Prediction Models for Postoperative Delirium Among Cancer Patients: A Scoping Review

Area of Science:

• Oncology
• Geriatrics
• Anesthesiology
• Medical Informatics

Background:

• Postoperative delirium (POD) is a significant concern in cancer patients, impacting recovery and outcomes.
• Accurate prediction models are crucial for early intervention and improved patient management.
• Current models for POD prediction in cancer patients require systematic evaluation.

Purpose of the Study:

• To systematically map existing postoperative delirium (POD) prediction models in cancer patients.
• To analyze study design, modeling techniques, performance evaluation, and reporting standards of these models.
• To assess the readiness of current POD models for clinical implementation.

Main Methods:

• Systematic mapping following the Arksey and O'Malley framework and PRISMA-ScR guidelines.
• Searched nine databases from inception to April 2026.
• Narrative synthesis of 32 included studies on POD prediction models in cancer patients.

Main Results:

• Thirty-two studies were included, with most showing a high risk of bias.
• Logistic regression was common; key predictors included age, operation time, and nutritional/inflammatory indicators.
• Model performance showed promising discrimination (AUC 0.690–0.973), but external validation and clinical utility assessment were often insufficient.

Conclusions:

• Existing POD prediction models for cancer patients demonstrate good discrimination but are not yet ready for routine clinical use.
• Limitations include inadequate external validation, inconsistent clinical utility assessment, and lack of model transparency.
• Future research should focus on standardized reporting, multicenter validation, and decision-curve analysis for safe clinical translation.