Postoperative delirium (POD) is a frequent and serious complication following heart valve replacement (HVR) performed with cardiopulmonary bypass (CPB). POD has been associated with significantly increased mortality and prolonged hospitalization.

A retrospective cohort study published in the Asian Journal of Psychiatry analyzed 1,076 adult patients who underwent HVR with CPB between January 2018 and December 2022. POD was assessed using the Confusion Assessment Method for the Intensive Care Unit (CAM-ICU). Perioperative factors were used to develop prediction models, including a traditional logistic regression–based nomogram and four machine learning models. The best-performing model was subsequently evaluated in an external validation cohort of 281 patients.

The overall incidence of POD was 25.1% (270/1076 patients). Patients who developed POD had significantly higher mortality compared with those without POD (11.9% vs 0.2%, P<0.001) and experienced longer hospital stays. Among the evaluated models, the Random Forest (RF) algorithm demonstrated the highest predictive performance in the internal test dataset, achieving an area under the receiver operating characteristic curve (AUROC) of 0.854 (95% CI 0.764–0.945), compared with 0.754 for the traditional nomogram. In the external validation cohort, the RF model maintained predictive performance with an AUROC of 0.793. Model interpretation using SHAP analysis identified postoperative awakening time (PAT), postoperative mechanical ventilation time (PMVT), and postoperative reintubation (POR) as the most influential predictors.

The Random Forest model demonstrated higher predictive performance than the traditional nomogram in this analysis. Key perioperative factors were identified as important contributors to POD risk after HVR.