Variance estimators for attributable fraction estimates consistent in both large strata and sparse data

Area of Science:

• Epidemiology
• Biostatistics
• Public Health

Background:

• Existing variance formulae for attributable fractions lack consistency in sparse data.
• Individually matched case-control studies often present sparse data challenges.

Purpose of the Study:

• To derive novel variance estimators for attributable fractions.
• To ensure consistency in both sparse data and large strata.
• To extend methods for effect modification and preventive exposures.

Main Methods:

• Employs Mantel-Haenszel estimation for variance estimation.
• Applies conditional maximum likelihood as an alternative.
• Derives extensions for complex epidemiological scenarios.

Main Results:

• Developed dually consistent variance estimators for attributable fractions.
• Demonstrated applicability to individually matched case-control studies.
• Provided extensions for effect modification and preventive exposures.

Conclusions:

• The proposed Mantel-Haenszel based estimators offer improved reliability for attributable fraction variance in sparse data.
• These methods enhance the analysis of case-control studies and related epidemiological research.
• The approach is versatile, accommodating various study designs and exposure scenarios.