The sociodemographic characteristics of the study sample are presented in . The characteristics of this sample are very similar to the overall patient population at the participating MMTP clinics with respect to age, race/ethnicity, current treatment duration, and methadone dose (data from clinic records regarding other sociodemographic and health indicator variables were unavailable to study staff for comparison). Several indicators suggest a high prevalence of service needs among the sample: almost half (47%) were unemployed; an average of slightly more than one type stressful life event in the prior six months; an average level of psychological distress that is more than one standard deviation higher than that observed in the general population of men in the U.S. (Derogatis, 1993
); and an HIV prevalence of 14%.
Table 1 Overview of the study sample, for the total sample as well as by those who received or did not receive additional off-site services; descriptive statistics are mean (and standard deviation, SD) for continuous variables and number (and proportion) for (more ...)
Compared to participants who reported not receiving additional, off-site services, recipients of such services had significantly more years of education, reported significantly more types of stressful life events and psychological distress, and were significantly more likely to be HIV-positive.
With respect to illicit drug use among the study sample, the most frequently reported was heroin, followed in decreasing order of prevalence by marijuana, cocaine, and then crack. Overall, about two-thirds of the participants used some type of illicit drugs at the baseline and 6-month follow-up timepoints (66% and 67% respectively), with a lower prevalence at the 12-month follow-up timepoint (62%). provides estimates of the prevalence of illicit drug use separated by type of illicit drug (and any illicit drug use) and timepoint among the entire sample as well as subgroups separated by receipt of additional, off-site services.
Illicit drug use versus time; estimates are proportion (and standard error, SE) of the following groups: the total sample (N = 356), recipients of additional off-site services (N =105), and non-recipients of additional off-site services (N = 251).
presents estimates of the causal effect (i.e., after propensity score matching and covariance adjustment for potential confounders) of receipt of additional, off-site services on the likelihood of abstaining from the indicated drug or any illicit drug use over various ensuing time periods. Beneficial effects of receipt of additional, off-site services was observed for all outcome measures during the first six months, with likelihood of abstaining from cocaine, heroin, and any illicit drug use exceeding the criterion level for significance. During the period encompassing six to twelve months following receipt of additional, off-site services, point estimates of the effects of receiving such services were in the beneficial direction all outcome measures, although only the increase in likelihood of abstaining from any illicit drug use was significant at the 95% confidence level. With respect to likelihood of abstaining from drug use during the entire, subsequent 12 month period, recipients of additional, off-site services were more than two and a half times as likely to abstain from cocaine and heroin use, and more than three and a half times as likely to abstain from any illicit drug use; estimates for abstaining from crack and marijuana use were also in the beneficial direction, but did not at the 95% confidence level.
Table 3 Effects of receiving additional off-site services on likelihood of abstaining from use of indicated drug among male MMTP patients; estimates are adjusted odds ratios (ORs) and 95% confidence intervals (95% CIs) after propensity score matching and covariance (more ...)
An eligibility criterion for the parent study required that potential participants current methadone treatment be at least 3 months. Since the time frame for measuring receipt of additional, off-site services was assessed at screening and covered a 6 month period, we examined the extent to which findings might be affected by receipt of additional, off-site services before initiation of methadone treatment. Only 5 of the participants in this study were on methadone for 6 months or less: 3 were non-recipients and 2 were recipients of additional off-site services. Results and inference remain unchanged if the 5 participants who were on methadone for 6 months or less are withheld from analyses.
Several additional steps were implemented to examine the validity of these findings. First, balance between the recipients and non-recipients of additional, off-site services after propensity score matching indicated that the groups were balanced (i.e., not significantly different) with respect to attributes for variables used for propensity score matching and covariance adjustment, including variables for which significant differences were observed before the propensity scoring matching process (e.g., level of education, SLEC score, etc.). Second, to address the potential problem of non-overlap in propensity scores between recipients and non-recipients of additional, off-site services, analyses were re-run while excluding service users whose propensity score fell outside of the range of propensity scores of non-service users. Third, instead of nearest neighbor matching, we employed caliper matching using widths that range from ±.02 to ±.10. Sensitivity analyses implementing the latter two steps indicate that the findings presented in remained remarkably consistent. In some instances, significant findings in no longer reached significance at the 95% confidence level. However, since estimates of the adjusted ORs in those instances changed no more than 20% (and less than 10% for the vast majority of estimates)—and in no case did the point estimate change from the observed beneficial direction (i.e., adjusted OR > 1) to a detrimental direction (i.e., adjusted OR < 1)—we posit that the essential relationships, and thus our conclusions, remain unchanged and that the lack of significance observed is due to the decrease in power stemming from the smaller sample sizes in models that exclude participants with non-overlapping propensity scores or no matches falling within the caliper width.