Peter Burney1, Cosetta Minelli2. 1. National Heart and Lung Institute, Imperial College, Emmanuel Kaye Building, 1b Manresa Road, London SW3 6LP, UK. Electronic address: p.burney@imperial.ac.uk. 2. National Heart and Lung Institute, Imperial College, Emmanuel Kaye Building, 1b Manresa Road, London SW3 6LP, UK.
Abstract
BACKGROUND: The impact of disease on population health is most commonly estimated by the population attributable fraction (PAF), or less commonly by the excess risk, an alternative measure that estimates the absolute risk of disease in the population that can be ascribed to the exposure. Using chronic airflow obstruction as an example, we examined the impact on these estimates of defining disease based on different "normal" values. METHOD: We estimated PAF and the excess risk in scenarios in which the true rate of disease was 10% in the exposed and 5% in the unexposed, and where either 50% or 20% of the population was exposed. Disease definition was based on a "lower limit of normal", using the 5th, 1st and 0.2nd centile of values in a "normal" population as thresholds to define normality. RESULTS: Where normality is defined by centiles of values in a "normal" population, PAF is strongly influenced by which centile is selected to define normality. This is not true for the population excess risk. CONCLUSION: Care should be taken when interpreting estimates of PAF when disease is defined from a centile of a normal population.
BACKGROUND: The impact of disease on population health is most commonly estimated by the population attributable fraction (PAF), or less commonly by the excess risk, an alternative measure that estimates the absolute risk of disease in the population that can be ascribed to the exposure. Using chronic airflow obstruction as an example, we examined the impact on these estimates of defining disease based on different "normal" values. METHOD: We estimated PAF and the excess risk in scenarios in which the true rate of disease was 10% in the exposed and 5% in the unexposed, and where either 50% or 20% of the population was exposed. Disease definition was based on a "lower limit of normal", using the 5th, 1st and 0.2nd centile of values in a "normal" population as thresholds to define normality. RESULTS: Where normality is defined by centiles of values in a "normal" population, PAF is strongly influenced by which centile is selected to define normality. This is not true for the population excess risk. CONCLUSION: Care should be taken when interpreting estimates of PAF when disease is defined from a centile of a normal population.