# Measuring Incidence Rates or Risks

From a closed cohort, we can estimate a risk or an incidence rate to measure disease occurrence. Calculation of a risk is complicated by the problem of competing risks (see Chapter 4). Because of competing risks, the population at risk will not remain constant in size over time, which means that some people will be removed from the population at risk before they have experienced the entire period of follow-up. Despite this problem, there are many cohort studies in which risks are estimated directly. Usually, the period of follow-up is short enough or the competing risks are small enough in relation to the disease under study that there is relatively little distortion in the risk estimates. In these studies, the risk in each cohort is calculated by dividing the number of new disease events by the total number of people who are being followed in the closed cohort. This approach was used to calculate the risk for cholera in Snow's analysis depicted in Table 5-1. Essentially the same approach was used in the study of vitamin A and birth defects described earlier, although the measure reported is the prevalence, rather than the risk, of birth defects.

It is problematic to measure risk directly in a dynamic cohort, in which new people are added to the cohort during the follow-up period. To get around this problem, the investigator can take into account the amount of time that each person spends in the population at risk and calculate an incidence rate by dividing the number of new disease events by the amount of person-time experienced by the population at risk. The same approach can be applied to a closed cohort, addressing the problem of competing risks.

In the calculation of an incidence rate, the ideal situation is to have precise information on the amount of time that each person has been in the population at risk. Often, this time is calculated for each person in terms of days at risk, although the final results may be expressed in terms of years after converting the time units.