Machine Learning-Based Prediction Models for Postoperative Delirium:A Systematic Review and Meta-Analysis

Abstract

Background The number of risk prediction models for postoperative delirium has increased yearly, but their quality and applicability in clinical practice and future research remain unclear. Aims This systematic review aimed to evaluate published studies on postoperative delirium risk prediction models and provide guidance for model establishment and improvement. Methods We searched PubMed, Embase, Cochrane Library, and Web of Science for eligible studies up to February 10, 2024. Included studies provided data for assessing the sensitivity and specificity of prediction models. Results We included 12 articles with 58 machine learning (ML) prediction models, covering 37,978 cases with 3,044 instances of postoperative delirium. The combined area under the receiver operating characteristic curve (AUC) for predicting postoperative delirium was 0.82 [95% CI, 0.79–0.85], with a sensitivity of 0.74 [95% CI, 0.69–0.78] and a specificity of 0.78 [95% CI, 0.73–0.82].Subgroup analysis showed that prediction models using random forests had a higher combined AUC of 0.90 [95% CI, 0.87–0.92]. Models for orthopedic surgeries and individuals aged over 60 had higher predictive value. Asian populations showed higher predictive value compared to European and American populations. Conclusions ML models perform well in predicting the occurrence of postoperative delirium, particularly random forest models.