You are on page 1of 12

Acta Psychologica Sinica ©2021 Chinese Psychological Society

2022, Vol. 54, No. 4, 371384 https://doi.org/10.3724/SP.J.1041.2022.00371

A developmental model of job burnout dimensions among primary


school teachers: Evidence from structural equation model and
cross-lagged panel network model
XIE Min1, LI Feng2, LUO Yuhan1, KE Li3, WANG Xia4, WANG Yun1
(1 State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University, Beijing 100875, China)
(2 Collaborative Innovation Center of Assessment for Basic Education Quality, Beijing Normal University at Zhuhai, Zhuhai 519087, China)
(3 Collaborative Innovation Center of Assessment for Basic Education Quality, Beijing Normal University, Beijing 100875, China)
(4 Beijing Institute of Education Science, Haidian District, Beijing 100080, China)

Abstract
Emotional exhaustion, depersonalization, and reduced personal accomplishment of teacher burnout are relatively independent but
also have mutual influences. Research into their developmental relationship is helpful in understanding the developmental process
and identifying the early symptoms of job burnout. A total of 3837 primary school teachers took part in this two-wave longitudinal
study with intervals of three years. The structural equation model and cross-lagged network model were used for analysis. The results
showed that the optimal development model of teacher burnout was “emotional exhaustion and reduced personal accomplishment at
T1 predict emotional exhaustion and reduced personal accomplishment at T2, respectively, and depersonalization at T1 predicts de-
personalization and reduced personal accomplishment at T2”. There was no gender difference or teaching experience difference in
the optional development model. The results emphasize the important role of depersonalization in the development of teacher burn-
out and have certain theoretical and practical significance for identifying the early symptoms of teacher burnout and for taking cor-
responding measures to effectively prevent further teacher burnout.
Key words primary school teachers, job burnout, structural equation model, cross-lagged panel network model, development model

tional exhaustion.
1 Introduction
These five models describe the longitudinal relationship
Burnout refers to a psychological syndrome of individuals among the three dimensions of burnout, but so far, none of
under long-term intense work pressure, including emotional them displays an obvious competitive advantage. For example,
exhaustion, depersonalization and reduced personal accom- some studies believe that the Leiter and Maslach models fit
plishment symptoms (Maslach, 2003). Among them, emotional their data relatively better (Toppinen-Tanner et al., 2002; Zang
exhaustion refers to feelings of overextending and draining Weiwei, 2007; Nguyen et al., 2018), and some researchers
emotional resources, depersonalization refers to negative, cal- support the Taris model (Diestel & Schmidt, 2010; Paker &
lous, or unfeeling responses to the job, and reduced personal Salmela, 2011). The Golembiewski model is considered to be
accomplishment refers to feelings of incompetency and reduc- less ideal in many studies; however, some researchers have
tion in productivity. Maslach and Schaufeli (1993) believe that studied the phenomena among IT practitioners and found that
the three dimensions of burnout do not arise and develop si- this model fits the data better (Gan & Gan, 2014). Burnout is
multaneously but have potential causality. There have been the result of long-term stress (Maslach, 2003), emotional ex-
many studies on the sequential development of burnout dimen- haustion is an emotional response, depersonalization is a cop-
sions and five representative dimension development models ing style, and reduced personal accomplishment can be consid-
have been created (Figure 1). Among them, Leiter and ered to be a cognitive evaluation. The focus of debate on these
Maslach’s (1988) model, Lee and Ashforth’s (1993) model and models mainly includes the following: (1) Is emotional exhaus-
Taris et al.’s (2005) model assert that emotional exhaustion tion the initial or final symptom of burnout? (2) What is the
signals the beginning of burnout and will affect depersonaliza- role of depersonalization in the development of burnout? (3) Is
tion. The difference between the three models lies in whether reduced personal accomplishment the final symptom of burnout
depersonalization affects reduced personal accomplishment and or will it continue to cause emotional burnout?
emotional exhaustion. Golembiewski’s (1983) model holds that According to Lazarus’s stress theory (1966), individuals
depersonalization is the initial symptom of burnout, while emo- faced with stress will have negative emotional reactions if they
tional exhaustion is the final symptom. van Dierendonck’s believe that coping cannot relieve their feeling of stress. If they
(2001) model holds that reduced personal accomplishment is think coping can relieve this feeling, then they will take cop-
the starting point of burnout, which eventually leads to emo- ing measures, such as positive problem solving or negative

Received Date: May 17, 2021


This work was supported by the Major project of National Social Science Foundation of China (20&ZD153).
Corresponding author: WANG Yun, E-mail: wangyun@bnu.edu.cn
The original article is in Chinese, translated by Eastrans and proofread by the author. The Chinese version shall always prevail in case of any discrepancy
or inconsistency between the Chinese version and its English translation.
Acta Psychologica Sinica

Figure 1. Five representative theoretical models.

withdrawal behavior. If coping is ineffective, then they will tion in personal accomplishment that may in turn lead to further
have negative emotional reactions or re-evaluate the current emotional exhaustion; reduced personal accomplishment is the
situation. Therefore, emotional exhaustion may occur in an product of depersonalization, which may also cause emotional
early stage simultaneously with coping or after ineffective exhaustion over time.
coping. Among the above five models, emotional exhaustion In addition, from the perspective of research methods,
serves as the beginning of burnout in three of the models and structural equation model was used to compare different de-
the end of burnout in two of the models. Depersonalization is a velopment models of burnout in cross sectional or longitudinal
psychological withdrawal method used to relieve stress. For data in most previous studies. Recently, some researchers have
example, Leiter and Maslach (1988) believe that depersonaliza- also tried to use item response theory to study the sequential
tion is a way to “try to deal with demanding customers and the development of burnout (de Vos et al., 2016; Brenner, 2020).
resulting fatigue”, but researchers believe that this is a dysfunc- Unfortunately, these research results are obtained through
tional coping style (Baker et al., 2000; Taris et al., 2005). Ac- cross-sectional data, which have some defects in causal infer-
cording to stress theory, depersonalization may stimulate seri- ence. In recent years, as an alternative to the model represent-
ous emotional reactions or lower cognitive evaluation, particu- ing the latent variables of mental constructs, the network model
larly in the sense that depersonalization affects reduced per- (NM) of mental constructs has been increasingly used by re-
sonal accomplishment, and consistency across time has been searchers (Cramer et al., 2012; Borsboom & Cramer, 2013;
confirmed in some studies (Huang et al., 2015). Among the Cramer et al., 2016; Fried et al., 2016; Fried et al., 2017). In the
above five models, depersonalization affects reduced personal network analysis model, causality is assumed to occur at the
accomplishment in three of the models and affects emotional item level, not at the latent variable level. Combining the net-
exhaustion in two of the models. Especially in the teacher work analysis model with the cross-lagged panel model
group of Taris et al. (2005), these two longitudinal relationship (CLPM), that is, the cross-lagged panel network model (CLPN),
paths have been verified simultaneously. In many studies, the can reveal the longitudinal process within and between psy-
simultaneous correlation between emotional exhaustion and chological constructs over time (Rhemtulla et al., 2019).
depersonalization is relatively high, while the correlation be- Teachers are one of the high incidence groups of job burnout
tween these two and reduced personal accomplishment is much (Wu et al., 2019). Teacher burnout will not only affect teachers'
lower. For example, some studies have found that the develop- own work and life (Huyghebaert et al., 2018) but also has ad-
ment process of reduced personal accomplishment is relatively verse effects on schools (Bi & Huang, 2005) and students
independent (Wang, 2006; de Vos et al., 2016), and emotional (Herman et al., 2018). It is important for the effective interven-
exhaustion affects depersonalization (Wang Fang, 2006; Bren- tion and maintenance of teacher mental health and quality to
ner, 2020), but other studies have found that reduced personal explore the developmental characteristics of teacher burnout
accomplishment is longitudinally associated with emotional and to determine the early symptoms. There are few longitudi-
exhaustion (McManut et al., 2002; Gan & Gan, 2014), and that nal studies on the development of teacher burnout dimensions
the longitudinal relationship between emotional exhaustion and in China, or even in foreign countries (Taris et al., 2005). In this
depersonalization is not high (Lee & Ashforth, 1993). There- study, the burnout of 3837 primary school teachers was meas-
fore, in theory, emotional exhaustion may be the beginning of ured by a longitudinal design. On the one hand, a structural
burnout, the end of burnout, or both. Depersonalization may be equation model is constructed to compare the fitting of the
caused by stress itself, or by fatigue, and may lead to a reduc- above five representative development models to the data; on
XIE Min et al.: A developmental model of job burnout dimensions among primary school teachers: Evidence from structural equation...

the other hand, a cross-lagged network model is established to 2.3 Data collection and analysis
explore the development relationship among the three dimen- (1) Questionnaire test. Both tests were electronic. The eval-
sions. The two methods are used to explore and verify each uation system was opened for one week and was conducted
other to obtain the optimal development model. The gender and anonymously. Teachers who voluntarily participated in the tests
teaching experience consistency of the optimal development could log into the system through computers or mobile phones
model among different groups is also investigated. at a convenient time.
(2) Data analysis. SPSS 17.0 statistical software was used
2 Methods
for data management and descriptive statistical analysis, Mplus
2.1 Participants 8 was used for structural equation model analysis, and R 4.0.3
The two-wave data were collected from 52 primary schools was used for cross-lagged network model analysis.
in Beijing with permission from the local education department. Structural equation model analysis. In addition to the above
A total of 5279 teachers took the first test in the middle-autumn five representative models, the comparative model also in-
term in 2014, 5184 teachers took the second test in the mid- cluded the basic model (i.e., T1 emotional exhaustion, deper-
dle-autumn term in 2017, and 3837 teachers took both tests sonalization and reduced personal accomplishment affected T2
(701 male and 3136 female). In the first test, the proportions of emotional exhaustion, depersonalization and reduced personal
teachers with less than 10 years, 10~20 years and more than 20 accomplishment respectively) and the full model (i.e., each of
years of teaching experience were 22.5%, 35.9% and 41.6%, the three dimensions at T1 affected all three dimensions at T2).
respectively. The proportion of teachers in charge of the class The comparison indices included chi-square, RMSEA, SRMR,
was about 54%; and the proportions of married, unmarried, and CFI, TLI, AIC and aBIC. Chi-square, RMSEA and SRMR are
divorced/widowed teachers were about 87.1%, 9.8% and 3.1%, all absolute fitting indices. The smaller the chi-square value,
respectively. the better the model fits. An RMSEA value of less than 0.05
This study examines the measurement equivalence of two indicates that the model fits well, while an RMSEA value of
groups of teachers who participated in the second test and did less than 0.08 is also acceptable. An SRMR value of less than
not participate in the second test in the first test from four as- 0.08 indicates that the model fits well. CFI and TLI are both
pects: measurement equivalence, dimension score distribution, relative fitting indices, and values greater than 0.9 indicate that
dimension score variance analysis, and item score variance the model fits well. AIC stands for Akaike information criterion.
analysis. The results showed that the two groups had high The smaller the AIC is, the better the model is. aBIC represents
equivalence in measurement; that is, the missing mechanism the BIC value calibrated by sample size. BIC stands for the
had nothing to do with the burnout level. Bayesian information criterion, and the smaller the aBIC value,
2.2 Measurements the better. The penalty of BIC is larger than AIC for consider-
This study adopted the Burnout Scale for primary and sec- ing the number of samples, especially when there are a large
ondary school teachers which was revised by the research number of samples, can effectively prevent the model complex-
group. On the basis of the MBI scale, the three-dimensional ity caused by high model accuracy (Hou et al., 2004).
scale for primary and secondary school teacher burnout devel- Cross-lagged network model analysis. The glmnet (Fried-
oped by Wu et al. (2003) and the four-dimensional scale for man et al., 2010), qgraph (Epskamp et al., 2012) and lavaan
primary and secondary school teacher burnout developed by (Rosseel, 2012) function packets were used. Taking each vari-
Wang and Xu (2004), combined with interviews with front-line able in the first test as the prediction variable and each variable
teachers, the Burnout Scale for primary and secondary school in the second test as the result variable, cyclic regularization
teachers was revised. The revised burnout scale includes three regression was conducted (adding constraint terms to prevent
dimensions, emotional exhaustion (9 items), depersonalization overfitting). Lasso regression was adopted in this study; that is,
(6 items) and reduced personal accomplishment (4 items, re- the constraint term was the sum of the absolute values of all
versed scoring). Items were rated on a 5-point Likert scale parameters. The centrality of variables was measured by ex-
ranging from 1 (never) to 5 (always). For each of the three cluding all autoregressive paths in the same dimension from
dimensions, the higher the score, the more obvious the symp- two aspects: cross-construct “predictability” and “influence”.
tom. This scale has been tested in many programs. Among pri- Predictability refers to the degree to which each variable is
mary school teachers, α ranged from 0.90 to 0.93, CFI ranged predicted by other structural variables in the network; simply
from 0.91 to 0.93, TLI ranged from 0.90 to 0.92, and RMSEA put, it is the proportion of the difference of a variable at T2
ranged from 0.07 to 0.08. explained by all variables of other structures at T1, with the
In the first test of this study, α = 0.91, CFI = 0.93, TLI = value ranging from 0 to 1. Influence refers to the degree to
0.92, RMSEA = 0.07, and SRMR = 0.05, and in the second test, which each variable predicts other structural variables in the
α = 0.93, CFI = 0.94, TLI = 0.93, RMSEA = 0.07, and SRMR = network. It is obtained by calculating the sum of squares of the
0.05, indicating good and robust reliability and validity. Be- standardized regression coefficients of a variable at T1 to all
sides, the fact that the revised burnout scale had strong longitu- variables of the other structures at T2. Since the standardized
dinal invariance in the two tests was verified. regression coefficients are not bounded at 1, there is no high
In addition, through the basic information questionnaire of boundary for this measurement (Rhemtulla et al., 2019).
teachers, information such as gender, teaching experience, mar-
3 Results
ital status, and whether they are teachers in charge of the class
were collected. The test showed that this scale had a strong 3.1 Descriptive statistical analysis
equivalence between male teachers and female teachers, as well The mean, standard deviation, and Pearson correlation coef-
as between different teaching experiences (0~10 years, 10~20 ficient of the three dimensions of teacher burnout in the two
years and more than 20 years). tests are shown in Table 1. In total, the emotional exhaustion
Acta Psychologica Sinica

Table 1
Mean, standard deviation and correlation coefficient of three dimensions in two tests
Dimension M SD 2 3 4 5 6
** * ** **
1 Emotional exhaustion T1 2.67 0.98 0.62 0.04 0.47 0.29 0.09**
** ** **
2 Depersonalization T1 1.79 0.66 0.16 0.31 0.40 0.20**
* **
3 Reduced personal accomplishment T1 2.50 0.99 0.04 0.12 0.25**
**
4 Emotional exhaustion T2 2.47 0.93 0.67 0.14**
5 Depersonalization T2 1.78 0.72 0.31**
6 Reduced personal accomplishment T2 2.22 0.84
* ** ***
Note. p < 0.05, p < 0.01, p < 0.001

levels of teachers in the two tests were relatively high, the re- duced personal accomplishment” path in the four models were
duced personal accomplishment levels were relatively low, and significantly greater than 0.1. The effect sizes of all paths in the
especially the depersonalization levels were very low, which Leiter and Maslach models were greater than 0.1, indicating
was consistent with previous studies (Wu et al., 2019). The that this model fit the data ideally at the level of explicit vari-
correlation between emotional exhaustion and depersonaliza- ables, but that the r value of the “T1 emotional exhaustion→T2
tion was relatively high, but longitudinally, the correlations depersonalization” path in this model was only slightly greater
between T1 emotional exhaustion and T2 depersonalization and than 0.1; the same was true of the effect size of this path in the
those between T1 depersonalization and T2 emotional exhaus- Taris model. In the Golembiewski model and Taris model, the r
tion were significantly lower. value of the “T1 reduced personal accomplishment→T2 emo-
3.2 Structural equation model analysis tional exhaustion” path was less than 0.1, but in adjusted
The fitting indices of the seven compared models with di- Model 1, the r value of this path was slightly greater than 0.1.
mensions as observed variables are shown in Table 2. Among Because the effect size of the “T1 depersonalization→T2
them, the basic model only contains the autoregressive path of reduced personal accomplishment” path was significantly
three dimensions, and the adjusted Model 1 is the model delet- greater than that of the “T1 emotional exhaustion→T2 de-
ing two insignificant paths in t test of “T1 depersonaliza- personalization” path except for the three autoregressive paths,
tion→T2 emotional exhaustion” and “T1 emotional exhaus- the adjusted Model 2 was constructed (Figure 2, the path of
tion→T2 reduced personal accomplishment” from the full “T1 depersonalization→T2 reduced personal accomplish-
model. Based on the RMSEA, SRMR, CFI and TLI values, the ment” was added to the basic model). The results showed that
Golembiewski model, Leiter & Maslach model, Taris model the r values of the four paths in this model were obviously
and adjusted Model 1 all met the statistical requirements, but greater than 0.1 and also fit the data well based on the fitting
the chi-square test of adjusted Model 1 was not significant. In indices. In the comparison of the seven development models
terms of the amount of information, the AIC value and the with dimensions as observed variables, it could be considered
aBIC value of the adjusted model were the smallest, slightly that the Leiter and Maslach model and adjusted Model 2 were
lower than those of the Leiter & Maslach model, the Taris ideal for data fitting.
model and the Golembiewski model. Structural equation model can include both measurement
Due to the large sample size of this study, the t test would model and structural model, making a great use of observation
amplify the significance of the regression coefficient. Therefore, information, allowing measurement errors of independent
the standardized regression coefficient β was used as the effect variables and dependent variables, and dealing with latent
size (Nieminen et al., 2013) and the r value obtained from its variables more accurately. The fitting index of the above mod-
transformation (Peterson & Brown, 2005) was used to deter- els with dimensions as latent variables is shown in Table 4. The
mine the size of the effect. According to Cohen (1988, p.83), if results showed that the Golembiewski model, Leiter and
the r value is greater than 0.1, then it can be considered to have Maslach model, Taris model and full model fit the data better.
a significant effect. The regression coefficient effects of the The chi-square value, AIC value and aBIC value of the Lee and
above four models with relatively good fitting are shown in Ashforth model and van Dierendonck model were obviously
Table 3. The results showed that the effect sizes of the three larger than those of the other models, and the data fitting was
autoregressive paths and the “T1 depersonalization→T2 re- relatively poor.

Table 2
Fitting index of each comparison model when dimension is taken as observed variable
Model χ2 df RMSEA SRMR CFI TLI AIC aBIC
Basic model 160.63 6 0.08 0.05 0.96 0.93 24076.50 24113.40
Golembiewski (1983) model 49.45 4 0.05 0.03 0.99 0.97 23969.32 24012.36
Leiter & Maslach (1988) model 25.35 4 0.04 0.02 0.99 0.98 23945.22 23988.27
Lee & Ashforth (1993) model 99.28 4 0.08 0.03 0.98 0.93 24019.15 24062.19
van Dierendonck (2001) model 145.37 4 0.09 0.05 0.97 0.90 24065.24 24108.29
Taris et al. (2005) Model 22.97 3 0.04 0.02 0.99 0.98 23944.83 23990.96
Adjusted Model 1 0.61 2 0.00 0.00 1 1 23924.48 23973.68
Adjusted Model 2 51.24 5 0.05 0.03 0.99 0.97 23969.11 24009.08
XIE Min et al.: A developmental model of job burnout dimensions among primary school teachers: Evidence from structural equation...

Table 3 Similarly, the effect size r was used to determine the effect
Regression coefficients and effect sizes of five models with dimen- magnitude of the regression coefficient (see Table 5). The re-
sions as observed variables
sults showed that the optimal development model was adjusted
Model Path b SE t β r Model 2. The effect sizes of the “T1 emotional exhaustion→T2
Golembiewski ee1→ee2 0.40 0.01 36.82 0.44 0.49 depersonalization” path in the Leiter-Maslach model and the
(1983) model Taris model were less than 0.1. The Golembiewski model was a
pa1→ee2 0.01 0.01 1.40 0.01 0.06
dp1→dp2 0.41 0.01 29.00 0.38 0.43 potential better development model. The r value of the “T1
dp1→pa2 0.21 0.02 10.45 0.16 0.21
reduced personal accomplishment→T2 emotional exhaustion”
path of this model was slightly less than 0.1, and it was slightly
pa1→pa2 0.18 0.01 13.77 0.21 0.26
higher than 0.1 in the full model. In Table 4, the fitting indices
Leiter & Maslach ee1→ee2 0.44 0.01 34.00 0.44 0.49
(1988) model of the Golembiewski model and those of adjusted Model 2
ee1→dp2 0.06 0.01 5.17 0.06 0.11 were relatively close.
dp1→dp2 0.36 0.0 24.33 0.36 0.41 In summary, the results of the model comparison using a
dp1→pa2 0.21 0.02 10.65 0.21 0.26 structural equation model showed that the optimal development
pa1→pa2 0.18 0.01 13.61 0.18 0.23 model of burnout among primary school teachers was adjusted
Taris et al. (2005) ee1→ee2 0.42 0.02 25.00 0.45 0.50 Model 2; that is, “emotional exhaustion and reduced personal
Mode l
dp1→ee2 0.04 0.03 1.50 0.03 0.08 accomplishment at T1 predict emotional exhaustion and re-
duced personal accomplishment at T2, respectively, and deper-
ee1→dp2 0.05 0.01 4.15 0.07 0.12
sonalization at T1 predicts depersonalization and reduced per-
dp1→dp2 0.38 0.02 19.25 0.35 0.40
sonal accomplishment at T2”.
dp1→pa2 0.21 0.02 10.75 0.17 0.22
3.3 Cross-lagged network model analysis
pa1→pa2 0.18 0.01 13.69 0.21 0.26 In the cross-lagged network path of all items, the autore-
Adjusted Model 1 ee1→ee2 0.44 0.01 34.08 0.47 0.52 gressive coefficient is generally larger than the cross-lagged
pa1→ee2 0.05 0.01 3.92 0.05 0.10 path coefficient. As the thickness of arrows corresponds to the
ee1→dp2 0.07 0.01 5.75 0.09 0.14 strength of the relationship, the autoregressive paths were
dp1→dp2 0.35 0.02 22.06 0.33 0.38 omitted in Fig. 3 and Fig. 4 to make the paths of the cross-
pa1→dp2 0.05 0.01 4.82 0.07 0.12
lagged relationship clearer.
As shown in Figure 3, there were obvious mutual influences
dp1→pa2 0.21 0.02 10.55 0.16 0.21
between Item 2 (EE2, “Feeling burnout and don't want to do
pa1→pa2 0.19 0.01 14.77 0.22 0.28
anything else by the end of the day”) and Item 7 (EE7, “Work
Adjusted Model 2 ee1→ee2 0.40 0.01 36.73 0.43 0.48 is exhausting me physically and mentally”) of the emotional
dp1→dp2 0.40 0.01 31.08 0.38 0.43 exhaustion dimension, between Item1 (DP1, “Some students
dp1→pa2 0.21 0.02 10.45 0.16 0.21 seem to me hopeless”) and Item 4 (DP4, “I don't care what's
pa1→pa2 0.18 0.01 13.85 0.21 0.26 going on with the students”) of depersonalization dimension,
and between Item 3 (PA3, “I think I make a positive impact on
others through my work”) and Item 4 (PA4, “I've done a lot of
good in this job”) of reduced personal accomplishment dimen-
sion. To reduce the redundancy of the network, we created
three sum scores of two variables for each group to reduce the
total set of items from 19 to 16 (Rhemtulla et al., 2019). The
new variables were named “Exhausted physically and men-
tally” (new EE2), “Some students are hopeless and don't care
what's wrong with them” (new DP1), and “Experiencing posi-
tive influence and value at work” (new PA3) (see Figure 4).
The strongest relationship paths were the direct influence of
“I don't care what my students think of me” (DP4) on “Experi-
Figure 2. Adjusted model 2. encing positive influence and value at work” (PA3) and the

Table 4
Fitting index of each comparison model when dimension is taken as latent variable
Model χ2 df RMSEA SRMR CFI TLI AIC aBIC
Basic model 7528.05 656 0.05 0.05 0.92 0.91 361321.05 361699.27
Golembiewski (1983) model 7430.24 654 0.05 0.04 0.92 0.91 361227.25 361611.61
Leiter & Maslach (1988) model 7437.23 654 0.05 0.04 0.92 0.91 361234.24 361618.60
Lee & Ashforth (1993) model 7485.55 654 0.05 0.04 0.92 0.91 361282.55 361666.92
van Dierendonck (2001) model 7527.66 654 0.05 0.05 0.92 0.91 361324.66 361709.03
Taris et al. (2005) Model 7436.94 653 0.05 0.04 0.92 0.91 361235.95 361623.39
Full model 7417.33 650 0.05 0.04 0.92 0.91 361222.33 361619.00
Adjusted Model 2 7437.25 655 0.05 0.04 0.92 0.91 361232.30 361613.50
Acta Psychologica Sinica

Table 5
Regression coefficients and effect sizes of five models with dimension as latent variable
模型 Model 路径 Path b SE t β r 模型 Model 路径 Path b SE t β r
Golembiewski ee1→ee2 0.44 0.01 31.29 0.50 0.55 Full model ee1→ee2 0.45 0.03 17.39 0.51 0.56
(1983) model pa1→ee2 0.03 0.01 2.70 0.03 0.08 dp1→ee2 –0.04 0.05 –0.87 –0.03 0.02
dp1→dp2 0.54 0.02 23.65 0.48 0.53 pa1→ee2 0.05 0.02 3.33 0.06 0.11
dp1→pa2 0.25 0.03 9.37 0.18 0.23 ee1→dp2 0.00 0.02 –0.10 0.00 0.05
pa1→pa2 0.18 0.02 12.00 0.23 0.28 dp1→dp2 0.53 0.04 12.95 0.47 0.52
Leiter & Maslach ee1→ee2 0.43 0.02 28.93 0.49 0.54 pa1→dp2 0.02 0.01 1.85 0.04 0.09
(1988) model ee1→dp2 0.00 0.02 0.13 0.00 0.05 ee1→pa2 –0.08 0.03 –2.78 –0.09 –0.04
dp1→dp2 0.53 0.03 18.35 0.47 0.52 dp1→pa2 0.36 0.05 7.26 0.26 0.31
dp1→pa2 0.25 0.03 9.26 0.18 0.23 pa1→pa2 0.17 0.02 10.06 0.22 0.27
pa1→pa2 0.18 0.02 12.27 0.23 0.28 Adjusted Model 2 ee1→ee2 0.43 0.01 31.00 0.49 0.54
Taris et al. (2005) ee1→ee2 0.42 0.02 17.67 0.48 0.53 dp1→dp2 0.53 0.02 23.22 0.47 0.52
Mode l dp1→ee2 0.02 0.04 0.55 0.02 0.07 dp1→pa2 0.25 0.03 9.26 0.18 0.23
ee1→dp2 0.00 0.02 –0.16 –0.01 0.05 pa1→pa2 0.18 0.02 12.27 0.24 0.29
dp1→dp2 0.54 0.04 15.11 0.48 0.53
dp1→pa2 0.25 0.03 9.30 0.18 0.23
pa1→pa2 0.18 0.02 12.27 0.24 0.29

Figure 3. Estimated CLPN of 3 dimensions of teacher burnout (19 items) with autoregressive paths omitted.
Note. a. EE1 represents the first item in the dimension of emotional exhaustion, DP1 represents the first item in the dimension of depersonalization,
and PA1 represents the first item in the dimension of reduced personal achievement; b. Thicker arrows indicate a stronger relationship; c. Green ar-
rows indicate positive effects and red arrows indicate negative effects; d. Figure 4, same note

direct influence of “Work often causes me insomnia and head- while the latter was an item of depersonalization, which was an
aches” (EE8) on “Exhausted physically and mentally” (EE2). important predictor.
The former belonged to the longitudinal process between psy- It can be seen in combination with the network paths of
chological constructs, while the latter belonged to the longitu- Figure 4, “I don't care what my students think of me” (DP4)
dinal process within psychological constructs. directly affects “Experiencing positive influence and value at
The centrality of the cross-lagged network model was esti- work” (PA3, reverse scoring) and “When students make pro-
mated (see Figure 5). The results showed that “Experiencing gress, I think all the effort is worth it” (PA1, reverse scoring)
positive influence and value at work” (PA3, reverse scoring) with large influence intensity. Although “Experiencing positive
and “I don't care what my students think of me” (DP4) were the influence and value at work” (PA3, reverse scoring) directly
centres of the network. The former was an item of reduced affects “It's easy for work to get me down” (EE7), “When stu-
personal accomplishment, which was an important outcome, dents make progress, I think all the effort is worth it” (PA1,
XIE Min et al.: A developmental model of job burnout dimensions among primary school teachers: Evidence from structural equation...

Figure 4. Estimated CLPN of 3 dimensions of teacher burnout (16 items) with autoregressive paths omitted.

Figure 5. Cross-construct predictability and influence of variables in the estimated CLPN.

reverse scoring) directly affects “Under pressure to do this job (DP1), and “Becoming more irritable than before since taking
because of facing people all the time” (EE4), “Feeling tired this job” (EE6) directly affects “I'm afraid this job will make
when get up in the morning” (EE3) directly affects “Some stu- me emotionless” (DP3), the influence intensity was obviously
dents are hopeless and don't care what's wrong with them” relatively small. The path relationship and intensity between
Acta Psychologica Sinica

these constructs confirmed that depersonalization had an im- shown in Figure 6. The Wald test values of the four paths were
portant impact on reduced personal accomplishment, while the 0.001~1.761 and were not significant; that is, the optimal de-
impact of reduced personal accomplishment on emotional ex- velopment model had gender consistency.
haustion and the impact of emotional exhaustion on deperson- The obtained optimal development model was compared
alization were relatively small at present. across groups with different teaching experiences. The devel-
In summary, the model fitting indices (Table 2 and Table 4) opment model of burnout with teaching experience of 0~10
of the structural equation model and the effect size (r value) of years, 10~20 years and more than 20 years is shown in Figure 7.
each relationship path (Table 3 and Table 5) showed that ad- The Wald test values were 0.613~3.391 in the 4 paths for
justed Model 2 (Figure 2) was the optimal development model teachers with teaching experience of 0~10 years and 10~20
at the observation and latent variable levels. The cross-lagged years, were 0.002~3.198 for teachers with teaching experience
network model found that there was an obvious longitudinal of 0~10 years and more than 20 years and were 0.063~1.549
relationship between depersonalization and reduced personal for teachers with teaching experience of 10~20 years and more
accomplishment at the item level, while there was an obvious than 20 years. They were not significant; that is, the optimal
longitudinal relationship within the emotional exhaustion con- development model had consistency among teachers with dif-
struct (Figure 4 and Figure 5). Therefore, this study finally ferent teaching experiences.
suggested that adjusted Model 2 could better describe the de-
4 Discussion
velopment process of burnout among primary school teachers;
that is, if exposed to pressure for a long time, emotional ex- 4.1 The optimal development model of burnout among
haustion and depersonalization were the beginning of burnout, primary school teachers
emotional exhaustion would affect emotional exhaustion after- The optimal development model obtained in this study is
wards, and when depersonalization cannot well relieve the “emotional exhaustion and reduced personal accomplishment at
feeling of pressure, teachers would have obviously reduced T1 predict emotional exhaustion and reduced personal accom-
personal accomplishment. plishment at T2 respectively, and depersonalization at T1 pre-
3.4 Consistency of the optimal development model in dicts depersonalization and reduced personal accomplishment
terms of gender and teaching experience at T2”. That is, the burnout of primary school teachers begins
The obtained optimal development model was compared with emotional exhaustion and depersonalization, and emotional
across groups between different genders. The development exhaustion affects further emotional exhaustion while deper-
model of burnout in male teachers and female teachers is sonalization directly affect reduced personal accomplishment

Figure 6. Development model of burnout in male and female teachers (male/female).


XIE Min et al.: A developmental model of job burnout dimensions among primary school teachers: Evidence from structural equation...

Figure 7. Development model of burnout in teachers with different teaching experience (0~10 years/10~20 years/more than 20 years).

over time. This optimal development model has consistency operate with teachers' requirements, which makes it easier for
between different genders and different teaching experiences in teachers to feel negative about the value of their work and to
this sample. minimize or lose their sense of accomplishment. Taris et al.
First, in the longitudinal relationship among the three di- (2005) found that this longitudinal path is significant among
mensions of burnout among primary school teachers, this study teachers, but not among cancer nursing workers.
supports the significant prediction of depersonalization on re- Second, the results of this study show that compared to the
duced personal accomplishment at the observed variable, latent high simultaneous correlation between emotional exhaustion
variable and item level. This longitudinal path is supported by and depersonalization, the delayed correlation is significantly
three of the five representative development models of burnout lower, which is consistent with previous studies (Lee & Ash-
(Golembiewski et al., 1983; Leiter & Maslach, 1988; Taris et forth, 1993). This study does not find an obvious longitudinal
al., 2005), and this effect has been proven to be stable over time relationship between these two dimensions, including the lon-
(Huang et al., 2015). As a coping mechanism of psychological gitudinal effect of emotional exhaustion on depersonalization
withdrawal (Taris et al., 2005), depersonalization can make and the longitudinal effect of depersonalization on emotional
individuals stay away from stress/stressors in the short term, exhaustion, which is inconsistent with previous studies (Taris
give individuals time to recover, and thus temporarily obtain et al., 2005). In this study, although the longitudinal effect of
better job performance (Golembiewski et al., 1983) or addi- emotional exhaustion on depersonalization is significant when
tional resources (Maslach & Leiter, 2017). However, in the we compared structural equation models with dimensions as
long term, it will have many adverse effects on individuals, observed variables (such as Leiter & Maslach model), its effect
such as the deterioration of interpersonal relationships and the is obviously smaller than that of other paths, and the longitudi-
formation of negative self-evaluation (Huang et al., 2010), nal relationship between these two dimensions is not obvious in
further aggravating the development of burnout. The primary the comparison of structural equation models with dimensions
school teachers are special in that they need to face the same as latent variables or in cross-lagged networks. Therefore, in
students in one semester, one year, or even several years, re- this study, the relationship between emotional exhaustion and
quiring more emotional labor. The value of teaching itself is depersonalization is closer to a “parallel” relationship; that is,
related to the development of students. When teachers deal with when faced with stress, the two may occur simultaneously, but
pressure in a depersonalized way, it is easy to widen the dis- there is no obvious interaction.
tance between teachers and students, causing additional contra- Third, this study does not find a significant longitudinal
dictions and conflicts. Students are even more reluctant to co- prediction of emotional exhaustion by reduced personal ac-
Acta Psychologica Sinica

complishment. Although this longitudinal path is not involved sonalization or reduced personal accomplishment remains to be
in four of the five representative models, some studies have further verified. However, the Lee & Ashforth model and van
revealed this path (McManut et al., 2002; Gan & Gan, 2014). In Dierendonck model are not suitable for describing the devel-
this study, although the correlation between reduced personal opment process of burnout among primary school teachers.
accomplishment and emotional exhaustion is obviously lower, According to the results of this study, it is suggested that add-
it can be seen from the cross-lagged network that there are still ing an assessment of the relationship between teachers and
additional connections between the items of these two dimen- students, colleagues and leaders in teacher evaluation would
sions, and in the comparison of latent variable structural equa- help to determine the symptoms of depersonalization and emo-
tion models, the effect size (r value) of this path (only in the tional exhaustion in time and to take effective measures to im-
Golembiewski model and the full model) fluctuates around 0.1. prove teachers' sense of work accomplishment to prevent fur-
It should be noted that the Leiter & Maslach model and the ther development of teacher burnout.
Golembiewski model are considered potentially better devel- 4.2 Limitations and prospects
opment models when the dimensions are used as observed First, in this study, compared with the very strong relation-
variables and latent variables for structural equation model ship of “T1 depersonalization→T2 reduction of personal ac-
comparison. Moreover, in the cross-lagged network, the influ- complishment”, the influence of reduced personal accomplish-
ence of reduced personal accomplishment items on emotional ment on emotional exhaustion and the influence of emotional
exhaustion items and the influence of emotional exhaustion exhaustion on depersonalization were both weaker. If there are
items on depersonalization items both existed, although the 3~4 wave tests, it may be better to determine whether these
intensity was relatively lower than that of the influence of de- relationships remain stable, are strengthened, or are weakened
personalization on reduced personal accomplishment. Accord- over time, which would more accurately determine the sequen-
ing to the Leiter & Maslach model, under long-term stress in- tial development of teacher burnout.
dividuals first produce emotional exhaustion reactions, then Second, the cross-lagged network model in this study de-
deal with emotional exhaustion through depersonalization, and scribes the development relationship of teacher burnout with a
finally produce negative evaluations regarding their job per- 3-year interval, which may not be applicable to the estimation
formance. The Golembiewski model holds that, under of some effects (DeBoeck & Preacher, 2016). For example, if it
long-term stress, individuals first respond in a depersonalized takes up to 1~2 years or even less for teachers to obviously
way, and that depersonalization will lead to a short-term and affect emotional exhaustion from reduced personal accom-
limited sense of accomplishment. When the emphasis on peo- plishment, then the 3-year time interval may suggest that the
ple conflicts with performance standards, emotional exhaustion influence has weakened. However, if it takes more than 3 years,
will occur. Both models use the influence of depersonalization then the 3-year time interval may not have obvious influence,
on reduced personal accomplishment, and the difference be- or some interventions in the middle may lead to obvious
tween the two models and the optimal development model weakening or enhancement of this influence. Therefore, to
obtained in this study lies in whether emotional exhaustion or solve the problem of the rationality of the time interval, it may
psychological withdrawal coping occurs first, or whether the be necessary to model through continuous time series (Voelkle
two symptoms occur simultaneously under pressure. According et al., 2012), which would not only require an expansion of the
to the stress theory of Lazarus (1966), when teachers face stress cross-lagged network model to meet the data requirements
for a long time, they may have strong negative emotional reac- (Rhemtulla et al., 2019), but also have some difficulties in
tions, such as emotional exhaustion, or they may cope nega- practice (Savicki & Cooley, 1994).
tively, such as through depersonalization. In this study, burnout Third, in this study, the number of teachers who took both
among primary school teachers starts with emotional exhaus- tests accounted for approximately 72.7% of the number of
tion symptoms and depersonalization symptoms occurring at teachers who took the first test, and the retest rate was not very
the same time, and depersonalization will significantly and high. During the 3-year interval, teachers left their jobs, trans-
directly affect reduced personal accomplishment. Additional ferred to other schools, or were still in sample schools but did
longitudinal studies are needed to verify whether reduced per- not participate in the second test for various reasons. The test
sonal accomplishment significantly affects emotional exhaus- results from the four aspects—measurement equivalence, di-
tion. The influence of emotional exhaustion on depersonaliza- mension distribution, dimension score variance, and item score
tion has been suggested to maintain low stability over time in variance—indicated that the two groups from the first test
enterprise employees. Whether this influence path is significant (those who participated in the second test and those who did
and stable across time for teachers also needs verification not participate in the second test) have high equivalence in
through additional longitudinal studies. measurement and the missing mechanism has nothing to do
In summary, this study uses a longitudinal design and uses with the burnout level. However, the development of burnout
structural equation model for model comparison and symptoms in these teachers and the reasons why they did not
cross-lagged network model for verification. The obvious lon- take the second test are also worth exploring. In further related
gitudinal relationship between depersonalization and reduced research, we would not only analyse longitudinal participants
personal accomplishment among primary school teachers and but also pay attention to teachers who have not completed lon-
the obvious longitudinal relationship among the internal items gitudinal tests and design investigations or case interviews to
of emotional exhaustion are clarified at the item level, latent better answer such questions as the impact of burnout on
variable, and observed variable level. This confirms the impor- teachers.
tant role of depersonalization in the development of burnout In addition, the participants of previous research on the se-
among primary school teachers. However, whether there is a quential development of burnout included medical workers,
longitudinal relationship of emotional exhaustion with deper- social service workers, enterprise employees, and teachers, but
XIE Min et al.: A developmental model of job burnout dimensions among primary school teachers: Evidence from structural equation...

so far, few studies have used the same research tools and Deboeck, P. R., & Preacher K. J. (2016). No need to be discrete: A
methods to compare the sequential development of burnout in method for continuous time mediation analysis. Structural Equation
different groups. Taris et al. (2005) collected data from teachers Modeling, 23(1), 61–75.
and tumour caregivers for analysis. Using the same method, it de Vos, J. A., Brouwers, A., Schoot, T., Pat-El, R., Verboon, P., &
was found that the higher the depersonalization, the higher the Näring, G. (2016). Early career burnout among Dutch nurses: A
process captured in a Rasch model. Burnout Research, 3(3), 55–62.
emotional exhaustion and the lower the personal accomplish-
Diestel, S., & Schmidt, K.-H. (2010). Direct and interaction effects
ment among the teachers, but this result was not found among
among the dimensions of the Maslach Burnout Inventory: Results
the tumour caregivers. Is it possible that the development of
from two German longitudinal samples. International Journal of
burnout is different in different occupational groups? This Stress Management, 17(2), 159–180.
question should be answered by more future related studies. In Epskamp, S., Cramer, A. O. J., Waldorp, L. J., Schmittmann, V. D., &
the process of further verification, network methods can play a Borsboom, D. (2012). qgraph: Network visualizations of relation-
greater role. Some researchers have suggested that some ships in psychometric data. Journal of Statistical Software, 48(4),
well-defined latent variables that are easy for the public to un- 1–18.
derstand can be used as observed variables to analyse the lon- Fried, E. I., Epskamp, S., Nesse, R. M, Tuerlinckx, F., & Borsboom, D.
gitudinal development relationship among dimensions (Schmi- (2016). What are “good” depression symptoms? Comparing the
ttmann et al., 2013) and the network development relationship centrality of DSM and non-DSM symptoms of depression in a net-
among many psychological constructs (Letina et al., 2019) work analysis. Journal of Affective Disorders, 189, 314–320.
through the time series network model after multiple measure- Fried, E. I., van Borkulo, C. D., Cramer, A. O. J., Boschloo, L., Scho-
ments. This method can analyse the interaction between differ- evers, R. A., & Borsboom, D. (2017). Mental disorders as networks
ent nodes (observation variables, typical behaviours, and of problems: A review of recent insights. Social Psychiatry and
Psychiatric Epidemiology, 52(1), 1–10.
symptoms) as well as different levels of dimensions (latent
Friedman, J., Hastie, T., & Tibshirani, R. (2010). Regularization paths
variables), and its developmental process over time, which is
for generalized linear models via coordinate descent. Journal of
more helpful in assisting researchers to formulate more targeted
Statistical Software, 33, 1–22.
solutions.
Gan, T., & Gan, Y. (2014). Sequential development among dimensions
5 Conclusions of job burnout and engagement among IT employees. Stress and
Health, 30(2), 122–133.
Structural equation model comparison and cross-lagged Golembiewski, R. T., Munzenrider, R., & Carter, D. (1983). Phases of
network model mutually verified that the adjusted Model 2 progressive burnout and their work site covariants: Critical issues in
obtained in this study is the optimal development model of OD research and praxis. The Journal of Applied Behavioral Science,
burnout among primary school teachers. The model holds that 19(4), 461–481.
burnout among primary school teachers begins with emotional Herman, K. C., Hickmon-Rosa, J., & Reinke, W. M. (2018). Empiri-
exhaustion and depersonalization, and emotional exhaustion cally derived profiles of teacher stress, burnout, self-Efficacy, and
affects further emotional exhaustion while depersonalization coping and associated student outcomes. Journal of Positive Be-
affects reduced personal accomplishment over time. There is havior Interventions, 20(2), 90–100.
gender consistency in this model, and it is also consistent Hou, K. T., Wen, Z. L., & Cheng, Z. J. (2004). Structural equation
model and its applications. Beijing: Educational Science Publishing
among teachers with different teaching experiences.
House.
References Huang, J., Bao, X. H., You, X. Q., & Zhou, L. M. (2010). The media-
tion effects of personal resources on relationship between the job
Bakker, A. B., Schaufeli, W. B., Sixma, H. J., Bosveld, W., & Van Di- demand-resource model and job burnout. Journal of Psychological
erendonck, D. (2000). Patient demands, lack of reciprocity, and Science, 33(4), 963–965.
burnout: A five-year longitudinal study among general practitioners. Huang, J., You, X.Q., Wang, Y. S., & Bao, X. H. (2015). A Longitudinal
Journal of Organizational Behavior, 21(4), 425–441. Analysis of the Developmental Process of Job Burnout. Journal of
Bi, Z. Z., & Huang, X. T. (2005). The role of burnout in the relationship Psychological Science, 38(4), 911–915.
between achievement motivation and turnover intention. Journal of Huyghebaert, T., Fouquereau, E., Gillet, N., Beltou, N., & Tellier, F.
Psychological Science, 28(1), 28–31. (2018). Effects of workload on teachers’ functioning: A moderated
Borsboom, D., & Cramer, A. O. J. (2013). Network analysis: An inte- mediation model including sleeping problems and overcommitment.
grative approach to the structure of psychopathology. Annual Re- Stress and Health: Journal of the International Society for the In-
view of Clinical Psychology, 9(1), 91–121. vestigation of Stress, 34(5), 601–611.
Brenner, J. (2020). Examining the stage progression of employee Lazarus, R. S. (1966). Psychological stress and the coping process.
burnout (Order No. 28148693). Available from ProQuest Disserta- New York: McGraw-Hill.
tions & Theses Global; Publicly Available Content Database. Lee, R. T., & Ashforth, B. E. (1993). A longitudinal study of burnout
(2451849425). among supervisors and managers: Comparisons between the Leiter
Cohen, J. (1988). Statistical power analysis for the behavioral sciences and Maslach (1988) and Golembiewski et al. (1986) Models. Or-
(2nd ed.). Hillsdale, NJ: Erlbaum. ganizational Behavior and Human Decision Processes, 54(3),
Cramer, A. O. J., van Borkulo, C., D., Giltay, E. J., van der Maas, H. L. 369–398.
J., Kendler, K. S., Scheffer, M., & Borsboom, D. (2016). Major de- Leiter, M. P., & Maslach, C. (1988). The impact of interpersonal envi-
pression as a complex dynamic system. PLoS One, 11(12), 1–20. ronment on burnout and organizational commitment. Journal of
Cramer, A. O. J., van der Sluis, S., Noordhof, A., Wichers, M., Gesch- Organizational Behavior, 9(4), 297–308.
wind, N., Aggen, S. H., … Borsboom, D. (2012). Dimensions of Letina, S., Blanken, T. F., Deserno, M. K., & Borsboom, D. (2019).
normal personality as networks in search of equilibrium: You can’t Expanding network analysis tools in psychological networks:
like parties if you don’t like people. European Journal of Personal- Minimal spanning trees, participation coefficients, and motif analy-
ity, 26(4), 414–431. sis applied to a network of 26 psychological attributes. Complexity
Acta Psychologica Sinica

(New York, N.Y.), 2019, 1–27. A network perspective on psychological phenomena. New Ideas in
Maslach, C. (2003). Job burnout: New directions in research and inter- Psychology, 31(1), 43–53.
vention. Current Directions in Psychological Science, 12(5), Taris, T. W., Le Blanc, P. M., Schaufeli, W. B., & Schreurs, P. J. (2005).
189–192. Are there causal relationships between the dimensions of the
Maslach, C., & Leiter, M. P. (2017). Understanding burnout: New Maslach Burnout Inventory? A review and two longitudinal tests.
models. In C. L. Cooper, & J. C. Quick (Eds.), The handbook of Work and Stress, 19(3), 238–255.
stress and health: A guide to research and practice (pp. 36−56, Toppinen-Tanner, S., Kalimo, R., & Mutanen, P. (2002). The process of
Chapter xxii). Wiley Blackwell. burnout in white-collar and blue-collar jobs: Eight-year prospective
Maslach, C., & Schaufeli, W. B. (1993). Historical and conceptual study of exhaustion. Journal of Organizational Behavior, 23(5),
development of burnout. In W. B. Schaufeli, C. Maslach, & T. 555–570.
Marek (Eds.), Professional burnout: Recent developments in theory van Dierendonck, D., Schaufeli, W. B., & Buunk, B. P. (2001). Toward
and research (pp. 1–16). Washington, DC: Taylor & Francis. a process model of burnout: Results from a secondary analysis. Eu-
McManus, I. C., Winder, B. C., & Gordon, D. (2002). The causal links ropean Journal of Work and Organizational Psychology, 10(1),
between stress and burnout in a longitudinal study of UK doctors. 41–52
The Lancet, 359, 2089–2090. Voelkle, M. C., Oud, J. H. L., Davidov, E., & Schmidt, P. (2012). An
Nguyen, H. T. T., Kitaoka, K., Sukigara, M., & Thai, A. L. (2018). SEM approach to continuous time modeling of panel data: Relating
Burnout Study of clinical nurses in Vietnam: Development of job authoritarianism and anomia. Psychological Methods, 17(2),
burnout model based on Leiter and Maslach's theory. Asian Nursing 176–192.
Research, 12(1), 42–49. Wang, F. (2006). Job burnout in university teachers: The causes and
Nieminen, P., Lehtiniemi, H., Vhkangas, K., Huusko, A., & Rautio, A. the internal relationship of the three dimensions (Unpublished doc-
(2013). Standardised regression coefficient as an effect size index in toral dissertation). Beijing Normal University.
summarising findings in epidemiological studies. Epidemiology Bi- Wang, F., & Xu, Y. (2004). Job burnout in elementary and high school
ostatistics & Public Health, 10(4), 1–15. teachers: Characteristics and relationship with social support. Acta
Peterson, R. A., & Brown, S. P. (2005). On the use of beta coefficients Psychologica Sinica, 36(5), 568–574.
in meta-analysis. Journal of Applied Psychology, 90(1), 175–181. Wu, X. C., Qi, Y. J., & Zang, W. W. (2019). Overall features and influ-
Rhemtulla, M., van Bork, R., & Cramer, A. O. J. (2019). encing factors of primary and secondary school teachers’ job burn-
Cross-laggedged network models. Multivariate Behavioral Re- out in China. Journal of South China Normal University (Social
search. Retrived from https://osf.io/r24q6/ Science Edition), (1), 37–42.
Rosseel, Y. (2012). Lavaan: An R package for structural equation mod- Wu, X. C., Zeng, L. J., Qin, X. G., & Zheng, Q. (2003). The current
eling. Journal of Statistical Software, 48(2), 1–36. situation and related factors of Chinese teacher’s burnout. Studies of
Savicki, V., & Cooley, E. J. (1994). Burnout in child protective service Psychology and Behavior, 1(4), 262–267.
workers: A longitudinal study. Journal of Organizational Behavior, Zang, W. W. (2007). The research on the relationship among occupa-
15(7), 655–666. tional stress, job burnout and coping style of elementary and high
Schmittmann, V. D., Cramer, A. O. J., Waldorp, L. J., Epskamp, S., school teachers (Unpublished master’s thesis). Beijing Normal
Kievit, R. A., & Borsboom, D. (2013). Deconstructing the construct: University.

You might also like