A prospective design was used to collect data from 132 HIV-positive individuals presenting for mental health care at an inner-city HIV-related mental health clinic located in a large metropolitan area of the Southeastern USA. Data were analyzed to assess associations between study variables and whether participants returned for care following an initial mental health assessment. Hierarchical logistic regression analyses were conducted to determine the extent to which the addition of measures of health beliefs and HIV-related stigma would enhance the predictive capacity of a dropout model that included demographics and health status measures. As health beliefs and HIV-related stigma were added to the model, the predictive capacity increased significantly (R2 = 0.29 to R2 = 0.63, p < 0.05). Those who did not return for care had higher levels of perceived barriers to mental health care, higher levels of HIV-related stigma, were more likely to be of non-white ethnicity and had lower T-cell counts. These findings indicate that mental health providers may need to ensure care systems are responsive to the social and cultural characteristics of clients, and the HIVepidemic itself, in order to effectively engage clients into care.