OBJECTIVES: Comorbidity often is an important confounder in health outcomes studies. Although a number of comorbidity measurements are available, little is known about their performance in studies in which health expenditures are the primary outcome. This study compares the performance of the Elixhauser Index and Charlson Index as well as remedial measurements of annual prescription use and physician visits in predicting health expenditures. METHODS: The study population included 19,793 members of a managed care organization who received a new prescription for an antihypertensive medication in 2002. Elixhauser and Charlson scores were derived from hospital claims occurring in the year preceding this encounter. The sum of physician's visits and prescription fills were calculated similarly. The primary prediction outcome was the log transformed sum of prescription, physician, and hospital expenditures in the year following the encounter. Performance was determined by fitting linear regression models and examining R2 statistics. RESULTS: A greater proportion of regression model variance was explained by the Charlson Index (R2=0.0989) than the Elixhauser Index (R2=0.0665). The remedial measurements of number of prescriptions filled (R2=0.1413) and number of physician's visits (R2=0.1295) per year appears better at predicting future health expenditures than the comorbidity indices examined. The addition of age and gender to the model increases the explained proportion of variance for each measurement. The addition of number of prescriptions, number of physician's visits, age, and gender maximizes the explained proportion of variance for the Charlson Index model (R2=0.2526) and Elixhauser Index model (R2=0.2135). CONCLUSION: The Charlson Index appears to outperform the Elixhauser Index as a comorbidity control in health expenditure studies. However, compared to the remedial measurements of number of prescriptions filled and number of physician visits per year these indices appear inferior.