Journal of Substance Abuse Treatment 25 (2003) 125 – 134
Regular article
Overview of 5-year followup outcomes in the drug abuse treatment outcome studies (DATOS) Robert L. Hubbard, Ph.D., M.B.A. *, S. Gail Craddock, M.S., Jill Anderson, M.S. Institute for Community-Based Research, National Development and Research Institutes, Inc. 940 Main Campus Drive, Suite 140 Raleigh, NC 27606, USA
Abstract Followup results from the Drug Abuse Treatment Outcome Studies (DATOS) 1-year and 5-year followups were used to describe the longterm outcomes of drug treatment and to further clarify the relationship between treatment duration and post-treatment outcomes in four treatment modalities: outpatient methadone, long-term residential (LTR), outpatient drug free (ODF), and short-term inpatient. Methods replicating those used in earlier analyses of the DATOS 1-year followup of 2,966 patients admitted to treatment in 1991 – 1993 and those of the Treatment Outcome Prospective Study patients admitted in 1979 – 1981 were employed. DATOS is a non-experimental longitudinal study conducted within the natural settings of 96 treatment programs in the U.S.A. The study followed patients during and after treatment at specified periods of time. Prevalence of drug use and behaviors were evaluated for the year prior to treatment; and the post-treatment time frames defined by the 1- and 5-year followups. In addition, the multivariate analytic technique of generalized estimating equations was used to examine the relationship of treatment duration and outcomes across both followups while also controlling for patient characteristics and pretreatment levels of behaviors. The 5-year stratified followup sample included 1,393 of the same individuals in the 1-year followup sample. Analyses were restricted to patients participating in both followups. Reductions in prevalence of cocaine use in the year after treatment (compared to the preadmission year) by patients were associated with longer treatment durations (particularly 6 months or more in LTR and ODF). In addition, reductions in illegal activity and increases in full-time employment were related to treatment stays of 6 months or longer for patients in LTR. The DATOS results from the 1-year and 5-year post-treatment followup combined suggest the stability of outcomes of substance abuse treatment. While results are generally consistent with the full 1-year followup, reduced sample size and bias of the sample toward patients with longer treatment retention may have attenuated the findings. D 2003 Elsevier Inc. All rights reserved. Keywords: Long-term outcomes; DATOS; Retention; Multiple logistic regression analyses; Treatment modality differences
1. Introduction The Drug Abuse Treatment Outcome Studies (DATOS) are the third in a series of evaluations in the United States that provide data to describe not only the outcomes of treatment immediately after termination but also outcomes over extended time frames. Each of the studies focused primarily on outcomes in the first years after treatment, although they also included longer term outcome assessments up to 5 years after treatment, and in the first of these national studies, up to 12 years after treatment. The major goals for the longer-term studies were to document the levels of drug use and other behaviors 5 years after treatment, investigate the effects of the initial treatment experience, and identify factors * Corresponding author. Tel.: +1-919-863-4600; fax: +1-919-863-4601. E-mail addresses:
[email protected] (R.L. Hubbard),
[email protected] (S.G. Craddock). 0740-5472/03/$ – see front matter D 2003 Elsevier Inc. All rights reserved. doi:10.1016/S0740-5472(03)00130-2
that could been seen as significant to the maintenance of positive treatment effects. The present paper addresses the first two aims for the 5year followup of patients involved in the DATOS research and provides a foundation to inform analyses targeting the third aim. The first national multi-program study, the Drug Abuse Reporting Program (DARP), followed a sample of 4,107 from a cohort of 27,214 patients admitted to treatment from 1969 to1972 (Simpson & Sells, 1982) with additional analyses focusing on 405 male opioid addicts up to 12 years after treatment. The second study, the Treatment Outcome Prospective Study (TOPS), modeled on DARP, followed samples of 4,270 of 9,989 patients admitted to treatment from 1979 –1981 (Hubbard et al., 1989). The 1979 cohort received 12 and 24 month followup interviews; the 1980 cohort was interviewed at 3 and 12 month followups; and, the 1981 cohort was interviewed from 3 to 5 years following treatment. Both studies documented that the large decreases
126
R.L. Hubbard et al. / Journal of Substance Abuse Treatment 25 (2003) 125–134
in opioid use and criminal involvement found in the first year after treatment continued for periods of 3 to 5 years after treatment (Hubbard, Marsden, Cavanaugh, Rachal, & Ginzburg, 1988; Hubbard et al., 1989; Sells & Simpson, 1980; Simpson, 1981; Simpson & Sells, 1982, 1990). While the analyses of the first year post-treatment found that treatment durations of 3 months or more were associated with more positive outcomes, the longer term outcomes were not as strongly associated with the duration of treatment in the index program. The review of these studies indicate that treatment was effective for opioid users entering the three major publicly-funded treatment modalities: outpatient methadone maintenance treatment (OMT), long term residential (LTR) and outpatient drug free programs (ODF), but the role of treatment in sustaining the behavior change diminished over time. The research indicated the need to investigate the factors that contribute to sustaining the decreases in drug use and other negative behaviors as well as the factors that support a more positive lifestyle. This paper presents an overview of outcomes within the context of patient and program diversity change previously described for DATOS (Etheridge, Hubbard, Anderson, Craddock, & Flynn, 1997). The 1-year and 5-year followups of DATOS patients occurred at critical points in the evolution of drug abuse treatment in the U.S. DATOS was the first national study of treatment examining patients in the three traditional modalities, OMT, LTR, and ODF, since the transition of funding authority to the states in 1981, the first since the epidemic of HIV/AIDS, and also the first to include patients in both private and public short-term inpatient (STI) programs. The inclusion of health and HIV risk measures in DATOS expanded the range of behaviors that may be affected by drug abuse treatment. DATOS is also the first national study of the effectiveness of typical communitybased programs in treating cocaine abuse. The design of the DATOS research incorporated measures, programs, and communities included in TOPS and permitted a unique contrast of outcomes in two eras with diverse patient populations and differing in health and social service systems. The findings reported in this paper are important both in their replication of previous results and their identification of the need for further investigation of emerging issues that support and sustain long-term recovery.
2. Methods The prospective epidemiological design for DATOS is described more fully in Flynn, Craddock, Hubbard, Anderson, and Etheridge (1997). The design enables investigation of correlates of treatment outcomes for subgroups of patients. Compared to the focus of the DARP and TOPS on general effectiveness and the association of retention and outcomes, the goal of DATOS emphasized the need to examine the complex associations of patient characteristics, treatment process, services environment, and outcome. The
sample selection, data collection, and measurement methods were chosen to support this goal. 2.1. Sample The original DATOS admission cohort included 10,010 patients who were interviewed in 96 programs from 11 cities at entry into treatment and during treatment and included programs from each of the four major treatment modalities (OMT, LTR, ODF, STI). The selection of programs was purposive to insure that patients could receive basic therapies and more comprehensive elements of treatment. The selection process was designed to maximize the goal of having the in-depth DATOS research findings provide information on patterns of behavior and relationships of patient characteristics and treatment factors with post-treatment outcomes. While recognizing the limitations for statistical representation, the conclusions were intended to broadly and reasonably generalize to patients treated in typical community-based programs in medium to large urban areas in the U.S. The sample for the 5-year followup was based on these considerations. The 5-year followup sample frame is the same as the 1year followup sample frame, described in detail by Flynn et al. (1997). The frame was restricted to patients who had completed both the Intake-1 and the Intake-2 questionnaires (n = 8,109) in order to have complete baseline data on all followup patients. Therefore, unlike the DARP and TOPS studies, most patients with less than 1 week of treatment were excluded. Patients from programs with fewer than 20 admissions were omitted from this frame resulting in data for 76 programs (19 LTR, 12 STI, 21 OMT, and 24 ODF). The eligible followup sample of 4,229 patients was stratified by program and treatment duration of less than or greater than 3 months (1 month for STI) and randomly selected to include a minimum of 1,000 patients for each treatment modality. In order to support analyses of variations in core therapies and comprehensive services, a larger proportion of patients completing at least 3 months of OMT, LTR, and ODF treatment or 1 month of treatment in the STI modality were selected for the sample. This selection ensured the availability of complete baseline information on patient impairment as well as treatment information obtained from during treatment interviews to support analyses of correlates of outcomes. The overall response rate for the 1-year followup was 74% (n = 3,147) with 26% (n = 1,082) who could not be located. Seventy percent (n = 2,966) of the followup sample were interviewed successfully, 1.5% (n = 64) were deceased, and 2.7% (n = 117) refused to be interviewed. To enable examination of the stability of the outcomes and the relationship of 1-year and 5-year outcomes, the design of the 5-year followup data collection was targeted for the 2,966 respondents to the 1-year followup as described above. Cases were excluded for both scientific and cost reasons: 619 cases in three cities were not followed due to small samples and the high cost of field work; 200 cases
R.L. Hubbard et al. / Journal of Substance Abuse Treatment 25 (2003) 125–134
for whom there was no locator information were dropped; and 106 others who were institutionalized or out of the interview area were excluded. The remaining 2041 patients from 18 OMT, 16 LTR, 20 OPF, and 8 STI programs in eight cities (Chicago, Miami, Minneapolis, Newark, New York, Phoenix, Pittsburgh, and Portland) formed the eligible sample of whom 1,618 were located (79%); 1,393 were interviewed (68%); 128 died (7%); and 65 refused (3%). Thus, of the patients targeted in the data collection, 73% were interviewed. The current analyses are focused on the 1,393 patients who were interviewed at the 5-year followup. While these patients may reflect a bias due to attrition in succeeding waves of data collection, a strength of this sample is the complete data available over the 5 years of followup for very heterogeneous subpopulations of patients. Three time frames are the subject of most analyses: (1) the pre-treatment year, which is the 12-month period prior to admission to treatment; (2) the post-treatment period of 12 months just prior to the 1year followup interview; and (3) the post-treatment period of 12 months just prior to the 5-year followup interview. Due to the timing of patients’ admission to treatment between 1991 and 1993 and the fieldwork scheduling, there is variation in the exact timeframes for which patients were interviewed as related to their admissions and discharges. 2.2. Measures The DATOS database contains multiple measures of outcomes and many explanatory measures that assess potential influences on those outcomes. Throughout the history of DARP, TOPS, and DATOS, the analyses have focused on basic, easily interpretable outcomes. A set of dichotomous variables clearly delineating the fundamental nature of outcomes has been developed, tested, and replicated. From this base, nine measures of outcomes covering diverse ranges of behaviors and functioning were selected to describe change between the pre-admission and followup years. The nine dichotomous measures (four on substance use and five on behavioral functioning) describe typical outcomes during the followup year. The measures of substance use are based on self-reports of heroin, cocaine (in any form), marijuana, and alcohol. Extensive levels of use that are clearly detrimental for the individual and community are defined as at least weekly use of heroin, cocaine, or marijuana, and 5 or more drinks in one sitting, at least once a week, for alcohol. Weekly or more frequent use over a 1-year period was highly correlated with clinical assessments of substance dependence. The prevalence rates of weekly or more frequent substance use at followup were also highly correlated with positive urine screening results (Flynn, Craddock, & Dunteman, 1995). The correlation of self-reported cocaine and opiate use with biologic measures (e.g., positive urine or hair test) was reconfirmed in the DATOS 5-year followup (Simpson, Joe, & Broome (2002), Broome, Joe, Flynn, & Simpson, 2002).
127
The measures describe functioning in a variety of areas including: criminal behavior, mental health, employment, sexual activities, and health. The dichotomous measures of outcome in these areas indicate the presence or absence of a specified level of a behavior in the followup year: predatory illegal acts (assault, robbery, burglary, larceny, forgery, stolen property), suicidal thoughts and/or attempts, less than fulltime work (working less than 35 h per week for 40 or more weeks a year), sexual behavior risk (sexual intercourse with two or more people without always using a condom), and health limitations (limitations in activities due to health for at least 90 days). The dichotomous measures provide the most easily interpretable clinical results and reduce the threat of spurious or illusory relationships due to scale, statistical, or measurement artifacts. All four measures of substance use and three of the five behavioral measures are exact replicas of those used in TOPS analyses (Hubbard et al., 1988, 1989). All measures are based on standard instruments used in research in other areas. The validity and reliability of selfreports of substance use, suicidal ideation, criminal behavior and employment measures was established in TOPS (Hubbard, Marsden, & Allison, 1984) and additional validation of self-reports of cocaine and other substance use was conducted for DATOS (Flynn et al., 1995). Two behavioral measures, sexual risk and health limitations, introduced in DATOS, are based on the data collection and analysis methodology established in other national studies of HIV risk for injecting drug users (Brown & Beschner, 1993) and health outcomes (Stewart & Ware, 1992). 2.3. Statistical analysis The analyses follow the basic plan used for DARP, TOPS, and DATOS, focusing first on clear descriptions and then on fundamental relationships. Later analyses can build upon this foundation. First, logistic regressions were used to assess differences in sample characteristics due to followup attrition. In a second set of analyses, paired t-tests were used to test prevalence rates of drug use and behaviors at the 1-year followup vs. the 5-year followup. Third, to provide direct tests of time in index treatment on each outcome, chi-square tests were done across time in treatment within each treatment modality. Statistical adjustments were applied to adjust p values downward to account for the large number of tests conducted (Benjamini & Hochberg, 1995). Finally, generalized estimating equations (GEE) were performed to test the overall impact of time in index treatment on the selected treatment outcome measures. The GEE models controlled for important patient background variables (age, marital status, education, race/ethnicity), source of treatment referral, more recent treatment participation, and the pre-treatment level of the outcome measure under consideration. The GEE models test the outcomes across both the 1-year and 5-year followup periods together, while controlling for the correlations arising from repeated measures over time. GEE is particularly suited to binary outcome measures in which
128
R.L. Hubbard et al. / Journal of Substance Abuse Treatment 25 (2003) 125–134
the assumption of normality is not met (Johnston,1999; Liang & >Zeger, 1986).
3. Results The analyses proceed systematically from the basic question of whether patients change after admission to treatment and how long this change persists. The relationship of treatment duration to change is then investigated and the descriptive findings are confirmed with multivariate analysis. Prior to the analysis for each question, the potential effects of bias due to attrition from the followup are discussed. Within each analysis a direct comparison is made between the results obtained for the TOPS (1979 –1981) and DATOS (1991 –1993) samples of patients. 3.1. Followup attrition Previous analyses reported in Hubbard, Craddock, Flynn, Anderson, and Etheridge (1997) compared the DATOS 1year followup respondents to the non-respondents in the followup sample frame. Significant results from those models will be reported here. In addition, analyses were conducted to determine if the 5-year followup respondents differed from the group of patients who were 1-year followup respondents, but were not respondents to the 5-year followup. (This second group of patients included both nonrespondents to the 5-year followup and those who were excluded from the eligible sample.) In these analyses the dependent variable indicated whether a 1-year followup respondent was also a respondent in the 5-year followup. Sets of logistic regressions were generated for each modality to examine the following independent variables: gender, race/ethnicity, age at the time of admission; marijuana, heroin, cocaine, and alcohol use (weekly or more frequent use in the year before treatment), days in treatment, referral source to the treatment, prior drug treatment, and character-
istics at admission including education, marital status and health insurance coverage. In OMT, previous results indicated that, compared to nonrespondents, those who completed 1-year followups had half the odds of being Hispanic, twice the odds of using heroin weekly or daily, about two thirds the odds of using cocaine, and were more likely to be in treatment. All results were significant at the p < .01 level (Hubbard et al., 1997). The OMT respondents to the 5-year followup showed no significant differences from the 1-year respondents lost to the 5year followup. In LTR, 1-year respondents had more days in treatment ( p < .01) than 1-year non-respondents, but no other variables showed significance. The 5-year followup respondents were more likely to have a criminal justice referral ( p < .05) and to have even more days in treatment than those 1-year respondents lost to the 5-year followup. In STI, 1-year followup respondents showed no significant differences from non-respondents on any variables in the model. However, 5-year followup respondents were more likely ( p < .01) to have a high school education, and private health insurance ( p < .01) than those 1-year respondents lost to the 5-year followup. The model comparing 1-year followup respondents and non-respondents in ODF was significant only in that Hispanics had three fifths the odds of being a respondent ( p < .01). Again, 5-year respondents were about half as likely as those who did respond at 1-year but were lost to 5-year followup, to be minorities ( p < .01). In addition, 5-year respondents were more likely to have had a high school education or GED at the time of admission to treatment. Overall, these results indicate that patients who responded to the 5-year followup were somewhat different from 5-year non-respondents in several areas. Consequently, the data presented on substance use and other behaviors in the 5-year followup are somewhat biased toward patients who remained in treatment longer (LTR), patients with more education (STI and ODF), patients with private health insurance (STI), and non-minorities (ODF). Table 1 reports percentages of these
Table 1 Characteristics of 1-year follow-up sample frame and 5-year followup respondents
More than 30 years of age Male African American or Hispanic High school or GED Married or living as married Criminal justice referral Previous drug treatment Private health Insurance Treatment duration more than 3 months
Outpatient methadone treatment
Long-term residential
Outpatient drug free
Short-term Inpatient
Frame (N = 1,203) %
Respondents (n = 432) %
Frame (N = 2,293) %
Respondents (n = 331) %
Frame (N = 2,000) %
Respondents (n = 364) %
Frame (N = 2,613) %
Respondents (n = 266) %
80.6 61.1 52.1 65.5 41.3 2.3 75.3 10.6
84.3 58.3 50.2 68.1 39.8 3.0 77.6 9.6
47.8 67.4 61.7 57.1 22.5 33.1 60.0 5.6
50.8 61.9 62.2 62.5 20.5 40.5 59.6 4.3
55.4 66.9 63.4 62.1 29.6 42.6 47.2 15.2
60.2 64.3 52.2 71.2 27.8 38.5 51.7 18.6
60.5 66.8 57.6 70.4 36.8 5.5 45.9 39.3
69.9 71.1 55.6 84.6 43.2 3.4 37.6 69.7
86.1
93.3
53.3
73.7
51.7
67.3
8.6
16.2
R.L. Hubbard et al. / Journal of Substance Abuse Treatment 25 (2003) 125–134
129
more frequent use of marijuana were similar across the three time periods. However, problem alcohol use (ie. 5 or more drinks at one sitting at least weekly) showed a significant decrease from the 1- to 5-year followup timeframe. In LTR programs the percentage of patients reporting at least weekly use was lower in both followups across all substances. Cocaine prevalence dropped from 65% in the preadmission year to 18% at the 1-year followup, but increased significantly to 26% ( p < .05) for the 5-year followup. Heroin use was reported by 17% prior to treatment, 3% at the 1-year followup, and 10% at the 5-year followup ( p < .05). The overall percentage of patients who reported problem alcohol use (as defined above) was reduced from 36% to 14% at the 1-year followup and stayed at about the same level (14%) for the 5-year followup period. Marijuana use was also reduced by more than half from pretreatment to the 1-year followup, although use increased a statistically insignificant 4 percentage points during the 5-year followup period. Among ODF patients, the percentage of weekly users of cocaine or marijuana dropped by more than half at the 1year followup. While the increase for marijuana at 5-year followup was significant ( p < .05), cocaine was not. Problem alcohol use (as defined above) declined from 33% to 15% at the 1-year followup and to 13% at the 5-year followup. Among STI patients, all substance use decreased at the 1year followup period and none changed significantly from those levels by the 5-year point. The trends in other behaviors from pre-treatment to 1-year post-treatment were less consistent in showing
characteristics for the original 1-year followup sample frame and for the 5-year followup respondents. 3.2. Change in substance use and behavior in the preadmission and followup years A fundamental issue in understanding treatment effectiveness is whether and to what extent patients modify their substance use and related behavior after receiving treatment. A second concern is the extent to which those changes persist through time. Table 2 reports on overall prevalence of behaviors of the respondents to the 5-year followup at three points in time: (1) the year prior to DATOS treatment; (2) the year following treatment, i.e. the 1-year followup; and (3) the 5th year following treatment, i.e. the 5-year followup. Paired t-tests are conducted to determine if there are overall differences in outcomes between the time periods of 1-year and 5-year followup for the patients participating in both interviews. Table 2 results show major reductions in most types of substance use across all modalities from the year before treatment to each followup time period. In OMT, the percent of subjects using heroin at least weekly in each of the followup years was at least one third of the percent that were using in the preadmission year. Nonetheless, the 5-year followup results indicated significantly greater use than was seen at 1 year. Weekly cocaine use decreased from 45% at admission to 22% and 21% in the two successive followup periods. Percentages of OMT patients reporting weekly or Table 2 Drug use and behaviors at intake, one-year follow-up and five-year followup Outpatient methadone treatment (n = 432)
Long-term residential (n = 331)
Outpatient drug free (n = 364)
Short-term inpatient (n = 266)
PreOne-year Five-year PreOne-year Five-year PreOne-year Five-year PreOne-year Five-year admission follow-up follow-up admission follow-up follow-up admission follow-up follow-up admission follow-up follow-up year % % % year % % % year % % % year % % % Heroin use Cocaine use Marijuana use Problem alcohol use Suicidal thought/ attempt Predatory illegal acts Sexual risk behavior Full-time work Health limitations
91.0 45.1
24.1 21.7
31.1* 20.9
17.3 65.4
2.5 18.2
9.7** 26.0**
7.7 37.4
3.9 13.8
5.5 17.3
9.0 61.7
3.8 18.2
4.2 16.2
15.8
13.1
13.9
24.5
11.9
16.3
27.0
11.6
18.7*
34.2
12.6
14.0
16.4
15.9
10.2*
36.3
14.4
14.2
32.5
14.9
13.2
51.3
21.1
16.5
14.6
12.3
14.8
22.7
10.3
10.3
21.7
11.3
15.4
30.8
15.4
16.9
31.2
14.3
12.7
40.1
15.3
15.7
24.3
9.8
8.5
16.3
7.7
5.6
27.3
12.7
17.6
48.5
26.7
26.7
33.3
25.4
17.9*
37.0
20.2
20.8
14.9
19.1
25.1*
10.6
25.5
36.4**
18.0
33.2
42.3*
51.4
43.9
54.3**
37.5
32.6
34.7*
29.3
21.8
13.3*
31.6
19.0
19.2
30.5
22.6
13.2**
Note: Differences between one-year and five-year follow-up percentages were tested using paired t-tests. * p < .05. ** p < .01.
130
R.L. Hubbard et al. / Journal of Substance Abuse Treatment 25 (2003) 125–134
improvement than seen in the substance use analyses. From the 1-year to 5-year followup periods, full-time work and crime showed significant improvements. Across all modalities, the proportion of patients working full-time increased significantly from the 1-year to the 5-year followup measurement periods. Predatory illegal activity declined sharply among patients from all four treatment modalities (31% to 14% for OMT, 40% to 15% for LTR, 24% to 10% for ODF, and 16% to 8% for STI) during the first year after treatment and stayed at the same level through the 5-year followup timeframe (all p < .01). With regard to health problems, LTR and STI patients showed continued and significant improvements from the 1-year to the 5-year followup.
followups indicated that most patients who improved their behavior did so for both time periods. Paired t-tests were conducted for differences between improvement rates in substance use at the 1-year vs. the 5-year followups. With the exception of lower improvement in the 5-year followup for heavy alcohol ( p < .05) and cocaine among LTR patients ( p < .05), rates were generally comparable across the 1-year and the 5-year followups. Thus, while there were small percentages of patients who either relapsed back to negative behaviors at the 5-year period or who improved their behavior from the 1-year to the 5-year followup, the earlier findings regarding general improvement at the 1-year followup (Hubbard et al., 1997) hold true for nearly all substance use outcomes at the 5-year followup.
3.3. Stability of substance use in followup years for patients in treatment at least 3 months To provide more meaningful comparative descriptions of change in substance use, the preceding analyses were conducted among subgroups of patients who reported weekly or daily use in the year prior to treatment. Comparing the proportions of those patients who no longer report weekly or daily drug use in each of the followup years focuses on just those with the specific problems at the time of the index admission (ie. excludes patients who never used the specific drug). The ‘‘improvement rate’’ is defined as the percentage of weekly/daily users of each specific drug who decreased their frequency of use or ceased use in the 1-year followup or at the 5-year followup. In general, these analyses showed stable reductions in substance use at both the 1-year and 5-year followups, when compared to the year prior to treatment admission (see Table 3). The combined rate, which is the percentage of patients reporting reduced drug use during both 1-year and 5-year
3.4. Substance use and behavior in the followup years by treatment duration The general finding from the DARP, TOPS, and the DATOS 1-year followups that treatment duration (e.g., time in treatment at the DATOS index program) of at least 3 months was associated with more positive outcomes supported an inference of treatment effectiveness. To investigate whether treatment duration effects were also present at the 5-year followup, analyses were performed to compare short and long-term treatment patients (see Table 4). Because the DATOS patients contacted for the 5-year followup had been in treatment longer than the patients in the 1-year followup sample, especially in the OMT and LTR modalities, there was a much smaller number of short-term treatment patients in the 5-year sample, and it was not possible to conduct all treatment duration analyses for each drug use subgroup.
Table 3 Drug use and behaviors in the 5-year follow-up year by modality and treatment duration
Heroin use Cocaine use Marijuana use Problem alcohol use Suicidal thoughts or attempts Predatory illegal activity Sexual risk behavior Full-time work Health limitations * p < .05. ** p < .01.
Outpatient methadone treatment
Long term residential
Outpatient drug free
Short term inpatient
1 year or less (n = 117) %
More than 1 year (n = 314) %
m2
Less than 6 months (n = 172) %
More than 6 months (n = 159) %
m2
Less than 3 months (n = 119) %
More than 3 months (n = 245) %
m2
Month or less (n = 187) %
More than 1 month (n = 79) %
m2
37.6 23.9 11.2
28.4 19.9 14.7
3.38 0.85 0.85
10.5 34.3 19.8
8.8 17.0 12.6
0.26 12.89* 3.13
7.6 20.2 17.7
4.5 15.9 19.2
1.46 1.01 0.12
4.84 17.11 13.37
2.53 13.92 15.38
0.74 0.42 0.19
7.7
11.2
1.13
18.0
10.1
4.20*
16.0
11.8
1.19
14.97
20.25
1.12
16.2
14.3
0.25
15.7
4.4
11.44*
21.0
12.7
4.30
19.79
10.13
3.69
12.8
12.7
0.00
21.5
9.4
9.10
7.6
9.0
0.21
5.88
5.06
0.07
19.7 24.8 35.0
16.9 25.2 34.4
0.43 0.01 0.02
26.7 27.9 15.1
26.6 45.6 11.3
0.00 11.10* 1.03
16.0 47.9 16.0
18.8 39.6 20.8
0.43 2.26 1.21
20.97 53.23 11.76
20.25 56.96 16.46
0.02 0.31 1.07
R.L. Hubbard et al. / Journal of Substance Abuse Treatment 25 (2003) 125–134
131
Table 4 Reduction in weekly or daily drug use from year prior to intake to followups, 1-year, 5-year and combined improvement rates
Heroin use Cocaine use Marijuana use Problem alcohol use
Outpatient methadone treatment
Long-term residential treatment
Outpatient drug free
Short term inpatient
1 year 5-year Both followup followup followups % % % n
1 year 5-year Both followup followup followups % % % n
1 year 5-year Both followup followup followups % % % n
1 year 5-year Both followup followup followups % % %
75.2
68.6
54.8
347 95.4
62.8
62.8*
43 n/a
n/a
n/a
19 n/a
n/a
n/a
63.9
69.8
50.3
169 81.5
69.5
62.3
151 79.6
80.7
68.2
88 72.3
73.9
58.5
58.1
71.0
48.4
62 77.1
65.6
54.1
61 68.8
54.7
46.9
64 69.7
71.1
61.8
63.6
83.3
57.6
66 79.8
77.5
66.3*
89 68.5
84.9
61.6
73 63.7
71.7
51.3
Note: Paired t-tests were conducted for differences between improvement rates in substance use at the 1-year vs. the 5-year follow-ups. ‘‘Improvement rates’’ are the percentages of patients who reduced their substance use from weekly or daily use in the year prior to treatment. * p < .05. ** p < .01.
For the ODF sample at 5 years, the percentages of substance use and other outcome behaviors were compared for subgroups of patients who remained in DATOS treatment more than 3 months (long-term) and for those staying at least 1 week but less than 3 months (short-term). Due to the longer treatment terms and the relatively smaller sample for the 5-year followup, treatment duration cut-points were 1 month for STI; 6 months for LTR; and 1 year for OMT. These treatment duration cut-points were chosen in accord with findings regarding retention times found to be associated with positive change (ODF and OMT) and in accord with programmatic retention objectives (STI) as well as program objectives in conjunction with study findings (LTR). (See Hubbard et al., 1997; Simpson, Joe, & Brown, (1997); Simpson, Joe, Broome, Hiller, Knight, & RowanSzal (1997)). There were few significant differences at the 5-year followup in levels of substance use between patients who had remained in the original DATOS treatment for shorter vs. longer durations. The exception was the LTR modality where patients who had been in treatment more than 6 months were significantly less likely to use cocaine or to be problem alcohol users (as defined above); less likely to report suicidal ideation and more likely to be employed than those who had had DATOS treatment durations less than 6 months ( p < .05). To further examine evidence of the association between initial treatment duration and outcomes in both 1-year and 5-year followup time periods, and to control for demographic and treatment background variables, a multivariate analysis for repeated measures was conducted, replicating the logistic regression analysis model used in previous DATOS (Hubbard et al., 1997) and TOPS research (Hubbard et al., 1989, pp. 37 –40). Dichotomous variables were created for 10 independent predictors including those shown in Table 1. These predictors were chosen because of known associations with key outcomes, minimum covariation, and limited
conceptual overlap with other predictors in the model. In the analyses for each of the outcomes, the level of behavior or substance use prior to treatment was also included as a predictor. The association of these predictors with outcomes (data not shown) differed across modalities and is not the focus here. In general, the type of substance use or behavior in the preadmission year was the strongest predictor of that behavior at the successive followup periods. However, most pertinent here is the association of treatment duration with outcome after these predictors were taken into account by statistical adjustment. In this regard, the original treatment duration categories were retained for both the LTR and ODF modalities as short-term (more than 1 week, but less than 3 months) and long-term (more than 3 months). However, for the OMT modality, there were too few cases in the original short-term group (n = 29) for a robust model using a 3-month comparison group so the short-term group was recategorized as 1 week to 6 months. We also created two additional treatment duration groups (3 to 6 months; and 6 to 12 months) for the OMT, RES, and ODF modalities because these durations have clinical significance and because there were sufficient numbers of patients in each category to support analysis. Because of the short duration of STI treatment, and previous research that clearly shows a lack of a relationship between treatment duration in STI and longer-term outcomes (Hubbard et al., 1997), we did not perform multivariate analysis for the STI modality. In the course of these analyses we also took the opportunity to consider the potential effects of ‘additional treatment’ (ie. treatment following the original DATOS episode) on outcomes. In these analyses, any exposure to treatment during the followup period was also included as a predictor. The inclusion of an additional treatment predictor in conjunction with the pretreatment predictors was chosen to provide a more rigorous test of the effect of time in DATOS treatment, and as an important step prior to specification of more complex structural equation and multilevel causal
132
R.L. Hubbard et al. / Journal of Substance Abuse Treatment 25 (2003) 125–134
modeling. This conservative analytic strategy enables a clearer statement of the likelihood that various lengths of original DATOS treatment were related to important clinical reductions in the odds of substance use and other outcome behaviors in the followup year. The odds ratios reported in Table 5 are calculations derived from SASR PROC GENMOD (Johnston, 1999) using the REPEATED option with the dependent variable (outcomes) repeated across both followup time periods for each patient. The base estimate of 1.00 is used for patients in treatment 1 to 13 weeks, except for OMT patients for which the base was treatment less than 6 months. Odds ratios less than 1.00 indicate decreased odds and greater than 1.00 indicate increased odds. The multivariate logistic regression analyses (Table 5) replicate the findings from earlier research and extend the time frames to which they apply. Compared with short-term OMT patients, OMT patients with 6 to 12 months of DATOS treatment had an odds ratio of 0.55 ( p < .05); and OMT patients with more than a year of treatment had an odds ratio of .29 ( p < .001) for weekly or more frequent heroin use at the 5-year followup point. This is consistent with the odds ratio that was reported in the TOPS analysis (odds = 0.23, p < .01) comparing to the 3-month duration to less than 3 months (Hubbard et al., 1989), and the previous DATOS research comparing the longer term patients still in treatment at the time of the 1-year followup to the less than 3-month patients (odds = 0.24, p < .01; Hubbard et al., 1997). These results suggest that remaining in methadone treatment is associated with reduction of heroin use for most patients, even after controlling for a variety of other explanatory factors. Within the same 5-year OMT followup sample, a small but significant association was shown suggesting lower cocaine use at followup among patients who stayed in treatment at least 6 months (odds = .55, p < .05), but no association between time spent in the index treatment episode and cocaine use was found in the earlier analysis
of the 1-year followup sample. There was no evidence of a relationship between original DATOS treatment duration and weekly or more frequent marijuana use among the 5year followup sample. However, among the OMT patients in the larger 1-year followup sample, the odds of weekly or more frequent marijuana use for longer term patients was 0.49 ( p < .05) compared to patients in treatment for the shortest period. Finally, no relationship was seen between treatment duration and followup use of alcohol. In fact, there was some evidence that the odds of problem alcohol use increased among OMT patients whose DATOS treatment had lasted 3 to 6 months, compared to those whose original treatment had been less than 3 months (odds = 2.73, p < .05). Among LTR patients, there was an association between original treatment duration and reduced cocaine use during both followup periods. Specifically, as compared with durations of 1 week to 3 months, treatment durations of 3 to 6 months and durations of more than 1 year showed odds of 0.30 and 0.17 respectively (both p < .001) of using cocaine at least weekly. Again, these findings replicate previous DATOS analyses on the 1-year followup sample (Hubbard et al., 1997). In comparison with short term patients in the LTR sample, the odds of reduced marijuana were 0.76 for the 3– 6 months group; 0.35 for the 6 – 12 months group and 0.52 for those in DATOS treatment more than 12 months. Only the odds for the 6 –12 months patients were significantly different from the short-term patients ( p < .01). Results showed significantly decreased odds of problem alcohol use after treatment for patients who remained in LTR treatment at least 6 months (odds = 0.18, p < .001 for 6 to 12 months; odds = 0.24, p < .01 more than 1 year). Thus, as seen with OMT patients, greater initial treatment duration in LTR was also associated with reductions in substance use for both the 1-year and 5-year followup periods. The multivariate analysis for the post-treatment followup repeated measures provided little evidence of an effect of
Table 5 Odds ratios of likelihood of substance use and behaviors in the followup years by modality longer duration of treatment relative to shorter duration
Heroin use Cocaine use Marijuana use Problem Alcohol use Suicidal thoughts or attempts Predatory illegal activity Sexual risk behavior Less than full-time work Health limitations
Outpatient methadone
Long-term residential
6-12 months
More than 1 year
3-6 months
6-12 months
More than 1 year
3-6 months
Outpatient drug free More than 6 months
0.55* 0.96 1.34 0.82 0.90 1.51 1.41 0.68 0.49
0.29*** 0.55* 0.88 0.69 0.91 1.00 1.05 0.76 0.75
—c 0.95 0.76 0.64 1.56 0.92 1.11 0.84 1.18
—c 0.30*** 0.35** 0.18*** 0.76 0.31** 0.89 0.45** 0.81
—c .17*** 0.52 0.24** 0.29 0.32* 0.89 0.32** 1.69
—c 0.70 1.50 1.22 0.76 1.04 1.14 1.17 1.05
—c 0.58 0.56 0.65 0.39* 0.58 1.34 0.80 2.12**
Note: Odds ratios were derived from generalized estimating equations (GEE) models using Proc GENMOD, SAS 8.01. * p < .05. ** p < .01. *** p < .001.
R.L. Hubbard et al. / Journal of Substance Abuse Treatment 25 (2003) 125–134
length of stay in ODF on reductions in drug or problem alcohol use. Previous models for the 1-year followup showed a pattern of significantly lower odds for the more than 6 months duration treatment group for all types of substance use considered. For the more than 6 months groups, the odds were 0.29 ( p < .01) for cocaine use, 0.29 ( p < .01) for marijuana use, 0.43 ( p < .01) for problem alcohol use. Multivariate analyses also examined the effect of time in index treatment on outcome behaviors other than substance use, as previously discussed. The results did not indicate a positive association between original treatment duration in OMT and any of these outcomes. In fact, the odds of illegal activities increased for patients leaving OMT after 3 to 6 months ( p < .05). Among LTR patients, an association with time in original DATOS treatment was found for two of the five outcome behaviors in the multivariate analysis (Table 5). The results of a multivariate analysis, including criminal justice involvement and pretreatment criminal involvement for patients staying more than 6 to 12 months show the odds of illegal activity in the year after treatment were reduced by ratios of 0.31 ( p < .01) and 0.32 for stays over 12 months ( p < .05). Multivariate analysis confirmed that stays of more than 6 months in LTR decreased the odds of less than full- time employment by about .45 ( p < .01) and 0.32 ( p < .01). For the 1-year followup, 6 months or more of ODF treatment was significantly ( p < .05) related to half the odds of having less than full-time employment, but this finding was not replicated for the combined 1- and 5-year followups. A decrease in the odds of illegal activity of 0.58 did not reach the .05 level of statistical significance in the multivariate analyses. The multivariate analyses also showed that stays of 6 months or more were associated with a decrease in the odds of suicidal ideation equivalent to a factor of .39 ( p < .05). A finding not present in previous research was the increased odds (odds = 2.12, p < .01) of health limitations for ODF patients staying in treatment more than 6 months.
4. Discussion The analysis of the 5-year followup outcomes showed reductions of 50% in prevalence of weekly or more frequent cocaine use from the preadmission year among OMT, LTR, and ODF patients. Multivariate analyses indicated that the odds of weekly or more use were significantly lower ( p < .001) for patients originally treated for 6 months or more in LTR. Patients reporting more than 1 year of OMT in their original DATOS episode had significantly lower odds ( p < .01) of weekly or daily heroin use than patients who left OMT within the first year. Other significant improvements at 5 years included reductions of 50% in illegal activity and 10% increases in full-time employment among OMT, LTR, and ODF patients. For LTR patients the decreased odds of illegal activity and
133
less than full time employment at 5-year followup were related ( p < .01) to original DATOS treatment stays of 6 months or longer. The results replicate findings for outcomes in the first year after treatment. These initial results from the overall analysis for each modality are again consistent with the persisting and positive effect of treatment on drug use and other key outcomes, especially in the LTR modality, where the most comprehensive and intensive services appear to be provided (Etheridge et al., 1997). Most importantly, evidence of the effectiveness of treatment for cocaine use in LTR and heroin use in OMT is demonstrated during both followups controlling for pre-treatment levels of use, initial source of referral and more recent treatment participation. There may be concerns about the proportions of patients who were interviewed at the 5-year followup and the exclusion of small samples in three cities. We do not think that these factors change the basic nature and pattern of the major results. It should be noted that the number of early dropouts differed among modalities. The higher rate of early dropout in ODF suggests that these results may be more applicable to patients with longer retention than to the total ODF entry cohort. No overall or systematic differences across modalities were found between respondents and patients who were not interviewed. Unless very complex patterns of interaction exist between attrition and outcomes, the key findings on stability of outcomes between the first and fifth year after treatment and the relationship of initial treatment duration with 5-year outcomes would likely be replicated regardless of the proportion of patients interviewed or the minimal bias found between respondents and patients who were not interviewed. It should be noted, however, that the study design virtually precluded consideration of those patients who dropped out of treatment within a week of program entry. The DATOS results from the 5-year post-treatment followup presented here add to the accumulating evidence suggesting the effectiveness of substance abuse treatment, particularly given the replication of results in the context of substantial change in the treatment system, new patterns of drug abuse and other environmental changes. However, we see again that the full potential of treatment is not achieved for patients who do not receive sufficient exposure to treatment. In general, there was little evidence of an association between treatment and the behavioral outcomes other than substance use. The only exception was in the LTR modality, possibly because this treatment is generally more intensive and aims to change a broader range of client behaviors. This result may also be associated with reductions in comprehensive services seen in OMT and ODF modalities (Etheridge et al., 1997) and the limited time spent in STI treatment. As described in other papers in this issue, there are many issues that need to be examined to determine the factors that contribute to the long-term recovery process and stable productive functioning. In that regard, the results available
134
R.L. Hubbard et al. / Journal of Substance Abuse Treatment 25 (2003) 125–134
from large-scale national studies like DATOS provide an opportunity to understand and disentangle the complex effects of environments, patient characteristics, and treatment on outcomes. These DATOS results make clear that many positive changes in drug use and behaviors seen at 1 year post-treatment are also in evidence at 5 years posttreatment for a continuing sample. While there is need to assess the possible additional influences on long term treatment outcome (e.g., additional treatment received), the persistent association between the duration of the index treatment episode with positive outcomes makes treatment retention a prudent clinical concern and emphasis. The other papers presented in this issue both expand on these core findings and focus analysis on several aspects of long-term recovery. Building on these findings, the research can move to the further investigation of those factors that contribute to long-term recovery for diverse patient subpopulations experiencing a range of post-treatment environments.
Acknowledgments This work was supported by National Institute on Drug Abuse (NIDA) Grant U01-DA10377-3 as part of the Cooperative Agreement on the DATOS. The project includes a Coordinating DATOS Research Center (Robert L. Hubbard, Principal Investigator at NDRI) and two Collaborating DATOS Research Centers (M. Douglas Anglin, Principal Investigator at UCLA, and D. Dwayne Simpson, Principal Investigator at TCU) to conduct treatment evaluation studies in association with NIDA (Bennett W. Fletcher, Principal Investigator at NIDA). Five-year follow-data collection was conducted by the National Opinion Research Center. Special thanks to Barry S. Brown, Ph.D. for his thoughtful reviews and suggestions on this manuscript. The interpretations and conclusions contained in this paper, however, do not necessarily represent the position of other DATOS Research Centers, NIDA, or the Department of Health and Human Services.
References Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society, B., 57, 289 – 300. Broome, K. M., Joe, G. W., Flynn, P. M., & Simpson, D. D. (2002).
Longitudinal outcome pattern for methadone patients. Addictions, this issue. Brown, B. S., & Beschner, G. M. (Eds.). (1993). Handbook on Risk of AIDS: Injection Drug Users and Sexual Partners. Westport, CT: Greenwood Press. Etheridge, R. M., Hubbard, R. L., Anderson, J., Craddock, S. G., & Flynn, P. M. (1997). Treatment structure and program services in the Drug Abuse Treatment Outcome Study (DATOS). Psychology of Addictive Behaviors, 11, 244 – 260. Flynn, P. M., Craddock, S. G., Hubbard, R. L., Anderson, J., & Etheridge, R. M. (1997). Methodological overview and research design for DATOS. Psychology of Addictive Behaviors, 11 (4), 230 – 243. Flynn, P. M., Craddock, S. G., & Dunteman, G. H. (1995). Cocaine Treatment Outcome Study: Overview and findings. Research Triangle Park, NC: Research Triangle Institute. Hubbard, R. L., Craddock, S. G., Flynn, P. M., Anderson, J., & Etheridge, R. M. (1997). Overview of 1-year follow-up outcomes in the Drug Abuse Treatment Outcome Study (DATOS). Psychology of Addictive Behaviors, 11 (4), 261 – 278. Hubbard, R. L., Marsden, M. E., Cavanaugh, E., Rachal, J. V., & Ginzburg, H. M. (1988). Role of drug abuse treatment in limiting the spread of AIDS. Reviews of Infectious Diseases, 10 (2), 377 – 384. Hubbard, R. L., Marsden, M. E., Rachal, J. V., Harwood, H. J., Cavanaugh, E. R., & Ginzburg, H. M. (1989). Drug abuse treatment: A natural study of effectiveness. Chapel Hill: The University of North Carolina Press. Hubbard, R. L., Marsden, M. E., & Allison, M. (1984). Reliability and validity of TOPS Data. RTI/1901/01 – 155. Research Triangle Park, NC: Research Triangle Institute. Johnston, G. (1999). Repeated measures analysis with discrete data using the SASR system. Cary, NC: SAS Institute SAS OnlineDoc, Version Eight. Liang, K.-Y., & Zeger, S. L. (1986). Longitudinal data analysis using generalized linear models. Biometrika, 73, 13 – 22. Sells, S. B., & Simpson, D. D. (1980). The case for drug abuse treatment effectiveness, based on the DARP research program. British Journal of Addiction, 75, 117 – 131. Simpson, D. D. (1981). Treatment for drug abuse: Follow-up outcomes and length of time spent. Archives of General Psychiatry, 38 (8), 875 – 880. Simpson, D. D., Joe, G. W., & Broome, K. M. (2002). A national 5-year follow-up of treatment outcomes for cocaine dependence. Archives of General Psychiatry, 59, 538 – 544. Simpson, D. D., Joe, G. W., Broome, K. M., Hiller, M. L., Knight, K., & Rowan-Szal, G. A. (1997). Program diversity and length of stay in treatment in DATOS. Psychology of Addictive Behaviors, 11 (4), 279 – 293. Simpson, D. D., Joe, G. W., & Brown, B. S. (1997). Length of stay in treatment and follow-up outcomes in DATOS. Psychology of Addictive Behaviors, 11 (4), 294 – 307. Simpson, D. D., & Sells, S. B. (1982). Effectiveness of treatment for drug abuse: An overview of the DARP research program. Advances in Alcohol and Substance Abuse, 2 (1), 7 – 29. Simpson, D. D., & Sells, S. B. (Eds.). (1990). Opioid addiction and treatment: A 12-year follow-up. Malabar, FL: Krieger Publishing Co. Stewart, A. L., & Ware, J. E. (1992). Measuring functioning and wellbeing: The Medical Outcomes Study approach. Durham, NC: Duke University Press.
