to me
f). My friends really try to help me 1 2 3 4 5 6 7
g). I can count on my friends when
1 2 3 4 5 6 7
things go wrong
h). I can told about my problems with
1 2 3 4 5 6 7
my family
i). I have friends with whom I can 1 2 3 4 5 6 7
309
share my joys and sorrows
j). There is a special person in my
1 2 3 4 5 6 7
life who cares about my feelings
k). My family is willing to help me
1 2 3 4 5 6 7
make decisions
l). I can talk about my problems with
1 2 3 4 5 6 7
my friends
4.2 Below is a list of statements dealing with your general feelings about yourself. Please
indicate how strongly you agree or disagree with each statement
Strongly Strongly
Disagree Agree
disagree agree
a). I feel that I am a person of worth, at least on
1 2 3 4
an equal plane with others
b). I feel that I have a number of good qualities 1 2 3 4
c). All in all, I am inclined to feel that I am a
1 2 3 4
failure
d). I am able to do things as well as most other
1 2 3 4
people
e). I feel I do not have much to be proud of 1 2 3 4
f). I take a positive attitude toward myself 1 2 3 4
g). On the whole, I am satisfied with myself 1 2 3 4
h). I wish I could have more respect for myself 1 2 3 4
i). I certainly feel useless at times 1 2 3 4
j). At times I think I am no good at all 1 2 3 4
4.3 Below is a list of some of the ways you may have felt or behaved. Please indicate how often
you’ve felt this way during the past week. Respond to all items
Rarely or Some or a Occasionally
All of the
none of the little of the or a moderate
During the past one week times (5-7
time (less time (1~2 amount of time
days)
than 1 day) days (3~4 days)
1). I am bothered by things that
1 2 3 4
usually don’t bother me
2). I did not feel like eating; my
1 2 3 4
appetite was poor
3). I felt that I could not shake off
the blues even with help from 1 2 3 4
my family
4). I felt that I was just as good as
1 2 3 4
other people
5). I have trouble keeping my mind
1 2 3 4
on what I was doing
6). I felt depressed 1 2 3 4
7). I felt that everything I did was
1 2 3 4
an effort
8). I felt hopeful about the future 1 2 3 4
9). I thought my life had been a 1 2 3 4
310
failure
10). I felt fearful 1 2 3 4
11). My sleep was restless 1 2 3 4
12). I was happy 1 2 3 4
13). I talked less than usual 1 2 3 4
14). I felt lonely 1 2 3 4
15). People were unfriendly 1 2 3 4
16). I enjoyed life 1 2 3 4
17). I had crying spells 1 2 3 4
18). I felt sad 1 2 3 4
19). I felt that people disliked me 1 2 3 4
20). I could not “get going” 1 2 3 4
4.4 The following questions is related to your relationship quality with your friends, please
indicate how much the statement is appropriate with your condition.
1). How do you think that your classmates are friendly with you?
□Not friendly □A little friendly □Friendly □Very friendly
2). Do your friends care about you?
□Not care □A little care □Care □Very care
3). How many do you have friends?
□One or two □Three or four □Five or six □So many
4). Do you satisfy with your friends?
□Very dissatisfied □Dissatisfied □Satisfied □Very satisfied
5). When you need help, can you find a trusted friend?
□Never □Rarely □Often □Always
6). Can you obtain happiness from group activities?
□Never □Rarely □Often □Always
4.5 Please read the following statement carefully, and then according to your personal condition,
select an appropriate number for each item.
Definitely Relatively
Relatively Exactly
not like not like Neutral
like me like me
me me
a). I often care what others think of
1 2 3 4 5
me
b). I am very sensitive to others’
1 2 3 4 5
criticism to me
c). I often feel uncomfortable when
1 2 3 4 5
talking before a crowd of people
d). It is a very difficult or me to tell
1 2 3 4 5
jokes seriously
4.6 During the past one month, how often do you have conflict with your father/mother in
terms of the following aspects (e.g. emotional, spoken, or physical)? Please fit the appropriate
number for each item, father and mother separately, according to your personal condition.
1=Never 2=1~2 times/month 3=about 1 time/week 4=several times/week 5=Almost every day
311
Father Mother
a). Your academic aspects (e.g. homework, academic record)
b). Housework (e.g. cleaning your room, Helping family members to do
housework )
c). Spending money (e.g. how do you spend money, what do you buy spending
money?)
d). Daily living arrangements(e.g. sleeping time, Eating habits, watching TV)
e). Body image (e.g. clothes, Hair styles)
f). Relationships with family members (e.g. respects, Courtesy, Relationship
between siblings)
g). Your privacy (watching you letter without your permission, Eavesdropping
your phone talking)
312

Effects of Online Social Netwo

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Effects of Online Social Netwo

Uploaded by

Copyright:

Available Formats

Effects of Online Social Networking Use on Inter-personal and Psychological

Well-being among Students in China: A Longitudinal Study

A Thesis Submitted in Partial Fulfilment

of the Requirements for the Degree of

The Chinese University of Hong Kong

All rights reserved

INFORMATION TO ALL USERS

All rights reserved.

Professor WONG Chi Sang Martin (Chair)

Professor SU Xuefen (Thesis Supervisor)

Professor MO Kit Han Phoenix (Thesis Co-supervisor)

Professor LAU Tak Fai Joseph (Thesis Co-supervisor)

Professor MAK Wing Sze Winnie (Committee Member)

Dr. IP Patrick (External Examiner)

Effects of Online Social Networking Use on Inter-personal and Psychological

Well-being among Students in China: A Longitudinal Study

Doctor of Philosophy in Public Health

The Chinese University of Hong Kong (May 2016)

Depression is the most widely reported psychiatric disorder among adolescents

worldwide. Social networking use as a new phenomenon has gained exponential

cross-sectional studies. However, the causal direction is still under-researched due to

limitations of the cross-sectional study design, especially in the area of social

networking addiction (‘SNA’). Moreover, the impacts of social networking use

concern, but remain under-researched particularly among adolescents. Current

validated instruments used. The mechanisms underlying the relationship between

conducting longitudinal studies among adolescents is highly warranted.

adolescents in Guangzhou, China. The phase II longitudinal study aimed to (1)

estimate the prevalence and incidence of depression, ‘SNA’, and IA among

predictive effects of intensity and emotional connection related to social networking

the impacts of social networking use intensity on psychosocial outcomes, including

friendship quality, perceived social support, self-esteem, adolescent-parents conflict

and social non-confidence; (5) explore the underlying mechanisms of psychosocial

use intensity and depression among adolescents.

univariate analyses, multilevel regression models, and mediation and moderation

analyses were used in this study.

Among the longitudinal sample, the prevalence of depression, ‘SNA’, and IA

nine-month period. Longitudinal analyses showed bi-directional relationships between

The developed instrument, SNAIS, showed acceptable psychometric properties.

prediction of emotional connection on incident depression was not statistically

By using changes in scores during the longitudinal period as variables, increased

increased social non-confidence (both SFUI and EFUI) and adolescent-mother

conflict (SFUI only).

adolescent-mother conflict) altogether mediated 78.77% of the effect of change in

SFUI score on change in depressive symptoms. The effect of change in EFUI on

Furthermore, the results of the mediational analyses revealed an interesting finding

suppressed the impact of change in social networking use intensity on change in

depressive symptoms, but the effect size was relatively small.

Depression and Internet addictive behaviors are present in epidemic proportions

effective intervention programs to prevent depression among adolescents. Future

confirm these findings. Translational research is warranted to use this information to

form effective intervention programs.

名廣州初中生社交網絡使用者。前瞻性隊列研究調查了 4237 名廣州初中生社交

和 7.7 每 100 人年。多水平模型顯示一般網絡成癮以及社交網絡成癮與抑鬱症的

因素（社交網絡使用強度的 AOR=1.017，娛樂功能使用強度的 AOR=1.046）

marathon full of challenges, frustrations, expectations, and excitement. I will carry

and support in various ways during my academic journey.

Firstly, I would like to give my deepest thanks to my supervisors, Prof. Xuefen

from literature reviews, to extracting research topic, writing a proposal, contacting

difficulties with strong confidence. Their hands-on supervision made me a much

guidance provided by my mentors.

classmates. Their support and companionship made my three-year study enjoyable

language of this thesis.

In addition, I would like to give my hearty thanks to professors, colleagues,

Lastly, but most importantly, my greatest gratitude goes to my beloved family