Prevention and Communication Research
Ask the Experts:
Interpreting estimates of personal risk

One of the inherent difficulties with risk communication is that risk estimates are based on populations, but individuals are most often interested in information that relates to their personal risk. What is the best way to explain risk, and how can individuals use population-based risk information to improve their personal health?

To many people, the word "risk" connotes something dangerous. As used by medical scientists, however, "risk" is just a synonym for "probability" or "chance." Statistical models based in frequentist probability theory are used to calculate the risk numbers that are found in journal articles, on the front pages of newspapers, and on the evening news.

The word "frequentist" above is important. Risks are computed by counting up the number (the frequency) of occurrences of a certain event over a certain period of time, and dividing this frequency by the number of people at the start of the time period who were "at risk" of experiencing the event. Here's a specific example. Let's say we start with 10,000 men aged 60 years of age, all of whom are heavy smokers and are free of lung cancer at that point in time. We observe that over a 5-year period, 100 of these men develop lung cancer, and 75 of these men die of the disease. The 5-year risk of developing lung cancer for 60-year-old men who smoke heavily would therefore be 100/10,000 (100 men got lung cancer, out of the 10,000 at the start of the time period who could have developed the disease.) Mathematically, 100/10,000 is the same as 1 per 100 or 1 percent. The 5-year risk of dying of lung cancer for 60-year-old men who smoke heavily would be 75/10,000 or 0.75 percent.

There are two important things to note in this brief example. First, a risk always has a time period attached to it (a 5-year risk will be higher than a 1-year risk, but lower than a 20-year risk). Second, the statistical models used to provide risk information cannot say anything about which individuals will be the specific ones to experience the event. In the example above, if we are talking to a 60-year-old man who smokes heavily, we can say with some certainty that "in a group of 10,000 men like you, 100 will get lung cancer in the next five years,” but we have no way of telling this man whether he will be in the group of 100 (out of 10,000) who will get lung cancer, or part of the 9,900 others who will not.

Thus, even though we often use the phrase "personal" or "individual" risk, risks are really just averages, taken over large numbers of people. If someone tells you about your individual risk of breast cancer or prostate cancer, that number is based on observation of a large number of individuals “similar” to you in some respect. The number is not "individual" in the sense that your cholesterol measurement or weight is individual.

This means that individuals have to be smart consumers of risk information. Just as the Dow Jones average is an average, and some people get very anxious and others pay no attention to changes in this average, disease risk estimates are averages. Some individuals will get very anxious about their "individual" risks, while others will not pay any attention to the information. A middle path between these two extremes may be best: risk averages contain relevant information for individuals, but there is a large degree of uncertainty in terms of what these averages mean for you personally. Will you be one of those who develops a particular disease in a particular time period, or will you be part of the group that remains disease-free? Statistical models and risk estimates cannot answer such questions. However, we can choose to use scientific information about behaviors that are associated with higher or lower risks of disease to live as healthy a life as we feel able, knowing that our individual lives are complex and unpredictable and not amenable to statistical modeling. No statistician or scientist can tell us with certainty what our future holds, but just because our personal future is unpredictable does not mean it is completely out of our control.
(Published February 2004)

Dr. Beverly J. Rockhill
Assistant Professor of Epidemiology,
University of North Carolina School of Public Health