We have seen that relative and absolute measures of association can provide slightly different information and serve different purposes. Consider for a moment that you are the minister of health and you can find only one of the following interventions. The first would replace an industrial chemical substance that has been found to cause cancer with a safe one. The second would completely remove a food additive that has also been found to cause cancer. Which one would you fund? This is evidently a very difficult question. In this lecture, I will introduce the concept of attributable risk, which might help you make such decisions. Attributable risk is a measure of the public health impact of an exposure on the exposed group. In other words, it quantifies the answer to the question, "If we remove the exposure, how much would the disease burden be reduced?" This information will be critical in prioritizing public health interventions. This might sound familiar to you and rightly so. Attributable risk is essentially the risk or incidence rate difference. When I speak of attributable risk though, instead of risk difference or incidence rate difference, I imply that there is a causal relationship between the exposure and the outcome. I also assume that there are no other sources of bias and the distribution of all other known and unknown factors that influence risk is the same in the exposed and the unexposed. Let's go back to the example we've used before to calculate risk difference. We have two groups of students who are all awake at 9:00 am. One group is exposed to the boring lecturer and the other group isn't. The risk of sleeping in the classroom in the exposed group is 0.4, and in the unexposed group, 0.2 over one hour. We assume that both groups are similar and importantly, that the boring lecturer is causing students to sleep. However, we should be fair and not blame him for all the sleeping that takes place in the classroom. Clearly, there is some incidence of the disease that is not due to the exposure. The students in Classroom B fell asleep without any exposure to the boring lecture. In general, the incidence of a disease in the unexposed group is also called background incidence and cannot be attributed to the exposure. The incidence of a disease in the exposed group is the sum of this background incidence plus the incidence due to the exposure. In the case of the classroom, the background risk is the risk among the unexposed, 0.2. The attributable risk is the risk difference. The risk among the exposed minus the risk among the unexposed, 0.4 minus 0.2 equals 0.2 over a one-hour period. In other words, a risk of 20 percent over a one-hour period among the exposed students can be attributed to the boring lecturer. Another concept linked to the attributable risk is the number needed to treat, which is the inverse of the attributable risk. In this example, where the attributable risk was 0.2, the number needed to treat was one over 0.2 equals five. This means that five people need to receive the intervention, staying away from the boring lecture to prevent one case of the disease. The number needed to treat is very relevant when testing the effectiveness of health interventions and treatments. Attributable risk can be expressed as a percentage, in which case, we call it attributable risk percent. Attributable risk percent is the proportion of disease among the exposed which can be attributed to the exposure, and therefore, could be prevented by eliminating the exposure. To calculate it, we divide the difference between the risk among the exposed and the unexposed by the risk among the exposed. Alternatively, we can use a simple formula. Risk ratio minus one over risk ratio. We can multiply by 100 to get a percentage. Instead of risk, we can also use incidence rate. The idea is the same. Looking back to the same classroom example, the attributable risk percent is 0.4 minus 0.2 and this divided by 0.4, which is 50 percent. The conclusion is that 50 percent of the cases of sleeping among the exposed can be attributed to the boring lecturer. Attributable risk and attributable risk percent are quite easy to calculate. They can be really helpful when you need to consider the effect of an exposure among the exposed group, which is something that happens all the time in public health.