A generalized linear models approach to spatial scan statistics for covariate adjustment.

Area of Science:

• Epidemiology
• Biostatistics
• Spatial Analysis

Background:

• The spatial scan statistic is widely used for cluster detection but has limitations in adjusting for confounding covariates.
• Existing methods may not adequately control for factors like race/ethnicity and age in spatial analyses.

Purpose of the Study:

• To propose a generalized linear models (GLM) approach for constructing spatial scan statistics.
• To enable robust adjustment for confounding covariates in spatial cluster detection.
• To provide a unified framework for various probability models within spatial scan statistics.

Main Methods:

• Developed a generalized linear models (GLM) framework for spatial scan statistics.
• Utilized log-likelihood ratio test statistics for hypothesis testing.
• Employed Monte Carlo hypothesis testing for evaluating statistical significance.
• Applied the method to Texas female breast cancer data.

Main Results:

• The GLM approach facilitates straightforward covariate adjustment in spatial scan statistics.
• Demonstrated the flexibility of the GLM framework across different probability models.
• Successfully applied the method to analyze spatial patterns of breast cancer stages, considering race/ethnicity and age.

Conclusions:

• The proposed GLM-based spatial scan statistic offers improved covariate adjustment capabilities.
• This method provides a flexible and unified framework for spatial cluster analysis.
• The approach is valuable for epidemiological studies requiring control for confounding variables.