A spatiotemporal quantile regression model for emergency department expenditures

Area of Science:

• Health economics
• Biostatistics
• Geographic information systems

Background:

• Emergency department (ED) care utilization and associated costs exhibit complex geographic and temporal trends.
• Analyzing medical expenditures requires methods that capture variations beyond the average, particularly at the extremes of the distribution.
• Existing models may not adequately address the spatiotemporal dynamics of healthcare spending.

Purpose of the Study:

• To develop and apply a novel spatiotemporal quantile regression model for analyzing ED-related medical expenditures.
• To identify distinct spatial patterns in expenditures across different quantiles of the cost distribution over time.
• To improve the understanding of healthcare cost variations in specific geographic areas.

Main Methods:

• Development of a hierarchical spatiotemporal quantile regression model.
• Incorporation of patient-level and region-level predictors.
• Utilizing intrinsic conditionally autoregressive priors for spatiotemporal random effects and an asymmetric Laplace distribution within a Bayesian framework.
• Efficient posterior sampling using conjugate full conditionals.

Main Results:

• The model identified distinct spatial patterns in ED medical expenditures that varied across different quantiles (e.g., median, 90th percentile) over time.
• Significant spatiotemporal variations were observed particularly in the extreme quantiles of expenditure, while mean expenditures showed less variation.
• The approach enhanced small-area estimation through maximum spatiotemporal smoothing.

Conclusions:

• Spatiotemporal quantile regression provides a powerful tool for analyzing the complex patterns of ED medical expenditures.
• Understanding variations in expenditure extremes is crucial for targeted healthcare cost management and policy.
• The developed Bayesian model offers an efficient and effective method for analyzing georeferenced health and financial data.