Variance estimation of partial population attributable fraction and its applications

The population attributable fraction (PAF) is defined to be the proportion by which the disease prevalence is reduced if the whole population is hypothesized to attain the same risk of disease as the individuals within the unexposed or lowest exposure category. Extensive research has been done to estimate PAF in situations where disease is dichotomous, risk is dichotomous or polytomous, and a stratifying factor is present in the population. Adjusted population attributable fractions do not sum to the adjusted summary attributable fraction for the same exposures. Where several exposure factors are to be considered simultaneously, the partial population attributable fraction (PPAF), originated from game theory, can be used to apportion the summary population attributable fraction for a set of exposure factors of interest.;The variance estimation of the PPAF adjusted for unmodifiables was provided using the delta, bootstrap, and jackknife method. The PPAF estimates give the possibility of ranking the risk factors according to their individual relevance for the disease load within the population under study. With the variance formula for the PPAF, the confidence intervals can be directly available. This will contribute to the policy and program decisions on risk factors of a disease.