Developing a Community-Specific Daily Weather Health Risk Index Across Australia Using Explainable Machine Learning

Area of Science:

• Environmental Health
• Biostatistics
• Machine Learning

Background:

• Weather conditions significantly impact human health.
• Effective communication of complex, interacting weather-related health risks is limited.
• Quantifying these risks is crucial for public health interventions.

Purpose of the Study:

• To develop and validate a daily weather-health risk index (WHRI) for communities across Australia.
• To quantify the joint health risks associated with multiple meteorological factors.
• To identify spatiotemporal patterns in weather-related mortality risks.

Main Methods:

• Collected daily mortality and meteorological data (2009-2019) for Australian communities.
• Utilized an explainable machine learning framework: eXtreme Gradient Boosting (XGBoost) with Shapley Additive exPlanations (SHAP).
• Constructed a community-specific daily WHRI.

Main Results:

• Temperature was the dominant factor contributing to mortality risks.
• Southern Australian communities exhibited higher weather-related mortality risks and WHRI than northern communities.
• WHRI showed distinct seasonal patterns, peaking in winter and lowest in summer.

Conclusions:

• Weather-related health risks display significant spatiotemporal heterogeneity.
• The developed WHRI effectively captures and quantifies dynamic risk patterns.
• Integrating WHRI into public health dashboards can support timely, evidence-based interventions for adverse weather events.