The estimation power of alternative comorbidity indices

Abstract

Objective: Health-care expenditures are strongly influenced by overall illness burden. Appropriate risk adjustment is required for correct policy analysis. We compared three risk adjustment methods: the Charlson comorbidity index (CCI), the chronic disease score (CDS), and the Agency for Healthcare Research and Quality's comorbidity index (AHRQCI) in terms of their estimation power in analyzing health-care expenditures.

Method: Data from the Thomson MarketScan Research Databases (Thomson Healthcare, Ann Arbor, MI) were used to estimate total health-care expenditures of migraine patients treated by a triptan. Seven distinct multivariate models were evaluated for model fit (CCI only, CDS only, AHRQCI only, CCI + CDS, CCI + AHRQCI, CDS + AHRQCI, and CCI + CDS + AHRQCI). The estimation power of these indices (alone and in combination) was evaluated using Bayesian and Akaike information criteria, log-likelihood scores, and pseudo R(2) values.

Results: Confirming results from previous studies, when comorbidity indices were considered individually the results were inconclusive. Statistically the best performance was observed in the model that included all three of the comorbidity measures (CCI + CDS + AHRQCI); however, the practical differences in the estimated values were small.

Conclusion: Low correlation between these comorbidity indices shows that it is possible to have potential risk factors that are not captured in the single comorbidity index. Each comorbidity measure considers different risks, and the collinearity of the three measures is not strong enough to preclude using them simultaneously in the same model.