People think positive tests in screening are more predictive of disease than they actually are

What is it about?

I examine whether patients and doctors realize that a woman with no symptoms of breast cancer, who takes part in a mass mammography screening program and tests positive, is much less likely than another woman who had a positive mammogram following symptoms, to actually have breast cancer. (Mass mammography screening means screening available to all women above a certain age). This is because the rate of cancer is much lower in a mass screening population; relevant rates of breast cancer are about 40 in 10,000 women in a screening population, versus about 1,000 out of 10,000 women with symptoms. In other words, very few women in a mass screening population can get a true positive, while the number of false positives adds up among all those tested, even when the probability of a false positive is low. In a population of women with symptoms, there are a lot more women who can get a true positive and fewer who can get a false positive, and so the chance that a positive is a true positive is higher. In fact, about 80% of women with symptoms and a positive mammogram actually have breast cancer; less than 20% of women with no symptoms and a positive mammogram in mass screening actually have breast cancer. I found nearly half of MBAs surveyed wrongly believed that the probability of cancer for the two women is exactly the same. About 90% of general practitioners knew they were different, but underestimated how different.

Featured Image

Why is it important?

Generally further diagnostic testing is recommended following a positive screening test because many positive screening results are false positives. Unfortunately, sometimes people choose aggressive treatment without waiting for the results of further testing; this is one avenue to ‘overtreatment’. This is true for any kind of screening program where the probability of the condition is very low in the population offered the screening test (although that low probability might be considered ‘high risk’). This research suggests such behavior may be because patients do not understand that the same test, with the same result, can mean different things to different people. This is not an intuitive fact.

Perspectives