A retrospective cohort study was performed to assess the relative risk of upper gastrointestinal (UGI) tract bleeding from two formulations of potassium chloride. Relevant information was obtained from 1980 through 1984 Medicaid billing data from the states of Michigan, Minnesota, Florida, and Ohio. After patients with a history of UGI tract bleeding prior to their first prescription for either of the two potassium chloride preparations under study were excluded, data were analyzed for 28,790 patients (143,512 patient-months) dispensed a microencapsulated formulation exclusively and 76,118 patients (560,341 patient-months) dispensed a wax-matrix formulation exclusively. The risk of UGI tract bleeding within 30 days after each prescription for the drug of interest was examined. After sampling from the undiseased study subjects and adjusting for multiple potential confounding variables using logistic regression, an odds ratio (95% confidence interval) of 0.67 (0.52 to 0.85) was observed.