Prediction Models for Postoperative Atrial Fibrillation After Cardiac Surgery: A Systematic Review and Critical...

Area of Science:

• Cardiology
• Medical Informatics
• Clinical Epidemiology

Background:

• Postoperative atrial fibrillation (POAF) is a common complication after cardiac surgery.
• POAF increases morbidity, hospitalization duration, and healthcare expenses.
• Existing risk prediction models for POAF often lack robust validation and clinical applicability.

Purpose of the Study:

• To systematically review and critically appraise prediction models for POAF post-cardiac surgery.
• To evaluate the methodological quality, validation strategies, and predictive performance of these models.

Main Methods:

• A comprehensive literature search was conducted following PRISMA 2020 guidelines.
• Data extraction utilized the CHARMS checklist, and methodological quality was assessed using PROBAST.
• Model performance focused on discrimination (C-statistic/AUC) and calibration, with an emphasis on validation strategies.

Main Results:

• 39 studies were included, with most models using logistic regression.
• Discrimination varied widely (0.60-0.98), and calibration reporting was inconsistent.
• A significant majority (82%) of studies were rated at high risk of bias, primarily due to issues with overfitting and limited validation.

Conclusions:

• Current POAF prediction models exhibit heterogeneous performance and are often hampered by high bias and insufficient validation.
• These limitations restrict the clinical utility of existing models.
• Future research must emphasize rigorous external validation, transparent reporting (TRIPOD), and methods to improve generalizability.