Even Doctors Misread Health Statistics

Is your physician's grasp of the risks of a drug or screening test what it should be?


Like people generally, most patients are not statistics whizzes, but we rightly expect physicians to understand and to be able to place in perspective any numbers that reflect on their patients' health and medical decisions. A distressing report in the journal Psychological Science in the Public Interest argues persuasively that many doctors simply don't, and that anxious patients who make bad healthcare decisions are among the consequences.

An example drawn from the report: a woman who has had a routine mammogram and is told the result is positive. "What does that mean?" she asks her physician. "What are the chances that I have breast cancer?" Her emotional reaction to a high percentage will surely be different than to a low one. In a study of women who had received a false positive on a mammogram, 1 in 8 said the anxiety it created affected their mood and ability to function months later.

The woman's doctor may give her a number, but he might be far off. "Many physicians do not know the probabilities that a person has a disease given a positive screening test," the report states. Last year, before one of the report's authors conducted a training session in risk communication, 160 gynecologists were provided with hypothetical but realistic percentages giving them all the data they needed to respond correctly to the woman's question. They were instructed to choose one of four possible answers. One in 5 of the doctors got it right (10 percent for this made-up case). One in 5 said that it was 1 percent. And the remaining 3 out of 5 said that it was either 81 or 90 percent. (After training, almost all answered a similar question correctly.)

The problem, according to the report, is not just that doctors and patients lack familiarity with statistics. It is also based on how numbers typically are presented in news reports and even in medical journals—as relative percentages or as "doubled" or "threefold" rather than in absolute terms. In 1995, the U.K. government warned in letters to physicians and public health officials that the latest generation of oral contraceptives raised the risk of possibly lethal blood clots in the legs or lungs twofold—100 percent. Many women stopped taking the pill, with abortions and unwanted pregnancies as a consequence. It was true that the risk of the newer pills was twice as great. But the absolute risk of a clot had risen from 1 in 7,000 with the older pills to 2 in 7,000 for the newer ones. A risk, in other words, equal to 1 in 3,500, still a vanishingly low figure.

The report makes compelling reading, and the authors' conclusions are worth pondering. Over the years, many a story has emerged from my clicking keyboard citing the increased risk from one drug as opposed to another, one procedure versus another, and countless other comparisons. I haven't always made the numbers as clear and meaningful as this report demonstrates they should be. I will try to keep it in mind.