Cross-sectional investigations of homelessness have many potential biases. Data from 2,452 individuals enrolled in a longitudinal cohort study of Baltimore, Maryland, residents recruited in 1988–1989 with a history of injection drug use were analyzed to identify the extent and determinants of homelessness. Proportions having ever experienced homelessness were compared across subgroups of injection drug users (IDUs) who were human immunodeficiency virus (HIV) negative, HIV positive, and HIV seroconverting. Logistic regression identified independent predictors of homelessness. In the cohort, 1,144 (46.7%) participants experienced homelessness during the course of the study. There were differences in prevalence of homelessness by serostatus: 42.4% (n=621) of participants who remained HIV negative were ever homeless, while 50.6% (n=346) of HIV-infected individuals and 58.9% (n=178) of those who seroconverted during the study were ever homeless (P<.001). Participants who consistently denied active injection drug use during follow-up were unlikely to experience homelessness (19%). Independent predictors of homelessness were male sex, HIV seroprevalence, and HIV seroconversion. Following participants over time captures more experiences of homelessness than cross-sectional studies and more accurately identifies risk characteristics. Our data suggest that homelessness is a significant problem among IDUs, especially those with HIV/AIDS. (Authors)