We developed a scheme using routinely available data to estimate the risk of human immunodeficiency virus (HIV) dementia in HIV-infected persons over time. We performed a longitudinal review of medical records from more than 100 medical facilities in 11 U.S. cities. A total of 19,462 HIV-infected persons without history of dementia enrolled in a multi-institution survey. The main outcome measure was the development of HIV dementia (1987 case definition) during the median follow-up period of 17 months (range, 1 to 72 months). Of 19,462 HIV-infected persons, HIV dementia developed in 880 persons (4.5%; 2.6% per person-year). The strongest predictors of HIV dementia were CD4+ T-lymphocyte count, anemia, and AIDS-defining infections or cancer. The 2-year probability of HIV dementia was highest for persons who had a CD4+ T-lymphocyte count of fewer than 100 cells/μL and an AIDS- defining illness or anemia or both (18.6 to 24.9%). Intermediate risk was observed in persons with CD4+ T-lymphocyte count of 100 to 199 cells/μL and an AIDS-defining illness or anemia or both or in persons with a CD4+ T- lymphocyte count of fewer than 100 cells/μL but without another risk factor (2-year probability, 10.4 to 15.2%). The 2-year probability that HIV dementia would develop was lowest (1.0%) for persons with CD4+ T-lymphocyte count of more than 200 cells/μL and no other risk factors. Risk stratification using routine clinical information provides information that may prove useful in patient care decisions.