Dr. David Casarett, chief medical officer at Penn Wissahickon Hospice, says there are some conversations that even the most seasoned doctors dread -- letting patients know how much time they have left to live.
Oncologists and palliative care physicians often work with cancer patients for months or years and grow close to them -- going to their granddaughter's wedding or exchanging birthday and anniversary cards. If treatment fails, those doctors then have to tell their patients the wrenching news.
"That's one of those times when I stop and take a really deep breath before I open that door," Casarett said. "Those are the discussions ... they are very difficult."
That difficulty may be at the heart of what experts describe as a tendency for health care providers to miscalculate how long patients with advanced, incurable cancer have left.
"Health care providers and physicians are really pretty bad at estimating prognoses," Casarett said, explaining that part of the issue is that there is no system of feedback in the same way there is if they miss a diagnosis.
"In general, not only are we not accurate, but we tend to be wrong in a more optimistic way, so we say six months to live when a person only has one," he added.
The implications of this are far-ranging. Without a clear sense of timing, patients often opt for aggressive, late-stage cancer treatments, getting chemotherapy just weeks before their death. Additionally, they are often left with too little time to prepare emotionally and consider their legacy.
But a new scoring system out of the U.K. is aimed at changing all that.
In a report published in the British Medical Journal, researchers introduced the Prognosis in Palliative Care Study (PiPS) model, which allows providers to plug factors -- including recent weight loss and current mental status -- into a computer interface to better gauge how long people with advanced cancer have left to live. A second model, dubbed "PiPS B" allows clinicians to also plug in blood test results, though such data is not always available in a clinical setting.
The researchers found the first model was at least as successful as clinicians at predicting whether patients had days, weeks or months left to live. And when a blood test was included, the test was significantly more accurate than any single doctor or nurse's prediction. This prompted the study's authors to conclude that the models have some definite advantages over existing ones, including clinicians' estimates.
"I should stress that it is not intended that patients should access this website themselves, since it is important that prognostic scores are not considered in isolation and such information should always be interpreted and sensitively communicated by a suitably trained health care professional," Dr. Patrick Stone of St. George's University of London, one of the study's authors, said in an email to HuffPost.
Indeed, in an accompanying editorial, Dr. Paul Glare, chief of Palliative and Pain Care Service at the Memorial Sloan-Kettering Cancer Center, said the way in which a doctor communicates a prognosis is just as important as actually getting it right. Many experts said they agree.
"Clinicians need better training in how to approach these conversations," said Dr. Tracey O'Connor, an assistant professor in the department of medicine at the Roswell Park Cancer Institute. "They take a great deal of social skill, and often they have gone through programs where they haven't been adequately prepared."
O'Connor said she thinks that training has improved in medical schools, but added that oftentimes physicians -- herself included -- only learn how to relay bad news by watching other doctors.
The authors of the new study say there are limitations to their research, describing it as a "first step" in the "incremental process of improving prognostic accuracy." They explain it needs to be adapted to a variety of platforms before it is practical to use at the bedside.
Outside experts like Glare said that though the new models are promising, he is still slightly disappointed that they are only as good as clinicians' predictions.
"We need to start thinking outside of the box and think of novel factors [researchers] could include in models to improve them, and get better than clinicians," he said.
In the meantime, physicians must balance the desires of patients and families to know how much time is left with the knowledge that end-of-life prognosis is an inexact science. They must also try to keep personal attachments and optimistic biases from coloring their judgment and accuracy.
"Prognosticating is very important because it allows us to get people into palliative care earlier when they get more benefits," O'Connor said. "Too often, when we refer them to hospice, they are already bed-bound, they haven't had time to prepare emotionally or legally -- to write what they want and say what they want. If we don't have accurate prognoses, we are not allowing people to prepare for death."