Table 6

Eight steps in our proposed new procedure (modified Mokdad) to calculate number of deaths attributable to a risk factor with multiple exposure categories, allowing for one or more neutral categories.

Let ω be the number of deaths attributable to a risk factor (e.g. overweight and obese); and ω_i be the number of deaths attributable to a specific exposure category i (e.g. overweight) of the risk factor.

Using the notations in Table 2,

Using the notations of Allison et al, equation 11 can be expressed as

Using the notations of Mokdad et al, and adding P_q (fraction of population underweight) and RR_q (relative risk for underweight), equation 11 becomes

Eight steps to estimate ω:

Step 1: Obtain M, total no. of deaths in the population during 1 y (from death records).

Step 2: Obtain f₀ [or P(R) or P₀], percentage of individuals in the population nonexposed (e.g. normal weight; never smokers) (from health surveys).

Step 3: Obtain f_i [or P(O_i) or P_i], percentage in separate exposure categories of the risk factor (e.g. overweight or obese; current occasional smokers or current daily smokers) (from health surveys).

Step 4: Obtain f_q [or P(Q) or P_q], percentage in a neutral category (e.g. underweight; former smokers) (from health surveys).

Step 5: Obtain R_i [or RR_i], relative risk of death for each separate exposure category (e.g. former smokers, current smokers) relative to none (e.g. never smokers) (from prospective cohort studies). When R_i is not available and when the event (e.g. death) is rare, relative risk can be approximated by HR (ie, h and q).

Step 6: Calculate PAF_i, population attributable fraction, using equation 11.

Step 7: Calculate ω_i using equation T5.

Step 8: Calculate ω using equation T4.

Choi Epidemiologic Perspectives & Innovations 2010 7:8   doi:10.1186/1742-5573-7-8