Machine Learning-Based Forecast of Hemorrhagic Stroke Healthcare Service Demand considering Air Pollution

Area of Science:

• Environmental Health
• Public Health
• Data Science

Background:

• Hemorrhagic stroke poses a significant public health burden.
• Understanding factors influencing healthcare demand is crucial for resource allocation.
• Air quality is increasingly recognized as a potential environmental determinant of stroke incidence.

Purpose of the Study:

• To forecast the demand for hemorrhagic stroke healthcare services.
• To evaluate the role of air quality and machine learning in this forecasting.
• To assess the impact of seasonality and lag effects on prediction accuracy.

Main Methods:

• Utilized data on hemorrhagic stroke cases, air quality, and meteorological factors from 2016-2017.
• Applied six distinct machine learning methods for demand forecasting.
• Incorporated seasonality and lag effects into the predictive models.

Main Results:

• The average area under the curve reached 0.7971, indicating good predictive performance.
• Forecasting accuracy was significantly higher during the warm season compared to the cold season.
• Including air pollution data demonstrably improved the machine learning model's performance.
• A partially linear relationship was observed between air pollutant concentrations and healthcare service demand.

Conclusions:

• Machine learning models integrating air quality data can effectively forecast hemorrhagic stroke healthcare demand.
• Short-term air pollutant concentrations are feasible predictors for healthcare service demand.
• The developed forecast model offers a valuable tool for advance warning and proactive healthcare management.