JMIR Publications

ABSTRACT

Background:

Unplanned readmissions increase unnecessary healthcare costs and reduce the quality of care. It is essential to plan the discharge care from the beginning of hospitalization to reduce the risk of readmission. Machine learning-based readmission prediction models can support patients’ preemptive discharge care services with improved predictive power.

Objective:

This study aims to develop a readmission prediction model for high-risk discharged patients that uses early hospitalization indicators to predict readmission and plan discharge care from the beginning of hospitalization.

Methods:

This retrospective study included the electronic medical records of 12,977 patients with one of the top six high-risk readmission diseases at a tertiary hospital in Seoul from January 2018 to January 2020. We used demographic, clinical, and nursing data to construct a prediction model. To improve the performance of the machine learning method, we performed 5-fold cross-validation and utilized Adaptive Synthetic Sampling to address data imbalance. A total of 6 models were implemented, and the final model was selected by evaluating the Accuracy, Precision, Recall, F1-score, and AUROC(Area Under a Receiver Operating Characteristic curve) of each model.

Results:

We constructed unplanned readmission prediction models by dividing them into Model 1 and Model 2. Model 1 used early hospitalization data (up to one day after admission), and Model 2 used all. Model 1, the Random Forest model, performed best, with the AUROC being 0.62, and Model 2, the CatBoost model, performed best, with the AUROC being 0.64.

Conclusions:

The machine learning-based readmission prediction model using nursing data developed from this study can be used as a clinician decision support system for early screening of high-risk discharged patients for readmission risk and setting discharge plans and interventions.