A Predictive Model for Male IDUs at Risk of HIV-1 Infection

Introduction: According to statistics from the Taiwan Centers for Disease Control, at the end of 2009 just over one-third (6,330 of 18,378, or 34.4%) of HIV-1-positive Taiwanese were injecting drug users (IDUs), representing the highest at-risk group. Our goal is to establish a predictive model for HIV-1 infection risk as a tool for increasing awareness among IDUs. Methods: In this case-control study, questionnaire responses from 182 male IDUs in Taiwanese prisons were collected in 2004 and 2005. A logistic regression was used to establish a risk prediction model. After summing the odds ratios (ORs) of all model variables, a receiver operating characteristic (ROC) curve was established to obtain a risk threshold value for predicting HIV-1 infection risk. A second questionnaire was administered to 671 Taiwanese male IDUs in 2007 and 2008 to test the accuracy of the predictive model. Results: We identified four predominant factors in the risk prediction model: education level, belief in a correlation between sexually transmitted diseases and HIV/AIDS infection, needle sharing, and drug solution sharing. According to ROC curve results, a total OR of 7.44 represents the most appropriate risk threshold value (sensitivity 95.6%, specificity 52.2%). For our predictive model, sensitivity was 72.2% and specificity 60.5%. Conclusions: The results suggest that the proposed risk predictive model may be used to support increased self-awareness in the form of a concise and economical risk evaluation questionnaire.