Comparing major comorbidity indices as predictors of all-cause mortality in the Veterans Affairs healthcare system
Abstract: Objective: The Charlson Comorbidity Index (CCI), the Elixhauser Comorbidity Index (ECI), and the Functional Comorbidity Index (FCI) are validated clinical measures of comorbidity, but direct comparisons between these measures have rarely been studied especially in high-risk patient populations, such as homeless individuals. The U.S. Department of Veterans Affairs (VA) offers large patient samples to compare these comorbidity measures as predictors of mortality using administrative and clinical records. We examined CCI, ECI, and FCI scores among veterans seeking VA healthcare services, including those experiencing homelessness, and compare their predictive value in relation to all-cause mortality risk. Study Design and Setting: Several VA databases from 2017–2021 were retrospectively linked and 4,701,711 U.S. veterans [308,553 with homelessness and 4,393,158 without homelessness] were evaluated over a median follow-up of 4.1 years, yielding 917,921 recorded deaths. Regression models were constructed, and Harrell’s Concordance Statistic (HCS) was calculated that assessed the ability of z-transformed comorbidity scores to discriminate ‘high-risk’ vs. ‘low-risk’ groups of patients for mortality risk, after adjustment for demographic and clinical characteristics. Results: In adjusted models, ECI (HCS: 0.76-0.77) and CCI (HCS: 0.75-0.76) were better able to discriminate ‘high-risk’ vs. ‘low-risk’ groups than FCI (HCS: 0.72-0.75) among homeless and non-homeless veterans. Compared to ECI and CCI, FCI was more strongly associated with homelessness. Conclusion: CCI and ECI may be more predictive of all-cause mortality risk than FCI, although FCI may be a useful measure of functioning in homeless populations.