[Version formatted for printing (PDF)]

Abstract

Predicting outcomes for individual patients entering substance abuse treatment has long been a clinical goal in the addictions field. Intake data from the Addiction Severity Index and other standardized scales were collected on 248 alcohol dependent/abusing patients entering an urban hospital treatment program. The outcome measure was frequency of drinking days in the past 30 days. Baseline data were used to identify predictors of posttreatment drinking frequency at two follow-up interviews (3 and 12 months postbaseline). Stepwise multiple regressions indicated that a set of baseline predictors accounted for similar and substantial proportions of outcome variance at the two follow-ups. When psychosocial predictors were combined with an index of alcohol use severity (which included drinking frequency), the proportions of variance explained were 31% and 28% at 3 and 12 months, respectively. Two psychosocial predictors were significant at both time periods, and thus most likely to be replicated in future research: a treatment motivation index (a combination of measures of commitment to treatment success and internal motivation to seek treatment) and an index of 12-step (self-help) participation (a combination of measures of frequency of 12-step meeting attendance and perceived helpfulness of 12-step participation). While the predictability of short-term (3 month) outcomes could help clinicians tailor treatment strategies to maximize patient motivation and reduce drinking behavior, the predictability of longer term (12 month) outcomes could help counselors plan aftercare programs, encourage self-help participation, and promote recovery-oriented activities to sustain initial treatment-induced gains.

^†This study was supported by grant no. R01AA10863 from the National Institute on Alcohol Abuse and Alcoholism.

Introduction

Predicting outcomes for individual patients entering substance abuse treatment has long been a clinical goal in the addictions field ^[1,2] . Despite extensive research ^[1,3,4] , studies attempting to identify preadmission predictors of treatment outcomes have encountered certain problems: 1) Treatment retention has been used as a proxy measure for postdischarge treatment outcomes ^[5,6] . 2) Outcomes have been exclusively short-term (e.g., 3 months), even though longer term outcomes (e.g., 1 year) may require different predictors ^[1,4] . 3) Smaller or homogenous samples have resulted in overlooking robust predictors ^[5] . 4) The range of baseline measures has been restricted, e.g., to only static predictors or to only measures of prior substance use ^[1] . 5) Heavy reliance on single-item predictors has raised concerns about the reliability of the measures and, therefore, the replicability of findings ^[4] . and 6) Predictor selection has not extended beyond bivariate to more informative multivariate analyses ^[4,7,8] .

This study, part of a larger research project on alcoholism treatment outcomes ^[9] , attempts to avoid these six limitations. Based on data from a multimodality treatment program, a naturalistic longitudinal design is used to examine preadmission predictors of drinking outcomes for alcoholics. The study investigates two types of predictors (i.e., measures of alcoholism severity and general psychosocial measures) in relation to short-term (3 months) and longer term (12 months) measures of drinking frequency.

Methods

Setting

Smithers Treatment Center (New York, NY) accepts patients who have an addiction problem, with public or private insurance coverage as well as self-pay status. During the period of the study, Smithers offered treatment programs at three levels of care, according to the level of care criteria recommended by the American Association of Addiction Medicine: regular outpatient (with sessions of 1.5 hr on two evenings per week; normal stay of 10–12 weeks), intensive outpatient (which met 3.5 hr a day; normal stay ranging from 3 weeks to 3 months), and inpatient rehabilitation (maximum stay of 28 days).

Eligibility Criteria

Study subjects were alcohol dependent/abusing patients (determined by DSM-IV diagnosis utilizing the Structured Clinical Interview for DSM-IV [SCID]) who were accepted for treatment at Smithers. The sample included patients who were newly accepted for treatment and also those who had been discharged from hospital detoxification or inpatient care and were applying for continuing treatment. However, applicants requiring hospitalization were excluded from the study because they were referred to other facilities.

Procedures

Eligible applicants were approached in the Smithers Evaluation Unit and invited to participate in the study in order of their appearance at the program. Interviewer availability determined the number of interviews conducted each day. Participation in the research study was voluntary based on written informed consent. Participants were interviewed at admission to the program and were followed up for interviews 3 and 12 months later; they received $30 for their time for each interview. The study protocol was approved by the Institutional Review Boards of the two collaborating organizations. The baseline study sample consists of 248 applicants who were interviewed and entered treatment during the 18-month period between April 1998 and September 1999.

Sample Follow-Up

Of the 248 patients in the baseline sample, 219 (88%) were located and interviewed at the 3-month follow-up and 173 (70%) at 12 months. When patients followed up were compared with those not located at 3 and 12 months, only for ethnicity was there evidence of significant attrition bias; whites were significantly less likely to be reinterviewed at 12-month follow-up than non-whites (p<0.05).

Measures

The study's baseline interviews were conducted as an extension of the regular intake assessment process at Smithers. These interviews, which averaged 2.5 hr, included the Addiction Severity Index (ASI) ^[10] , which served as the primary source of this study's measures. Drawing on its individual items, the ASI provides composite scores for the severity of patient dysfunction in seven problem areas (medical, employment, alcohol, drug, legal, family/social, and psychiatric). It also provides measures of sociodemographic items such as age, sex, ethnicity, and education.

In addition, the baseline interview included six previously reported scales (see Table 1 for examples of items): Treatment Motivation Questionnaire (TMQ; separate subscales for internal and external sources of motivation) ^[11] , an adapted subset of items from the URICA ^[12] , the Beck Depression Inventory (BDI) ^[13] , an abbreviated Hamilton Depression Scale ^[14] , a spirituality scale ^[15] , and an abbreviated 12-Step Participation Scale ^[15,16] . Level of treatment entered by the patient (i.e., a three-category nominal variable: inpatient, intensive outpatient, and regular outpatient) was also included as a predictor variable.
 

The study's two follow-up interviews included the follow-up version of the ASI. The current analysis focuses on one core outcome measure, obtained from the ASI at baseline, and at both 3 and 12 months after admission: self-reported number of days in the past 30 days on which the patient consumed any alcohol. The measure does not make any adjustments for special circumstances, such as time that the patient may have spent in a controlled environment (e.g., inpatient program or hospitalization for detoxification).

Table 1 provides psychometric data on the new indices: their component items, coefficient alpha, and their zero-order or partial correlations with the outcome measures. Where appropriate, measures were adjusted to facilitate statistical analyses. For example, they were modified via recodes, log transformations (base e), and imputation (mean substitution) to avoid problems of item reversal, skewed distributions, reduced sample size, and/or missing data.

Data Analysis Plan

As summarized in Tables 1 2 3, the data analysis included zero-order correlations, partial correlations, and multiple regressions. The regression analyses, which were run with SPSS Version 10 ^[17] , used a combination of forced and stepwise entry of predictors. Statistical tests were two-tailed (alpha<0.05). Parallel statistical analyses were performed on the 3- and 12-month data.

The study divided measures into two domain-based categories: first, alcoholism severity (one domain), and second, psychosocial characteristics (all other domains). Because the outcome drinking measures (i.e., drinking frequency at 3 and 12 months) were themselves measures of alcoholism severity, the study initially investigated baseline severity measures (including drinking frequency) as predictors of drinking outcomes. These measures were used to create the study's control variable in the multivariate analyses.

The reports of the regression analysis also include the results of the initial regressions in the stepwise sequence. These contained only the alcoholism severity measure as a predictor, whereas the final regressions included psychosocial measures as well as the alcoholism severity measure as predictors. Comparisons between initial and final regressions indicated the incremental contribution of the psychosocial predictors.

Results

Baseline Characteristics of the Sample

The 248 subjects in the sample included a mix of working and middle-class patients, as well as those socioeconomically disadvantaged. The majority were male (72%); there were more African Americans (44%) than whites (35%) or Hispanics (17%). Almost one quarter (23%) had less education than a high school diploma or GED, another quarter (26%) had a high school diploma (or its equivalent), and the remaining one half (51%) had more education than a high school diploma. Their ages ranged from 19 to 72 (mean=39.8; sd=8.8). Over one half (58%) had some form of government-related insurance (usually Medicaid); one third (35%) had private insurance; and 7% were self-pay; 44% of the subjects were employed either full- or part-time.

The 248 subjects were distributed among the three treatment programs at Smithers: regular outpatient (28%), intensive outpatient (18%), and inpatient rehabilitation (54%).

Substance Use

At intake, the mean number of days that subjects reported drinking alcohol in the last 30 was 15.4 (sd=9.5). Use of additional psychoactive drugs was prevalent (68%), with 53% also using cocaine, 16% opiates, 24% cannabis, and 12% other drugs at least one day in the past 30 days.

Frequency of drinking diminished substantially and significantly between intake and the two follow-ups. Three- and 12-month drinking days were 4.5 (sd=8.6) and 4.8 (sd=9.0), respectively.

Alcohol Severity as a Predictor

As noted, three items measuring baseline alcoholism severity predicted posttreatment drinking outcomes at both the 3- and 12-month follow-ups (Table 1 ). These items were number of drinking days in past 30, number of dollars spent on alcohol in past 30 days, and number of days with alcohol problems in past 30. Their high intercorrelations justified the construction of the three-item index of alcoholism severity (coefficient alpha=0.77). The percentage of outcome variance accounted for by this alcohol severity composite index was the same for the 3- and 12-month data (i.e., 6%).

 [Top of Page]

Psychosocial Measures as Predictors

Table 2 3 also reports results of the regressions for the 3- and 12-month drinking outcomes, which included the control on baseline alcoholism severity. The total percentage of outcome variance accounted for by all independent variables was substantial and similar for both time periods: 31% at 3 months and 28% at 12 months. The increment in variance accounted for (i.e., change in R square) by the addition of the psychosocial predictors to alcohol severity was 25% at 3 months and 22% at 12 months (both p<0.001).

When treatment level was added as a predictor to the 3- and 12-month regressions, it was not significant for either time period.

Discussion

This study of intake predictors of drinking outcomes of substance abuse treatment sought to avoid prior methodological problems, i.e., avoid proxy measures of drinking outcomes, short-term outcomes exclusively, small homogeneous samples, a narrow range of baseline measures, single-item predictor measures, bivariate analyses exclusively, and attrition bias. In the latter instance, although a lower proportion of whites than non-whites were reinterviewed at 12 months, the lack of association between race and drinking frequency at 12 months suggests that the study's findings were not distorted by retrieval bias.

Given the reasonable expectation that predictability would decline between 3 and 12 months, the small reduction in variance accounted for by baseline measures (from 31% to 28%) suggests that, notwithstanding an intervening treatment episode, historical and other baseline characteristics remain associated with drinking outcomes for an extended period.

 [Top of Page]

Six of 10 psychosocial predictors emerged as significant in the 3-month regressions, and five of eight in the 12-month regressions were significant. Most of the significant predictors were multi-item measures, specifically, three of the six in the 3-month regression (i.e., treatment motivation, 12-step participation, medical problem severity) and all five in the 12-month regression.

Although known at intake, treatment modality was not strictly a baseline characteristic because it included treatment effects (i.e., the effect of treatment intensity). Nonetheless, the failure of treatment level to uniquely predict outcomes or affect the significance of other predictors underscores the generality of the findings across treatment conditions and thus supports the study's use of a heterogeneous sample.

 [Top of Page]

Study Limitations

Some other types of drinking outcome measures (e.g., those involving continuous drinking, heavy drinking, abstinent periods, etc) that have been used to advantage in other studies ^[4] were not used in this study.

Fourth, outcomes were not measured beyond 12-months postadmission, which is longer than most studies, but less than the 2 or 3 years of several other recent studies ^[4,22] .

References:

¹  Ehrman RN, Robbins SJ, Cornish JW, Results of a baseline urine test predict levels of cocaine use during treatment, Drug Alcohol Depend, 62 (2001) 1–7.

²  McLellan AT, Luborsky, Woody GE, O'Brien CP, Druley KA, Predicting response to alcohol and drug abuse treatments, Arch Gen Psychiatry, 40 (1983) 620–625.

³  Long CG, Williams M, Hollin CR, Alcoholism treatment: intake variables as outcome predictors, Addict Res, 6 (1998) 295–305.

⁴  McKay JR, Weiss RV, A review of temporal effects and outcomes predictors in substance abuse treatment studies with long-term follow-ups, Eval Rev, 25 (2001) 113–161.

⁵  Mammo A, Weinbaum DF, Some factors that influence dropping out from outpatient alcoholism treatment facilities, J Stud Alcohol, 54 (1993) 92–101.

⁶  McCaul ME, Svikis DS, Moore RD, Predictors of outpatient treatment retention: patient versus substance use characteristics, Drug Alcohol Depend, 62 (2001) 9–17.

⁷  Lehman WEK, Simpson DD Prediction of year 12 outcomes, Opioid addiction and treatment: A 12-year follow-up, Simpson DD, Sells SB Krieger, Malabar, 1990, pp. 203–220.

⁸  McKay JR, Merikle E, Mulvaney FD, Weiss RV, Koppenhaver JM, Factors accounting for cocaine use two years following initiation of continuing care, Addiction, 96 (2001) 213–225.

⁹  Magura S, Staines G, Kosanke N, Rosenblum A, Foote J, Deluca A, Bali P. Predictive validity of the ASAM patient placement criteria for naturalistically matched vs. mismatched alcohol dependent patients. American Journal on Addictions, .

¹⁰  McLellan AT, Luborsky L, Cacciola J, Evans F, New data from the Addiction Severity Index: reliability and validity in three centers, J Nerv Ment Dis, 173 (1985) 412–423.

¹¹  Ryan RM, Plant RW, O'Malley S, Initial motivations for alcohol treatment: relations with patient characteristics, treatment involvement, and dropout, Addict Behav, 20 (1995) 279–297.

¹²  Willoughby FW, Edens JF, Construct validity and predictive utility of the stages of change scale for alcoholics, J Subst Abuse, 8 (1996) 275–291.

¹³  Beck AT, Ward CH, Mendelson M, Mock J, Erbaugh J, An inventory for measuring depression, Arch Gen Psychiatry, 4 (1961) 561–571.

¹⁴  Steer RA, McElroy MG, Beck AT, Correlates of self-reported and clinically assessed depression in outpatient alcoholics, J Clin Psychol, 39 (1983) 144–149.

¹⁵  Laudet A, Magura S, Vogel H, Knight E, Support, mutual aid and recovery from dual diagnosis, Community Ment Health J, 36 (2000) 457–476.

¹⁶  Magura S, Laudet A, Mahmood D, Rosenblum A, Vogel H, Knight E, The role of self-help processes on achieving abstinence in dual recovery, Addict Behav,  in press.

¹⁷   10.0 Syntax Reference Guide,  SPSS Inc., 1999.

¹⁸  Humphreys K, Huebsch PD, Finney JW, Moos RH, A comparative evaluation of substance abuse treatment: V. Substance abuse treatment can enhance the effectiveness of self-help groups, Alcohol Clin Exp Res, 23 (1999) 558–563.

¹⁹  Markowitz F, DeMasi M, Carpinello S et al. The role of self-help in the recovery process. Paper presented at the 6th Annual National Conference on State Mental Health Agency Services Research and Program Evaluation, Arlington, VA. 1996.

²⁰  Brown BS, O'Grady KE, Farrell EV, , Factors associated with frequency of 12-step attendance by drug abuse clients, Am J Alcohol Drug Abuse, 27 (2001) 147–160.

²¹  Sterling RC, Gottheil E, Glassman SD, Weinstein SP, Serota RD, Lundy A, Correlates of employment: a cohort study, Am J Drug Alcohol Abuse, 27 (2001) 137–146.

²²   , Matching alcoholism treatments to client heterogeneity: Project Match three-year drinking outcomes, Alcohol Clin Exp Res, 22 (1998) 1300–1311.

²³  Staines GL, Magura S, Foote J, DeLuca A, Kosanke N, Polysubstance use among alcoholics, J Addict Dis, 20 (2001) 53–69.