Machine Learning-Based Demand Prediction for Emergency Medical Services

Area of Science:

• Operations Research
• Data Science
• Public Health

Background:

• Emergency Medical Services (EMS) are vital for out-of-hospital care.
• Optimizing EMS operations requires accurate demand prediction and resource allocation.
• Existing digital twins for EMS lack efficiency and integration of temporal factors like environmental conditions.

Purpose of the Study:

• To enhance an existing EMS digital twin using machine learning for improved demand forecasting.
• To integrate external time-variant factors and reduce computational processing time.
• To enable more efficient EMS deployment and resource optimization.

Main Methods:

• Development and comparison of multiple machine learning models.
• Integration of external time-variant factors into the predictive model.
• Evaluation of model performance against historical data and a baseline using error metrics (RMSE).

Main Results:

• The machine learning-enhanced model significantly improved predictive accuracy, reducing RMSE from 8.7 to 2.5 events/hour.
• Substantial reduction in computational time, enabling day-by-day predictions.
• Demonstrated potential for real-time decision-making in EMS resource allocation.

Conclusions:

• Machine learning offers a more efficient and accurate approach to EMS demand forecasting.
• The enhanced model provides valuable insights for policy development and resource alignment with demand dynamics.
• This approach can lead to improved emergency response outcomes and better out-of-hospital patient care.