Vol. 124 May 1, 2015 Why Can’t Physicians Prognose Better?


“It’s tough to make predictions, especially about the future.
-Yogi Berra


In 2013 seven Italian scientists were convicted of manslaughter and sentenced to six years in prison for not predicting an earthquake. The group of them, the Major Risks Committee (MRC), reassured the populace of L’Aguila which had just experienced a “swarm” (a geo-seismic term) of small tremors that the likelihood of a major earthquake was so small that “no action was needed”. Three weeks later a 6.3 earthquake in the region killed 309 people. Relatives of 29 of the fatalities pressed charges.  Such an outcome demonstrates the dangers of declaring the “unlikely” as “impossible”. In our medical world of close to “zero tolerance for risk” and a “demand for certainty” the risk of incorrect predictions can be daunting.

With all our training, scientific knowledge, and experience you would think that  physicians would be pretty good at predicting survival, aka “making a prognosis”. Several comparison studies have shown that physicians are not really any better than their patients in estimating survival time. Prognosis reliability does vary by diagnosis type and “closeness to the end” (in retrospect, of course, because we can never be sure that we are near the end until the end is here). For example, BOTH physicians and patients are overly optimistic in predicting the time left for lung cancer patients. Physicians’ prognosis of death are correct nearly 90% of the time in non-cancer nursing home patients when death occurs within 7 days later. That percentage drops to 13-16% when actual death occurs 3 – 6 weeks later.  Physicians also tend to be even more overly optimistic than patients about the expected quality of life near the end.

One Harvard explanation of this is as follows:
“Similar to other forecasting experts, physicians face different [non-monetary] costs depending on whether their best forecasts prove to be an overestimate or an underestimate of the true probabilities of an event. We provide the first empirical characterization of physicians loss functions. We find that even the physicians subjective belief distributions over outcomes are not well calibrated, with the loss characterized by asymmetry in favor of over-predicting patients’ survival. We show that the physicians’ bias is further increased by (1) reduction of the belief distributions to point forecasts, (2) communication of the forecast to the patient, and (3) physicians own past experience and reputation.”
 In other words, a physician’s gut feeling is often just “guesswork.”

All of us have heard the story of a patient who was told that he had “a year to live” and the patient went on fishing for cod or bluefish or whatever for another six years. Some of us have suffered the opposite experience. I can not forget the infant boy we diagnosed with hemophilia in the 1970’s and reassured the parents that we, along with modern medical science, could promise him a near-normal life with the use of factor VIII infusions. The patient enjoyed a healthy six years and even learned to infuse himself when only 8 years old. He became one of the first to contract HIV from factor VIII, and he died a teen ager with resistant, intractable pneumonia; a personally wrenching failure of a promise (prognosis) made by modern medicine. Technology and research have corrected that cause, but the memory lingers on.

Studies to identify specific characteristics or elements that could be used to more accurately state a survival prognosis have revealed a mixed bag. None is reliable enough for general clinical use.  If we did have a scientific consensus then the often complex, complicated negotiations of end-of-life care would be a lot easier for both physicians and patients and their families. It would perhaps be less costly for Medicare also.

Of course, despite the decades of study and technological advances we are no better at predicting the actual day a baby is born either … especially within a week of the actual delivery. At least we have “the nine-month consensus” to limit our predictive unreliability for birth dates.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: