Outcome-specific Charlson Comorbidity Indices for Predicting Poor Inpatient Outcomes Following Noncardiac Surgery Using Hospital Administrative Data

Background:A need exists for adapting existing perioperative risk stratification methods such as the Charlson Comorbidity Index (CCI) for application with hospital administrative data in noncardiac surgery populations.Objective:Develop and validate outcome-specific CCIs for predicting inpatient mortality, and cardiac and renal morbidity in noncardiac surgery patients using hospital administrative data.Methods:We used hospital administrative data from the 2010 and 2011 California State Inpatient Database (SID) to develop (derivation cohort: 2010 SID, n=177,280) and validate (validation cohort: 2011 SID, n=179,145) 3 outcome-specific CCIs. Along with the 17 CCI comorbidities, the clinical importance and weighted point scores for age, male sex, race, emergent admission, and high-risk surgery were also determined from the coefficients of a logistic regression model. Cumulative outcome-specific CCI, CCI, and age-adjusted CCI (AACCI) scores were calculated for each patient. Receiver-operator characteristic curve analyses were used to determine the prognostic accuracy (area under the curve) of each outcome-specific CCIs, the CCI, and the AACCI. Risk was stratified according to cumulative point scores for each outcome-specific CCI, and posttest probabilities for each risk category were calculated.Results:All outcome-specific CCIs showed good performance as a prognostic tools (area under the curve>0.800 for all) and performed better than the CCI and AACCI. We attached clinical relevance to a given cumulative point score by determining posttest probabilities for each outcome-specific index.Conclusions:We successfully adapted and validated 3 outcome-specific CCIs for use in noncardiac surgery patients based on ICD-9 and hospital admission data. Further validation of these outcome-specific CCIs is warranted.