Correcting for baseline differences in the comparison of rates

Area of Science:

• Biostatistics
• Epidemiology
• Statistical Modeling

Background:

• Comparing event rates across groups often requires adjusting for confounding covariates to avoid bias.
• Traditional stratification methods (e.g., direct standardization) struggle with simultaneous adjustment of multiple confounders, especially with limited sample sizes.
• Stratification can also lead to residual confounding due to the categorization of continuous variables.

Purpose of the Study:

• To present a statistical procedure for comparing occurrence rates across exposure or treatment groups.
• To address the limitations of stratification by employing a more flexible statistical approach.
• To provide a method for adjusting for one or more confounding covariates simultaneously.

Main Methods:

• The study utilizes a multiple logistic regression model for statistical adjustment.
• This approach allows for simultaneous adjustment of multiple confounding covariates.
• A detailed numerical example is provided to illustrate the application of the proposed method.

Main Results:

• Multiple logistic regression effectively adjusts for confounding covariates in rate comparisons.
• This method overcomes the feasibility and residual confounding issues associated with stratification.
• The proposed procedure offers a robust alternative for comparative rate analysis.

Conclusions:

• Multiple logistic regression provides a superior method for comparing event rates when adjusting for multiple confounders.
• The presented statistical procedure is applicable in epidemiological and clinical research for bias reduction.
• A computer program is available to facilitate the implementation of this advanced statistical technique.