Application and interpretation of attributable fraction measures from clinical practice to policy: a narrative review
10.5124/jkma.26.0040
- Author:
Sue Kyung PARK
- Publication Type:Continuing education column
- From:Journal of the Korean Medical Association
2026;69(5):411-418
- CountryRepublic of Korea
- Language:Korean
-
Abstract:
Attributable fraction (AF) measures, including AF and population attributable fraction (PAF), quantify the potential reductions in disease burden if specific risk factors were eliminated. Although relative risk describes the strength of association, it does not directly address the clinically and policy-relevant question of how much disease is preventable. This article presents an integrated framework for interpreting AF and PAF across clinical and population contexts, focusing on applications to cancer prevention.Current concepts: AF, defined at the level of exposed individuals, helps explain causal contribution and the preventable proportion in clinical settings. PAF that is based on exposure prevalence among those with the disease (Pc) reflects the causal composition of disease among patients, whereas PAF based on exposure prevalence in the population (Pe) estimates potential impacts of population-level interventions. These measures are grounded in a counterfactual framework and require causal interpretation. This review presents key formulations, including Levin’s and Miettinen’s approaches, together with their underlying assumptions. By distinguishing target populations and question types—causation, prevention, and prioritization—AF and PAF provide complementary insights at the individual, patient group, and population levels.Discussion and conclusion: AF measures offer a unified framework linking causal inference with clinical decision-making and public health policy. AF supports patient-level counseling and intervention prioritization, whereas PAF informs population-level prevention and resource allocation. Appropriate interpretation requires careful consideration of causal assumptions, target populations, and the presence of multiple risk factors. Overall, AF and PAF are essential tools for translating epidemiologic evidence into actionable clinical and policy decisions.
