Are the Charlson and Elixhauser Comorbidity Indices Reliable Predictors of Postoperative Delirium in Abdominal Surgery?

Abstract Background Postoperative delirium (POD) is a common complication of surgery. It is associated with a number of detrimental effects, including mortality and healthcare costs. We sought to determine whether common comorbidity indices are predictors of POD. Methods Using the Medical Information Mart for Intensive Care (MIMIC)-IV database, we identified 8022 abdominal surgery procedures across 7212 adult patients. We calculated both the Charlson comorbidity index (CCI) and the Elixhauser comorbidity index (ECI) for each procedure and used logistic regression to predict postoperative delirium, which was defined as delirium within 30 days following the procedure. Results Models based on either the CCI and ECI were predictive of postoperative delirium (area under the receiver-operator characteristic curve (AUC-ROC) of 0.622 and 0.652, respectively). However, the addition of other factors known to be associated with delirium improved model performance (AUC-ROC of 0.680). Conclusions Both the CCI and ECI are predictors of postoperative delirium in patients undergoing abdominal surgery. Addition of factors known to be associated with delirium renders additional predictive value and should be included in models that predict postoperative delirium.