Adjusting for population shifts and covariates in space-time interaction tests

Area of Science:

• Epidemiology
• Biostatistics
• Spatial Analysis

Background:

• Statistical tests for epidemic patterns rely on space-time clustering measures.
• Standard methods like Knox's test are biased by population shifts and background variable interactions.
• Existing variants, such as Kulldorff and Hjalmars' test, also exhibit bias.

Purpose of the Study:

• To address the limitations of current statistical tests for epidemic pattern analysis.
• To develop an unbiased method for detecting space-time event clustering.
• To account for population shifts and covariate effects in epidemic studies.

Main Methods:

• Proposes an alternative approach to Knox's test that conditions on observed spatial and temporal margins.
• Utilizes Metropolis sampling of permutations for unbiased analysis.
• Incorporates controls for covariate clustering effects.

Main Results:

• Demonstrates that Kulldorff and Hjalmars' variant of Knox's test is generally biased.
• The proposed alternative method is shown to be unbiased under more general conditions.
• The new approach effectively accounts for exposure shifts and covariate clustering.

Conclusions:

• Existing statistical tests for epidemic patterns, including variants of Knox's test, are susceptible to bias.
• The novel Metropolis sampling approach offers an unbiased and more reliable method for space-time cluster detection.
• This method enhances the accuracy of epidemiological surveillance and outbreak detection.