METHODS We sought to describe the trends in survival from 1983 to 1989 among persons with the acquired immunodeficiency syndrome (AIDS) and to examine the relative effects on the natural history of AIDS of zidovudine use and demographic and clinical characteristics. This longitudinal, observational, population-based study used data from the Maryland Human Immunodeficiency Virus Information System, a data base that links information from the Maryland AIDS Registry with data on public and private health care claims, vital statistics, and hospital, long-term care, and ambulatory care records. RESULTS The median survival after diagnosis among persons with AIDS (n = 1028) was 140 days longer for those given their diagnoses between 1987 and 1989 than for those given their diagnoses between 1983 and 1985 (450 vs 310 days). Among the 714 persons in whom AIDS was diagnosed after April 1987 (when zidovudine became available), two-year survival was greater among men than women (P less than 0.03), among persons less than 45 years old than among older persons (P less than 0.001), among non-Hispanic whites than among minorities (P less than 0.001), and among persons whose category of human immunodeficiency virus transmission was homosexual contact than among those with heterosexual, transfusion-related, or less common modes of transmission (P less than 0.02). In all the analyses the groups with the longer survival were more likely to have received zidovudine. The median survival among those who received zidovudine was 770 days, as compared with 190 days among those who never received the drug (P less than 0.001). By proportional-hazards analysis, zidovudine therapy was the factor most strongly associated with improved survival. CONCLUSIONS For Maryland residents with AIDS there has been an improvement in survival since 1987. Zidovudine therapy and perhaps other aspects of care associated with it have contributed substantially to the improved survival.