This systematic review provides the first comprehensive synthesis of the literature on the application of deep learning (DL) models to the task of predicting ICU readmission, offering insights into current approaches and their open challenges, helping healthcare professionals in assessing their clinical applicability.

A high risk of bias was observed, considerable variability across study settings, and highly heterogeneous performance.

Reproducibility was poor, and most approaches relied on US-based datasets, raising concerns about their generalizability, Model explainability was also limited.

Taken together, these observations suggest that the quality of the existing studies is currently inadequate, hindering both the evaluation and clinical applicability of DL approaches to this important area.

A systematic review of studies were performed, describing DL models for the prediction of ICU readmission risk, according to the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) statement.

This systematic review was registered in the International Prospective Register of Systematic Reviews (PROSPERO) under ID: CRD420251000813.

PubMed, Embase, Scopus, and Web of Science was searched for articles using MeSH terms and free words (Supplementary Table 1, date of systematic search: March 4th, 2025).

Read More