Journal of Substance Abuse Treatment 35 (2008) 148 – 155

Regular article

Predictors of changes in alcohol-related self-efficacy over 16 years

John McKellar, (Ph.D.)⁎, Mark Ilgen, (Ph.D.), Bernice S. Moos, (B.A.), Rudolf Moos, (Ph.D.)
Department of Veterans Affairs, Center for Health Care Evaluation, Palo Alto Health Care System, Menlo Park, CA 94025, USA
Stanford University School of Medicine, Stanford, Menlo Park, CA 94305, USA
Received 4 June 2007; received in revised form 17 September 2007; accepted 18 September 2007

Abstract

Self-efficacy is a robust predictor of short- and long-term remission after treatment. This study examined the predictors of self-efficacy in
the year after treatment and 15 years later. A sample of 420 individuals with alcohol use disorders was assessed five times over the course of
16 years. Predictors of self-efficacy at 1 year included improvement from baseline to 1 year in heavy drinking, alcohol-related problems,
depression, impulsivity, avoidance coping, social support from friends, and longer duration of participation in Alcoholics Anonymous (AA).
Female gender, more education, less change in substance use problems, and impulsivity during the first year predicted improvement in self-
efficacy over 16 years. Clinicians should focus on keeping patients engaged in AA, addressing depressive symptoms, improving patient's
coping, and enhancing social support during the first year and reduce the risk of relapse by monitoring individuals whose alcohol problems
and impulsivity improve unusually quickly. Published by Elsevier Inc.
Keywords: Alcohol; Self-efficacy; Longitudinal

1. Introduction Self-efficacy is hypothesized to determine whether, and

Alcohol dependence is often a chronic and relapsing
condition, and a full understanding of this disorder is only
possible when it is examined over the long term (McLellan,
Lewis, O'Brien, & Kleber, 2000). Consistent with this
conceptualization, many current psychosocial treatments of
drug and alcohol dependence reflect the importance of
monitoring and treating substance use over a prolonged
period (McKay, 2001, 2005). These treatment approaches
focus on the long-term monitoring of identified risk factors
to help patients decrease the likelihood of relapse. One factor
commonly targeted in long-term monitoring interventions is
self-efficacy or an individual's confidence in avoiding
problematic alcohol use (Bandura, 1982, 1997; DiClemente,
1986; Moos, 2007).
⁎ Corresponding author. Department of Veterans Affairs, Center for
Health Care Evaluation, Palo Alto Health Care System, 795 Willow Road
(MPD 152), Menlo Park, CA 94025, USA. Tel.: +1 650 493 5000x23366.
E-mail address: john.mckellar@va.gov (J. McKellar).
under what circumstances, individuals will experience a
relapse (DiClemente, 1986; Maisto, Carey, & Bradizza,
1999; Marlatt & Gordon, 1985). Consequently, many drug
and alcohol treatments are explicitly designed to target and
improve self-efficacy (Witkiewitz & Marlatt, 2004). Broad
support for the role of self-efficacy as a determinant of
positive treatment outcome has been found in alcohol- and
drug-dependent patients (Allsop, Saunders, & Phillips, 2000;
Goldbeck, Myatt, & Aitchison, 1997; Miller & Longabaugh,
2003; Rychtarik, Prue, Rapp, & King, 1992; Stephens,
Wertz, & Roffman, 1993). For example, Ilgen, McKellar,
and Tiet (2005) compared measures of self-efficacy to other
measures of alcohol use and psychosocial functioning; high
self-efficacy was the single strongest predictor of abstinence
from drug and alcohol use 1 year after treatment. Self-
efficacy also predicts positive treatment outcome at longer
intervals ranging from 3 to 16 years (Moos & Moos, 2006).
The consistent relationship between higher self-efficacy
and better substance use disorder (SUD) treatment outcomes
has led some researchers to explore whether specific
treatment (e.g., cognitive–behavioral therapy vs. 12-step-

0740-5472/08/$ – see front matter. Published by Elsevier Inc.

doi:10.1016/j.jsat.2007.09.003
 J. McKellar et al. / Journal of Substance Abuse Treatment 35 (2008) 148–155
oriented therapy) differentially influences self-efficacy.
Although self-efficacy tends to increase during formal
treatment (Ilgen et al., 2005; Stephens, Wertz, & Roffman,
1995), such increases do not appear to be influenced by
treatment type (DiClemente et al., 2001; Litt, Kadden,
Stephens, et al., 2005; Project MATCH Research Group,
1997; Stephens et al., 1993). In a recent study, (Glasner-
Edwards et al. 2007) found that patients with SUD with
concomitant depression randomized to cognitive–behavioral
therapy or 12-step facilitation showed comparable improve-
ment in self-efficacy and that improvement in self-efficacy
was related to better outcomes for both groups.
To the best of our knowledge, however, only a handful of
studies have explored predictors of posttreatment self-
efficacy other than type of treatment, and no studies of
which we are aware have examined long-term predictors of
the course of self-efficacy. Stephens et al. (1995) investi-
gated predictors of self-efficacy for patients with marijuana
use disorder and found that lower frequency of marijuana
use, less temptation to use, higher likelihood of coping
effectively with temptation, less perceived stress, and less
contact with other substance users were associated with
higher self-efficacy at the end of treatment. In a study of
2,000 male patients in the Department of Veterans, Ilgen,
McKellar, and Moos (2007) found that more years of
education, less severe substance-related problems, higher
confidence in abstinence as judged by a treatment provider,
and greater engagement with skills training during treatment
predicted more self-efficacy 1 year after treatment. Connors,
Tonigan, and Miller (2001) found that more participation in
Alcoholics Anonymous (AA) predicted posttreatment self-
efficacy in both outpatient and aftercare patients. An
additional study of the same data set found that the pattern
of results held for Type A and Type B alcohol-dependent
individuals (Bogenschutz, Tonigan, & Miller, 2006).
Thus, there is a consistent relationship between self-
efficacy and better substance use outcomes immediately
following treatment, in the year after treatment, and at longer
term follow-up. Further, preliminary evidence from a small
number of studies suggests that the determinants of higher
self-efficacy include more education, less severe SUD, less
perceived stress, better use of coping skills, more engage-
ment in skills training during treatment, and greater
engagement with AA (Bogenschutz et al., 2006; Connors
et al., 2001; Ilgen et al., 2007; Stephens et al., 1995). These
apparent short-term determinants of self-efficacy are quite
consistent with cognitive social learning theory (Bandura,
1977) and relapse prevention theory (Marlatt & Gordon,
1985; Witkiewitz & Marlatt, 2004).
(Bandura, 1977) identified several sources of self-
efficacy, including past experiences with the behavior (e.g.,
prior attempts to quit or cut down on substance use, which
may be reflected by less severe current problems), vicarious
experiencing and verbal persuasion or encouragement (e.g.,
exposure to supportive abstinent role models that often occur
in group therapy or self-help groups), and level of arousal/
149
impulsivity and distress. More recent versions of the model
by (Bandura, 1977) suggest that general social support can
also enhance self-efficacy. Marlatt and Gordon (1985)
proposed that self-efficacy is increased when patients are
taught to identify high-risk situations and cope with them
effectively. In addition to demographic variables, we follow
these ideas and view the potential predictors of posttreatment
self-efficacy as reflecting cognitive social learning variables
and treatment participation variables.
The extent to which these findings apply to the longer
term course of self-efficacy in individuals with an alcohol or
drug use disorder is unknown. As treatments shift focus to
target self-efficacy over the long term (McKay, 2005), better
data are needed on the factors that influence self-efficacy
over longer periods. We also do not know whether the
findings apply to women because most participants in prior
studies were men. Work by Skutle (1999) suggests that,
compared to men, women appear to possess higher self-
efficacy to resist using substance in the presence of positive
emotional states but lower self-efficacy to resist using
substances in the presence of negative emotions.
This study uses a sample of treatment-seeking indivi-
duals with alcohol use disorders to investigate predictors of
self-efficacy 1 year after baseline and then over a period of
15 more years. The analyses focus on how demographic
variables (e.g., education, age), social cognitive variables
(SUD severity, depression, impulsivity, and coping and
social support), and treatment participation (duration of
formal treatment and AA) predict 1-year self-efficacy and
the trajectory of self-efficacy over the ensuing 15 years,
using cognitive social learning theory as a framework. On
the basis of the limited number of prior studies, we
expected that higher education, less severe substance use
and depression, and more coping skills would predict
higher 1-year self-efficacy. Consistent with the updated
model of self-efficacy by (Bandura, 2004), we also
expected that more social support would be associated
with higher self-efficacy at 1-year follow-up. Finally,
because it reflects less resistance to temptation, we thought
that impulsivity would predict lower self-efficacy.
2. Methods
The sample was composed of individuals with alcohol
use disorders who had never received professional
treatment for these disorders prior to baseline. These
individuals were initially recruited after contacting an
information and referral center or detoxification program
for information about treatment for alcohol-related pro-
blems. After providing informed consent, 628 eligible
individuals completed a baseline assessment, which
represented 64% of patients approached for the study
(for more information about the initial data collection
process, see Finney & Moos, 1995). Data were collected
by trained research staff through a combination of mailings
150 J. McKellar et al. / Journal of Substance Abuse Treatment 35 (2008) 148–155
and telephone follow-up. Individuals who entered the
study had an alcohol use disorder, as determined by one or
more alcohol use problems, dependence symptoms, drink-
ing to intoxication in the past month, and/or perception
of alcohol abuse as a significant problem. At 1, 3, 8, and
16 years after entering the study, participants were
contacted to complete a follow-up assessment.
A total of 121 of the 628 baseline participants (19%) had
died by the 16-year follow-up. Of the remaining 507
participants, 420 (83%) completed the 1-year follow-up
and at least two of the assessments at 3-, 8-, and 16-year
follow-ups. These 420 individuals comprise the core sample
for the analyses reported here. Compared to surviving
individuals who did not complete the 1-year assessment,
those who were followed up at 1 year were more likely to be
female (51% vs. 39%; χ2 = 3.9, p b .05) and Caucasian
(82% vs. 72%; χ2 = 3.8, p b .05) and to report more years of
education (mean of 13.2 years vs. 12.5 years), F(1, 505) =
7.5, p b .01. On average, individuals with complete 1-year
data reported fewer drinks per drinking day (12.1 vs. 15.6), F
(1, 505) = 7.5, p b .01, and fewer alcohol problems (10.3 vs.
13.2), F(2, 505) = 12.3, p b .01, than did those without
complete data.
2.1. Measures
In addition to obtaining demographic information, we
asked respondents about their alcohol use and alcohol-
related problems, psychosocial functioning, and participa-
tion in professional treatment and AA.
2.2. Social cognitive variables
2.2.1. SUD severity
Questions from the Health and Daily Living Form (HDL;
Moos, Cronkite, & Finney, 1992) were used at baseline and
1-year follow-up to assess change in alcohol consumption
and problems related to alcohol use. Average amount of
alcohol consumed on drinking days was assessed by
responses to three questions asking how often and what
amount per week (never, less than once, once or twice, three
to four times, nearly every day) participants had consumed
wine, beer, and hard liquor in the last month. On the index of
current drinking problems, respondents rated how often (on a
5-point scale varying from 0 = never to 4 = often) in the last
6 months they had experienced each of nine problems (e.g.,
with health, job, money, family arguments) as a result of
drinking (α at baseline = .80). Despite covering an array of
problem areas, factor analysis of the drinking problems
measure suggests a unitary factor solution with all items
loading ≥.4.
2.2.2. Distress and personality
Depression was assessed with a measure derived from the
Research Diagnostic Criteria and included in the HDL
(Moos et al., 1992). Respondents rated how often (on a
5-point scale with 0 = never and 4 = often) they
experienced each of nine symptoms of depression in the
last month (α = .92) at baseline and 1-year follow-up.
Impulsivity was assessed at baseline and 1-year follow-up
by the impulsivity scale of the Differential Personality
Inventory (Jackson & Messick, 1986). Respondents rated
from 1 (strongly disagree) to 4 (strongly agree) their level of
agreement with statements focusing on lack of planning
(e.g., “I like to take time for planning when I do
something”—reverse coded) or impulsive behavior (“I
usually act on the first thought that comes into my head”;
α at baseline = .74).
2.2.3. Coping
Coping was assessed by two 6-item subscales, which
measured problem solving and avoidance (α at baseline = .75
and .57, respectively). The subscales were composed of
4-point items ranging from no to fairly often drawn primarily
from the Coping Responses Inventory (Moos, 1993) that
were completed at baseline and at 1-year follow-up.
2.2.4. Social support
This domain was assessed by asking participants about
friend-related resources (α = .83) and stressors (α = .75) at
baseline and 1 year on several 5-point items drawn from the
Life Stressors and Social Resources Inventory (Moos &
Moos 1994).
2.3. Participation in treatment and AA
At the 1-year follow-up, individuals were asked
whether or not they had participated in professional
treatment or AA for their drinking habits or drinking-
related problems since baseline. If participants answered
“yes” for an episode of treatment, they were asked to
record the number of weeks. In addition, participants
were asked to record the number of weeks for each
episode of participation in AA. Based upon prior
findings (Moos & Moos, 2006), the number of weeks
of treatment was placed into one of four categories (0,
1–8, 9–26, 27+ weeks) and the number of weeks of AA
was also placed into one of four categories (0, 1–16,
17–32, 33+ weeks).
2.4. Primary outcome: Alcohol-related self-efficacy
Alcohol-related self-efficacy or confidence to resist
alcohol consumption was assessed with 14 items (α =
.93 at baseline) from the Situational Confidence Ques-
tionnaire (Annis & Graham, 1988). The items covered
situations involving negative and positive emotions,
interpersonal conflict, and testing one's self-control.
Each item was rated on a 6-point scale varying from
not at all confident to very confident; participants
completed this measure at baseline and at 1-, 3-, 8-,
and 16-year follow-ups.
 J. McKellar et al. / Journal of Substance Abuse Treatment 35 (2008) 148–155
2.5. Analytic plan
We modeled the 420 individual trajectories (status at
1 year and slope across 1, 3, 8, and 16 years) of alcohol-
related self-efficacy over the 16-year assessment period
using mixed-effects regression, as described in Singer and
Willett (2003) and Bryk and Raudenbush (2001). One of
the strengths of mixed-effects regression models is that
trajectories can be estimated for individuals who do not
provide data at all assessment periods. The estimates of
parameters for individuals with incomplete data were
calculated with all available data for each person, plus
information about the parameters of similar individuals, and
are appropriately associated with error due to uncertainty. In
this way, mixed-effects regression models use all available
data to estimate trajectories.
For each individual, a linear trajectory was modeled,
yielding three parameter estimates: (a) the individual's level
of alcohol-related self-efficacy at 1 year; (b) the individual's
slope across 1, 3, 8, and 16 years, a measure of change over
time with positive slope indicating an increase in self-
efficacy and negative slope indicating a decrease in self-
efficacy; and (c) error, a measure of how well the linear
model fits that person's data. The models were constructed
so that intercepts corresponded to 1-year self-efficacy.
Predictors of self-efficacy were either based upon baseline
to 1-year change scores or mean centered (for the duration of
formal treatment and AA) to aid interpretation of the
intercept as the value of the outcome of self-efficacy at
average values.
To ease interpretation of our results and capitalize upon
the longitudinal nature of the data, we organized our analyses
temporally. The initial set of analyses predicted the
individual's level of alcohol-related self-efficacy at 1 year.
The predictors were baseline demographic variables, base-
line to 1-year change (improvement) in SUD severity,
depression and impulsivity, and coping and social support, as
well as the duration of participation in treatment and AA in
the first year. The next set of analyses involved using the
same predictors to predict the trajectory of an individual's
alcohol-related self-efficacy from the 1-year to the 16-year
follow-up data. Baseline alcohol-related self-efficacy was
used as a covariate in all analyses.
Due to the lack of prior research predicting long-term
self-efficacy, we chose to take an exploratory rather than a
confirmatory approach. Specifically, we assessed the
predictive significance of each of the indices described
above by examining their relationship to self-efficacy
outcomes in separate multilevel regression analyses. By
casting a wide net, we increased the number of statistical
tests performed, thereby increasing the probability of a
Type I error in the service of reducing Type II error. The
criterion for significance in analyses predicting 1-year
alcohol-related self-efficacy was set to .01, and the
criterion for predicting the 1- to 16-year slope of self-
efficacy was set to .05.
151
3. Results
3.1. Correlation between alcohol-related self-efficacy and
abstinence at 16 years
To provide a context for understanding the analyses that
follow, we calculated a set of correlations (Kendall's tau-b)
between self-efficacy at baseline and at 1-, 3-, 8-, and
16-year follow-up and abstinence at 16-year follow-up.
Greater self-efficacy was consistently related to a higher
likelihood of abstinence, with significant correlations
ranging from .06 at baseline to .15 at 1-year follow-up to
.54 at the 16-year follow-up.
3.2. Predictors of alcohol-related self-efficacy at 1 year
The largest increase in alcohol-related self-efficacy
occurred between baseline and 1 year (Table 1). The
means and standard deviations of the baseline and treatment
values of predictors of self-efficacy are shown in Table 2.
Several baseline to 1-year change variables predicted
1-year level of alcohol-related self-efficacy (Table 3).
Greater improvement in amount of heavy drinking and in
drinking problems, depression, impulsivity, avoidance
coping, and friend-related resources all predicted higher
alcohol-related self-efficacy at 1 year. A longer duration
of attendance of AA in the first year also predicted higher
1-year self-efficacy.
3.3. Predictors of the trajectory of alcohol-related self-
efficacy over 16 years
There was significant variability in the trajectory of self-
efficacy from the 1-year to the 16-year follow-up. Thus, for
each patient, we can identify characteristics that are
associated with improvement (positive slope) or deteriora-
tion (negative slope) of self-efficacy. Two baseline demo-
graphic variables and two baseline to 1-year change
variables significantly predicted changes in alcohol-related
self-efficacy (Table 3). More education and female gender
were associated with improvement in self-efficacy from the
1-year to the 16-year follow-up. Greater improvement in
alcohol problems and impulsivity from baseline to 1-year
predicted deterioration in self-efficacy from the 1-year to the
16-year follow-up.
Table 1
Means and standard deviations of controlled drinking self-efficacy over
16 years
Period M SD
Intake 64.7 25.5
Year 1 79.1 22.8
Year 3 81.8 23.0
Year 8 84.7 21.1
Year 16 81.6 22.9
152 J. McKellar et al. / Journal of Substance Abuse Treatment 35 (2008) 148–155

Table 2 are scaled to the metric of the predictors. However,

Baseline demographic and psychosocial variables and duration of treatment comparing the deviance of the model containing a predictor
in the first year
to a benchmark model with no predictors (the unconditional
Baseline/First-year values growth model) yields a change in deviance statistic, ΔD, that
Variables M or % SD follows a chi-square distribution (df = 2). The size of the
Demographic variables change in the deviance statistic can be used as an effect size
Education (years) 13 2.3 estimate to identify the predictors that provide the most
Age (years) 33 8.7 improvement from the unconditional growth model (Table 3,
Female gender 51% n/a last column). For example, duration of AA participation in
Employed (part-time or higher) 45% n/a
Married 76% n/a
the first year and improvement (decrease) in avoidance
Baseline social cognitive variables coping were the strongest predictors of self-efficacy at
SUD severity 1 year, with ΔD values of 2,005 and 1,442, respectively.
Heavy drinking 6.5 7.4
Alcohol-related problems 9.1 6.4
Distress and personality 4. Discussion
Depression (0–36) 20.9 8.9
Impulsivity (0–30) 14.7 4.5
Mean levels of self-efficacy increased notably during
Coping
Problem-solving action (6–24) 15.8 4.4 the initial year after seeking help for alcohol-related
Avoidance coping (6–24) 14.7 3.9 problems and, on average, remained high over the
Social support subsequent 15 years. In addition, there was significant
Friend stressors (0–16) 5.7 2.5 individual variability in self-efficacy over time. At 1 year,
Friend resources (0–24) 17.3 4.7
higher self-efficacy was predicted by greater improvement in
Treatment participation in first year
Duration of formal treatment 12.9 20.8 frequency of heavy drinking and drinking problems; more
Duration of AA 17.5 22.4 improvement in depression, impulsivity, avoidance coping,
and social support; and a longer duration of participation in
AA between baseline and 1 year. Examination of long-term
3.4. Relative magnitude of regression coefficients in models patterns revealed that women and individuals with more
education reported continued increases in self-efficacy over
It is not possible to compare the magnitude of the time. However, individuals who experienced the greatest
regression coefficients between models because the effects initial improvements in alcohol-related problems and

Table 3
Baseline demographics, baseline to 1-year change scores, and treatment participation predicting alcohol-related self-efficacy at 1 year and across 16 years using
mixed-effects regressions
Alcohol-related self-efficacy Alcohol-related self-efficacy
Test of deviance/“Δ”
at 1 year across time to 16 years
from unconditional
Variables β p β p model
Baseline demographics
Gender (1 = female) 1.04 ns 0.417 ⁎ b.019 χ2(2) = 14.1, p b .01
Age 0.207 ns 0.012 ns
Education (years) 0.99 ns 0.08 ⁎ b.037 χ2(2) = 15.2, p b .01
Baseline to 1-year change scores in social cognitive variables
SUD severity
Heavy drinking 0.357 ⁎⁎ b.002 −0.011 ns
Alcohol-related problems 0.827 ⁎⁎ b.000 −0.031 ⁎ b.018 χ2(2) = 43, p b .01
Distress and personality
Depression 0.565 ⁎⁎ b.000 −0.01 ns χ2(2) = 75, p b .01
Impulsivity 0.638 ⁎⁎ b.003 −0.048 ⁎ b.021 χ2(2) = 49, p b .01
Coping
Problem-solving action 0.392 ns −0.029 ns
Avoidance coping 0.924 ⁎⁎ b .000 −0.026 ns χ2(2) = 1,442, p b .01
Social support
Friend stressors 0.452 ns −0.04 ns
Friend resources 0.856 ⁎⁎ b.000 −0.013 ns χ2(2) = 199, p b .01
Treatment participation in first year
Duration of formal treatment 1.47 ns 0.026 ns
Duration of AA 3.93 ⁎⁎ b.000 0.149 ns χ2(2) = 2,005, p b .01
⁎ p b .05.
⁎⁎ p b .01.
 J. McKellar et al. / Journal of Substance Abuse Treatment 35 (2008) 148–155
impulsivity were less likely to maintain high levels of self-
efficacy over the long term.
Changes in self-efficacy were most evident during the
first year after seeking help for problem drinking, and a
number of factors predicted the change. In contrast to prior
studies, this study investigated whether improvement in
domains measured during the first year predicted the level of
self-efficacy at the 1-year follow-up, even after accounting
for baseline self-efficacy. Despite this methodological
difference, our findings are consistent with prior studies
that found improvements in severity of SUD, depression, and
coping to predict greater self-efficacy. Thus, individuals who
have more severe problems and fewer personal resources at
baseline are not destined to continue to have lower self-
efficacy. In fact, substantial improvements in these domains
are associated with higher subsequent self-efficacy. Our
findings that improvements in impulsivity and social support
predict greater self-efficacy at 1 year are novel and help
extend knowledge in this area of research.
In addition to these patient-related factors, a longer
duration of contact with AA was associated with a larger
increase in self-efficacy during the first year. In fact,
duration of AA participation was the strongest single
predictor of self-efficacy at 1 year. This is consistent with a
growing body of research describing an association between
self-help participation and higher subsequent self-efficacy
(Ilgen et al., 2007; Morgenstern, Labouvie, McCrady,
Kahler, & Frey, 1997). As noted by Connors et al.
(2001), some of the primary goals of AA and other self-
help groups related to the admission of powerlessness
appear to play down the importance of self-efficacy.
Nevertheless, participation in AA is associated with
enhanced self-efficacy, perhaps due to AA members'
support for and modeling of abstinence (Alcoholics
Anonymous, 1939; Moos & Moos, 2006). AA involvement
may also be associated with self-efficacy because most
measures of self-efficacy focus on the degree of confidence
about not drinking heavily rather than on why a person feels
confident about not drinking heavily.
For longer term interventions to be effective, it is not
enough to know who will do well initially; it is also
necessary to identify individuals at risk for future problems.
The present findings suggest that individuals who report the
greatest initial improvements in alcohol-related problems
and impulsivity may have difficulty maintaining a high level
of self-efficacy. This may reflect the fact that individuals
with the sharpest initial improvements had the most severe
problems and, thus, at baseline, were at greatest overall risk
for relapse. Alternatively, rapid initial improvement may not
allow enough time to develop some of the core behavioral
skills that play an important role in maintaining self-efficacy
(Witkiewitz & Marlatt, 2004). In any case, our results
indicate that identification and ongoing monitoring of
individuals whose alcohol problems and impulsivity
improve quickly are important to solidify gains and reduce
the risk for future problems.
153
Female gender and more education were associated
with a gradual increase in self-efficacy over the follow-up
interval. Prior research has found that women have higher
levels of self-efficacy to resist drinking when experiencing
pleasant emotional states (Skutle, 1999). Because contin-
ued behavior rehearsal is likely to increase and stabilize
self-efficacy, women's initial abilities to manage responses
to positive emotional states may grow more pronounced
with practice over time. Similarly, better educated
individuals may have a greater ability to consolidate initial
gains and to gradually build on these strengths over the
long term.
This study has implications for how providers can
increase their patients' self-efficacy immediately after
treatment and over the longer term. Bolstering a patient's
self-efficacy during treatment, and in the year after treatment,
should involve teaching problem-solving coping techniques,
reducing depression, and helping patients focus on long-term
rewards associated with decreasing alcohol use. Encouraging
attendance in AA also appears to improve self-efficacy in the
short term and may help patients increase the number of
friends who are supportive of their recovery. Attempts to
influence patients' self-efficacy over longer periods may
require increased focus on patients who have less education
and on those who report apparent rapid improvement in
impulsivity early in treatment. To counteract these risk
factors, providers might create low-intensity, long-term
methods of monitoring these patients (e.g., via telephone)
and/or encourage longer term engagement with AA or other
self-help groups.
The current results can be viewed using cognitive
social learning theory as a guide. For example, more
education and provision of coping skills training during
treatment may influence self-efficacy by increasing the
coping resources of a patient with SUD. Less severe
SUD at treatment entry may reflect initial attempts to
reduce substance use and, as a form of successful past
behavior, increase the later development of self-efficacy.
Less depression may reflect patients' success in managing
their SUD and, thus, portend that they will do well in the
future. Finally, attendance at AA or other self-help groups
may increase self-efficacy by providing an opportunity
for patients to hear other member's success stories
(vicarious experiencing).
This research has several limitations. First, this was a
naturalistic study and not a randomized controlled trial;
hence, the degree to which variations in predictors caused
changes in subsequent self-efficacy is unknown. How-
ever, effort was made to capitalize on the longitudinal
nature of the data by organizing the analyses temporally
and minimizing the use of purely contemporaneous
prediction. The sample consisted of participants who
had never previously engaged in SUD treatment; thus, we
do not know whether the present findings apply to
individuals with prior alcohol and/or drug treatment
histories or to non-treatment-seeking populations. Addi-
154 J. McKellar et al. / Journal of Substance Abuse Treatment 35 (2008) 148–155
tionally, our sample represented only 83% of those who
were still alive and initially participated in the baseline
assessment. Because preliminary analyses identified
differences between individuals with and without com-
plete data, it is possible that our sample was not fully
representative of the initial baseline sample. However, we
utilized an analytic technique that maximized the use of
the available data and our ability to generalize the
resulting findings.
Finally, our measures of participation in AA and formal
treatment reflected help as it was provided in the real
world. Treatment likely varied in the degree to which it
incorporated principles of relapse prevention and an
explicit focus on self-efficacy. As described earlier,
treatment type is generally unrelated to measures of
posttreatment self-efficacy (e.g., DiClemente et al., 2001),
but it remains possible that the content of the treatment
provided to the present sample may not have focused
explicitly on self-efficacy. Future research is needed to
examine how treatments that explicitly target self-efficacy
influence (e.g., McKay et al., 2004) the level and stability
of self-efficacy over time.
Despite these limitations, this study extends the
developing literature on predictors of self-efficacy in the
first year after seeking help for alcohol problems and is
the first to identify determinants of the long-term
trajectory of change in self-efficacy. As treatments shift
their focus to the ongoing management of SUDs (McKay,
2005), better data are needed to characterize the course
and stability of factors that play a primary role in
predicting subsequent substance use. Identifying patients
who have difficulty solidifying initial gains in self-
efficacy could help treatment providers tailor long-term
interventions to focus on those at elevated risk for a
future relapse.
Acknowledgments
This project was supported in part by National Institute on
Alcohol Abuse and Alcoholism Grant AA12718 and by the
Department of Veterans Affairs Health Services Research
and Development Service. The views expressed herein do
not necessarily express those of the Department of Veterans
Affairs. Some of the findings were presented at the 30th
annual meeting of the Research Society on Alcoholism
(2007) in Chicago, IL.
References
Alcoholics Anonymous. (1939). Alcoholics Anonymous. New York, NY:
Works Publishing Inc.
Allsop, S., Saunders, B., & Phillips, M. (2000). The process of
relapse in severely dependent male problem drinkers. Addiction, 95,
95−106.
Annis, H. M., & Graham, J. M. (1988). Situational Confidence
Questionnaire (SCQ-39) user’s guide. Toronto, Ontario, Canada:
Alcoholism and Drug Addiction Research Foundation.
Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral
change. Psychological Review, 84, 191−215.
Bandura, A. (1982). The assessment and predictive generality of self-
percepts of efficacy. Journal of Behavioral Therapy and Experimental
Psychiatry, 13, 195–199.
Bandura, A. (1997). Self-efficacy: The exercise of control. New York, NY:
W.H. Freeman.
Bandura, A. (2004). Health promotion by social cognitive means. Health
Education and Behavior, 31, 143−164.
Bogenschutz, M. P., Tonigan, J. S., & Miller, W. R. (2006). Examining the
effects of alcoholism typology and AA attendance on self-efficacy as a
mechanism of change. Journal of Studies on Alcohol, 67, 562−567.
Bryk, A. S., & Raudenbush, S. W. (2001). Linear models: Applications and
data analysis methods (2nd ed.). Thousand Oaks, CA: Sage.
Connors, G. J., Tonigan, J. S., & Miller, W. R. (2001). A longitudinal model
of intake symptomatology, AA participation and outcome: Retrospective
study of the Project MATCH outpatient and aftercare samples. Journal
of Studies on Alcohol, 62, 817−825.
DiClemente, C. C. (1986). Self-efficacy and the addictive behaviors. Jour-
nal of Social and Clinical Psychology, 4, 302−315.
DiClemente, C. C., Carbonari, J. P., Walker-Daniels, J., Donovan, D. M.,
Bellino, L. E., & Neavans, T. M. (2001). Self-efficacy as a secondary
matching hypothesis: Causal chain analysis. In R. Longabaugh, & P. W.
Wirtz (Eds.), Project MATCH hypotheses: Results and causal chain
analyses, Vol. 8. (pp. 239−259) (Project MATCH Monograph Series
NIH Pub No. 01-4238). Bethesda, MD: National Institute on Alcohol
Abuse and Alcoholism.
Finney, J. W., & Moos, R. H. (1995). Entering treatment for alcohol abuse: A
stress and coping model. Addiction, 90, 1223−1240.
Glasner-Edwards, S., Tate, S. R., McQuaid, J. R., Cummins, K., Granholm,
E., & Brown, S. A. (2007). Mechanisms of action in integrated
cognitive-behavioral treatment versus twelve-step facilitation for
substance-dependent adults with comorbid major depression. Journal
of Studies on Alcohol and Drugs, 68, 663−672.
Goldbeck, R., Myatt, P., & Aitchison, T. (1997). End-of-treatment self-
efficacy: A predictor of abstinence. Addiction, 92, 313−324.
Ilgen, M., McKellar, J., & Moos, R. (2007). Personal and treatment-related
predictors of abstinence self-efficacy. Journal of Studies on Alcohol, 68,
126−132.
Ilgen, M., McKellar, J., & Tiet, Q. (2005). Abstinence self-efficacy and
abstinence one year after substance use disorder treatment. Journal of
Consulting and Clinical Psychology, 73, 1175−1180.
Jackson, D. N., & Messick, S. (1986). The differential personality inventory.
London: Authors.
Litt, M. D, Kadden, R. M., & Stephens, R. S. (2005). Coping and self-
efficacy in marijuana treatment: Results from the marijuana treatment
project. Journal of Consulting and Clinical Psychology, 73, 1015−1025.
Maisto, S. A., Carey, K. B., & Bradizza, C. M. (1999). Social learning
theory. In K. E. Leonard, & H. T. Blane (Eds.), Psychological theories of
drinking and alcoholism. The Guilford substance abuse series (2nd ed.,
pp. 106–163). New York, NY: Guilford Press: Guilford Press.
Marlatt, G. A., & Gordon, J. R. (1985). Relapse prevention: Maintenance
strategies in the treatment of addictive behaviors. New York, NY:
Guilford.
McKay, J. R. (2001). Effectiveness of continuing care interventions for
substance abusers: Implications for the study of long-term treatment
effects. Evaluation Review, 25, 211−232.
McKay, J. R. (2005). Is there a case for extended interventions for alcohol
and drug use disorders? Addiction, 100, 1594−1610.
McKay, J. R., Foltz, C., Leahy, P., Stephens, R., Orwin, R. G., & Crowley, E.
M. (2004). Step down continuing care in the treatment of substance
abuse: Correlates of participation and outcome effects. Evaluation &
Program Planning, 27, 321−331.
McLellan, A. T., Lewis, D. C., O'Brien, C. P., & Kleber, H. D. (2000). Drug
dependence, a chronic medical illness: Implications for treatment
insurance, and outcomes evaluation. Journal of the American Medical
Association, 284, 1689−1695.
 J. McKellar et al. / Journal of Substance Abuse Treatment 35 (2008) 148–155
Miller, W. R., & Longabaugh, R. (2003). Summary and conclusions. In T. F.
Babor, & F. K. Del Boca (Eds.), Treatment matching in alcoholism
international research monographs in the addictions (pp. 207–221). New
York, NY: Cambridge University Press: Cambridge University Press.
Moos, R. H. (1993). Coping Responses Inventory manual. Odessa, FL:
Psychological Assessment Resources.
Moos, R. H. (2007). Theory-based processes that promote the remission of
substance use disorders. Clinical Psychology Review, 27, 537−551.
Moos, R. H., Cronkite, R., & Finney, J. (1992). Health and Daily Living
Form manual (2nd ed.). Redwood City, CA: Mind Garden.
Moos, R. H., & Moos, B. S. (1994). Life stressors and social resources
inventory manual, psychological assessment resources. Odessa,
FL: Psychological Assessment Resources.
Moos, R. H., & Moos, B. S. (2006, Feb). Rates and predictors of relapse
after natural and treated remission from alcohol use disorders. Addiction,
101, 212−222.
Morgenstern, J., Labouvie, E., McCrady, B. S., Kahler, C. W., & Frey, R. M.
(1997). Affiliation with Alcoholics Anonymous after treatment: A study
of its therapeutic effects and mechanisms of action. Journal of
Consulting and Clinical Psychology, 65, 768−777.
155
Project MATCH Research Group. (1997). Matching alcoholism treatments
to client heterogeneity: Project MATCH posttreatment drinking out-
comes. Journal of Studies on Alcohol, 58, 7−29.
Rychtarik, R. G., Prue, D. M., Rapp, S. R., & King, A. C. (1992). Self-
efficacy, aftercare and relapse in a treatment program for alcoholics.
Journal of Studies on Alcohol, 53, 435−440.
Singer, J. D., & Willett, J. B. (2003). Applied longitudinal analysis:
Modeling change and event occurrence. Oxford: Oxford University
Press.
Skutle, A. (1999). Association between gender and marital status and
confidence in remaining abstinent among alcohol abusers in treatment.
Addiction, 94, 1219−1225.
Stephens, R. S., Wertz, J. S., & Roffman, R. A. (1993). Predictors of
marijuana treatment outcomes: The role of self-efficacy. Journal of
Substance Abuse, 5, 341−354.
Stephens, R. S., Wertz, J. S., & Roffman, R. A. (1995). Self-efficacy and
marijuana cessation: A construct validity analysis. Journal of Consulting
and Clinical Psychology, 63, 1022−1031.
Witkiewitz, K., & Marlatt, A. G. (2004). Relapse prevention for alcohol and
drug problems. American Psychologist, 59, 224−235.