Stuart Lipsitz
Department of Biostatistics, Bioinformatics and Epidemiology
Medical University of South Carolina
Charleston, South Carolina
Confidence Intervals for Population Attributable Risk from Complex
Surveys
Population attributable risk (PAR) combines information on prevalence and
relative risk to provide an estimate of the proportion of disease in the
population that is attributable to a particular exposure. Methods for
assessing the uncertainty associated with the PAR by calculating 95%
confidence intervals (CI) are not currently available for data from complex
surveys. In this talk, we provide a simple, theoretically valid method to
obtain confidence intervals for PAR based on the Bonferroni inequality.
The method is demonstrated using the cross-sectional National Health and
Nutritional Examination Survey (NHANES) III and the NHANES Epidemiologic
Follow-up Study (NHEFS), a prospective cohort (1971-1992). Based on our
proposed method, researchers analyzing complex national surveys should be
able to routinely assess uncertainty in their PAR estimates by including
confidence intervals in published work.
Back to Colloquium Series