220 JOURNAL OF STUDIES ON ALCOHOL / MARCH 2005
Individual and Partner Predictors of Recovery from
Alcohol-Use Disorder over a Nine-Year Interval: Findings
from a Community Sample of Alcoholic Married Men*
MARY J. MCAWEENEY, PH.D.,† ROBERT A. ZUCKER, PH.D., HIRAM E. FITZGERALD, PH.D.,† LEON I. PUTTLER, PH.D.,
AND MARIA M. WONG, PH.D.
Department of Psychiatry and Addiction Research Center, University of Michigan, Ann Arbor, Michigan
ABSTRACT. Objective: Numerous studies have focused on the pre-
dictors of recovery in persons with alcohol use disorder (AUD). Most
have been retrospective and have measured only predictors of short-term
recovery after the completion of treatment. This prospective study evalu-
ates the role of psychological and social factors in a community sample
of both alcoholics and their partners in predicting recovery over a 9-
year interval. Method: Alcoholic diagnostic status and life functioning
of 134 community-recruited, initially coupled men meeting criteria for
a 3-year Diagnostic and Statistical Manual of Mental Disorders, Fourth
Edition, AUD diagnosis at baseline were assessed over the 9-year pe-
riod. Their partners also were assessed. Prediction involved comparing
those who still had an AUD diagnosis at 9-year follow-up against those
who no longer met AUD criteria. Results: Predictors of recovery in-
cluded number of experiences with treatment, education, number of
A LTHOUGH THERE HAS BEEN a moderate amount
of research on the predictors of recovery from an al-
cohol use disorder (AUD), most studies have been retro-
spective and/or have measured only predictors of short-term
recovery after the completion of treatment (Bischof et al.,
2000; Humphreys and Moos, 2001; Kaskutas et al., 2002;
McCrady et al., 1999, 2002; Sobell and Sobell, 1998;
Walters, 2000; Weisner et al., 2003). Common time points
for follow-up are 6 months to 1 year after treatment. Few
studies have used community samples, and even fewer have
looked at the longer-term outcomes that would be indica-
tive of a sustained pattern of recovery. To our knowledge,
no studies have examined the influence of partner variables
Received: April 27, 2004. Revision: October 11, 2004.
*This study was part of the first author’s postdoctoral research fellow-
ship at the University of Michigan and was supported by National Institute
on Alcohol Abuse and Alcoholism (NIAAA) grant T32 AA07477. This re-
search was also supported by NIAAA grant R37 AA07065 to Robert A.
Zucker and Hiram E. Fitzgerald.
†Correspondence should be sent to Mary J. McAweeney, Substance Abuse
Resources and Disabilities Issues Program, Wright State School of Medi-
cine, 3171 Research Blvd., Kettering, OH 45420, or via email at:
mary.mcaweeney@wright.edu. Hiram E. Fitzgerald is with the Department
of Psychology, Michigan State University.
220
years of intervening recovery over the follow-up period, partner baseline
AUD status and partner’s social support network. Alcoholics’ initial se-
verity of drinking did not predict long-term outcome. Furthermore, re-
covered men’s partners decreased their AUD in the interim, whereas
nonremitters’ partners increased their AUD. The findings highlight the
transitions in and out of AUD, with 62% of the men having stable drink-
ing patterns of either remission or unremitting AUD diagnosis over the
entire follow-up period. Conclusions: This study demonstrated the im-
portance of interpersonal factors in maintaining AUD or promoting re-
covery. Consideration of partner characteristics and the marital context
as factors in the recovery process is essential. Future research should
examine the predictors of recovery in women, in adolescents and in ra-
cial groups other than white. (J. Stud. Alcohol 66: 220-228, 2005)
in predicting spouses’ long-term recovery. Consequently,
the exact nature of how the psychological and social fac-
tors of both the alcoholic and his or her partner influence
the recovery process over an extended period remains
unclear.
We addressed this issue by focusing on the 9-year pre-
diction of recovery within a community sample of married
men who have AUD. We also examined the potential in-
fluence of partners in the initiation and/or maintenance of
continued drinking or recovery. Knowledge of these pre-
dictors may suggest ways to design better interventions for
alcoholics and other drug abusers (Connors et al., 2001;
Epstein and McCrady, 1998; Humphreys et al., 2004;
McCrady et al., 2002; Moos and Moos, 2003a, 2004) as
well as to develop practical and useful techniques for pre-
venting relapse (Bischof et al., 2000; Connors et al., 1996;
Humphreys and Tucker, 2002).
In one of the few community-based, longitudinal studies
reported in the literature, Humphreys et al. (1997) followed
395 (50.1% men) untreated alcoholics over 8 years and
correctly classified 70% of persons in recovery, based on a
psychosocial model. This nontreatment study included sub-
jects who, over the course of follow-up, may or may not
have been involved in formal or informal treatment.
 McAWEENEY ET AL.
Humphreys et al. (1997) found social/community variables
to be more predictive of recovery than demographic vari-
ables. Significant social/community variables included ac-
tive membership in a religious organization, extent of
outpatient treatment sessions, family relationships, and at-
tendance at Alcoholics Anonymous (AA) meetings. Only
one demographic variable, gender, predicted recovery, with
women being more likely to recover.
These findings are consistent with results found in other
studies involving persons who were either treated or un-
treated for their alcoholism (Bischof et al., 2000; Finney
and Moos, 1991; Moos and Moos, 2003b; Sobell et al.,
1993, 1996, 2000; Timko et al., 2000). In a recent study
involving initially untreated individuals—participants who
at baseline had not received professional treatment for their
AUD but were recruited during their initial contact with a
treatment system—Moos and Moos (2003b) found that more
participation in AA or other treatment over the 8-year fol-
low-up predicted remission. Also, they found that partici-
pants who consumed more alcohol and were intoxicated
more often, those who had a heavier drinking pattern and
those who had increased tolerance were more likely to be
nonremitted at 8-year follow-up. In contrast, Humphreys et
al. (1997) found that addictive behavior factors such as
alcohol consumption and alcohol dependence at baseline
did not predict recovery.
Humphreys et al. (1997) noted some optimism, with half
of the sample free of alcohol-related problems at 8-year
follow-up. Moos and Moos (2003b) reported similar rates
of remission, with variation dependent on the number of
risk factors at baseline. Subjects with more than nine risk
factors had a remission rate of 22%, and those with two or
fewer risk factors had a remission rate of 76% after 8 years.
Similarly, Finney and Moos (1991) reported at 10-year fol-
low-up that 42% of the sample had remitted from alcohol
and had done so for at least a 2-year period. Individuals
were classified as remitted if they had missed no work; had
not been rehospitalized for alcoholism; had consumed less
than 3 ounces of ethanol per day, on average, in the past
month; and had had no problems associated with drinking
in the past year. Yet, in all of these studies, a sizable por-
tion of the samples reported severe alcohol problems
throughout the extended follow-up intervals. These find-
ings underscore the importance of the need for continued
research in this area.
The existing literature indicates that recovery involving
continued practice of lifestyle change reinforces and per-
petuates a desire to maintain the change, in a positive spi-
ral (Kaskutas et al., 2002; Prochaska et al., 1992; Vaillant,
1995). For instance, in both natural and treatment settings,
a pattern of decreased drinking or total abstinence from
drinking can produce a rise in self-esteem, a lessening of
depression and a greater confidence to resist drinking
(Blomqvist, 1999; Hall et al., 1991; Humphreys et al., 1994).
221
Prochaska and colleagues (1992) have highlighted the need
to view recovery as a behavior change process, and they
have suggested the value of studying this process over a
lengthy period that begins with minimal behavior change
and progresses through the maintenance phase of recovery.
The present study utilized a psychosocial model similar
to that of Moos and Moos (2003b) and Humphreys et al.
(1997). First, the model included personal and social vari-
ables of the alcoholics. Second, we included personal and
social variables of the partners of those with AUD, mea-
sured at baseline. The importance of including partner vari-
ables in predicting outcomes in other chronic illnesses, such
as pain and myocardial infarction, has been well documented
(Flor et al., 1987; Helder et al., 2002; Schwartz et al., 1997;
Waltz et al., 1988), but their inclusion is not common in
the longitudinal alcoholism literature. Third, based on the
Prochaska et al. (1992) theory of change, we included in-
tervening variables in the prediction model—namely, num-
ber of experiences with formal treatment, amount of AA
participation, number of years in recovery and number of
years smoking. The inclusion of these indicators is consis-
tent with a perspective that views continued practice of
decreased drinking or total abstinence and other healthy
behaviors (i.e., participation in AA, nonsmoking) as rein-
forcers that perpetuate the desire to change and maintain
the change (Kaskutas et al., 2002; Prochaska et al., 1992;
Vaillant, 1995). The model proposes that, the more action
taken, the higher the likelihood of recovery. Thus, these
factors should be important, chronologically intermediate
variables in the model.
Personal variables used included level of education and
seven symptom and help-seeking behavior variables (de-
pressive symptoms, antisocial behavior, years of smoking,
severity of drinking, number of years in remission, amount
of participation in formal treatment and attendance at AA).
Partner variables included depressive symptoms, antisocial
behavior and severity of drinking. Although other variables,
such as partner education and smoking, were available to
us, the addition of more partner variables would have re-
sulted in a substantial loss of power. Social variables in-
cluded extent of social support and family cohesion for
both family members. All predictors have substantial sup-
port in the literature (Bischof et al., 2000; Grant, 1998;
Humphreys et al., 1997; Kaskutas et al., 2002; McCrady et
al., 1999; Moos and Moos, 2004; Zucker et al., 1994a).
We hypothesized that alcoholics who are in recovery at
follow-up will have more education, higher family cohe-
sion, fewer depressive symptoms, less antisocial behavior,
less severe drinking, a larger social support network, more
years of intervening recovery, more experiences with treat-
ment and fewer years of smoking, when compared with a
group of men who continue to meet diagnostic criteria.
In addition, the partners of those who recover will more
likely be free of an AUD at baseline, report higher family
222 JOURNAL OF STUDIES ON ALCOHOL / MARCH 2005
cohesion, have a larger social support network, report fewer
depressive symptoms and have less antisocial behavior than
those partnered with men who continue to meet diagnosis
criteria at follow-up.
Method
Sample and procedure
Participants were all men (N = 134) and their partners
from an ongoing longitudinal study of families at high risk
for AUD (Zucker et al., 2000) who received past 3-year
Diagnostic and Statistical Manual of Mental Disorders,
Fourth Edition, (DSM-IV; American Psychiatric Associa-
tion, 1994) AUD diagnosis at baseline and provided data at
the follow-up. Based on the initial design, all men were
coupled (97% married) and were parents of young chil-
dren. Thirty-three women (25%) met criteria for a 3-year
DSM-IV AUD diagnosis at baseline. The mean age (SD)
for men was 32.5 (4.5), and the mean age for women was
30.3 (4.1). The sample was 96% white. This group of 134
represented 84% of the men originally involved in the study
and diagnosed with AUD at baseline. Using baseline mea-
sures, individuals assessed at follow-up were not different
when compared with nonparticipants on all predictors ex-
cept higher family cohesion (p < .01).
Participants were recruited in two ways. Eighty percent
(n = 107) were recruited through the courts, which involved
contacting all drunk drivers in a four-county area who had
a blood alcohol concentration (BAC) of .15% or higher (or
.12% if multiple arrests), were coupled and were parents of
children. Twenty percent (n = 27) were recruited via a com-
munity canvass in the neighborhood where the court alco-
holics lived (see Zucker et al., 2000, for details). We were
able to examine predictors of recovery over the subsequent
9-year interval by comparing those men who had an AUD
(abuse or dependence) diagnosis at follow-up against those
who no longer met AUD criteria.
The majority of the men (87%) had a dependence diag-
nosis at baseline. Thus, combining the two groups (those
with an abuse disorder and those with a dependence disor-
der) eliminated comparisons between subgroups that would
have been too small for fine-grained analyses. In addition,
preliminary analyses between the abuse and dependence
groups showed no differences across the independent vari-
ables. After the initial recruitment and assessment, indi-
viduals participated in multisession assessments every 3
years (see Zucker et al., 2000).
Measures
Data collection involved demographics and two other
sets of variables: information about family and social sup-
port resources, and psychological and addictive behavior
factors, including a baseline index of alcohol-problem
severity.
Family functioning. The Moos Family Environment Scale
(FES) consists of 10 scales measuring family environments
as perceived by the family members (Moos, 1990; Moos
and Moos, 1994). The cohesion scale used in this report
assesses the quality of intimate social relationships within
the family, including the extent of commitment, concern
and support provided by family members to one another.
The FES has adequate psychometric properties and good
convergent validity involving both observational and self-
report measures among antisocial alcoholic families, non-
antisocial alcoholic families and nonalcoholic families
(Sanford et al., 1999).
Depressive symptoms. The Beck Depression Inventory
is a widely used measure of depressive symptoms validated
through extensive study (Arnau et al., 2001; Beck et al.,
1988). This self-report instrument assesses cognitive, emo-
tional, motivational and physical manifestations of depres-
sion and is available in a long form (21 items) and a short
form (13 items). At the time of baseline data collection, we
used the short form (for details regarding psychometric prop-
erties, see Beck et al., 1988).
Antisocial behavior. The Antisocial Behavior Checklist
(Zucker et al., 1994a,b) is a 45-item questionnaire that as-
sesses the frequency of aggressive and antisocial activities
in childhood (e.g., lying to parents, being suspended from
school) and adulthood (e.g., being fired for absenteeism,
defaulting on a debt). The scores for each item range from
0 to 3, with higher scores reflecting behavior that is more
antisocial. A series of reliability and validity studies with a
variety of clinical and nonclinical populations has shown
adequate test-retest reliability and discrimination between
alcoholic and nonalcoholic adults (Moses et al., 1993;
Zucker et al., 1994a,b).
Social support. The Norbeck Social Support Question-
naire comprises items that evaluate structural support (size
of network) (Norbeck et al., 1981; Norbeck, 1984). Re-
spondents generate a list of significant people in their life.
The size of total network summary variable was used for
this index.
Drinking and drug history. The Drinking and Drug His-
tory Questionnaire (DDHQ; unpublished) is an extensive
self-administered tool incorporating items from the 1978
National Institute on Drug Abuse survey (Johnston et al.,
1979), from the American Drinking Practices survey
(Cahalan et al., 1969), and from the Veterans Administra-
tion Medical Center Research Questionnaire for Alcoholics
(Schuckit, 1978). These items have been used extensively
in a variety of survey and clinical settings. They provide
data on the quantity, frequency and variability of alcohol
consumption, drug use and tobacco use, and they ask about
consequences and troubles related to the use of these sub-
stances. Use and consequences of alcohol were used for
 McAWEENEY ET AL.
diagnostic purposes (see below). The question “Have you
smoked cigarettes in the past 12 months?” with answers
ranging from “never” to “regularly now,” was used to as-
sess smoking status over the entire follow-up.
AUD diagnosis. Subject AUD status was assessed by
the Diagnostic Interview Schedule (Robins et al., 2000);
the Short Michigan Alcohol Screening Test (SMAST; Selzer
et al., 1975) and the DDHQ (discussed above). Based on
information collected by all three instruments, a diagnosis
of AUD (lifetime as well as in the past 3 years) was made
at baseline and in every subsequent wave by a trained cli-
nician using DSM criteria. The availability of these three
sources of information allowed us to code DSM, Third Edi-
tion, Revised (DSM-III-R; American Psychiatric Associa-
tion, 1987) and later DSM-IV diagnoses at all data points.
Given the volume of material collected in the overall pro-
tocol as well as the spacing between sessions—separated
in some instances by 2-3 months—it was unlikely that re-
spondents would recall their specific replies, because mul-
tiple information sources were used. In cases of discrepant
information, we used the data represented by the majority
of information sources in establishing the diagnosis. Interrater
reliability for diagnosis was assessed by having a second
trained clinician independently review a subset of subjects
(n = 26), and the reliability was excellent (kappa = .81).
Experiences with treatment or AA. Every assessment in-
cluded two items from the SMAST regarding “receiving
help for a drinking problem” or “attending self-help groups.”
Because we lacked information about duration, we com-
puted an index of treatment experiences and one for par-
ticipation in AA by summing all the dichotomized yes/no
answers. Summing responses from the baseline assessment
phase plus the three 3-year follow-up assessments provided
two variables, each with a 0-4 range.
Lifetime drinking problem index. The Lifetime Alcohol
Problems Score (LAPS) (Zucker, 1991; Zucker et al., 1997)
is a measure of lifetime alcohol problem severity and com-
prises three components: less primacy, variety and life per-
cent components. The LAPS was designed to assess
differences in the magnitude and extent of drinking prob-
lems over the life course and was derived from the DDHQ.
The primacy component is the squared inverse of the age
at which the respondent reported first drinking enough to
get drunk. The variety component involves the number of
different drinking symptoms reported. Finally, the life per-
cent component involves a measure of the interval between
the most recent and the earliest drinking problems, cor-
rected for current age. Subscores were standardized and
summed to form the final score. This measure effectively
distinguishes between alcoholics and nonalcoholics; distin-
guishes among levels of severity of DSM-III-R alcohol de-
pendence; and correlates with a wide range of external
measures of alcohol-related difficulty, such as blood alco-
hol level at arrest and treatment involvement. The LAPS
223
was used as a measure of drinking severity for the men
instead of AUD diagnosis, because we were predicting AUD
status. Partner AUD status, however, was used as an index
of severity of drinking at baseline, because it more accu-
rately represented whether the partner was or was not par-
allel in diagnosis to the subject.
Years of intervening recovery. AUD status was not con-
tinuous for all men over the follow-up interval. We com-
puted an index of “years of intervening recovery” by
summing all the yearly intervals during which the AUD
diagnosis was not present (range 0-9). For example, if a
subject never met diagnosis criteria during the 9 years fol-
lowing baseline measurement, his index would be 9. If a
subject always met 3-year diagnosis during every interval,
his index of recovery would be 0.
Statistical power and missing data
Tabachnick and Fidell (1996) suggest a rule of thumb
for a sample size adequate for providing medium power to
detect moderate effects (Cohen’s d) for logistic regression
of N > 50 + 8m (where m is the number of predictors).
With 15 independent variables, a sample of 134 provides
adequate power (.8) to detect moderate effects (.4), which
is the magnitude of effect sizes observed in earlier studies
(Bischof et al., 2000; Cohen and Cohen, 1983; Humphreys,
et al., 1997; McCrady et al., 1999). Although other vari-
ables, such as partner education or partner participation in
treatment, were available to us, the addition of more vari-
ables would have resulted in a substantial loss of power.
We checked to see the effect of these variables on the out-
come variable and discovered at the bivariate level that
they bore no relationship to recovery. There were 55 miss-
ing data points out of a possible 3752 (<2%), which were
estimated using SPSS, version 10.0 (SPSS Inc., Chicago,
IL) computations, with the mean of the variable substituted
for missing observation.
Results
Overview of analyses
Table 1 presents the means, the standard deviations and
the results of the analyses of group differences between
those in recovery (46%) and those who continued to meet
diagnosis criteria (54%) across baseline and proximal pre-
dictors. Recovery, defined as the past 3-year interval of no
diagnosis, is a common term often cited as characterizing
securely abstinent former abusers (Finney and Moos, 1991;
Vaillant, 1995). A Bonferroni procedure was conducted to
correct for the number of tests completed.
Table 2 presents demographic and other characteristics
of the alcoholics and their partners at baseline and at fol-
low-up. Notably, significant shifts in partner AUD status
224 JOURNAL OF STUDIES ON ALCOHOL / MARCH 2005

TABLE 1. Characteristics differentiating recovered from continued AUD TABLE 2. Characteristics of recovered and continued AUD males and
males and their partners at 9-years post baseline (N = 134 couples) their partners at baseline and follow-up (N = 134 couples)
Diagnosis status at 9-year follow-up Continued
Recovery group AUD group Significance
Recovered group Continued AUD
Variables (n = 61, 46%) (n = 73, 54%) tests
(61 couples) group (73 couples)
Predictors Mean (SD) Mean (SD) t Variables at baseline
Age (years), mean (SD)
Block one baseline Alcoholics (n = 134) 32.0 (3.8) 33.0 (5.1) t = -1.23, 1/133 df
Self-Predictors Partners (n = 134) 30.4 (4.2) 30.2 (3.9) t = .280, 1/133 df
Alcohol problem Substance Use, % (n/N)
severity 11.2 (1.9) 11.4 (1.9) -0.72 Partners w/positive
Education (in years) 13.6 (1.9) 12.7 (2.0) 2.41* AUD 18 (11/61) 30 (22/73) χ2 = 2.62*, 1 df
Depressive symptoms 2.2 (2.6) 3.3 (2.9) -2.30 Variables at follow-up
Family cohesion 7.3 (1.6) 7.2 (1.6) 0.45 Partners w/positive
Number of social AUD, % (n/N) 11 (7/61) 33 (24/73) χ2 = 8.82†, 1 df
supports 6.8 (3.4) 5.9 (2.8) 1.16 Married to same baseline
Antisocial behavior 21.2 (13.2) 23.4 (14.2) -0.91 partner, % (n/N) 84 (51/61) 81 (59/73) χ2 = .67, 1 df
Block two baseline
Partner Predictors Note: AUD = alcohol use disorder.
Alcohol problem *p < .05; †p < .01.
severity 9.9 (1.6) 10.4 (1.3) -1.8
Depressive symptoms 2.9 (3.0) 3.2 (3.3) -0.53
Family cohesion 7.1 (2.2) 7.1 (1.9) 0.22
Number of social Prediction of 9-year recovery
supports 8.2 (3.7) 6.8 (2.7) 2.61*
Antisocial behavior 11.3 (7.6) 12.8 (7.4) -1.12 Recovery was predicted using logistic regression. Al-
Block three
Mediating Predictors though we have a special interest in evaluating the contri-
Years of smoking 7.3 (8.1) 11.0 (8.7) -2.45* bution of partner predictors to the recovery process, given
Treatment experiences 1.6 (2.3) 1.1 (2.0) 1.541 that the dominant model in the field is one of personal
Years of intervening
recovery 3.5 (2.3) 0.9 (1.7) 7.1† causation, we entered the six male personal predictors mea-
Participation in AA 1.9 (1.9) 2.0 (2.0) -1.09 sured at baseline first, followed by the block of five partner
predictors. Finally, because the number of successful re-
Notes: AUD = alcohol use disorder; AA = Alcoholics Anonymous.
*p < .05; †p < .01. covery years, experiences with treatment, participation in
AA and number of years smoking are more proximal vari-
ables and they change over time, we entered them last.
Because no interactions were significant, none was included
occurred over the 9 years. The partners of those in the in the model.
recovery group went from 18% to 11% with an AUD (a The full model was significant. The R2 change also was
39% decrease), while the partners of those in the AUD significant when Block 2 and Block 3 were entered. The
group went from 30% to 33% with an AUD (a 10% in- exponentiated b weights (ExpB) express the increase in odds
crease) 9 years later—a statistically significant difference of recovery per each unit increase in the independent vari-
between the two changes in proportions (p < .01). able when all other independent variables are held con-
Table 3 highlights changes in AUD status over the 9 stant. For example, among the alcoholic men, for each
years. As mentioned above, most of the men at baseline incremental year of recovery, the odds of recovery increased
had a dependence diagnosis. In terms of drinking patterns 1.72 times. The relationship between recovery and having
over time, some in the AUD group did not have a continu-
ous diagnosis throughout the entire 9 years, and, likewise,
some in the recovery group were not in remission the en- TABLE 3. Changes in AUD status over a 9-year period (n = 134 initially
tire time following baseline. Forty-one percent (n = 55) of alcoholic men)
the men meeting diagnosis criteria at baseline continued to 3-year 6-year 9-year
Group classifications Baseline follow-up follow-up follow-up
show an AUD diagnosis over the following 3-, 6- and 9-
year intervals. A smaller portion, 15% (n = 20), fluctuated Unremitting AUD entire X X X X
9 years (n = 55)
between having the diagnosis and not having it over the 9 Developmentally
years. Twenty-one percent (n = 28) of the men who were limited (n = 28) X O O O
diagnosed at baseline never met diagnosis criteria at any Later remission A (n = 19) X X O O
Later remission B (n = 12) X X X O
point following baseline, while 23% (n = 31) of the men Fluctuating A (n = 6) X O O X
had later remission at the 6- or 9-year assessments. In total, Fluctuating B (n = 7) X X O X
62% of the men had stable drinking patterns—either remis- Fluctuating C (n = 5) X O X X
Fluctuating D (n = 2) X O X O
sion or an unremitting AUD diagnosis over the entire 9-
year follow-up period. Notes: AUD = alcohol use disorder; X = positive diagnosis; O = no AUD.
 McAWEENEY ET AL. 225

a partner with an AUD diagnosis was opposite (i.e., b weight
is -.310, odds ratio .734). To better interpret this odds ratio,
the reciprocal or inverse was calculated. Computing the
reciprocal for values less than 1.0 results in odds ratios
greater than 1.0, leading to a more intuitive interpretation
(Pedhazur, 1997). If an alcoholic was married to a woman
without a positive AUD diagnosis, the odds of recovery
were 1.37 greater. The five significant predictors are iden-
tified in Table 4.
Discussion
We have identified a set of predictors for long-term re-
covery among a community sample of men with AUD, and
our findings show that a complex web of personal and so-
cial variables contributes to this process. These variables
include a baseline measure of higher achievement (educa-
tion); two health-inducing interpersonal factors (having a
spouse without AUD who also has a large social support
network); a more sustained involvement in treatment (more
intervening treatment experiences); and the help of a his-
tory of prior success (more intervening years of recovery).
The proportion of correct classification in our model (79%)
is impressive, given the 9-year interval over which predic-
tion was made, and it is similar to the rate Humphreys et
al. (1997) found using a comparable model (70%). Consis-
tent with prior findings (Finney and Moos, 1991; Moos
and Moos, 2004; Weisner et al., 2003), individuals who
were able to remit from alcohol and/or have more treat-
ment experiences over the follow-up period were more likely
to recover. As Prochaska and associates (1991) have dis-
cussed, this finding may be related to increases in self-
esteem and a greater confidence to resist drinking.
The recovery rate we found (46%) is similar to rates
found in prior research that included diverse populations:
an 8-year follow-up of community-sampled alcoholics
(Humphreys et al., 1997); a 20- to 30-year follow-up of
both college-educated and Core City (urban sample) AUD-
diagnosed males (Vaillant, 1995, 2003); and a 10-year fol-
low-up of a treatment sample (Finney and Moos, 1991).
Nonetheless, many men continued to drink throughout the
study period, demonstrating the seriousness of the illness
and the value of discovering what promotes its continuation.
Numerous studies have identified the benefits of AA
and the influence of social networks within it (Humphreys
and Moos, 2001; Moos and Moos, 2003b, 2004; Timko et
al., 2000). Interestingly, we found no relationship between
recovery and participation in AA or in the severity of drink-
ing at baseline. Similar to our findings, Humphreys et al.
(1997) found no relationship between the severity of baseline
alcohol symptoms (and consumption) and remission at 8-
year follow-up; others have found a strong association be-
tween severity of drinking and the likelihood of continued
drinking (Moos and Moos, 2003b).
Our finding provides further evidence to support the role
of nondrinking variables in the initiation and maintenance
of recovery. It also suggests that the floridity of symptoms
at a particular time point is not necessarily a precursor of
future outcomes. This, in fact, is one of the ways that lon-
gitudinal studies inform our understanding of clinical phe-
nomena and provide a different view of adaptation than the
one seen by the clinician, whose purview is a relatively

TABLE 4. Logistic regression analysis predicting recovery (N = 134 couples)

Odds ratio,
exponentiated
b weights
Predictors b (SE) (ExpB) R2 Change in R2
Block 1: Personal Predictors
Education .250 (.143) 1.28*
Alcohol severity .020 (.191) 1.02 .135,
Depressive symptoms -.183 (.123) 0.83 58% correctly –
Antisocial behavior .025 (.026) 1.02 classified
No. of social supports .007 (.088) 1.00
Family cohesion -.187 (.183) 0.83
Block 2: Partner Predictors
AUD Status -.310 (.207) (.734) 1.37*R
Depressive symptoms -.004 (.116) 1.00 .205,
Antisocial behavior .002 (.046) 1.00 65% correctly .065*
No. of social supports .165 (.096) 1.27* classified
Family cohesion -.054 (.155) 0.95
Block 3: Mediating Predictors
Years of recovery .547 (.126) 1.72† .475,
Treatment experiences .527 (.720) 1.69* 79% correctly .27†
AA Involvement -.092 (.709) 0.91 classified
Years of smoking -.030 (.030) 0.97

Notes: AUD = alcohol use disorder; AA = alcoholics Anonymous; R denotes reciprocal.

*p < .05; †p < .001; the model (χ2 = 23.1) was significant at the p < .001, 79% correctly classified.
226 JOURNAL OF STUDIES ON ALCOHOL / MARCH 2005
narrow time window. At the same time, it would be useful
to know what characteristics lead some studies to find a
severity/long-term outcome relationship and others not. We
suspect that in those in which the social environment is
less fluid—conceivably in older, possibly more socially
damaged alcoholics—such an association is present. There
is no way to test this hypothesis here, but it should be
examined in future work.
Partner influence
One of the goals of this study was to examine the pos-
sible influence of partners on their spouses’ drinking over
time—specifically the facilitation of continued drinking or
the initiation and maintenance of recovery. We found that
partners of those in recovery had a greater number of so-
cial supports at baseline when compared with partners whose
spouses continued to drink at follow-up. In other words,
partners of those in recovery appear to be more integrated
into a broader social network, and their social supports in
turn may serve as a buffer for their spouses’ drinking. This
phenomenon may be similar to the buffering effect found
in the myocardial infarction literature (Waltz, 1988).
Moreover, the wives of the men with continued AUD at
follow-up were more likely to be alcoholic at baseline and
more likely to continue to meet AUD diagnosis criteria at
follow-up. In fact, a significant shift in partner diagnosis
took place over this interval, with a decrease in partners
with AUD diagnosis in the recovery group and an increase
of partners with AUD in the continued AUD group. Our
partner findings are consistent with the results of Moos et
al. (1982), who found that spouses of relapsed alcoholics
drank more alcohol and participated in fewer social
activities.
The relationship between the alcoholic and his partner
appears to have had a reciprocal influence on each other’s
drinking. The measures here are likely indicators of under-
lying microprocesses that influence each other over time,
resulting in some people recovering and some not. Such
factors might include the quality of microinteractions be-
tween partners (leading to divorce or not), the level of nega-
tivity associated with the drinking, the specific nature of
their peer relationships and their social networks (Jacob et
al., 2001; Leonard and Jacob, 1997; McCrady et al., 2002).
These findings highlight the dynamic nature of the com-
plex relationship between couples, as pointed out years ago
by Moos et al. (1982) and recently by Kahler et al. (2003)
and McCrady et al. (2002).
Variation and stability in drinking patterns
The variations in the drinking patterns among these male
alcoholics over time illustrate the complex process of re-
covery. A majority of the participants were stable, with
62% continuing either unremitting AUD or remittance from
shortly after baseline. Some men, however, moved in and
out of remission throughout the 9 years of follow-up. These
fluctuations in drinking patterns and sustained diagnosis are
similar to those of Kerr et al. (2002) and Finney and Moos
(1991). Finney and Moos reported that 37% of their treat-
ment sample was abstinent throughout a 6-year period, while
40% reported some mixture of abstinence and no problem
drinking. Of those remitted at the 2-year follow-up, 77%
continued to remit 8 years later. Overall, 59% of Finney
and Moos’ sample reported a stable drinking pattern (either
stable problem or nonproblem drinking)—similar to the rate
found in this study (62%).
Limitations
The variables we have identified may be used to predict
recovery, but some limitations are noted. This was not an
intervention study, and it lacked details about treatments
received, including AA. It would have been useful to as-
sess the content and process of treatment for those who
received formal treatment. Information about the content of
care, especially whether such programs as 12-step groups
were encouraged or whether marriage counseling was pro-
vided, would have enabled us to examine whether some of
the predictors changed more or less because they were not
addressed adequately in treatment. Similarly, we did not
consider other factors of AA, such as working the steps,
reading 12-step literature, the quality of the relationship
with a sponsor and the number of friends active in 12-step
programs. Finally, the present study addressed only factors
predicting the recovery of an essentially white sample of
alcoholic married men and their spouses and can only be
generalized to a similar population.
Future directions
We suggest three areas of continued research. First, more
in-depth research investigating the nature of the relation-
ship between alcoholic men and their partners in maintain-
ing recovery has the potential to assist in guiding the
development of intervention efforts and prevention mea-
sures. Prospective studies that characterize the details of
that relationship, with numerous assessments of both the
alcoholic and his partner, would be informative. Second, a
better understanding of what drives natural recovery at the
individual level may assist clinicians in helping those who
choose not to participate in formal treatment or AA yet
may be reached with nonthreatening approaches. Third, it
is equally important to examine predictors of recovery in
other subgroups (women, adolescents, racial groups other
than whites). The personal and social environments of these
diverse groups are different from those of the white alco-
holic men we studied here. Investigation in these areas is
 McAWEENEY ET AL.
likely to inform treatment activity by increasing our under-
standing of the course of recovery and its variability in
these different social environments.
References
AMERICAN PSYCHIATRIC ASSOCIATION. Diagnostic and Statistical Manual of
Mental Disorders (DSM-III-R), Washington, DC, 1987.
AMERICAN PSYCHIATRIC ASSOCIATION. Diagnostic and Statistical Manual of
Mental Disorders (DSM-IV), Washington, DC, 1994.
ARNAU, R.C., MEAGHER, M.W., NORRIS, M.P. AND BRAMSON, R. Psycho-
metric evaluation of the Beck Depression Inventory-II with primary
care medical patients. Hlth Psychol. 20: 112-119, 2001.
BECK, A.T., STEER, R.A. AND GARBIN, M.G. Psychometric properties of the
Beck Depression Inventory: Twenty-five years of evaluation. Clin.
Psychol. Rev. 8: 77-100, 1988.
BISCHOF, G., RUMPF, H.-J., HAPKE, U., MEYER, C. AND JOHN, U. Mainte-
nance factors of recovery from alcohol dependence in treated and un-
treated individuals. Alcsm Clin. Exp. Res. 24: 1773-1777, 2000.
BLOMQVIST, J. Treated and untreated recovery from alcohol misuse: Envi-
ronmental influences and perceived reasons for change. Subst. Use
Misuse 34: 1374-1406, 1999.
CAHALAN, D., CISIN, I.H. AND CROSSLEY, H.M. American Drinking Practices:
A National Study of Drinking Behavior and Attitudes, Rutgers Center
of Alcohol Studies, Monograph No. 6, New Brunswick, NJ, 1969.
COHEN, J. AND COHEN, P. Applied Multiple Regression/Correlation Analy-
sis for the Behavioral Sciences, 2nd Edition, Mahwah, NJ: Lawrence
Erlbaum, 1983.
CONNORS, G.J., LONGABAUGH, R. AND MILLER, W.R. Looking forward and
back to relapse: Implications for research and practice. Addiction 91
(Supplement): S191-S196, 1996.
CONNORS, G.J., TONIGAN, J.S. AND MILLER, W.R. A longitudinal model of
intake symptomatology, AA participation and outcome: Retrospective
study of the Project MATCH outpatient and aftercare samples. J. Stud.
Alcohol 62: 817-825, 2001.
EPSTEIN, E.E. AND MCCRADY, B.S. Behavioral couples treatment of alcohol
and drug use disorders: Current status and innovations. Clin. Psychol.
Rev. 18: 689-711, 1998.
FINNEY, J.W. AND MOOS, R.H. The long-term course of treated alcoholism:
I. Mortality, relapse and remission rates and comparisons with com-
munity controls. J. Stud. Alcohol 52: 44-54, 1991.
FLOR, H., KERNS, R.D. AND TURK, D.C. The role of spouse reinforcement,
perceived pain, and activity levels of chronic pain patients. J.
Psychosomat. Res. 31: 251-259, 1987.
GRANT, B.F. Age at smoking onset and its association with alcohol con-
sumption and DSM-IV alcohol abuse and dependence: Results from
the national longitudinal alcohol epidemiologic survey. J. Subst. Abuse
10: 59-73, 1998.
HALL, S.M., WASSERMAN, D.A. AND HAVASSY, B.E. Relapse prevention. In:
PICKENS, R.W., LEUKEFELD, C.G. AND SCHUSTER, C.R. (Eds.) Improving
Drug Abuse Treatment. NIDA Research Monograph No. 106, DHHS
Publication No. (ADM) 91-1754, Washington: Government Printing
Office, 1991, pp. 279-292.
HELDER, D.I., KAPTEIN, A.A., VAN KEMPEN, G.M.J., WEINMAN, J., VAN
HOUWELINGEN, J.C. AND ROOS, R.A.C. Living with Huntington’s dis-
ease: Illness perceptions, coping mechanisms, and spouses’ quality of
life. Int. J. Behav. Med. 9: 37-52, 2002.
HUMPHREYS, K. AND MOOS, R. Can encouraging substance abuse patients
to participate in self-help groups reduce demand for health care? A
quasi-experimental study. Alcsm Clin. Exp. Res. 25: 711-716, 2001.
HUMPHREYS, K., MOOS, R.H., AND COHEN, C. Social and community re-
sources and long-term recovery from treated and untreated alcohol-
ism. J. Stud. Alcohol 58: 231-238, 1997.
227
HUMPHREYS, K., MOOS, R.H. AND FINNEY, J.W. Two pathways out of drink-
ing problems without professional treatment. Addict. Behav. 20: 427-
441, 1994.
HUMPHREYS, K. AND TUCKER, J.A. Toward more responsive and effective
intervention systems for alcohol-related problems (editorial). Addic-
tion 97: 126-132, 2002.
HUMPHREYS, K., WING, S., MCCARTY, D., CHAPPEL, J., GALLANT, L., HABERLE,
B., HORVATH, A.T., KASKUTAS, L.A., KIRK, T., KIVLAHAN, D., LAUDET,
A., MCCRADY, B.S., MCLELLAN, A.T., MORGENSEN, J., TOWNSEND, M.
AND WEISS, R. Self-help organizations for alcohol and drug problems:
Toward evidence-based practice and policy. J. Subst. Abuse Treat. 26:
151-158, 2004.
JACOB, T., LEONARD, K.E. AND HABER, J.R. Family interactions of alcohol-
ics as related to alcoholism type and drinking condition. Alcsm Clin.
Exp. Res. 25: 835-843, 2001.
JOHNSTON, L.D., BACHMAN, J.G. AND O’MALLEY, P.M. Drugs and the Class
of ’78: Behaviors, Attitudes, and Recent National Trends, DHEW Pub-
lication No. (ADM) 79-877, Washington: Government Printing Of-
fice, 1979.
KAHLER, C.W., MCCRADY, B.S. AND EPSTEIN, E.E. Sources of distress among
women in treatment with their alcoholic partners. J. Subst. Abuse Treat.
24: 257-265, 2003.
KASKUTAS, L.A., BOND, J. AND HUMPHREYS, K. Social networks as media-
tors of the effect of Alcoholics Anonymous. Addiction 97: 891-900,
2002.
KERR, W.C., FILLMORE, K.M. AND BOSTROM, A. Stability of alcohol con-
sumption over time: Evidence from three longitudinal surveys from
the United States. J. Stud. Alcohol 63: 325-333, 2002.
LEONARD, K.E. AND JACOB, T. Sequential interactions among episodic and
steady alcoholics and their wives. Psychol. Addict. Behav. 11: 18-25,
1997.
MCCRADY, B.S., EPSTEIN, E.E. AND HIRSCH, L.S. Maintaining change after
conjoint behavioral alcohol treatment for men: Outcomes at 6 months.
Addiction 94: 1381-1396, 1999.
MCCRADY, B.S., HAYAKI, J., EPSTEIN, E.E. AND HIRSCH, L.S. Testing hy-
pothesized predictors of change in conjoint behavioral alcoholism treat-
ment for men. Alcsm Clin. Exp. Res. 26: 463-470, 2002.
MOOS, R.H. Conceptual and empirical approaches to developing family-
based assessment procedures: Resolving the case of the Family Envi-
ronment Scale. Fam. Process 29: 191-208, 1990.
MOOS, R.H., FINNEY, J.W. AND GAMBLE, W. The process of recovery from
alcoholism: II. Comparing spouses of alcoholic patients and matched
community controls. J. Stud. Alcohol 43: 888-909, 1982.
MOOS, R.H. AND MOOS, B.S. The Family Environment Scale Manual, 3rd
Edition, Palo Alto, CA: Consulting Psychologists Press, 1994.
MOOS, R.H. AND MOOS, B.S. Long-term influence of duration and inten-
sity of treatment on previously untreated individuals with alcohol use
disorders. Addiction 98: 325-337, 2003a.
MOOS, R.H. AND MOOS, B.S. Risk factors for nonremission among initially
untreated individuals with alcohol use disorders. J. Stud. Alcohol 64:
555-563, 2003b.
MOOS, R.H. AND MOOS, B.S. Long-term influence of duration and fre-
quency of participation in Alcoholics Anonymous on individuals with
alcohol use disorders. J. Cons. Clin. Psychol. 72: 81-90, 2004.
MOSES, H.D., ZUCKER, R.A. AND FITZGERALD, H.E. Moderators of the ef-
fects of father alcohol problems on child behavior problems. Paper
presented at the biennial meetings of the Society for Research on Child
Development, New Orleans, LA, 1993.
NORBECK, J.S. The Norbeck Social Support Questionnaire. Birth Def. Orig.
Art. Series 20: 45-57, 1984.
NORBECK, J.S., LINDSEY, A.M. AND CARRIERI, V.L. The development of an
instrument to measure social support. Nurs. Res. 30: 264-269, 1981.
PEDHAZUR, E.J. Multiple Regression in Behavioral Research: Explanation
and Prediction, 3rd Edition, Fort Worth, TX: Harcourt Brace College,
1997.
228 JOURNAL OF STUDIES ON ALCOHOL / MARCH 2005
PROCHASKA, J.O., DICLEMENTE, C.C. AND NORCROSS, J.C. In search of how
people change: Applications to addictive behaviors. Amer. Psychol. 9:
1102-1114, 1992.
ROBINS, L.N., COTTLER, L.B., BUCHOLZ, K.K., COMPTON, W.M., NORTH, C.S.
AND ROURKE, K.M. (Eds.) Diagnostic Interview Schedule for the DSM-
IV, St. Louis, MO: Washington University School of Medicine, 2000.
SANFORD, K., BINGHAM, C.R. AND ZUCKER, R.A. Validity issues with the
family environment scale: Psychometric resolution and research appli-
cation with alcoholic families. Psychol. Assess. 4: 315-325, 1999.
SCHUCKIT, M.A. Research Questionnaire, San Diego, CA: Alcoholism Treat-
ment Program, V.A. Medical Center, University of California, San
Diego, 1978.
SCHWARTZ, L., SLATER, M.A. AND BIRCHLER, G.R. The role of pain behav-
iors in the modulation of marital conflict in chronic pain couples. Pain
71: 227-233, 1997.
SELZER, M.L., VINOKUR, A. AND VAN ROOIJEN, L. A self-administered Short
Michigan Alcoholism Screening Test (SMAST). J. Stud. Alcohol 36:
117-126, 1975.
SOBELL, L.C., CUNNINGHAM, J.A. AND SOBELL, M.B. Recovery from alcohol
problems with and without treatment: Prevalence in two population
surveys. Amer. J. Publ. Hlth 86: 966-972, 1996.
SOBELL, L.C., ELLINGSTAD, T.P. AND SOBELL, M.B. Natural recovery from
alcohol and drug problems: Methodological review of the research
with suggestions for future directions. Addiction 95: 749-764, 2000.
SOBELL, L.C., SOBELL, M.B., TONEATTO, T. AND LEO, G.I. What triggers the
resolution of alcohol problems without treatment? Alcsm Clin. Exp.
Res. 17: 217-224, 1993.
SOBELL, M.B. AND SOBELL, L.C. Guiding self-change. In: MILLER, W.R.
AND HEATHER, N. (Eds.) Treating Addictive Behaviors, 2nd Edition,
New York: Plenum Press, 1998, pp. 189-202.
TABACHNICK, B.G. AND FIDELL, L.S. Using Multivariate Statistics, 3rd Edi-
tion, New York: HarperCollins College, 1996.
TIMKO, C., MOOS, R.H., FINNEY, J.W. AND LESAR, M.D. Long-term out-
comes of alcohol use disorders: Comparing untreated individuals with
those in alcoholics anonymous and formal treatment. J. Stud. Alcohol
61: 529-540, 2000.
VAILLANT, G.E. The Natural History of Alcoholism: Revisited, Cambridge,
MA: Harvard Univ. Press, 1995.
VAILLANT, G.E. A 60-year follow-up of alcoholic men. Addiction 98: 1043-
1051, 2003.
WALTERS, G.D. Spontaneous remission from alcohol, tobacco, and other
drug abuse: Seeking quantitative answers to qualitative questions. Amer.
J. Drug Alcohol Abuse 26: 443-460, 2000.
WALTZ, M., BADURA, B., PFAFF, H. AND SCHOTT, T. Marriage and the psy-
chological consequences of a heart attack: A longitudinal study of
adaptation to chronic illness after 3 years. Social Sci. Med. 27: 149-
158, 1988.
WEISNER, C., MATZGER, H. AND KASKUTAS, L.A. How important is treat-
ment? One-year outcomes of treated and untreated alcohol-dependent
individuals. Addiction 98: 901-911, 2003.
ZUCKER, R.A. Scaling the developmental momentum of alcoholic process
via the Lifetime Alcohol Problems score. Alcohol Alcsm, Supplement
No. 1, pp. 505-510, 1991.
ZUCKER, R., BOYD, G. AND HOWARD, J. (Eds.) The Development of Alco-
hol Problems: Exploring the Biopsychosocial Matrix of Risk. NIAAA
Research Monograph No. 26, NIH Publication No. 94-3495, Rockville,
MD: Department of Health and Human Services, 1994a.
ZUCKER, R.A., DAVIES, W.H., KINCAID, S.B., FITZGERALD, H.E. AND REIDER,
E.E. Conceptualizing and scaling the developmental structure of be-
havior disorder: The Lifetime Alcohol Problems score as an example.
Devel. Psychopathol. 9: 453-471, 1997.
ZUCKER, R.A., FITZGERALD, H.E. AND NOLL, R.B. Drinking and Drug His-
tory: Revised Edition, Version 4, East Lansing, MI: Department of
Psychology, Michigan State University, 1990, unpublished question-
naire.
ZUCKER, R.A., FITZGERALD, H.E., REFIOR, S.K., PUTTLER, L.E., PALLAS, D.M.
AND ELLIS, D.A. The clinical and social ecology of childhood for chil-
dren of alcoholics: Description of a study and implications for a dif-
ferentiated social policy. In: FITZGERALD, H.E., LESTER, B.M. AND
ZUCKERMAN, B.S. (Eds.) Children of Addiction: Research, Health, and
Public Policy Issues, New York: RoutledgeFalmer, 2000, pp. 4-65.
ZUCKER, R.A., NOLL, R.B., HAM, H., FITZGERALD, H.E. AND SULLIVAN, L.S.
Assessing antisociality with the Antisocial Behavior Checklist: Reli-
ability and validity studies, Ann Arbor, MI: Department of Psychiatry,
University of Michigan, and East Lansing, MI: Department of Psy-
chology, Michigan State University, 1994b, unpublished manuscript.