You are on page 1of 7

RESEARCH REPORT doi:10.1111/j.1360-0443.2008.02295.

Mediators of a successful web-based smokeless


tobacco cessation program

Brian G. Danaher, Keith Smolkowski, John R. Seeley & Herbert H. Severson


Oregon Research Institute, Eugene, OR, USA

ABSTRACT

Aim To examine self-efficacy and program exposure as possible mediators observed treatment effects for a web-based
tobacco cessation intervention. Design The ChewFree trial used a two-arm design to compare tobacco abstinence at
both the 3- and 6-month follow-up for participants randomized to either an enhanced intervention condition or a basic
information-only control condition. Setting Internet in US and Canada. Participants Our secondary analyses
focused upon 402 participants who visited the web-based program at least once, whose baseline self-efficacy rating
showed room for improvement, who reported that they were still using tobacco at the 6-week assessment, and for
whom both 3- and 6-month follow-up data were available. Intervention An enhanced web-based behavioral smoke-
less tobacco cessation intervention delivered program content using text, interactive activities, testimonial videos and
an ask-an-expert forum and a peer forum. The basic control condition delivered tobacco cessation content using static
text only. Measurements Change in self-efficacy and program exposure from baseline to 6 weeks were tested as simple
and multiple mediators on the effect of treatment condition on point-prevalence tobacco abstinence measured at 3- and
6-month follow-up. Findings While both participant self-efficacy and program exposure satisfied the requirements for
simple mediation, only self-efficacy emerged as a mediator when we used the more robust test of multiple mediation.
Conclusions Results confirm the importance of self-efficacy change as a probable underlying mechanism in a suc-
cessful web-based behavioral intervention. While program exposure was found to be a simple mediator of tobacco
abstinence, it failed to emerge as a mediator when tested with self-efficacy change in a multiple mediator test suggesting
that self-efficacy and program exposure share a complex, possibly reciprocal relationship with the tobacco abstinence
outcome. Our results underscore the utility of searching for mediators in research on web-based interventions.

Keywords Internet, mediators, smokeless, tobacco cessation, web.

Correspondence to: Brian G. Danaher, Oregon Research Institute, 1715 Franklin Boulevard, Eugene, OR 97403, USA. E-mail: briand@ori.org
Submitted 6 December 2007; initial review completed 25 February 2008; final version accepted 29 April 2008

INTRODUCTION We describe secondary analyses of data from the


ChewFree.com project that explore mechanisms that
Given its prevalence and health consequences, smokeless might explain the observed treatment effect. We explore
tobacco (ST)—which includes chewing tobacco and two a priori putative mediators: participant self-efficacy
snuff—remains a significant public health problem [1,2]. and the extent to which participants viewed the content
The reach and convenience of web-based behavioral of the web-based intervention (program exposure).
interventions seem particularly appropriate because
clinic-based ST cessation programs are generally not
Self-efficacy
available—particularly to chewers located in rural set-
tings. The web-based ChewFree randomized controlled Components of the ST cessation intervention were based
trial (RCT) found that participants assigned to an on Social Cognitive Theory [4,5] as it has been applied to
enhanced (highly interactive) condition were significantly tobacco abstinence [6,7]. The intervention was designed
more likely to be tobacco abstinent at both 3- and to encourage participants to use strategies that address
6-months post-enrollment than participants who were behavior, cognition and environment [8,9] and thereby
assigned to a basic control (textual information) condition enhance participant self-efficacy and increase tobacco
[3]. abstinence. Other research has found self-efficacy to act

© 2008 The Authors. Journal compilation © 2008 Society for the Study of Addiction Addiction, 103, 1706–1712
Mediators of a web-based tobacco cessation program 1707

as a mediator across a wide variety of behavioral changes enhanced intervention condition (n = 1260) or the basic
[10–14], including for tobacco cessation [15–19]. control conditions (n = 1263) did not differ in terms of
these baseline participant characteristics.
Program exposure Consistent with many studies of web-based behavioral
interventions [22], the ChewFree trial experienced sub-
One defining characteristic of research on web-based
stantial attrition at the follow-up assessments [3]: 52% at
behavior change interventions is its focus on participant
3 months, 55% at 6 months and 66% attrition when we
engagement—a key ingredient of which involves analy-
required participation at both 3 and 6 months.
sis of program exposure data that are collected unobtru-
sively as participants visit the program website. Two key
exposure measures include the frequency and the dura- MEASURES
tion of participants’ online visits to view program
Tobacco abstinence outcome
content [20]. Participants in web-based programs are
typically able to control how much or little they use a Self-reported measures of smokeless tobacco use, ciga-
website and several such tobacco cessation studies have rette smoking and pipe or cigar smoking were obtained
shown that participants make relatively few visits to the at all assessment points by asking about 7-day point pre-
site [20]. More frequent and longer visits are presumed valence use of tobacco products (ST, cigarettes, pipes
to enable more learning of coping skills and therefore and cigars). Repeated point prevalence of self-reported
increase self-efficacy. Alternatively, more exposure might tobacco abstinence at both the 3- and 6-month assess-
reflect stronger motivation for behavior change. In ment was our dependent variable.
either case, we would expect exposure to be related posi-
tively to outcome. Self-efficacy

Participant self-efficacy was measured at baseline as well


as at follow-up assessments in both conditions by asking:
METHODS
‘how confident are you that you will not be using smoke-
ChewFree RCT less tobacco a year from now?’, using a five-point Likert
scale: 0 = not at all, 2 = somewhat and 4 = completely.
The ChewFree trial used a two-arm design to compare
Change in self-efficacy was computed by subtracting the
tobacco abstinence at both the 3- and 6-month follow-up
rating obtained at baseline from the rating obtained at
for participants randomized to either: (i) an enhanced
the 6-week assessment. An increase in self-efficacy across
intervention condition (text, graphics, interactive activi-
these two time-points would be described by a positive
ties, testimonial videos and an ask-an-expert forum and a
number, a reduction in self-efficacy would be indicated by
peer forum); or (ii) a basic information-only control con-
a negative number and no change would be represented
dition (online self-help ST cessation booklet, overview of
by a zero.
cessation resources and annotated list of other helpful
websites for tobacco cessation). One of the aims of the
Program exposure
study was to test for mediators that might represent
underlying mechanisms of observed changes. Each participant’s username and password were
obtained during their log-in which permitted unobtrusive
Participant recruitment and characteristics tracking of the number and duration of each visit to
access program content [20]. We considered only expo-
The ChewFree participant recruitment campaign used
sure data for the period prior to the 6-week assessment.
news releases to print and broadcast media, advertising
After we established that standardized versions of
on Google and in newspapers and magazines, placement
number and duration of visits were highly correlated
of ChewFree.com links on other websites and direct mail
(r = 0.67), we used a composite exposure variable defined
to health care and tobacco control professionals [21].
as the mean of the Z-score transformations of visits
Participant eligibility criteria included: use of ⱖ 1 ST can/
(number) and duration (minutes).
week for at least 1 year; interest in quitting ST; ⱖ 18 years
of age; ability to read English; resident of United States or
Participant selection and characteristics
Canada; use of personal e-mail account ⱖ 1/week; will-
ingness to provide name, home address and telephone For the mediation analysis we used a subsample of Chew-
number; ability to read and write English; and completion Free participants for whom data were available on all of
of an Informed Consent Statement approved by the Insti- the key variables. We included participants whose web-
tutional Review Board of Oregon Research Institute. server data indicated that they had visited the web-based
A total of 2523 participants randomized either to the program for at least a single occasion. Because we were

© 2008 The Authors. Journal compilation © 2008 Society for the Study of Addiction Addiction, 103, 1706–1712
1708 Brian G. Danaher et al.

2523 Randomized

1260 Assigned to receive 1263 Assigned to receive


Enhanced Condition Basic Condition

1198 Visited program 1175 Visited program


(95.1%) webpages (93.0%) webpages

962 Baseline 993 Baseline


(76.3%) self-efficacy < 4 (78.6%) self-efficacy < 4

473 Self-efficacy rating 575 Self-efficacy rating


(37.5%) on 6 week Assessment (45.5%) on 6 week Assessment

301 Tobacco use indicated 465 Tobacco use indicated


(23.9%) on 6 week Assessment (36.8%) on 6 week Assessment

143 Completed both 3- and 259 Completed both 3- and


Figure 1 Final sample of 402 partici-
(11.3%) 6-month Assessment (20.5%) 6-month Assessment
pants by condition and inclusion criteria

interested in the possible mediation effects of participant minutes of waking (54%), current smoking (5.5%), ST
improvement in self-efficacy, we included participants quit attempt in last year (69.4%) and readiness to quit
who could improve their self-efficacy, i.e. those with a ST (mean = 7.8, SD = 1.7; based on an adaptation of the
baseline self-efficacy rating of less than 4 (ceiling level of contemplation ladder [26] that used a 10-point Likert
‘complete confidence’). In order to enhance the probabil- scale with 0 = no thought of quitting; 2 = think I need
ity that change in self-efficacy predated the change in to consider quitting someday; 5 = think I should quit
tobacco use rather than reflected it—thus satisfying but not quite ready; 8 = starting to think about how
the temporal precedence requirement for mediation to change my patterns of using chew or snuff; and
[23,24]—we limited our analysis to those participants 10 = taking action to quit (for example, cutting down,
who reported that they were still using tobacco at the enrolling in program).
6-week assessment. Finally, we included only those indi-
viduals for whom both 3- and 6-month follow-up data Statistical analyses
were available. These inclusion criteria yielded a sub-
sample of 402 participants (enhanced condition n = 143 Condition comparisons
and basic condition n = 259) used in the current analyses We used contingency table analysis to compute the odds
(see Fig. 1) of whom 11.4% (46 of 402) were tobacco ratios (ORs) for the association between treatment condi-
abstinent at both the 3- and 6-month follow-up assess- tion and tobacco abstinence outcome (repeated point
ments. The imbalance in number of participants by con- prevalence at both 3- and 6-month assessments).
dition is associated with the fact that fewer participants
in the enhanced condition reported using tobacco at 6
Mediation analyses
weeks and, further, that the enhanced condition experi-
enced more overall attrition [3]. We initially used the widely used ‘causal steps’ mediation
The subset of 402 participants did not differ by condi- test outlined by Baron & Kenny [27]. Referencing the
tion at baseline in terms of age (mean = 37.1, SD = 9.2), paths depicted in Fig. 2, we used linear regression to
gender: (male = 96.5%), marital status (69.7% married); determine whether treatment condition was related sig-
rurality: (39.3% rural using RUCAS measure [25]), days nificantly to tobacco abstinence (direct effects; path c);
their ST can lasted (mean = 2.2, SD = 1.5), years of ST whether condition was related significantly to each
use (mean = 18.8, SD = 9.1), use of ST within first 30 mediator (paths a1 and a2); whether each mediator was

© 2008 The Authors. Journal compilation © 2008 Society for the Study of Addiction Addiction, 103, 1706–1712
Mediators of a web-based tobacco cessation program 1709

Tobacco 6-week follow-up assessment than the basic condition


Condition c Abstinence [mean = 0.43 and standard deviation (SD) = 0.98 versus
3 & 6 months mean = 0.71 and SD = 0.91; t = -2.85, d.f. = 310.42,
P = 0.005, two-tailed, assumed unequal variances;
Self-efficacy d = 0.30]. Similarly, the enhanced condition had greater
Change program exposure than the basic condition (mean = 0.59
and SD = 1.16 versus mean = -0.32 and SD = 0.50;
a1 b1 t = -8.87, d.f. = 171.82, P = 0.000, two-tailed, assumed
Tobacco unequal variances; d = 1.10).
c1’
Condition Abstinence
c2’
3 & 6 months
a2 b2
Simple mediation

We examined each of the putative mediators alone in a


Program test for simple mediation using Baron & Kenny’s [27]
Exposure
causal steps strategy (using paths described in Fig. 2).
The effect of condition on tobacco abstinence was sig-
Figure 2 Causal steps mediation test
nificant (path c; b = 0.108, P = 0.030). For program
exposure, we found significant effects of condition on
related significantly to tobacco abstinence (paths b1 and program exposure (path a2; b = 0.477, P = 0.000) and
b2) controlling for treatment; and whether the prediction for exposure on tobacco abstinence (path b2; b = 0.133,
of abstinence by condition became non-significant when P = 0.008). The effect of condition on abstinence
each mediator was entered separately with condition as a was reduced markedly and became statistically non-
predictor of tobacco abstinence (path c′). significant when controlled for the effects of program
Next, we tested for simple and multiple mediation using exposure (path c2′; b = 0.039, P = 0.304). For change in
the innovative non-parametric bootstrapping procedure self-efficacy, we found significant effects of condition on
recommended by MacKinnon [28] and articulated further self-efficacy (path a1; b = 0.139, P = 0.005) and self-
by Preacher & Hayes [29,30] designed to estimate the efficacy on abstinence (path b1; b = 0.051, P = 0.002).
sampling distribution of the indirect effect. The indirect The effect of condition on outcome was markedly reduced
effect is represented by the product of the coefficients (e.g. and became statistically non-significant when controlled
a1b1 in Fig. 2). This bootstrapping procedure assumes that for the effects of self-efficacy change (path c1′; b = 0.039,
the distribution of the measured variables approximates P = 0.304). Thus, self-efficacy and program exposure
that of the population while it avoids making the often- were both identified as simple mediators.
tenuous assumption that the indirect effect is distributed Next we tested for simple mediation using Preacher &
normally. For all analyses tobacco abstinence was coded as Hayes’ [29,30] bootstrapping methodology for indirect
1 = abstinent and 0 = ST use, and condition was coded as effects based on 5000 bootstrap resamples to describe the
1 = enhanced condition and 0 = basic condition. confidence intervals of indirect effects in a manner that
makes no assumptions about the distribution of the indi-
RESULTS rect effects. Interpretation of the bootstrap data is accom-
plished by determining whether zero is contained within
Tobacco abstinence
the 95% CIs (thus indicating the lack of significance).
A total of 11.4% (46 of 402) of participants were tobacco Results for simple mediation showed that both change in
abstinent at both the 3- and 6-month assessments: self-efficacy and program exposure were simple media-
16.1% (23 of 143) in the enhanced condition and 8.9% tors. Thus the causal steps test and the bootstrapping test
(23 of 259) in the basic condition. Contingency table agreed that self-efficacy and program exposure were
analysis revealed a significant effect benefiting the simple mediators (see Table 1).
enhanced condition [c2 = 4.72, P = 0.030, OR = 1.97,
95% confidence interval (CI) = 1.06–3.65]. These results Multiple mediation
for the subset of 402 participants are consistent with the
Next, we used the non-parametric bootstrapping proce-
results reported for the full sample of 2523 (intent-to-
dure for multiple mediation which indicated similarly
treat analysis) reported in the outcome results paper [3].
that self-efficacy was a significant mediator of the effect
of condition on abstinence (controlling for program
Changes in putative mediators
exposure) but program exposure was not found to be a
The enhanced condition showed significantly more significant mediator (controlling for self-efficacy)—its
improvement in self-efficacy from baseline to the confidence interval contained zero. The contrast testing

© 2008 The Authors. Journal compilation © 2008 Society for the Study of Addiction Addiction, 103, 1706–1712
1710 Brian G. Danaher et al.

Table 1 Simple and multiple mediation of the indirect effects of Limitations


treatment condition on tobacco abstinence (at 3- and 6-month
follow-up assessment) through changes in self-efficacy and The current study had some limitations including our use
program exposure (n = 402; 5000 bootstrap samples).
of a restricted sample that satisfied the inclusion criteria
BCa* 95% CI for the current analyses. This sample of 402 participants
Point represented 15.9% of the original randomized sample of
estimate Lower Upper 2523, a reduction caused largely by our requirements to
include self-efficacy data at 6 weeks in addition to tobacco
Simple indirect effects
Self-efficacy 0.0129 0.0028 0.0318 use data for both the 3- and 6-month assessments. Our
Program exposure 0.0334 0.0005 0.0800 use of a subset of original participants is predicated on
Multiple indirect effects our focus to perform secondary analyses of the data [34],
Self-efficacy 0.0122 0.0024 0.0309 whereas we describe results using the entire sample of
Program exposure 0.0293 -0.0030 0.0777 participants in our outcome report [3]. Although our
Total 0.0415 0.0080 0.0896 sample was sufficiently large to permit mediation analy-
Contrast: self-efficacy -0.0171 -0.0666 0.0204 ses [35], it is also important to explore alternative impu-
versus program exposure
tation methodologies that increase the number of cases
that could be used for mediation analyses.
*BCa = bias corrected and accelerated bootstrapping confidence intervals
that include corrections for both median bias and skew [42]. Confidence
It is also important to note that we did not collect data
intervals containing zero are interpreted as not significant. on the extent that participants may have used other web-
based programs, although we have no reason to assume
thatanysuchprogramuseoccurreddifferentiallybycondi-
the two putative mediators was not significant, indicating
tion. Another possible limitation is our use of self-reported
that the magnitude of these indirect effects could not be
point prevalence abstinence data for the key measures of
distinguished.
tobacco abstinence. Although the Society for Research on
Nicotine and Tobacco (SRNT) has acknowledged that bio-
DISCUSSION
chemical verification may not be appropriate for research
As successful web-based tobacco cessation research of low-intensity, low-demand tobacco cessation interven-
projects emerge, there is an increasing opportunity to tions [36], it has also recommended that cessation studies
explore possible mechanisms for observed treatment use prolonged abstinence measures [37]. Additional
effects. The current paper sought to examine putative threats to the ability of any mediation analysis to identify
mechanisms underlying a successful web-based the ‘true indirect pathway’ [38] must also be acknowl-
smokeless tobacco cessation trial. We examined two edged, including the possibility that: (i) putative mediators
putative mediators—self-efficacy change and program may, in fact, be highly correlated proxies of other, unmod-
exposure—using both simple and multiple mediation eled variables; and (ii) measurement reliability across vari-
analysis techniques. Self-efficacy change emerged as a ables might be markedly different.
mediator in all tests which is consistent with the Social We were also unable to establish the reliability of our
Cognitive Theory explanation that behavior change global measure of self-efficacy because it was based on a
programs—including those delivered via the web—are single item. It is noteworthy, however, that we were able
effective, at least in part, because they encourage positive to identify a mediation effect even given the potential
changes in participant self-efficacy [4,5,31,32]. attenuation of that effect due to the limited reliability of
Program exposure emerged as a mediator using the our measure of risk [27]. Future research should con-
causal steps test for simple mediation but it fell out as a sider including additional items that tap different aspects
mediator when considered in combination with self- of self-efficacy [39] (e.g. confidence in being able to avoid
efficacy change in a test of multiple mediation. Program relapse when experiencing stress, when using alcohol, in
exposure is a complicated matter—consider that more is the presence of other chewers). It is also important to
not always better, and there is likely to be minimal value acknowledge that other putative underlying mechanisms
derived from using more of an ineffective program—but should be measured and analyzed in future research. For
it continues to have key importance for research on example, additional factors addressed by the enhanced
web-based interventions as it is related significantly to condition included use of putative risk/protection
outcome. Future research should explore alternate ways mechanisms (e.g. stress management strategies, setting a
to define program exposure and seek to identify whether quit date, telling friends and family, planning for tough
some participants benefit from very little exposure, situations). In contrast to earlier findings of an RCT that
whereas others follow a more linear dose–response tested self-help methods for ST cessation [40], none of
relationship [33]. these additional risk/protective mechanisms were found

© 2008 The Authors. Journal compilation © 2008 Society for the Study of Addiction Addiction, 103, 1706–1712
Mediators of a web-based tobacco cessation program 1711

to be significant in the ChewFree study. Future research linear chain of underlying mechanisms is difficult when
should also explore the contribution of different putative considering the temporal requirements for separating the
mechanisms (e.g. exposure to symbolic modeling of reciprocal impact of attitude change and behavior
content via use of video testimonials, participation in web change. This is especially true when the essential trial
forums). involves participants engaged in a web-based interven-
tion with relatively few opportunities for assessment.
Strengths
Clearly, much remains in terms of improving our under-
Our paper has several noteworthy strengths, including standing of the mechanisms that help to explain positive
our use of precautionary steps to attenuate the possibility findings in research on web-based behavior change inter-
that putative mediators might reflect prior behavior ventions. Our results point clearly to the conclusion that
change rather than predict that change [23,41]. For outcome is influenced by both changes in self-efficacy and
example, we limited our analyses to participants who had degree of program exposure. Future research should
not yet quit at the 6-week assessment. These consider- explore alternative ways to measure program exposure,
ations resulted in the exclusion of 170 participants who the extent to which some participants benefit from very
reported that they were tobacco abstinent at the 6-week little exposure, whereas others follow a more linear dose–
follow-up (‘early quitters’) and who accounted for 65.6% response relationship [33], and to understand more
(170 of 259) of all participants abstinent at both the 3- clearly the interplay between program exposure and par-
and 6-month assessments. Without an assessment earlier ticipant self-efficacy.
than 6 weeks we are unable to determine the extent to
which these early quitters also experienced an improve- Declarations of interest
ment in their self-efficacy or program exposure which
None.
was followed by lasting tobacco abstinence. As a second
precaution to address the temporal precedence require-
ment of mediation, we limited our measures of program Acknowledgements
exposure to reflect website visits that occurred before the We extend our thanks to Kristopher J. Preacher, Andrew
date of each participant’s 6-week assessment. Hayes and Edward Lichtenstein for their review of earlier
drafts of this report. This work was funded, in part,
CONCLUSION by a grant from the National Cancer Institute
(R01-CA84225).
We had hypothesized that being able to stop using
tobacco would be related to using a web-based tobacco
References
cessation intervention that included components based
on Social Cognitive Theory (e.g. peer modeling through 1. Alguacil J., Silverman D. T. Smokeless and other nonciga-
video testimonials and web forums) designed to increase rette tobacco use and pancreatic cancer: a case–control
study based on direct interviews. Cancer Epidemiol Biomark-
self-efficacy. In addition, we expected that a more engag-
ers Prev 2004; 13: 55–8.
ing program would also encourage greater program 2. USDHHS. Report on Carcinogens (RoC), 11th edn. Washing-
exposure which, in turn, would lead to improvements in ton, D.C: US Department of Health and Human Services,
self-efficacy and tobacco abstinence. Our results indicate Public Health Service, National Toxicology Program; 2005.
that program exposure and self-efficacy change do, Available at: http://ntp.niehs.nih.gov/ntp/roc/eleventh/
indeed, act as a simple mediators but results from the profiles/s176toba.pdf (accessed 12 January 2007).
3. Severson H. H., Gordon J. S., Danaher B. G., Akers L. A.
multiple mediation test underscored the complexity of
ChewFree.com: evaluation of a Web-based cessation
the interplay between these two factors. Although the program for smokeless tobacco users. Nicotine Tob Res 2008;
timing of our assessments did not permit us to test with 10: 381–91.
confidence whether program exposure preceded self- 4. Bandura A. Self-Efficacy The Exercise of Control. New York:
efficacy change, it is plausible that program exposure WH Freeman; 1997.
5. Bandura A. Health promotion by social cognitive means.
would occur before self-efficacy would improve. However,
Health Educ Behav 2004; 31: 143–64.
it also possible that individuals with elevated self-efficacy 6. Brandon T. H., Herzog T. A., Irvin J. E., Gwaltney C. J.
might be predisposed to remain engaged with an inter- Cognitive and social learning models of drug dependence:
vention in order to review program content thoroughly. implications for the assessment of tobacco dependence in
In addition, it is also likely that when efficacious individu- adolescents. Addiction 2004; 99: 51–77.
7. Niaura R. Cognitive social learning and related perspectives
als use a carefully tailored web-based intervention that
on drug craving. Addiction 2000; 95: S155–63.
permits rapid access to program content, they might be 8. Marlatt G. A. Relapse prevention: theoretical rationale
very efficient at finding what they need thus resulting and overview of the model. In: Marlatt G. A., Gordon J. R.,
in briefer exposure. Any attempt to identify a clear-cut editors. Relapse Prevention: Maintenance Strategies in the

© 2008 The Authors. Journal compilation © 2008 Society for the Study of Addiction Addiction, 103, 1706–1712
1712 Brian G. Danaher et al.

Treatment of Addictive Behaviors. New York: Guilford Press; Measuring participant rurality in Web-based interventions.
1985, P. 3–70. BMC Public Health 2007; 7: 228.
9. Shiffman S., Kassel J., Gwaltney C., McChargue D. Relapse 26. Biener L., Abrams D. B. The Contemplation Ladder: valida-
prevention for smoking. In: Marlatt G. A., Donovan D. M., tion of a measure of readiness to consider smoking cessa-
editors. Relapse Prevention: Maintenance Strategies in the tion. Health Psychol 1991; 10: 360–5.
Treatment of Addictive Behaviors, 2nd edn. New York: Guil- 27. Baron R. M., Kenny D. A. The moderator–mediator variable
ford Press; 2005, P. 92–129. distinction in social psychological research: conceptual,
10. Linde J. A., Rothman A. J., Baldwin A. S., Jeffery R. W. The strategic, and statistical considerations. J Pers Soc Psychol
impact of self-efficacy on behavior change and weight 1986; 51: 1173–82.
change among overweight participants in a weight loss 28. MacKinnon D. P. Contrasts in multiple mediator models.
trial. Health Psychol 2006; 25: 282–91. In: Rose J., Chassin L., Presson C. C., Sherman S. J., editors.
11. Strecher V. J., DeVellis B. M., Becker M. H., Rosenstock I. M. Multivariate Applications in Substance Use Research: New
The role of self-efficacy in achieving health behavior Methods for New Questions. Mahwah, NJ: Lawrence
change. Health Educ Q 1986; 13: 73–92. Erlbaum; 2000, p. 141–60.
12. Lorig K. R., Ritter P. L., Laurent D. D., Plant K. Internet- 29. Preacher K. J., Hayes A. F. Asymptotic and Resampling Strat-
based chronic disease self-management: a randomized trial. egies for Assessing and Comparing Indirect Effects in Multiple
Med Care 2006; 44: 964–71. Mediator Models. Manuscript under editorial review; 2007.
13. Reynolds K. D., Buller D. B., Yaroch A. L., Maloy J. A., Cutter Available at: http://www.comm.ohio-state.edu/ahayes/
G. R. Mediation of a middle school skin cancer prevention indirect2.pdf (accessed 21 September 2007).
program. Health Psychol 2006; 25: 616–25. 30. Preacher K. J. & Hayes A. F. Asymptotic and resampling
14. Li F., Fisher K. J., Harmer P., McAuley E. Falls self-efficacy as strategies for assessing and comparing indirect effects in
a mediator of fear of falling in an exercise intervention for multiple mediator models. Behav Res Methods 2008; 40:
older adults. J Gerontol B Psychol Sci Soc Sci 2005; 60: 879–91.
34–40. 31. Ajzen I. Perceived behavioral control, self-efficacy, locus of
15. Solomon L. J., Bunn J. Y., Pirie P. L., Worden J. K., Flynn B. S. control, and the theory of planned behavior. J Appl Soc
Self-efficacy and outcome expectations for quitting among Psychol 2002; 32: 1–20.
adolescent smokers. Addict Behav 2006; 31: 1122–32. 32. Rimal R. N. Closing the knowledge–behavior gap in health
16. Gwaltney C. J., Shiffman S., Balabanis M. H., Paty J. A. promotion: the mediating role of self-efficacy. Health
Dynamic self-efficacy and outcome expectancies: prediction Commun 2000; 12: 219–37.
of smoking lapse and relapse. J Abnorm Psychol 2005; 114: 33. Christensen H., Mackinnon A. The law of attrition revisited.
661–75. J Med Internet Res 2006; 8: e20.
17. Dijkstra A., Wolde G. T. Ongoing interpretations of accom- 34. Fergusson D., Aaron S. D., Guyatt G., Hebert P. Post-
plishments in smoking cessation: positive and negative self- randomisation exclusions: the intention to treat principle
efficacy interpretations. Addict Behav 2005; 30: 219–34. and excluding patients from analysis. BMJ 2002; 325:
18. Strecher V. J., Shiffman S., West R. Moderators and media- 652–4.
tors of a Web-based computer-tailored smoking cessation 35. Fritz M. S., MacKinnon D. P. Required sample size to detect
program among nicotine patch users. Nicotine Tob Res the mediated effect. Psychol Sci 2007; 18: 233–9.
2006; 8: S95–101. 36. Benowitz N. L., Jacob P., Ahijevych K., Jarvis M. J., Hall S.,
19. Piper M. E., Federmen E. B., McCarthy D. E., Bolt D. M., LeHouezec J. et al. Biochemical verification of tobacco use
Smith S. S., Fiore M. C. et al. Using mediational models to and cessation. Nicotine Tob Res 2002; 4: 149–59.
explore the nature of tobacco motivation and tobacco treat- 37. Hughes J. R., Keely J. P., Niaura R. S., Ossip-Klein D. J., Rich-
ment effects. J Abnorm Psychol 2008; 117: 94–105. mond R. L., Swan G. E. Measures of abstinence in clinical
20. Danaher B. G., Boles S. B., Akers L., Gordon J. S., Severson trials: issues and recommendations. Nicotine Tob Res 2003;
H. H. Defining participant exposure measures in Web-based 5: 13–25.
health behavior change programs. J Med Internet Res 2006; 38. Little T. D., Card N. A., Bovaird J. A., Preacher K. J., Crandall
8: e15. C. S. Structural equation modeling of mediation and mod-
21. Gordon J. S., Akers L., Severson H. H., Danaher B. G., Boles eration with contextual factors. In: Little T. D., Bovaird J. A.,
S. M. Successful participant recruitment strategies for an Card N. A., editors. Modeling Contextual Effects in Longitudi-
online smokeless tobacco cessation program. Nicotine Tob nal Studies. Mahwah, NJ: Lawrence Erlbaum Associates;
Res 2006; 8: S35–41. 2007, p. 207–30.
22. Eysenbach G. The law of attrition. J Med Internet Res 2005; 39. Condiotte M. M., Lichtenstein E. Self-efficacy and relapse in
7: e11. smoking cessation programs. J Consult Clin Psychol 1981;
23. Kraemer H. C., Stice E., Kazdin A., Offord D., Kupfer D. How 49: 648–58.
do risk factors work together? Mediators, moderators, and 40. Severson H. H., Andrews J. A., Lichtenstein E., Gordon J. S.,
independent, overlapping, and proxy risk factors. Am J Psy- Barckley M. S., Akers L. A self-help cessation program for
chiatry 2001; 158: 848–56. smokeless tobacco users: comparison of two interventions.
24. Stice E., Presnell K., Gau J., Shaw H. Testing mediators of Nicotine Tob Res 2000; 2: 363–70.
intervention effects in randomized controlled trials: an 41. MacKinnon D. P., Fairchild A. J., Fritz M. S. Mediation analy-
evaluation of two eating disorder prevention programs. sis. Annu Rev Psychol 2007; 58: 593–614.
J Consult Clin Psychol 2007; 75: 20–32. 42. Efron B., Tibshirani R. An Introduction to the Bootstrap. New
25. Danaher B. G., Hart L. G., McKay H. G., Severson H. H. York: Chapman and Hall; 1993.

© 2008 The Authors. Journal compilation © 2008 Society for the Study of Addiction Addiction, 103, 1706–1712

You might also like