Of 247 treatment-seeking NEP clients invited to participate, 245 (99%) consented and completed the baseline study visit. Of these 245 participants, 76% were African-American, 69% were male and 91% were unemployed. The median age was 42 (interquartile range: [IQR]: 37–48), and 44% had not finished high school. HIV prevalence was 19%, similar to previous studies of the Baltimore NEP population (Riley et al., 2002
). Almost all were already aware of their HIV-serostatus. A comparison between sociodemographic characteristics of the study sample and the overall population of Baltimore NEP attenders during the study period revealed no significant differences (results not shown). There were no seasonal differences between control and intervention conditions, and no observed differences in participant characteristics among the 10 NEP sites.
By design, 128 (52%) were randomized to the intervention and 117 (48%) to the control condition. The number of subjects in each group was not equal since randomization was conducted by NEP site, not by individual, and was based on the number of drug treatment slots available in locations nearest to a NEP site. There were no baseline differences between the intervention and control groups with respect to any sociodemographic or behavioral characteristics, or self-reported barriers to accessing drug treatment (); however, the median age at first injection was slightly older among controls compared to clients randomized to the intervention arm (23.9 versus 21.5 years, respectively; p = 0.019; ). The majority (87%) stated that they required transportation to attend drug treatment, but this proportion did not differ between intervention and control groups.
Characteristics of IDUs enrolled in the treatment retention intervention study (2002–2004)
Overall, 34% entered treatment within 7 days of the referral from NEP (intervention: 40% versus control: 26%, p = 0.03). Since the follow-up period was very short, there were no dropouts or post-randomization exclusions.
Among the 128 subjects randomized to the intervention, 104 (81.3%) utilized some degree of case management services. The median duration of case management time received within the 7-day period was 25 min (IQR: 15–80), among a total of 201 contacts (median number of contacts per person: 2; IQR: 1–3). During these contacts, case managers provided primarily transportation assistance (23%) and counseling (23%) and assistance with social services, medical care, housing and employment.
In univariate analyses (), factors associated with a greater odds of entering treatment were having been randomized to receive case management (Odds ratio [OR]: 1.84; 95% confidence interval [CI]: 1.07–3.16). Having two or more contacts with a case manager prior to the intake visit at the drug treatment program (OR: 2.47; 95% CI: 1.33–4.59), having received more time with a case manager or being driven to treatment by a case manager (OR: 4.94, 95% CI: 2.19–11.4) were significantly associated with entering treatment. Participants who were older or had access to a car were marginally more likely to enter treatment.
Correlates of entry into drug treatment: univariate logistic regression
The mean distance participants were required to travel to the treatment program they were referred was 5 miles (±5 miles). Surprisingly, subjects who entered treatment travelled slightly farther than those who did not (mean # of miles: 6.2 versus 4.4; OR = 1.09 per mile, p = 0.04). Similarly, subjects who were required to travel farther to the drug treatment program than they had intended were significantly more likely to enter treatment. Treatment entry did not differ according to gender, race, HIV serostatus, marital status, education, employment, living arrangement, alcohol dependence, prior treatment experience, injection history, overdose, age at first injection, barriers to drug treatment, depressive symptoms or treatment readiness score.
In a multivariate “intention to treat” model (i.e., not taking into account the amount of case management clients actually received), those randomized to case management were 87% more likely to enter treatment within 7 days, even after adjustment for farther travel, access to a car, age and clustering by NEP site (AOR: 1.87, 95% CI: 0.91–3.86); however, this association was marginally significant (p = 0.06) ().
Independent predictors of drug treatment entry within 7 days: “Intention to Treat” and “As treated” models
We also conducted an ‘as treated’ analysis taking into account the ‘dose’ of case management each client actually received (). In the final multivariate ‘as treated’ model, after adjusting for age, randomization assignment, and clustering by NEP site, having received more case management time was independently predictive of treatment entry (). In particular, participants who received 30 min or more of case management within 7 days of the baseline visit were 33% more likely to enter treatment. Additionally, persons with access to a car were almost three times more likely to enter drug treatment. Even after adjusting for these factors, persons who were required to travel farther than intended remained more likely to enter treatment (adjusted odds ratio [AOR]: 2.89; 95% CI: 1.59–5.26). In an effort to explain this association, we examined an interaction term between access to a car and traveling farther than intended. This interaction term was marginally significant in the final model (AOR: 0.17, p = 0.06), suggesting that those living farther away without access to a car were 83% less likely to enter treatment (results not shown). A potential interaction between traveling farther than intended and being driven to treatment, however, was not significant (p = 0.77). The inclusion of variables related to income, education, employment, treatment clinic site or treatment modality (i.e., methadone or LAAM) did not improve the overall model fit.
In an effort to identify the mechanism through which case management operated to facilitate treatment entry, we also offered the variable that indicated whether or not case managers had driven clients to the treatment program into both the ‘intention to treat’ and ‘as treated’ models in . Inclusion of this variable did not appreciably change parameter estimates; however, in the ‘intention to treat’ model, being randomized to the case management intervention became insignificant (AOR: 1.13; 95% CI: 0.59–2.16, p = 0.72) and the variable indicating whether or not case managers drove clients to treatment was highly significant (AOR: 4.99; 95% CI: 1.98–12.56). Similarly, in the “as treated” model, case management dose became insignificant (AOR: 1.03, p = 0.58) and the variable indicating whether or not case managers drove clients to treatment was significant (AOR: 3.89, p = 0.03). This suggests that the ‘active ingredient’ of case management was the provision of transportation to the treatment program. Adjustment for age at first injection did not appreciably alter any of the odds ratios presented.