BACKGROUND:  Charlson and Elixhauser comorbidity indices are frequently used in observational studies for patient comorbidity risk adjustment. Cross sectional hospital billing databases may underestimate patient comorbidities when compared to longitudinal databases such as administrative claims. OBJECTIVES:   To compare Charlson and Elixhauser comorbidity indices and the individual comorbid categories calculated from a cross sectional hospital billing data with longitudinal claims data. METHODS:  Data source included two data files with linked patients: IMS Health Real-World Data Hospital (billing data) and IMS Health Real-World Data Adjudicated Claims (claims data). Patients who had a total hip/knee replacement surgery from 07/2011 to 06/2015 were identified using billing data. The hospital admission date was termed as the index date and administrative claims data for a one year pre-index period were obtained. Charlson and Elixhauser comorbidity indices and their disease component categories were calculated from claims across all care settings (gold standard) and compared to the scores calculated with the cross-sectional hospital billing data. Charlson and Elixhauser indices from the two databases were compared using generalized estimating equations (GEE) model. Sensitivity and specificity of assessing individual comorbid conditions was also examined. RESULTS: A total of 11,708 patients (age: 59.1 ± 5.8 years) were identified. Their average Charlson comorbidity index was 0.47 (95% CI: 0.45-0.48) from hospital bill data, and 1.09 (1.06-1.12) from claims data. Average Elixhauser comorbidity indices were 1.72 (1.69-1.74), and 2.95 (2.92-2.99) from two databases respectively. Average sensitivity for individual Charlson categories was 0.321 (range: 0.100–0.710), and 0.374 (0.070-0.802) for Elixhauser categories. Conditions with low sensitivity were cerebrovascular, liver disease, AIDS/HIV, cancer, and psychosis. Average specificity was 0.999 for Charlson, and 0.997 for Elixhauser categories. CONCLUSIONS: Comorbidity indices were underestimated by 57% for Charlson and 42% for Elixhauser when calculated from hospital billing data. Individual comorbidity categories had high specificity, but low sensitivity.