Enhancing heatstroke prediction accuracy with interpretable machine learning: a multi-center data-driven approach

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Public Health Surveillance

Background:

• Heatstroke is a critical public health concern with high mortality rates.
• Accurate and timely diagnosis of heatstroke is essential for effective patient management.
• Existing diagnostic methods may benefit from enhanced predictive tools.

Purpose of the Study:

• To develop and validate an interpretable machine learning model for heatstroke prediction.
• To utilize clinical and laboratory data for forecasting heatstroke risk.
• To enhance early identification and management of heatstroke patients.

Main Methods:

• A gradient boosting machine (GBM) model was developed using data from 24 hospitals (2021-2022).
• Model performance was evaluated using area under the receiver operating characteristic curve (AUROC) and calibration plots.
• SHapley Additive exPlanations (SHAP) were used for model interpretability.

Main Results:

• The GBM model achieved an AUROC of 0.971 on training data and 0.836 on validation data.
• Creatine kinase (CK)-MB was identified as the most significant predictor in the GBM model.
• Decision curve analysis indicated substantial net benefits for the GBM model.

Conclusions:

• The developed machine learning model demonstrates robust predictive capabilities for heatstroke.
• This interpretable model can assist clinicians in identifying and managing at-risk patients.
• The findings support the integration of ML tools into clinical practice for heatstroke surveillance.