Aim: To examine the validity of race/ethnicity-specific comorbidity adjustment scores in estimating in-hospital mortality. Materials & methods: Using 2007-2014 data from the State Inpatient Databases (SID), we compared the performance of derived race/ethnicity-specific composite scores to the existing scores and binary Elixhauser comorbidity measures at estimating in-hospital mortality. Results: In the overall validation sample (N = 9,564,277), our index (c = 0.80; 95% CI: 0.79-0.80) discriminated better than the van Walraven score (c = 0.79; 95% CI: 0.79-0.79), SID 29 (c = 0.78; 95% CI: 0.78-0.79) and SID 30 (c = 0.78; 95% CI: 0.78-0.78), but was not superior to the binary indicators (c = 0.80; 95% CI: 0.80-0.80). Similar findings were observed in individual populations of White and Black patients. All models showed weak calibration. Conclusion: Race/ethnicity-specific indexes discriminated slightly better than existing composite measures at modeling in-hospital mortality in individual subgroups of race/ethnicity.
Keywords: administrative data; comorbidity; in-hospital mortality; modeling; race and ethnicity.