Attributable Risk

The attributable risk statistic provides an estimate of the proportion or number of events that can be explained by a particular risk factor. Epidemiologists frequently use attributable risk calculations to determine the population impact associated with a disease, [Page 57]behavior, or condition. The U.S. Surgeon General's estimate that smoking accounts for up to 400,000 deaths annually in the United States is an example of an attributable risk inference.

In the context of cohort studies, attributable risk is also referred to as risk difference, in this case quantifying the excess risk in exposed versus unexposed groups. Attributable risk can be calculated in several ways. When accompanied by a relative risk statistic, the attributable risk is equal to the rate of events in the unexposed group × (relative risk-1). A more widely applicable formula requires the creation of a 2 × 2 contingency table, as illustrated in Table 1 with a hypothetical sample of smokers and nonsmokers.

In Table 1, A through D represent the number of cases in each study cell. The event rate among smokers (3,000/10,000) is three times higher than among nonsmokers (1,000/10,000). Half the sample is composed of smokers, and 20% develop cancer over the course of the study. Attributable risk uses both the group event rates and the prevalence of the exposure (smoking) for calculation purposes. The attributable risk value of .20 tells us that if smokers in the study became nonsmokers, the incidence of lung cancer would decrease by 20 per 100 individuals. This represents a potential 66% decrease in lung cancer cases.

Important facts about attributable risk:

• The attributable risk statistic alone does not imply that a causal relationship exists between the exposure factor and the event.
• Because attributable risk uses information about exposure prevalence, the attributable risk values between two exposure factors that each double the risk of an event such as cancer can differ dramatically if one exposure (e.g., working in coal mines) is much more rare than a more common exposure (e.g., smoking).
• Attributable risk can be used to calculate the population attributable risk by use of the following formula: attributable risk × rate of exposure in the population.
• The proportion of events potentially eliminated in a population by changing the exposure rate to that of the unexposed group is often referred to as the attributable proportion.
• By combining attributable risk with the financial costs of a health event, researchers can estimate the health care expenses associated with an exposure factor and calculate the health care savings achievable by modifying a risk factor in a population.