Objective:

To systematically review and quantitatively synthesize the evidence base for machine learning-based readmission prediction after total hip and knee arthroplasty (THA/TKA).

Approach:

• Literature Search: A systematic search was conducted across PubMed, Embase, Cochrane Library, and Web of Science from inception to December 31, 2025, focusing on studies developing or validating ML models for THA/TKA readmission.
• Data Extraction and Analysis: Model performance (C-statistic/AUC) was extracted, and study quality was assessed using the PROBAST + AI tool. A multivariate random-effects meta-analysis was performed to pool C-statistics and quantify heterogeneity.

Key Findings:

• Fifteen studies (57 distinct models) were included, with a pooled C-statistic of 0.76 (95% CI: 0.71–0.81).
• Extreme heterogeneity (I2 = 99.9%) limited the clinical utility of the pooled estimate.
• Single-center models showed higher performance (0.86) compared to multicenter models (0.65).
• THA-specific models yielded higher estimates than TKA-specific models.
• Advanced ML algorithms did not consistently outperform traditional logistic regression.
• High risk of bias was identified in the majority of studies due to analytical shortcomings and lack of model recalibration.

Interpretation:

The evidence for ML-based readmission prediction after THA/TKA is characterized by extreme heterogeneity and high methodological bias.

Limitations:

• Extreme heterogeneity among studies.
• High risk of bias in most studies.
• Lack of model recalibration and calibration metrics.

Conclusion:

The current body of evidence for ML-based readmission prediction after THA/TKA is characterized by extreme heterogeneity and high methodological bias.

Sources:

• Systematic Review Registration