The Worst Fortune Tellers
Why doctors are so bad at predicting how long their patients will live.
A year ago, U.K. officials released convicted Lockerbie bomber Abdelbaset Ali Mohmet al-Megrahi to Libya on humanitarian grounds, after doctors predicted that he had just three months to live. The British cancer expert whose opinion heavily influenced the decision, Karol Sikora, recently owned up to his miscalculation, telling a London paper, "There was always a chance he could live for ten years, twenty years. ... But it's very unusual." Lockerbie victims'families blasted the decision to release al-Megrahi, and the issue was placed on a short list of U.S. concerns in a recent meeting between President Obama and new British Prime Minister David Cameron. This week, Sikora acknowledged his flawed assessment to the U.K.'s Observer. "If I could go back in time I would have probably been more vague and tried to emphasize the statistical chances and not hard fact," he said, later adding, "Initially I thought he had 18 months, but when I saw the data, the blood tests and X-ray reports and spoke to the prison doctor, who had observed the pace of the disease, I thought it would be much quicker."
Which raises the obvious question: How did Dr. Sikora get it so wrong?
Doctors are notoriously poor at predicting life span, even when they're dealing with the terminally ill (defined as a prognosis of six months or less). In the case of al-Megrahi, the governments of the United Kingdom and Libya were both interviewing experts to determine how much longer the prisoner had to live. Under normal circumstances, however, it takes some badgering to even get a doctor to offer a prognosis. Doctors prefer not to prognosticate for three reasons: We don't like to be wrong; we don't want to take away hope for survival or good quality of life in the time that remains; and we just aren't adequately trained to do it. And our reluctance to make such guesses means that when we do try to predict the future, we're pretty lousy at it.
Almost every doctor can tell you a story of a patient living months or even years longer than expected. An oncologist colleague of mine, for example, sent a patient with prostate cancer to hospice with "a prognosis of less than six months. … Almost two years later, I saw him at his son's wedding." This was no doubt a joyous occasion. Yet these results are not typical: In the few studies that have investigated physicians' abilities to prognosticate death, doctors usually overestimate survival time by at least twice as much. There are many theories to explain this, chief among them that doctors tend to be overly optimistic with patients they've known and treated for a long time.
Call us actuarially-challenged. Data on survival are available, especially for chronic and well-studied conditions like cancer or heart disease. Al-Megrahi, the Lockerbie bomber, is reported to have metastatic (advanced) prostate cancer, a diagnosis with a median survival of 18 months, but one with a variance as wide as half a decade or more. Since doctors typically avoid making predictions, these tools are infrequently dusted off and put to use. Our collective reluctance to offer patients a prognosis makes us less accurate in the rare instances we actually do it.
In his seminal book Death Foretold: Prophecy and Prognosis in Medical Care, Nicholas Christakis, a medical doctor and sociologist, argues that medical science has given the processes of diagnosis and treatment disproportionate emphasis in the educational curricula of doctors. He examined major medical textbooks that have been used by medical students (and practicing physicians) for decades and determined that the relative percentage of space for each disease entity devoted to prognosis diminished with each subsequent edition, often to a paragraph or less. Furthermore, he writes that avoiding prognosis is a professional norm for doctors at all levels of training. In our research, teaching, and communication, we focus almost exclusively on the ever-expanding sciences of diagnosis and treatment, leaving prognosis almost entirely to the side.
Making predictions about the lives of al-Megrahi and other cancer patients can be particularly tricky. William Dahut, clinical director of the Center for Cancer Research at the National Cancer Institute, blames "a general lack of understanding of the specific biology of the cancer as well as a general lack of understanding of the biology of the individual." Doctors and scientists often refer to an individual's biology as "host factors," making allowances for the fact that patients are indeed different—in immunity, resilience, and attitude. The difficulty in accounting for such differences is another reason that predictive accuracy is so low.
About the only time that doctors seem to accurately be able to prognosticate impending death is near the very end of life. Predictions at this stage range from hours to days. To paraphrase Justice Potter Stewart, a doctor can tell you when a patient is very near death because "we know it when we see it." This usually involves a patient's choice (or a family member's, if the patient is incapacitated) to forgo life-sustaining technology like a mechanical ventilator ("breathing machine").
John Henning Schumann is an internist at the University of Chicago. He blogs at GlassHospital.com.
Photograph of Abdelbaset Ali Mohmet al-Megrahi by Manoocher Deghati/AFP/Getty Images.