How do machine learning algorithms perform in predicting hospital choices? evidence from changing environments.

Area of Science:

• Health economics
• Econometrics
• Machine learning

Background:

• Machine learning (ML) models often outperform traditional econometric models in prediction tasks.
• The performance of these models can vary depending on the stability of the choice environment.

Purpose of the Study:

• To evaluate the relative performance of ML algorithms versus econometric models in predicting hospital demand.
• To assess how changes in the choice environment, specifically due to natural disasters, affect model performance.

Main Methods:

• Comparative analysis of machine learning algorithms and econometric models.
• Evaluation using hospital demand data, considering scenarios with stable and disrupted patient choice sets.

Main Results:

• Machine learning models demonstrated superior predictive accuracy in stable choice environments.
• The performance advantage of ML models diminished when natural disasters substantially altered patients' available hospital choices.
• Econometric models provided significant additional information during periods of major environmental change.

Conclusions:

• While ML is powerful for prediction, its advantage decreases with significant disruptions to choice sets.
• Econometric models remain valuable for understanding demand shifts in dynamic and unpredictable environments.