Key result
Random Forest model predicts hospital readmission with an F1-score of ~0.54 using clinical utilization metrics.
Why the study?
Hospital readmission is a key measure of healthcare quality and safety, and predictive analytics may help identify patients at greater risk of readmission to improve healthcare delivery.
Observational (n=25,000)
Machine learning models, particularly Random Forest, can predict hospital readmission using clinical and utilization variables, though performance is moderate and requires further refinement.
Answers are grounded in this paper and cite their sources.
Moderate ML readmission prediction warrants clinical caution; leaves open need for prospective validation and refinement.
Hospital readmission is a key measure of health care quality, safety and hospital performance. This study aimed to investigate hospital readmission rates using predictive analytics to determine which patients were at greater risk of readmission and help inform healthcare services. In this study used a structured hospital readmission dataset with 25,000 patients and 17 variables. Readmission was the dependent variable, with the independent variables being age, length of stay, number of laboratory tests, number of medications, number of outpatient visits, number of inpatient visits, number of emergency visits, medical speciality, diagnosis categories, glucose test, A1C test, medication change and diabetes medication status. There were no missing data or duplicate records in the dataset. Descriptive statistics revealed a readmission rate of 47.02% of patients, suggesting a high readmission rate. Machine learning models, such as Logistic Regression, Decision Tree and Random Forest were trained and tested based on accuracy, precision, recall, F1-score, and ROC-AUC. Random Forest emerged as the top model with the highest F1-score (0.5428) and Logistic Regression with the highest ROC-AUC (0.6434). Laboratory procedures, number of medications, time in hospital, age, diagnosis categories, medical speciality, and past inpatient visits were identified as important features in predicting readmission. The results indicate that predictive analytics can help hospitals identify patients at risk, plan discharge, organise follow-up care and allocate resources, which can lead to better patient-centred care.
No takes yet. Share an insight, caveat, or question.
Sandeep Kulkarni (2025) conducted an observational in Hospital readmission (n=25,000). Machine learning predictive models was evaluated on Prediction of hospital readmission. A Random Forest machine learning model predicted hospital readmission with an F1-score of 0.5428, identifying the number of laboratory procedures, medications, and length of stay as the top predictors.
Synapse has enriched 10 related papers. Consider them for comparative context: