Attributable fractions

Question 1

Attributable risk

• Formula
• Interpretation

Answer 1

• Risk1 - Risk0

- If exposure disappeared, x cases for every 100/1000 exposed.

Question 2

Attributable fraction

• Formula
• Interpretation

Answer 2

• Risk1-Risk0/Risk1

- Among exposed, % of cases are attributable to the exposure.

Question 3

Population attributable risk

• Formula
• Interpretation

Answer 3

• Risk_tot-Risk0
• Attributable Risk * P(Exposure)
• Among 100/1000 population, there are x excess cases attributable to exposure.

Question 4

Population attributable fraction

• Formula
• Interpretation

Answer 4

• Pop Attributable Risk/risk_total
• Risk_tot-Risk0/Risk_tot
• [P(Exp)(RR-1)]/[P(Exp)(RR-1)+1]

Question 5

Population attributable fraction in case-controls (OR), or when we have an adjusted RR?

Answer 5

= P(CE) x (RR-1)/RR
where P(CE) is fraction of cases that are exposed

Question 6

Assumptions of population attributable risk

Answer 6

1. Exposure must be causally related to the disease
2. Plausible intervention exists to achieve immediate and sustained reduction in exposure from the level in the exposed group to the unexposed group
3. Intervention has no other effects aside from reducing exposure

Question 7

Link between crude RR and the PAF in case of confounding?

Answer 7

Opposite direction

Question 8

Direction of bias of PAF when the exposure-confounder association and the disease-confounder

Answer 8

PAF is underestimated

When the directions of confounder associations are opposite, the PAF is overestimated

Question 9

How to stratify for confounding in PAF?

Answer 9

= (% of cases in stratum A * PAF_stratum) + (% of cases in stratum B * PAF_stratum)