Comorbidity index was successfully validated among men but not in women

Abstract

Objective: To validate the prognostic accuracy of a previously proposed comorbidity index using information of a different and separate population-based cohort.

Study design and setting: We assessed the predictive accuracy of a comorbidity index to predict mortality by looking at calibration and discrimination in the development cohort as well as in a new cohort for validation. Calibration of the model was assessed by comparing predicted and current mortality in the new cohort by means of Hosmer-Lemeshow test (HL). Discrimination of the models was analyzed using the area under the receiver operating characteristic curves (ROC).

Results: In the development cohort, we have not detected differences between the predicted and the observed mortality in both, men (HL=7.7, P=0.46) and women (HL=11.7, P=0.16). The discrimination of the model accounted 81% in men and 79% in women. In the validation cohort, we obtained a good calibration among men (HL=10.1, P=0.43) but not in women (HL=21.4, P=0.01). The discrimination was quite similar to the development cohort in both sexes (ROC area=80% in men, ROC area=78% in women).

Conclusion: The comorbidity index has good calibration and discrimination and was successfully validated in a different population-based cohort among men but not among women.