Doctors overestimate survival times for terminal patients

February 17th, 2000

Doctors who refer terminally ill patients to hospice care are systematically overoptimistic, according to a study by University of Chicago researchers published in the February 19, 2000, issue of the British Medical Journal. Physicians, on average, predicted that their dying patients would live 5.3 times longer than they actually did. In only 20 percent of cases were the doctors' predictions accurate.

Such prognostic inaccuracy may lead to unsatisfactory end-of-life care. Patients are left to make clinical, social, and financial decisions based on inaccurate information. For example, referral to hospice or other forms of palliative care is delayed, resulting in prolonged emphasis on aggressive but futile medical treatment, insufficient pain control, unnecessary expense, and decreased patient and family satisfaction.

"Achieving a good death, one that is consistent with the patient's wishes, requires some advance warning and that is just not happening," said Nicholas Christakis, MD, PhD, associate professor in the departments of medicine and sociology at the University of Chicago and director of the study.

"Physicians are superbly trained in diagnosis and treatment but they know less about, ignore, and often actively avoid prognosis. They can't or won't make predictions about a patient's future and as a result many patients die deaths they deplore in locations they despise."

In the first large, prospective study of this issue, Christakis and Elizabeth Lamont, MD, a fellow in the Robert Wood Johnson clinical scholars program at the University of Chicago, followed the progress of every patient enrolled at five outpatient hospices in Chicago during 130 consecutive days in 1996.

As soon as they were notified about the arrival of a new patient, the researchers contacted the referring doctor to conduct a four-minute telephone survey and elicited the physician's best guess as to how long that patient would live.

Then they followed each patient's progress until death. They collected data regarding 343 different physicians and 468 patients who had died by June 30, 1999.

When an accurate prediction was defined as anywhere between one-third less to one-third more than actual survival, 63 percent of prognoses were overestimates, 20 percent were correct, and 17 percent were underestimates.

If that definition was relaxed to include any predictions ranging from one-half to two-times actual survival--for example, guessing anywhere from one week to four weeks for a patient who survived two weeks--most doctors were still overly optimistic. In 55 percent of the 468 cases, the doctors predicted that their patients would live more than twice as long as they really did.

Actual survival averaged only 24 days. Three months is considered ideal for hospice care.

Analysis of the doctors provided few clues about who develops prognostic skills. Surgeons did somewhat better than doctors in general practice, who did slightly better than cancer specialists, who did better than other specialists. But neither board certification, self-rated optimism, number of recent hospice referrals, nor number of medically similar patients in the past year separated the precise prognosticators from the Pollyannas or the pessimists.

Surprisingly, the better the physician knew the patient, the more likely he or she was to err. "Physicians," explains Christakis, "do not want to believe that a patient they know well is going to do poorly."

The fact that the errors were consistent, tending toward rampant optimism, suggests that some improvements may not be that difficult. Disinterested doctors, with less contact with the patient and less personal involvement may provide more accurate prognoses and could be called in to provide predictive second opinions.

But the real problem lies deeper than that, says Christakis, who recently published a book lamenting physicians' inability to provide accurate prognoses stressing the importance of accurate predictions (Death Foretold: Prophesy and Prognosis in Medical Care).

"Prognosis is an essential part of medicine," he insists, "in many cases just as important as diagnosis and treatment."

"How long have I got, doc?" is not an unusual question from a patient. General practitioners hear it an average of six times a year and oncologists face that question 100 times a year. Yet the topic--rarely even indexed in textbooks--has been the subject of less than four percent of published studies and is rarely stressed in medical training. In fact, physicians are often taught not to make predictions, but to focus instead on providing hope.

Providing the right kinds of hope can be beneficial and comforting, noted Christakis, yet too much optimism near the end of life may mean patients never see the end coming, never prepare for it, and fight vainly against it. "At some point," suggested Christakis, "patients might benefit more from having their doctors focus on the hope for a good death."

Terminal patients tend to have frequent contact with the medical profession for months, he said, yet patients and their families complain of being uninformed about appropriate end-of-life arrangements. They often have dying experiences that are, "to put it mildly, suboptimal in fundamental ways," said Christakis.

More than 80 percent of Americans die in healthcare institutions rather than at home. An estimated 40 to 70 percent of dying patients unnecessarily suffer pain, 25 to 35 percent impose significant financial and personal burdens on their families, and 10 to 30 percent express preferences about the dying process that are disregarded by their healthcare providers.

Much of this suffering and expense could be avoided by an honest attempt to provide each patient with an accurate prognosis. Death is a normal and unavoidable life passage, notes Christakis. Physicians strive to delay it, but "we need to stop confusing the drive to avert death clinically with the desire to avoid it rhetorically."

Support for this study was provided by the Soros Foundation Project on Death and Dying in America, the Robert Wood Johnson Clinical Scholars Program, and the American Medical Association Education and Research Foundation.

Summary: Doctors who refer patients to hospice care are systematically overoptimistic. They predicted that their dying patients would live 5.3 times longer than they actually did. In only 20 percent of cases were the doctors' predictions accurate. Such prognostic inaccuracy may result in unsatisfactory end-of-life care.