Ensemble Machine Learning Models for Predicting Patients With High Usage: Model Validation and Economic Impact Analysis

1

The study evaluated ensemble machine learning models to predict high health care utilization among diabetes patients.
2

Boosted tree models demonstrated the highest predictive performance for inpatient length of stay and emergency department visits.
3

The models achieved multiclass area under the receiver operating curve scores of 0.6877 for length of stay and 0.7601 for ED visits.
4

Economic analysis indicated a potential cost reduction of SGD $152 million through the use of the boosted tree model.
5

Ensemble models can support targeted interventions and inform planning in diabetes-related population health programs.

1