What question did this study set out to answer?

February 5, 2026Scientific ReportsOpen Access

Predicting non-emergency healthcare use in Australia using machine learning on longitudinal household data

View Full Paper

Ask AI

Bookmark

Authors

This analysis predicts future healthcare use in the population, indicating potential interventions for cost control.

To identify and predict non-emergency healthcare use in Australia using machine learning methods on longitudinal data.
Analyzed three waves of the HILDA survey containing health-related information.
Applied four machine learning methods: Random Forest, Gradient Boosting Decision Trees, Extreme Gradient Boosting, and multilayer perceptron neural networks.
Evaluated predictive performance using accuracy, sensitivity, specificity, and AUC measures.
Conducted model calibration using Brier score and explained model predictions with LIME and SHAP.
Identified strong predictors of healthcare use, including age, socio-economic status, and private insurance status.
Gradient Boosting Decision Trees outperformed logistic regression in prediction accuracy.
Logistic regression achieved an AUC of 0.69, while ML models had AUCs between 0.68 and 0.76.
Sensitivity ranged from 86% to 89%, while specificity was between 40% and 44%.

Lee et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark

February 5, 2026Scientific ReportsOpen Access

View Full Paper

Ask AI

Bookmark

Authors

This analysis predicts future healthcare use in the population, indicating potential interventions for cost control.

To identify and predict non-emergency healthcare use in Australia using machine learning methods on longitudinal data.
Analyzed three waves of the HILDA survey containing health-related information.
Applied four machine learning methods: Random Forest, Gradient Boosting Decision Trees, Extreme Gradient Boosting, and multilayer perceptron neural networks.
Evaluated predictive performance using accuracy, sensitivity, specificity, and AUC measures.
Conducted model calibration using Brier score and explained model predictions with LIME and SHAP.
Identified strong predictors of healthcare use, including age, socio-economic status, and private insurance status.
Gradient Boosting Decision Trees outperformed logistic regression in prediction accuracy.
Logistic regression achieved an AUC of 0.69, while ML models had AUCs between 0.68 and 0.76.
Sensitivity ranged from 86% to 89%, while specificity was between 40% and 44%.

Lee et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark