Misspecified poisson regression models for large-scale registry data: inference for 'large n and small p'

Area of Science:

• Epidemiology
• Biostatistics

Background:

• Register-based epidemiology commonly uses Poisson regression to link exposures to event rates.
• Large sample sizes ('large n') with numerous covariates ('small p') present unique statistical modeling challenges.

Purpose of the Study:

• To explore modeling strategies for time-varying covariates and their effects in large-scale epidemiological studies.
• To address issues of over-significance in hypothesis testing and missing confounder information.
• To propose methods for enhancing the reliability of conclusions from potentially misspecified models.

Main Methods:

• Discusses sensitivity analysis techniques.
• Explains estimation of average exposure effects using aggregated data.
• Introduces a semi-parametric bootstrap for robust standard errors.

Main Results:

• Methods are illustrated using Danish national registry data on antipsychotic treatment and diabetes incidence.
• The study demonstrates approaches to handle statistical complexities in real-world epidemiological data.

Conclusions:

• The proposed methods aid in drawing more reliable conclusions from epidemiological studies with large datasets and complex covariate structures.
• Sensitivity analysis and robust error estimation are crucial for valid inference in register-based epidemiology.