This study used a longitudinal data set of 4,957 patients and 39,455 patient-months of observation, drawn from 17 months of data on patients with symptomatic HIV disease identified from the New York State Medicaid Program. Multivariate regression analysis was used to evaluate the effect of AZT on the use of several medical services. The results show that AZT produced substantial reductions in resource use. Patients who took AZT were hospitalized fewer days per month and, including the cost of the drug, had Medicaid expenditures that were several hundred dollars lower per month. The reduction in total expenditures is largely accounted for by a reduction in inpatient expenditures and an increase in pharmacy expenditures. These effects were temporary, however, as the impact of the drug decreased with each month of use. There were no declines in expenditures and utilization after approximately 7 to 9 months of AZT use.