Difficulties with regression analyses of age-adjusted rates

Area of Science:

• Epidemiology
• Public Health
• Biostatistics

Background:

• Observational studies frequently compare regional population rates to evaluate policy impacts.
• Age-adjusted rates are commonly regressed on predictors for effect estimation in public health.

Purpose of the Study:

• To identify potential biases in commonly used regression methods for policy effect assessment.
• To propose alternative regression approaches for more accurate effect estimation.

Main Methods:

• The study examines regression techniques applied to age-adjusted and crude population rates.
• Covariance-adjusted estimates derived from age-adjusted rates are analyzed for bias.
• Analysis includes regression of crude rates with age as a covariate.

Main Results:

• Regression of age-adjusted rates often results in biased estimates of the true regression coefficient.
• Analysis of crude rates, incorporating age as a covariate, can yield unbiased effect estimates.
• Alternative regression methodologies are explored for improved accuracy.

Conclusions:

• Standard regression methods using age-adjusted rates in observational studies can be biased.
• Utilizing crude rates with age as a covariate is a more reliable approach for unbiased effect estimation.
• Further consideration of various regression methods is recommended for epidemiological research.