OBJECTIVES: To validate an alternative weighting algorithm of the Charlson Comorbidity Index (CCI) for the prediction of health care expenditures and utilization in previously hospitalized patients. METHODS: Data from the Medical Expenditure Panel Survey (MEPS) Panel 12 (2007-2008) were retrieved for this retrospective cohort study. Two CCI scores were calculated for patients who were hospitalized in 2007: one based on the original weights (Charlson-CCI) and the other based on the weights updated by Quan et al. (Quan-CCI) [both were developed to predict mortality]. Adjusted R2 from linear regression models were used to estimate log-transformed healthcare expenditures (COST) in 2008. Odds ratios and c statistics from logistic regression models were used to compare the predictive power of the risk of hospitalizations  (≥ 1 admission), risk of emergency department visits (≥ 1 visit), and high expenditures (≥ 90th percentile of COST) in 2008. RESULTS:  Seven hundred patients who had been previously hospitalized were included in the study. The mean (SD) age was 52.5 (15.3) years, and 65% were female. In the linear regressions, the Charlson-CCI explained more variance in COST than the Quan-CCI (adjusted R2 = 20.7% vs. 19.9%), adjusting for age and sex. The Charlson-CCI was a better predictor of the risk of emergency department visits (c=0.600) than the Quan-CCI (c=0.571). Compared with the Quan-CCI, the Charlson-CCI showed better discriminatory power for the prediction of high-expenditure individuals (c=0.770 vs. 0.743) and the risk of hospitalizations (c=0.589 vs. 0.581). The Quan-CCI did not significantly predict high-expenditure individuals (OR=1.15; 95% CI=0.99-1.33) or the risk of hospitalizations (OR=1.14; 95% CI=0.99-1.30). CONCLUSIONS: In a group of previously hospitalized patients, the original CCI exhibited better discrimination for the prediction of healthcare expenditures, hospitalizations, and emergency department visits. The weights updated by Quan et al. were developed to predict mortality and may have limited utility in predicting health care utilization.