Comparing comorbidity adjustment scores for predicting in-hospital mortality using administrative data

Abstract

The Charlson Comorbidity Index (CCI) is a method for classifying comorbidities as a single measure on a scale, based on ICD-9-CM codes over administrative databases. This study aims to compare different variants of the CCI and their relation with in-hospital mortality, using different weights associated to each comorbidity. Within the 9,613,563 hospitalizations in the 15-year studied period, the area under the ROC curve was higher considering the original Charlson weights, when compared to other more recent proposals. Also, all the indexes had an increased association with in-hospital mortality throughout time. For recent years this association is stronger, demonstrating an increased applic ability of the Charlson index in administrative databases. The validity of the coding algorithms strongly depends on the completeness and accuracy of diag nostic coding, particularly considering secondary diagnoses. The Charlson index can be a valuable tool for longitudinal studies, but important differences among weights, through years, and for different main diagnoses, should be considered and discussed.