Adaptive Fisher's method using weakly geometric grid for combining p-values with application to COVID-19 surveillance

Area of Science:

• Epidemiology
• Biostatistics
• Public Health

Background:

• Effective COVID-19 surveillance requires detecting regional case increases.
• Classical statistical methods struggle with the rare events and varying sparsity of early COVID-19 outbreaks.
• Existing p-value combination methods are limited to detecting either moderately sparse or ultra sparse signals, but not both.

Purpose of the Study:

• To develop a novel nonparametric p-value combination method for COVID-19 surveillance.
• To create a method adaptable to the full spectrum of signal sparsity in early outbreak detection.
• To enhance the detection of significant case increases in regional COVID-19 surveillance.

Main Methods:

• A modified Fisher's method employing a weakly geometric system-based search strategy was developed.
• The method adapts to varying signal sparsity, addressing limitations of existing p-value combination techniques.
• An efficient algorithm was created for calculating the p-value of the proposed method.

Main Results:

• The modified Fisher's method demonstrates theoretical and numerical power across the entire spectrum of signal sparsity.
• The approach is robust in combining approximated p-values, even with a large number of p-values relative to sample sizes.
• In early US COVID-19 surveillance, the method consistently detected outbreaks across regions with diverse sparsity levels.

Conclusions:

• The proposed method offers a novel, powerful nonparametric strategy for COVID-19 surveillance.
• It effectively uncovers diverse COVID-19 spread patterns where competing methods fail due to sparsity limitations.
• This adaptable approach improves the detection of significant case increases in public health surveillance.