SHM Converge Scientific Abstract ePoster Gallery-Determining the Effectiveness of Machine Learning Models for Predicting Hospital Length of Stay: A Systematic Review

Pdf Summary

This systematic review evaluated whether machine learning (ML) models can accurately predict hospital length of stay (LoS) in U.S. hospitals. Using PubMed and a search window from January 2020 to January 2024, the authors screened 658 articles and included 24 studies meeting criteria focused on general internal medicine, LoS, ML models, and U.S. hospitals.<br /><br />The review highlights LoS as an important hospital metric tied to illness severity, cost, resource use, discharge planning, and readmission risk. Accurate LoS prediction could help clinicians make earlier decisions, improve scheduling, reduce physician workload, support patient counseling, and optimize resource allocation.<br /><br />Overall, the studies showed strong potential for ML in this area. Many reported prediction accuracy above 89%, though performance varied by model and by the length of stay being predicted. Short-term LoS predictions, generally under 7 days, were more accurate than long-term predictions, which became less reliable as stay duration increased. Random forest was frequently reported as one of the most accurate and robust models, including when data were incomplete or uncategorized.<br /><br />The review also notes important limitations, especially the need for larger and more diverse datasets and concern about bias in algorithms. The authors conclude that ML models could meaningfully transform hospital operations and patient care, but further programming improvements and external validation are needed before widespread clinical adoption.

Asset Subtitle

Mukul Sharda

Meta Tag

Author List Janavi Wagle, Mukul Sharda, Nathaniel Verhagen, Pinky Jha, Ritisha Mathur, Rupesh Prasad, Saathvik Gowda, Sanjay Bhandari, Saujas Sharma, Sharath Kommu, Shaunak Raikar

Category Research

Concept Machine Learning Model

Concept Hospital Length of Stay

Concept Systematic Review

Concept Model Accuracy

Concept Inclusion Criteria

Distinguished Non-Finalist

Presenter Organization Medical College of Wisconsin

Presenting Author Mukul Sharda

Track Technology in Hospital Medicine

Keywords

machine learning

hospital length of stay

systematic review

U.S. hospitals

prediction accuracy

random forest

general internal medicine

resource allocation

readmission risk

external validation

Machine Learning Model

Hospital Length of Stay

Systematic Review

Model Accuracy

Inclusion Criteria