A probability-based approach for predicting HIV infection in a low prevalent population of injection drug users

Abstract

This article proposes a method for estimating HIV risk in low-HIV-prevalent populations. Allard's risk probability model was used to compute individual risk scores. Based on a sample of 3854 injection drug users (IDUs) who were confidentially tested for HIV at five methadone treatment clinics in Los Angeles County, the following self-reported risk behaviors were used to derive an individual IDU risk score: (i) frequency of injection, (ii) frequency of using uncleaned needles, (iii) number of people sharing a needle, (iv) frequency of needle sharing, and (v) type of needle sharing practice. The overall HIV prevalence for the IDU sample was 2%. The risk score was strongly associated with HIV seropositivity (chi-square = 16.1, p < 0.0001), but only one of the individual IDU risk behaviors (needle cleaning) was significantly associated with HIV seropositivity (chi-square = 10.9, P < 0.001). In addition, the risk score was strongly associated with HIV serostatus for both males and females. For females, however, none of the individual IDU risk behaviors were associated with HIV serostatus. Our findings indicate indicate that when predicting HIV infection in a low-prevalence population, the probability-based risk score makes a statistically significant contribution over individual IDU risk behaviors.