Predictors of adolescent A.A. affiliation. (Alcoholics Anonymous)

From: Adolescence

Alcoholics who receive treatment in in-patient settings are routinely referred to Alcoholics Anonymous upon discharge, yet not all affiliate with A.A. The characteristics of A.A. affiliators have been explored in the past to further improve discharge planning, but to date no studies have described the characteristics of adolescents who affiliate with A.A. The sample used in this study was 70 adolescents who had completed in-patient treatment and were contacted as part of a follow-up survey. Half of the group had affiliated with A.A. A discriminant analysis was used to predict affiliation, and the study found that affiliators were more likely to have had prior treatment, had friends who did not use drugs, had less parental involvement while in treatment, and more feelings of hopelessness. Possible explanations for these findings are discussed as well as areas for further research.

Thousands of teenagers are treated each year in in-patient or residential settings for drug and alcohol dependency. The majority of these programs use the traditional treatment model, also known as the Minnesota Model (Littrell, 1991). This model emphasizes addiction as a disease where one achieves recovery through abstinence, and that one must participate in a support group such as Alcoholics Anonymous (AA) in order to maintain this abstinence. Patients attend AA while hospitalized and are encouraged to continue their attendance upon discharge. AA itself has seen a tremendous growth in membership over the last decade. Currently there are over one million members in the United States alone (About AA, 1990) with approximately 3% of its members being under 21 years of age.

While not every adolescent who receives treatment will affiliate with AA, it is important to study the characteristics of those who do. Because referral to AA is a routine discharge plan, it may help mental health practitioners to know which of their clients may be more likely to join this self-help program. This subject is particularly pertinent because it has not been investigated with the adolescent population.

Various studies have researched A.A. affiliation with the goal of describing who joins AA in the hope that treatment providers could utilize "a more informed basis for treatment planning" (O'Leary, Calsyn, Haddock, & Freeman, 1980, p. 137). An outline of these studies can be found in Table 1.

It is virtually impossible to define the population of abstinent alcoholics, both those in AA and those not so affiliated. One of the strictest traditions of AA is that of anonymity of its members. As can be seen, the samples used in most of these studies were mainly hospitalized, adult males. Using currently hospitalized patients may pose problems because of long-term cognitive impairment (Gorski & Miller, 1982) and what Marlatt (1985) refers to as the abstinence violation effect, which is intense shame, guilt, and hopelessness that A.A. members may feel upon relapse and re-entry into treatment. This effect may bias the responses of the recently relapsed patients who are readmitted. Of all these studies, only Hurlburt, Gade and Fuqua (1983) specified race. Four of the studies used AA to recruit participants, leading to a self-selected sample. The study by Alford (1980) was a follow-up that had a somewhat better sample in that all the respondents had received the same intervention.

As a few of the researchers pointed out, it is hard to ascertain whether respondents had the characteristics studied prior to affiliation with AA or if AA caused them to develop. An example of this is Mindlin's (1964) findings of "less socially ill at ease" and "less loneliness."

At least four of the studies (Fontana, Dowds, & Bethel 1976; Greg-son & Taylor, 1977; Reilly & Sugarman, 1967; Trice, 1959) are similar in their finding that attributes could be categorized as external personality characteristics, such as use of more external sources of authority, religiosity, and formalistic thinking.

Many findings among the studies were contradictory; for example, Boscarino (1980) found less alcohol-related problems, while O'Leary et al. (1980) and Vaillant (1983) found more alcohol-related problems. Hurlburt et al. (1983) found that the AA affiliates were "less emotional" and Reilly and Sugarman (1967) found them to be more sensitive and concerned with acceptance. (See Table 1.)

This review of the literature also indicates that the area of AA affiliation has mainly been researched among adult samples. It was found that females, extroverts, and those who have suffered more alcohol-related problems are more likely to affiliate with A.A.

While A.A. affiliation has not been studied with adolescents, treatment outcome studies provide some relevant information. Hoffman [TABULAR DATA FOR TABLE 1 OMITTED] and Kaplan (1991) found in a study of treated alcoholic adolescents that those whose parents had participated in treatment were more likely to participate in a support group, and the less the teens' peers used drugs, the more likely they were to achieve abstinence.

On the basis of the previous research, the following hypothesis was proposed: Adolescent A.A. affiliation may be predicted by gender (female), the number of prior treatment experiences (indicating more alcohol-related problems), parents' involvement in treatment, fewer peers who use drugs/alcohol, and lower levels of hopelessness (depression).

The purpose of this study was to describe characteristics of treated adolescents who affiliate with Alcoholics Anonymous, and of those who choose not to. A discriminant analysis was conducted to determine if the hypothesized set of variables accurately predict AA affiliation.

METHOD

This study was part of a larger outcome study conducted by a telephone interview survey. The telephone interview is an appropriate method for obtaining sensitive information that a respondent might otherwise be reluctant to give face-to-face (Mayer & Greenwood, 1980). It also provided a greater response than would be received in a mailed questionnaire. Numerous attempts were made to locate the respondents in order to decrease self-selection bias. In this study, the interviewers were neutral parties, having no interest in whether the respondent was "successful." Maisto and Connors (1988) highly recommend this approach.

Sample Population and Data Collection

The population for this study consisted of all patients who were admitted to an adolescent residential treatment facility between May, 1989 and November, 1990. All patients in the study were diagnosed as dependent on drugs and/or alcohol by a psychiatrist, according to DSM III-R criteria. They may or may not have had a secondary behavioral health diagnosis which was treated concurrently. The age range is 12 to 21, with an average age of 15.1. Most of the subjects are from families of middle to upper class incomes; the treatment is funded by third-party insurance.

The interviewers were social work graduate students who received brief training on the purpose and method of interviewing the adolescents. The treatment program supplied the investigator with the names and last known phone numbers of the population to be surveyed. Also provided was treatment background data, such as age, length of stay in treatment, drug of choice, and to whom the patient was discharged. The total population was 155; 80 could not be located, five refused to participate in the survey, and 70 agreed to be interviewed. Of the 70 respondents, 60% were female and 39% were male.

The interviewer introduced him or herself, indicating that an outcome study of the treatment program's former patients was being conducted, and that the interviewer was an independent consultant. Respondents were all assured of the confidentiality of their answers. Verbal consent was then solicited. Finally, the interviewer asked the respondent to be as honest as possible since the answers were very important.

The survey itself asked standard questions regarding abstinence, or if relapsed, the drugs or alcohol used. Also included were variables regarding life satisfaction, mood states, social and family support, self-esteem, thoughts regarding drug/alcohol use and addiction, and self-help activities.

Measures

Six variables were included in this study: A.A. affiliation, gender, prior treatment experiences, parents' involvement in treatment, peers' use of drugs or alcohol, and feelings of hopelessness. They were measured as follows:

A.A. affiliation. This was measured in two ways: first as a dichotomous variable, as to whether the respondent was attending AA, and second as a continuous variable, asking how many times a week they attend meetings.

Gender. This was a dichotomous variable asked at the beginning of the survey.

Number of treatment experiences. Prior treatment experiences were solicited, for both chemical dependency and behavioral health treatment. Because the respondent could interpret these questions in many ways (was previous "treatment" school-based, an out-patient group or individual therapy, or in-patient care?), it was left to respondents to decide what they considered "previous treatment." Both questions were then combined to make a continuous variable.

Parents' involvement in treatment. Respondents were asked if the family participated while they were in treatment. Other measures asked if the various components of family treatment were helpful (individual family therapy, group family therapy, and a family workshop). Regardless of how respondents viewed the helpfulness of these components, those who answered other than "didn't participate" or "missing" were considered as having had parental participation. Those who answered "yes" to the participation question and were involved in individual and group family therapy or the family workshop were considered to have parental involvement. This is because not every family was given the opportunity to participate in the workshop and some families do not go to the workshop because it is too far.

Peers' use of drugs/alcohol. Respondents were asked about friends who regularly use drugs or alcohol. Their responses were classified as "none," "only one," or "a few."

Hopelessness. The hopelessness scale used in the study is a measure of depression which manifests as isolation, loneliness, and introversion, the opposite of affiliator behaviors found in Mindlin (1964) and Hurlburt et al. (1983). This was measured by the Hopelessness Scale for Children developed by Kazdin, Rodgers, and Colbus (1986). The scale contains 27 items wherein the respondent indicates "yes" or "no" to each item. A cumulative index is generated varying from 17 to 34, higher scores indicating greater hopelessness.

This scale was originally evaluated on 262 child psychiatric inpatients, ages 6 to 13. The reliability of the scale measured by internal consistency, yielded a coefficient alpha of .97 and a Spearman-Brown split-half reliability of .96. To determine the validity of the scale, its scores were correlated with measures of depression, self-esteem, and social behavior. Resulting correlations indicated significant relations in the predicted direction. Hopelessness was found to be positively correlated with depression (r = .58) and negatively correlated with self-esteem (r = -.61) and social skills (r = -.39) (Kazdin et al., 1986).

Several components of instrument construction and data gathering in this survey lend to high reliability. There were no open-ended questions in this survey, since they may lead to problems with coding error as well as interviewer interpretation (Sudman & Bradburn, 1989). The questions were formulated in language that adolescents would understand.

There are some limitations to the process: The retrospective nature of the data may negatively impact reliability. Many of the respondents had been out of treatment for a good while, and may have had difficulty recalling what drug they may have used or how many AA meetings they attended. Accuracy of on-the-spot data collected in a telephone interview may be limited.

RESULTS

The following is a descriptive analysis of the variables used in this study. Relevant background and demographic information regarding the subjects are presented in Table 2.

Table 2

Descriptive Statistics of the Adolescent Sample

Gender
Male 27 (39%)
Female 42 (60%)

Mean Age 15.5

AA Attendance

Yes 31 (44%)
No 39 (55%)

Prior Treatment

None 34 (50%)
One 22 (32%)
Two 12 (18%)

Family Participation

None 7 (12%)
Some 19 (31%)
All 35 (57%)

Friends' Drug Use

None 30 (43%)
One 17 (24%)
More than one 23 (33%)

Hopelessness Scale (Higher numbers reflect greater hopelessness)

Range 17-34
Mean 21
Mode 20

Note: Percentages may not add up to 100 due to missing data.

A correlation matrix was calculated for all the independent variables (prior treatment, family participation, sex, hopelessness, and friends who use). It was found that all variables had very low correlations with one another, as shown on Table 3, indicating that different constructs were indeed being measured. Thus, they were appropriate for use as predictors in the proposed discriminant analysis.

Next, a stepwise discriminant analysis was performed. This is the method of choice when it is not known how well the proposed variables discriminate between the groups (Klecka, 1980). The stepwise method enters the variables into the predictive equation, one at a time, with the strongest discriminator going in first, as determined by the computer program. The results are presented in Table 4.

The multivariate aspects of the model can be examined by using the canonical discriminant functions (Hair, Anderson, & Tatham, 1987). The canonical correlation is .4842 (p [less than] .006), and by squaring the correlation it can be seen that 23% of the variance in the person's decision to affiliate can be explained by this model that contains four of the independent variables.

Table 3

Correlation Matrix of Predicted Variables

family gender hopelessness friends use
particip-
ation
prior
treatment -.15 -.06 .03 -.11
family
particip. -.09 .16 -.28
gender .02 -.19
hopelessness .08
Table 4

Discriminant Analysis of AA Affiliation

Summary Table

Variable Wilks' Lambda Sig. Min. D Sig.
Squared

Friends Use .882 .008 .531 .008
Family
Participation .826 .005 .836 .005
Hopelessness .792 .005 1.06 .005
Prior
Treatment .766 .006 1.21 .006

Within Groups Centroids and Correlations

Function 1

Group Centroids:

1. Affiliators -.65
2. Nonaffiliators .46

Within Groups Correlations:

Friends who use .66
Prior Treatment -.47
Hopelessness -.32
Family Participation .23

Group centroids can be used to interpret the discriminant function from an overall perspective (Hair et al., 1987). A group centroid is reported since it is "the imaginary point which has coordinates that are the group's mean on each of the variables" (Klecka, 1980, p. 16). They represent the mean of the individual Z-scores for each group. As can be seen, the two groups' centroids differ a great deal, with the affiliator group being larger, indicating more variation within this group.

The within-groups correlations are reported since they show the relationships between the variables in the function - standardized (Klecka, 1980). These scores are then interpreted with the group centroid to determine their contributions to the discriminant function.

Using the stepwise method, it was found that friends who use drugs was the greatest discriminator of the chosen variables between the groups. The next best discriminator was prior treatment, then hopelessness, and finally, family participation in treatment. Sex was not included in the final results since it washed out in the analysis.

According to this data, two variables did predict affiliation as hypothesized: Affiliators are more likely to have friends who use little or no drugs, and they have experienced prior treatment. Contradictory to the hypothesis, however, are the findings that affiliators are more likely to experience feelings of hopelessness and they have received less family participation in their treatment process.

A classification table was produced to assess the predictive accuracy of the function; results appear in Table 5. The 72.9% of correctly classified cases may contain a slight upward bias (Hair et al., 1987). The potential for upward bias is usually identified by using a hold-out sample; however, the sample size in this study was too small to allow for this procedure.

Table 5

Classification Table

Actual Group Number of Predicted Group Membership
Cases Affiliation Non-affiliation

AA affiliation 31 61.3% (19) 38.7% (12)
Non-AA affiliation 39 17.9% (7) 82.1% (32)

Percent of Cases Correctly Classified: 72.86% (100) (19+32/70)

Correct Group Classification:

1. Affiliators 61.3%
2. Nonaffiliators 82.1%

Proportional Chance Criterion: 50.6%

In interpreting the classification data, it is important to compare the percentage of correctly classified cases with the a priori chance of classifying individuals correctly without the discriminant function. Using the proportional chance model one can accurately predict 50.6% of the sample to be AA affiliators or non-AA affiliators. The classification table shows that these variables predict nonaffiliators to an even greater degree. The discriminant function predicts group membership 22% more accurately than does the proportional chance model. Nonaffiliators are more likely to have friends who use drugs, have not experienced prior treatment, are more hopeful, and have had more family participation.

DISCUSSION

There are many limitations to this study in that it was conducted with a small population and that a limited amount of information was obtained from the sample. However, some interesting conclusions can be drawn.

Unfortunately, there were some limitations in the data collection procedures. No other information regarding demographics or other behavioral health diagnoses was obtained. There was also no pretreatment information or measurement to know what kinds of experiences patients brought to treatment, such as prior A.A. affiliation, that may have impacted the study.

Maiso and Connors (1988) recommend that multiple measures be taken, and from more than one source. The data from this study is limited to self-report; it would have made the data more reliable to have corroborating interviews with another informed party, such as a parent. A study by Winters, Stinchfield, Henly, and Henly (1991) did find, however, that adolescents do give consistent reports of their drug use.

While the nonresponse rate appears high (55%), actually only five respondents declined to be interviewed. The rest could not be contacted. This is not unusual in an alcoholic population (Moos, Finney, & Cronkite, 1990). Several studies have addressed the concern with the attrition rate in this population; that is, how difficult it is to find the treated alcoholic. Having only those who are easily located in the final sample analysis may lead to limited inferences since they are usually more stable and exhibit better functioning (Mackenzie, Funderburk, Allen, & Stefan, 1987; Moos et al., 1990). At the most, this study could be generalized only to adolescent addicts/alcoholics who were similar in their demographics (coming from employed families where insurance paid for the treatment), received the same model of treatment, and agreed to participate in a follow-up telephone survey. Further, a great deal is unknown about the respondents.

This study, like the others cited, is influenced by self-selection. What is different however, is that all respondents in the study received the same intervention (treatment with recommendation to attend AA) and then proceeded to affiliate or not affiliate. The respondents were all living in the community at the time of the survey, which eliminates the complicating difficulties of studying hospitalized patients. They include both females and males, and they were all adolescents.

Referrals to Alcoholics Anonymous for continued support upon discharge from treatment are consistently found in treatment plans in most alcoholism treatment centers across the United States. Not all clients follow this advice, or if they do, attendance may be minimal. The characteristic that best predicted affiliation, having friends that do not use drugs, may itself be an effect of the affiliation. Adolescents, once in A.A., may choose to socialize with similar peers they meet there.

The finding that prior treatment is associated with affiliation confirms previous research but can also be of interest to clinicians. It would be interesting to know if this prior treatment was a less restrictive alternative (i.e., outpatient therapy, or if respondents were recidivists). Having a "repeat patient" in treatment can be discouraging, yet this reveals that these youth may be more likely to join A.A. upon discharge than are peers who have not experienced any other treatment.

The finding that affiliators were more hopeless was surprising and conflicts with previous research on support group involvement (Powell, 1990). One explanation may be that the hopelessness scale asked many questions regarding the respondents' views of the future. In A.A., members are told to "live one day at a time" and not focus on the future since they cannot predict or control it. Therefore, to make assertions about a positive future may go against the "here and now" approach of Alcoholics Anonymous.

Finally, it is of interest that family participation was not predictive of affiliation. Hoffman and Kaplan (1991) found that family participation in treatment and in self-help groups after treatment was highly correlative with adolescent abstinence and A.A. affiliation. Alternatively, perhaps parents seek out treatment since they feel helpless to provide the type of parenting skills their adolescents need. While they participate in treatment, many may not make enough therapeutic gains to provide the kind of support their children want or need. The affiliator adolescents may find the support of a self-help program more relevant to their needs. It would be interesting to study how these adolescents perceive support, their parental support, and the type of support they receive through involvement with A.A.

Further research is clearly needed in this area of affiliation since this study shows that affiliators exhibit considerable variation that is difficult to explain. Perhaps a longitudinal study would provide more information, not only of the characteristics of affiliators, but how the process of affiliation is carried out.

CONCLUSIONS

This study of A.A. affiliation was better able to predict characteristics of adolescents who did not affiliate with A.A. than with those who did: Those who had friends who used drugs, had no prior treatment experience, had greater parental involvement while in treatment, and were more hopeful, were less likely to affiliate with A.A. The difficulty with predicting the affiliator group may be because there is more variation in this group.

This study has important implications for practice in that it is an initial attempt to describe adolescents who affiliate with A.A. Referral to Alcoholics Anonymous for adolescent clients who are alcoholic or substance abusers is a standard treatment practice of social workers and addictions counselors. These adolescents are not a homogeneous group, however, and it is important for clinicians to know which clients may benefit most from this type of referral.

An interesting variable that may have an impact on A.A. affiliation that was not considered in the study is the effect of drug or alcohol use itself. Obviously, self-help groups are supportive of abstinence and may attract only those who are abstinent.

A crosstabulation was conducted utilizing the variable of affiliation with self-reported drug and or alcohol use over the prior four weeks. The results are presented in Table 6.

While it is clear that the affiliators were more likely to be abstinent and the nonaffiliators were more likely to have used drugs, not quite half of the nonaffiliator group was abstinent. The abstainers had reached the overall treatment goal (no use of drugs or alcohol), yet only slightly more than half chose to follow through on the referral to affiliate with A.A.

These findings indicate that the original research question still needs exploration. Perhaps it should be expanded to answer more specifically: Who benefits from the Minnesota Model of treatment? Of those who are "treatment successes" (i.e., the abstainers), what qualities or characteristics predict who will benefit from A.A. affiliation? A larger sample of abstaining affiliator and nonaffiliators than was available in this study would be needed to explore these questions.

Table 6

Crosstabulation of Use and Affiliation

Affiliators Non-affiliators

No use of drugs/
alcohol 24 19
(77.4%) (49%)

Has used drugs/
alcohol 7 20
(22.6%) (51%)

The findings of this particular study raise some further questions, particularly those findings that did not go in the direction of the predicted hypothesis. As indicated earlier, it would be interesting to know exactly what kinds of treatment experiences these adolescents had prior to their inpatient treatment. Issues of perception of parental support and how this relates to affiliation could be explored in further detail. Finally, one might want to expand the analysis of personality or behavioral characteristics to include not only hopelessness but locus of control, including issues related to secondary control. If A.A. affiliators are given the message to not worry about the future, that they are powerless over the events of their life, and that they need to trust a higher power, perhaps what is then being measured is secondary control rather than feelings of hopelessness.

REFERENCES

About AA. (1990). AA world service office, Grand Central Station. New York, New York.

Alford, G. (1980). Alcoholics Anonymous: An empirical outcome study. Addictive Behaviors, 5, 359-370.

Boscarino, J. (1980). Factors related to "stable" and "unstable" affiliation with Alcoholics Anonymous. The International Journal of Addictions, 15, 839-848.

Canter, F. (1966). Personality factors related to participation in treatment by hospitalized male alcoholics. Journal of Clinical Psychology, 22, 114-116.

Fontana, A., Dowds, B., & Bethel, M. (1976). AA and group therapy for alcoholics: An application of the world hypotheses scale. Journal of Studies on Alcohol, 37, 675-682.

Gorski, T., & Miller, M. (1982). Counseling for relapse prevention. Independence, MO: Independence Press.

Gregson, R., & Taylor, G. (1977). Prediction of relapse in men alcoholics. Journal of Studies on Alcohol, 38, 1749-1759.

Gynther, M., & Billiant, P. (1967). Marital status, readmission to hospital, and intrapersonal and interpersonal perceptions of alcoholics. Quarterly Journal of Studies on Alcoholism, 28, 52-58.

Hair, J.S. Anderson, R.E., & Tatham, R.L. (1987). Multivariate data analysis. New York: Macmillan.

Hoffman, N., & Kaplan, R. (1991). One-year outcome results for adolescents: Key correlates and benefits of recovery. CATOR Report, 1-21.

Hurlburt, G., Gade, E., & Fuqua, D. (1983). Personality differences between Alcoholics Anonymous members and nonmembers. Journal of Studies on Alcohol, 45, 170-171.

Kazdin, A., Rodgers, A., & Colbus, D. (1986). The hopelessness scale for children: Psychometric characteristics and concurrent validity. Journal of Consulting and Clinical Psychology, 54, 241-245.

Klecka, W. (1980). Discriminant analysis. Beverly Hills: Sage.

Littrell, J. (1991). Understanding and treating alcoholism, Vol. 1. Hillsdale, NJ: Erlbaum.

Mackenzie, A., Funderburk, F., Allen, R., & Stefan, R. (1987). The characteristics of alcoholics frequently lost to follow-up. Journal of Studies on Alcohol, 48, 119-123.

Maisto, S., & Connors, G. (1988). Assessment of treatment outcome. In D. Donovan, & G. Marlatt (Eds.), Assessment of addictive behavior. New York: Guilford.

Marlatt, G. (1985). Cognitive factors in the relapse process. In G. Marlatt, & J. Gordon (Eds.), Relapse prevention: Maintenance strategies in the treatment of addictive behaviors. New York: Guilford.

Mayer, R., & Greenwood, E. (1980). The design of social policy research. Englewood, NJ: Prentice-Hall.

Mindlin, D. (1964). Attitude toward alcoholism and toward self: Differences in three alcoholic groups. Quarterly Journal of Studies on Alcoholism, 25, 136-141.

Moos, R., Finney, J., & Cronkite, R. (1990). Alcoholism treatment: Context, process and outcome. New York: Oxford University Press.

Reilly, D., & Sugarman, A. (1967). Conceptual complexity and psychological differentiation in alcoholics. The Journal of Nervous and Mental Disease, 144, 14-17.

O'Leary, M., Calsyn, D., Haddock, D., & Freeman, C. (1980). Differential alcohol use patterns and personality traits among three Alcoholics Anonymous attendance level groups: Further considerations of the affiliation profile. Drug and Alcohol Dependence, 5, 135-144.

Sudman, S., & Bradburn, N. (1989). Asking questions. San Francisco: Jossey-Bass.

Trice, H. (1959). The affiliation motive and readiness to join Alcoholics Anonymous. Quarterly Journal of Studies on Alcohol, 20, 313-320.

Vaillant, G. (1983). The natural history of alcoholism. Cambridge, MA: Harvard University Press.

Winters, K., Stinchfield, R., Henly. G., & Schwartz, R. (1991). Validity of adolescent self-report of alcohol and other drug involvement. The International Journal of the Addictions, 25, 1379-1395.

Melinda Hohman, Ph.D., Assistant Professor, San Diego State University, School of Social Work.

Reprint requests to Craig Winston LeCroy, Ph.D., Professor, School of Social Work, Arizona State University, 2424 E. Broadway, Suite 100, Tucson, Arizona 85719.

COPYRIGHT 1996 Libra Publishers, Inc.
This material is published under license from the publisher through the Gale Group, Farmington Hills, Michigan. All inquiries regarding rights should be directed to the Gale Group.