Background: The Chronic Disease Score is a risk-adjustment metric based on age, gender, and history of dispensed drugs. We compared four versions of the score for their ability to predict hospitalization among members of eight health maintenance organizations nationwide.
Methods: The study included 29,247 women age 45 years and older. Logistic regression models were constructed using rank quintile and rank decile indicators for each of four scores as predictors of hospitalization during the year after 1 October 1995. Discrimination and model fit were compared using several model properties including the C statistic and the odds ratio comparing highest with lowest quantiles.
Results: All Chronic Disease Score versions performed similarly, with the version that predicts total healthcare cost, proposed by Clark et al. (Med Care 1995;33:783-795), performing somewhat better than the other three. The overall risk of hospitalization was 12%. Individuals with higher quantile ranks had a higher risk of hospitalization. Among the Chronic Disease Score versions, the risk of hospitalization ranged from 4% for the lowest decile to 27-29% for the highest decile. Odds ratios comparing the highest with the lowest deciles ranged from 8.9 to 10.2.
Conclusions: The Chronic Disease Score predicts hospitalization and therefore may be a useful indicator of baseline comorbidity for control of confounding.