RTI uses cookies to offer you the best experience online. By clicking “accept” on this website, you opt in and you agree to the use of cookies. If you would like to know more about how RTI uses cookies and how to manage them please view our Privacy Policy here. You can “opt out” or change your mind by visiting: http://optout.aboutads.info/. Click “accept” to agree.
Confidence intervals for population attributable fractions using complex survey data
Vaish, A., & Khavjou, O. (2017). Confidence intervals for population attributable fractions using complex survey data. In Proceedings of the 2017 Joint Statistical Meetings (pp. 1351-1357). American Statistical Association. http://www.asasrms.org/Proceedings/y2017/files/593887.pdf
This paper describes an easy-to-implement method for producing confidence intervals (CIs) for population attributable fraction (PAF) statistics using survey data. The PAF is used by epidemiologists and policymakers to assess how much of the disease burden in a population could be reduced if the exposure to certain risk factors were eliminated. The PAF is defined as p(rr-1)/rr , where p denotes the proportion of cases exposed to a risk factor and rr denotes the model-based relative risk comparing the proportion of cases among the exposed group with the proportion of cases among the unexposed group. The rr is obtained by modeling the log of the prevalence of the disease as a linear function of covariates where exposure to the risk factor is included as one of the model covariates. The proposed methodology is based on the Taylor series linearization and properly accounts for survey design features in estimating the variances and covariances of the estimated quantities. The methodology is implemented using the VARGEN procedure of SUDAAN® version 11.0 software on 2013 Behavioral Risk Factor Surveillance System data to produce state-by-age group PAF estimates and CIs of hypertension with diabetes as the risk factor.
