A simple method of determining confidence intervals for population attributable risk from complex surveys

Area of Science:

• Epidemiology
• Biostatistics
• Public Health

Background:

• Assessing uncertainty in population attributable risk (PAR) estimates using confidence intervals (CIs) is crucial for public health research.
• Existing software for complex sample surveys often lacks readily available methods for calculating CIs for PAR.

Purpose of the Study:

• To develop and demonstrate a simple, theoretically valid method for calculating 95% confidence intervals (CIs) for population attributable risk (PAR) in complex sample surveys.
• To enable researchers to routinely provide a population perspective and a valid measure of uncertainty for PAR estimates.

Main Methods:

• A novel method for obtaining CIs for PAR was developed using the Bonferroni inequality.
• The method was validated through simulation in a (2 x 2) table and applied to a cohort study calculating CIs for the PAR of coronary heart disease death using proportional hazards regression.

Main Results:

• The developed method provides a straightforward and theoretically sound approach to calculating CIs for PAR.
• The Bonferroni inequality-based method effectively generates CIs for PAR estimates in complex survey data.

Conclusions:

• This method enhances the ability of researchers analyzing complex surveys to quantify the uncertainty associated with PAR estimates.
• The findings support the routine use of this method to provide a population perspective and reliable uncertainty measures in epidemiological studies.