Detecting the spatial clustering of exposure-response relationships with estimation error: a novel spatial scan...

Area of Science:

• Environmental Epidemiology
• Spatial Statistics
• Biostatistics

Background:

• Spatial clustering of exposure-response relationships (ERR) is crucial for disease control, identifying sensitive regions, and targeted interventions.
• Existing spatial scan statistics struggle with multivariate spatial data containing estimation errors, like ERR vectors from regression models.
• The increasing availability of such complex spatial datasets necessitates novel cluster-detection tools.

Purpose of the Study:

• To develop a novel spatial scan statistic capable of handling multivariate spatial datasets with estimation error.
• To propose a two-stage analytic process for detecting spatial clustering of exposure-response relationships.
• To validate the performance of the new method using a real-world example and simulations.

Main Methods:

• Extended the classic scan statistic to create the estimation-error-based scan statistic (EESS).
• Developed a two-stage analytical process for practical application in detecting ERR spatial clustering.
• Validated EESS using a published motivating example and a simulation study.

Main Results:

• The EESS effectively detects spatial clusters in datasets with estimation error.
• Detected clusters accurately reflect underlying heterogeneity in exposure-response relationships.
• The method provides more precise ERR estimates by accounting for spatial heterogeneity.

Conclusions:

• The developed EESS is a valuable tool for analyzing spatial clustering in complex environmental health data.
• EESS addresses limitations of traditional methods when dealing with estimation errors.
• This approach enhances the identification of environmental risk factors and informs public health strategies.