All Content

End-of-Life Predictions


The July 26 New England Journal of Medicine contained several articles of interest for hospitalists: A trial of steroids for bronchiolitis, mortality associated with type B aortic dissections, cardiovascular outcomes in patients using rofecoxib, implications of our social networks on obesity, and a terrific review of methicillin-resistant Staph aureus in soft-tissue infections.

But all these were trumped in the media by Oscar the cat.1

National Public Radio, CNN, Fox, and the BBC all covered the story about the cantankerous cat that could predict death in a community nursing home’s third-floor dementia unit with Grim Reaper-like proficiency. The 2-year-old cat was adopted and raised from kittenhood by staff members at the Steere House Nursing and Rehabilitation Center in Providence, R.I.

Nursing staff began noticing that Oscar would curl up in bed next to patients who had only hours to live. Oscar was so adept at predicting death that the nursing staff began using his prophesies to alert family members of an impending demise. Oscar rounds daily, waiting for closed doors to open when necessary and employing somewhat less-sophisticated methods than his human colleagues—namely observing and sniffing. However, this “cat scan” appears to serve as an unprecedented augur, accurately predicting the death of 25 nursing home residents.

Having a strong, long-lasting relationship with a patient appears to be an important predictor of inaccurate prognosis.

Why all the fuss about this predictive puss in boots? To be sure it’s an intriguing story. But why does the public care so much that Oscar the cat can now be found in Wikipedia—and why should the average hospitalist care about a prognosticating cat?

The answer lies in the medical profession’s dreadful lack of ability to predict patients’ survival at the end of life. Multiple studies have shown physicians are terrible at predicting survival, most often in an overly optimistic manner. A paper in the British Medical Journal examined the ability of 343 doctors to estimate the survival of 468 patients at the time of referral to hospice.2 Only 20% of physicians were able to accurately prognosticate the time of death to within 33% of the actual survival. While median survival was 24 days doctors, on average, predicted patients would live 5.3 times longer than they did. Interestingly, the longer the doctor knew the patient the more likely they were to predict wrongly.

Most seriously ill patients value prognostic information more than they do information about therapeutic options. It is a central determinant in how patients and families make end-of-life decisions. The fact that we seem to systematically provide overly optimistic predictions has several important downstream effects.

Unduly optimistic predictions likely delay referral to palliative care providers and hospice. Doctors generally believe patients should receive hospice care for several months before death. However, patients typically receive this care for less than a month and in many cases only a few hours. Our inability to realistically predict survival most likely plays into this discrepancy and negatively affects our patients’ quality of life.

Second, our consistently overestimated predictions likely translate into patients making inappropriate or counterproductive end-of-life decisions.

We’ve all experienced the patient who continues to push for obviously futile care despite our attempts to nudge them toward a palliative approach. Yet it’s easy to see why patients who believe they have six months to live would push for more aggressive treatments. If they knew they had six weeks to live, they might choose more palliative options.

Finally, hospitalists are increasingly under the gun to decrease length of stay (LOS) and contain costs. It is estimated that more than 25% of Medicare costs (approximately $88 billion a year) are associated with care in the last year of life, much of it in the hospital. While several studies have shown that early palliative care intervention reduces LOS and cost of care, overly optimistic predictions make it less likely these timely referrals will be made.

The development of the hospitalist model, with its fractured approach to care, might provide an opportunity to improve end-of-life prognostication. Having a strong, long-lasting relationship with a patient appears to be an important predictor of inaccurate prognosis. It isn’t hard to imagine that a provider with a long relationship with a patient might have a more difficult time acknowledging a patient’s poor outcome or unwittingly be less likely to squelch a patient’s hope with a poor prognosis.

We can take advantage of the inherent discontinuities of the hospitalist model as well as the severity and immediacy of the patient’s acute illness to not only proffer an end-of-life prediction but contextualize it for the patient. This, of course, needs to be done in a sensitive manner that recognizes our brief role in their care and, as such, is most often best managed in concert with the patient’s primary care provider.

The promise of the hospital medicine movement is that we can do it better and cheaper. This is a tall order indeed. While the hospitalist model has improved efficiency and quality, future improvements will require us to adopt and develop new efficiencies and better systems of care.

It is in this vein that I believe we can and should be able to improve end-of-life care. An early salvo in this front can be a dedicated and systematic push to improve end-of-life prognostication and its myriad downstream effects. This will require a conscientious effort, more formal education, and better predictive tools. As hospitalists, we are perfectly positioned to lead these efforts. The other alternative may be to adopt more cats into our multidisciplinary team. TH

Dr. Glasheen is associate professor of medicine at the University of Colorado at Denver and Health Sciences Center, where he serves as director of the Hospital Medicine Program, Inpatient Clinical Services in the Department of Medicine, and the Hospitalist Training Program, and as associate program director of the Internal Medicine Residency Program.


  1. Dosa DA. A day in the life of Oscar the cat. N Engl J Med. 2007;357(4):328-329.
  2. Christakis NA, Lamont EB. Extent and determinants of error in doctors’ prognoses in terminally ill patients: prospective cohort study. BMJ. 2000;320:469-472.

Next Article:

   Comments ()