Artificial Intelligence Based Prediction of Emergency Department Revisits and hospitalization: A Systematic Review and Meta-Analysis

Abstract

Background and aim: Short-term emergency department readmission is one of the major challenges in resource management and quality of clinical care. The aim of this study was to evaluate the performance of artificial intelligence models in predicting patients' frequent emergency department visits and to investigate the effect of the data type and algorithm used on the sensitivity and specificity of prediction.

Method: The international databases PubMed, EMBASE, and Web of Science were searched using keywords aligned with the study objective between from January 2015 to 10 December  2025. In the present study, 27 studies were included in the meta-analysis. Data analysis was performed using the stata.v17 software and analysis was performed using a random effects model (DerSimonian–Laird) to determine sensitivity and specificity.

Result: The overall sensitivity of the models was 0.61 [95% CI; 0.54 to 0.68] and the overall specificity was 0.83 [95% CI; 0.78–0.89]. Subgroup analysis showed that the type of AI algorithm had a significant effect on diagnostic accuracy; Ensemble and XGBoost models performed better.

Conclusion: Artificial intelligence models can be a useful tool for identifying high-risk ED patients, especially using multi-source data and hybrid algorithms.