Big Data and End-of-Life Care: Promise and Peril

There is no controversy in the United States about the facts. Too much money is spent doing too many things to people they do not want when they are within months of dying. Hospice care, the comprehensive team-based provision of palliative care to patients and families, has been proven to be nearly twice as good as hospital, home health, or nursing home care during this period. ¹ Therefore, the integration of palliative care into standard care, and the timely referral for hospice care should be a quality issue like handwashing, vaccination, and antibiotic use.

There are two systematic sources of error that prevent that from happening. First, physicians are 300–500% more optimistic about prognosis than observed. ² Second, physicians, nurses, patients, and families are systematically overoptimistic about the benefit of a particular treatment for a particular treatment—a phenomenon termed the “therapeutic illusion.” ³ Clearly, there needs to be an evidence-based countermeasure to all this optimism if patients are to get access to best possible care.

Big Data, the slang term for analyzing heretofore unmanageable amounts of data to gather meaningful information, is now here. Health systems and health plans have piles of data about individual patients. The promise of analyzing such data is to produce patient-specific information that will help guide the provision of the right care to the right person at the right time. An example of that appears in this issue. ⁴ There are both promises and perils of this approach.

The promise is obvious. Everyone would like a sign when the last months are approaching. Very few advocate for doing futile care. But even the best predictive modeling is only the equivalent of a laboratory test or an imaging study. It must be interpreted by a seasoned clinician with critical thinking skills. That is the peril. It may be shocking to put this in print, but some healthcare professionals do not demonstrate critical thinking skills. Imagine such a person walking into the patient's hospital room, or into the office examination room with eyes on a computer screen to say, “The computer says you'll be dead in six months—its time to quit.” Physicians have been pulling their hair out for years over clinicians who only look at chest radiographs to diagnose pneumonia or laboratory results to diagnose kidney failure. But this takes it to a whole new level.

But there is one greater peril, in our view. Most of the efforts at developing approaches to mine Big Data for prognostic information is being done by private, proprietary, for-profit companies. In contrast to the article in this issue, there will be no general scrutiny in the public view. We will not know whether the way they are using the tool is accurate, or whether the drive to reduce cost overshadows the accuracy of the information.