OBJECTIVES: Comorbidity scores are frequently used for controlling confounding in observational studies using administrative claims. Several comorbidity measures have been developed and evaluated for predicting a variety of outcomes.  However, scores are often used with outcomes other than those included in validation studies.  Few publications have compared the performance of different scores in predicting different types of outcomes. The objective of the present study was to compare the Charlson Comorbidity Index (CCI), a frequently used diagnosis-based comorbidity score, and Rx Risk, a prescription-based comorbidity score, as predictors of three outpatient outcomes. METHODS: A retrospective analysis was conducted using Mississippi Medicaid medical and pharmacy claims data for the period January 2010 to December 2011. Inclusion criteria were continuous enrolment during the observation period, not dual-eligible, and having both medical and pharmacy claims in both calendar years. CCI scores and Rx Risk Scores were calculated using 2010 claims. Scores were evaluated as predictors of outpatient visits, total outpatient costs and total pharmacy costs in 2011. Costs were log transformed. A base general linear model with age, gender, race and ethnicity was developed. Predictive ability of each comorbidity score was measured as the change in R when the score was added to the base model. RESULTS: Rs for the base model were visits - 0.07, outpatient costs - 0.01, and pharmacy costs - 0.03.  CCI and Rx Risk improved prediction for visits and pharmacy costs (CCI Rs; 0.10, 0.05;  Rx Risk Rs; 0.13, 0.07).  CONCLUSIONS: Although CCI is often used for outpatient outcomes, Rx Risk provides a better measure of comorbidity when the dependent variables are outpatient utilization or costs. The CCI was developed for predicting mortality during hospitalization.  These results indicate that comorbidity scores developed for predicting outpatient outcomes would be better for controlling for comorbidity in outpatient based studies.