OBJECTIVE: To identify the factors predicting outpatient drug costs and quantify their magnitude. METHODS: Data were collected for the capitated patient populations enrolled at the sixty-one ambulatory internal medicine, pediatric medicine, and family practice clinics located within the 21 outpatient sites of the Henry Ford Health System. Medicare risk patients were excluded since in these programs often have limited prescription benefits. The data included drug costs, diagnoses, and demographic information. Many factors including age, gender, HIV, diabetes, transplant, congestive heart failure, coronary artery disease, and percent Medicaid were identified as possible contributors towards these per member per month (PMPM) drug costs. The data for each of these factors were input into a multivariate linear regression model as the independent variables, while prescription PMPM costs were designated as the dependent variable. The regression model was created using the SPSS? software program. RESULTS: Five factors found to be statistically significant predictors of PMPM drug cost were the percentage of patients: between ages 40 to 64 (P<0.000); with diabetes (P<0.044); receiving HIV drug treatment (P<0.034); who were female (P<0.025); with a transplant diagnosis (P<0.000). The best-fit model was $PMPM = [(41.88 ? % 40-64 years) + (71.77 ? % of diabetic patients) + (775.6 ? % on HIV drugs) + (2299 ? % of transplant patients) + (11.30 ? % of female patients)] - .0233 (adjusted r2 > 0.957). Age of 65 or greater was not a significant predicator because Medicare enrollees were not included. CONCLUSION: Age of 40 to 64, diabetes, HIV, transplant, and female gender are the five factors strongly predictive of drug costs in this capitated population. This model can be used to risk adjust and predict the pharmaceutical budgets of the various sites within an integrated health system.