The Best Use of the Charlson Comorbidity Index With Electronic Health Care Database to Predict Mortality

Background: The most used score to measure comorbidity is the Charlson index. Its application to a health care administrative database including International Classification of Diseases, 10th edition (ICD-10) codes, medical procedures, and medication required studying its properties on survival. Our objectives were to adapt the Charlson comorbidity index to the French National Health Insurance database to predict 1-year mortality of discharged patients and to compare discrimination and calibration of different versions of the Charlson index. Methods: Our cohort included all adults discharged from a hospital stay in France in 2010 registered in the French National Health Insurance general scheme. The pathologies of the Charlson index were identified through ICD-10 codes of discharge diagnoses and long-term disease, specific medical procedures, and reimbursement of specific medications in the past 12 months before inclusion. Results: We included 6,602,641 subjects at the date of their first discharge from medical, surgical, or obstetrical department in 2010. One-year survival was 94.88%, decreasing from 98.41% for Charlson index of 0-71.64% for Charlson index of >= 5. With a discrimination of 0.91 and an appropriate calibration curve, we retained the crude Cox model including the age-adjusted Charlson index as a 4-level score. Conclusions: Our study is the first to adapt the Charlson index to a large health care database including > 6 million of inpatients. When mortality is the outcome, we recommended using the age-adjusted Charlson index as 4-level score to take into account comorbidities.