Risk prediction models for postoperative delirium in adult patients undergoing cardiac surgery: a systematic review and meta-analysis

Postoperative delirium (POD), a frequent complication following cardiac surgery, is associated with adverse clinical outcomes. Despite the development of numerous prediction models for estimating the risk of POD, the overall performance and methodological quality of these models are not well understood. To systematically review and meta-analyze the performance of prediction models for postoperative delirium in adult patients undergoing cardiac surgery, with a particular emphasis on model discrimination and the identification of key predictors. This study included studies that developed or validated multivariable prediction models for POD in adults undergoing cardiac surgery. Ten databases were searched from inception to July 10, 2025. Data extraction followed a standardized form based on the Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modelling Studies (CHARMS) checklist. Risk of bias and applicability were assessed using the Prediction Model Risk of Bias Assessment Tool (PROBAST). Meta-analysis was performed using R software. A total of 22 studies comprising 27 prediction models were included. All studies were rated as having a high risk of bias. The pooled area under the curve for modeling cohorts was 0.821 (95% CI: 0.778–0.858; 95%PI: 0.562–0.943), indicating a robust discrimination despite substantial heterogeneity (I²=96.7%, τ2 = 0.3692). Seven significant predictors were identified, including age, history of cerebrovascular disease, cardiopulmonary bypass time, mechanical ventilation time, American Society of Anesthesiologists classification, operative time, and Acute Physiology and Chronic Health Evaluation II (APACHE II) score. Although, existing prediction models for POD in patients undergoing cardiac surgery demonstrate promising performance, the evidence is limited by high risk of bias and heterogeneity across studies. There remains a need to improve methodological rigor, such as multicenter prospective studies and external validation. This systematically review provides a reference for the development and validation of subsequent models for adults undergoing cardiac surgery. Although with modest discrimination, future models should focus on the construction of the preoperative models, which provide more opportunities for early prevention. PROSPERO: CRD420251081560.