The correlates and predictive validity of HIV risk groups among drug users in a community-based sample: methodological findings from a multi-site cluster analysis (2001)

Outreach and intervention with out-of-treatment drug users in their natural communities has been a major part of our national HIV-prevention strategy for over a decade. Intervention design and evaluation is complicated because this population has heterogeneous patterns of HIV risk behaviors. The objectives of this paper are to: (a) empirically identify the major HIV risk groups; (b) examine how these risk groups are related to demographics, interactions with others, risk behaviors, and community (site); and (c) evaluate the predictive validity of these risk groups in terms of future risk behaviors. Exploratory cluster analysis of a sample of 4445 out-of-treatment drug users from the national data set identified eight main risk subgroups that could explain over 99% of the variance in the 20 baseline indices of HIV risk. We labeled these risk groups: Primary Crack Users (29.2%), Cocaine and Sexual Risk (12.8%), High Poly Risk Type 2 (0.3%), Poly Drug and Sex Risk (10.9%), Primary Needle Users (24.1%), High Poly Risk Type 1 (1.4%), High Frequency Needle Users (19.8%), and High Risk Needle Users (1.6%). Risk group membership was highly related to HIV characteristics (testing, sero-status), demographics (gender, race, age, education), status (marital, housing, employment, and criminal justice), prior target populations (needle users, crack users, pattern of sexual partners), and geography (site). Risk group membership explained 63% of the joint distribution of the original 20 HIV risk behaviors 6 months later (ranging from 0.03 to 37.2% of the variance individual indices). These analyses were replicated with both another 25% sample from the national data set and an independent sample collected from a new site. These findings suggest HIV interventions could probably be more effective if they targeted specific subgroups and that evaluations would be more sensitive if they consider community and sub-populations when evaluating these interventions. (Author' s abstract)