Population Attributable Fraction

The proportion of disease cases in an entire population that would not occur if a specific exposure were eliminated, accounting for its prevalence.

Source: Levin 1953

The population attributable fraction is the share of all disease cases in a whole population that would not occur if a harmful exposure were removed, taking into account how common the exposure is. It reveals, for public health, how much disease could in principle be prevented by eliminating an exposure across the population, not just among the exposed. A common exposure with a modest effect can carry a larger population fraction than a rare one with a strong effect, which guides where broad prevention pays off. Population-wide preventable burden is what it shows. Rothman and colleagues (2008) describe this measure.

Source: Rothman et al. 2008

The population attributable fraction is calculated by combining the relative effect of the exposure on the disease with the prevalence of the exposure in the population, giving the proportion of total cases attributable to the exposure. A common form uses the exposure prevalence and the relative risk to derive the fraction. So the population attributable fraction is calculated from both how strongly the exposure raises risk and how widespread it is, since a harmful but rare exposure accounts for few cases overall, which is why prevalence enters the calculation alongside the strength of the association to give the exposure's impact across the whole population.

Source: Levin 1953

The population attributable fraction is important because it conveys the public health significance of an exposure by showing how much of the total disease burden in a population it accounts for, which depends on both its harmfulness and its prevalence. A modestly harmful but widespread exposure can cause many cases. So the population attributable fraction is important for prioritising public health action, since it indicates where removing an exposure would prevent the most disease across the population, combining the strength of the association with the exposure's frequency, information that relative measures alone do not provide and that guides where prevention efforts would have the greatest impact.

Source: Levin 1953

The population attributable fraction is the proportion of cases in the whole population, exposed and unexposed together, attributable to the exposure, while the attributable fraction in the exposed is the proportion of cases among the exposed only that is due to the exposure. The population measure also incorporates the exposure's prevalence, so a rare exposure yields a small population fraction even if it strongly affects the exposed. So the two differ in the group considered and in whether prevalence enters, with the population attributable fraction reflecting the exposure's overall impact across the population and the exposed fraction its impact among those exposed.

Source: Levin 1953

The limitations of the population attributable fraction include that it assumes the association is causal, so it misleads if confounding or bias distorts the estimate; that it depends on accurate estimates of both the effect and the exposure prevalence; and that fractions for multiple exposures need not sum to one hundred per cent, since causes overlap. So the population attributable fraction is interpreted with care, relying on the exposure being a genuine cause and on sound estimates, since treating a non-causal association as attributable overstates the preventable burden, and its value as a guide to prevention depends on these assumptions holding in the population considered.

Source: Levin 1953