Prediction Models for Postoperative Delirium After Cardiovascular Surgery: A Systematic Review and Critical Appraisal

Area of Science:

• Cardiology
• Neurology
• Medical Informatics

Background:

• Postoperative delirium of cardiovascular surgery (PODOCVS) increases complications and mortality.
• Early identification via prediction models can improve outcomes.
• Existing models' quality and applicability for cardiovascular surgery (CS) patients are unclear.

Purpose of the Study:

• To systematically review and critically evaluate prediction models for PODOCVS.
• Assess model development, performance, and applicability.
• Guide healthcare professionals in selecting appropriate models.

Main Methods:

• Systematic search of multiple databases up to August 12, 2024.
• Inclusion of studies developing and validating multivariate predictive models for PODOCVS in adult CS patients.
• Quality assessment using CHARMS checklist and PROBAST tool.

Main Results:

• 24 studies described 67 prediction models for PODOCVS.
• Incidence of delirium ranged from 3.6% to 36%.
• Common predictors included age, EF value, bypass time, and ICU stay; logistic regression was frequent.

Conclusions:

• Models showed good discriminatory ability but high risk of bias (PROBAST).
• High or unclear risk of bias in analysis domains due to underreporting.
• Utility and generalizability are uncertain; models are not recommended for clinical use.