Key takeaways:
- End-of-life discussions are much easier when they start as “serious illness discussions” and patients are still stable.
- Online calculators can help you make a prognosis.
- End-of-life signs signal when patients have only days to live.
“A GOOD DEATH,” which is what most people want for themselves and their loved ones, refers to patient- and family-centered end-of-life care that controls symptoms while providing emotional and spiritual support. Because many hospitalists practice in facilities where they don’t have palliative care, they end up being the clinicians who need to help patients have as good a death as possible.
“As hospitalists, we are really well-positioned to identify serious illness and have skilled conversations about what’s next in someone’s care,” said Kencee Graves, MD, chief medical quality officer at University of Utah Health in Salt Lake City. During a recent presentation, Dr. Graves said that when she talks to patients about prognosis, she gives them ranges: hours to days, days to weeks, weeks to months, months to years, rather than exact times. “I also talk about what that prognosis looks like in time and in practice. What is their life going to look like?”
But when should you broach that conversation? And what should you say (or not say) to help steer patients and families toward more comfort-oriented, less aggressive care, if warranted? Dr. Graves and a colleague—Elizabeth Gundersen, MD, associate professor of hospice & palliative medicine at the University of Colorado School of Medicine in Aurora, Colo.—recently spoke about end-of-life care in the hospital in a Society of Hospital Medicine webinar.
The two presenters, who are both hospitalists and palliative physicians, covered how to make a prognosis and how to recognize and manage symptoms in the hospital at end of life.
Make it a “serious illness discussion”
Dr. Gundersen began her portion of the presentation by pointing out that she prefers the phrase “serious illness discussion” to “end-of-life discussion.” That’s because “serious illness”—defined as anything that has an impact on the patient’s quality or quantity of life—reframes when physicians should begin having this conversation.
“End-of-life care can start happening months to years before the actual end of life,” Dr. Gundersen pointed out. “One of my biggest objectives is finding out how to move end-of-life care more upstream so that it’s really serious illness care.”
“When you have a patient being readmitted for a second time in a year, consider that a pivotal event.”
Elizabeth Gundersen, MD
University of Colorado School of Medicine
She illustrated that point by discussing a bread-and-butter case for most hospitalists: a 74-year-old woman who is in the hospital for the first time for heart failure. The patient is responding well to therapy and is relieved to be doing better. But she is also worried about what the future holds.
“Is it even appropriate to have a conversation with her at this stage about what lies ahead?” Dr. Gundersen asked. “I think the answer is ‘yes.’ ”
That’s because, she explained, so much of the end-of-life care that hospitalists deliver isn’t aligned with what patients hope to receive.
According to CDC data released in 2016, 70% of Americans say they want to die at home. In fact, only about 30% of them do, while another roughly 30% die in a hospital and 20% die in a nursing home. Around 8% die in a hospice facility, and 12% die in other places.
“There is so much literature about how many patients would love their care teams to know their end-of-life preferences, but very few actually have those discussions,” said Dr. Gundersen. “That’s largely because these conversations are scary and hard to have.”
As for the trajectories of serious illness, cancer patients tend to have high functional status until they are close to their end of life, when they experience a sharp and sudden decline. Patients who die from progressive chronic illness such as COPD or CHF instead have a slowly-trending downward trajectory, while those with dementia and frailty have persistently low functional status that continues to dwindle over time.
Given those trajectories, “you can’t go wrong having these discussions earlier,” Dr. Gundersen noted. “It’s much easier to have such conversations when things are stable vs. when something urgent or emergent is going on.”
Less than a year?
It’s certainly time to have the serious illness discussion when you think patients may have less than a year to live.
How do you gauge when that is? According to Dr. Gundersen, asking yourself if you’d be surprised if the patient died within a year is a “moderately validated question.” Doctors, she noted, tend to over-estimate prognosis. “So if you think they could die in a year, chances are it’s possible they could die in the next several months.”
She also steered clinicians toward ePrognosis, a Web site maintained by the University of California, San Francisco, that has well-validated prognostic calculators.
Other indications that it’s time to start a conversation about serious illness: when a patient’s condition changes, when patients express concerns or uncertainty, or when your own “spidey sense,” as Dr. Gundersen called it, kicks in.
“We all know when patients are moving in the wrong direction,” she pointed out. “Even before I did my hospice and palliative care fellowship, I was really good at identifying when patients were going to die during their current admission.”
Poor prognosis indicators
Indicators of poor prognosis include comorbidities, infections, malnutrition, weight loss and incontinence. Dr. Gundersen also cautioned physicians to pay close attention to patient readmissions.
“When you have a patient being readmitted for a second time in a year, consider that a pivotal event,” she noted.
“Decreasing patients’ pill burden is actually a big piece of quality of life for a lot of patients when they have only months to live.”
Kencee Graves, MD
University of Utah Health
Declining functional status is another big clue, particularly in cancer patients. (Functional status is defined as someone’s ability to perform life functions including key activities of daily living.)
“Functional status is the most important prognostic indicator in patients with cancer,” she said. “Patients with solid tumor cancers have a 70% decline in their functional status in their last three months of life, so it’s really profound.”
One “quick and easy way” to assess if patients are declining, Dr. Gundersen said, is to ask them or their loved ones how much time per day they spend in a bed or chair.
“If their response is greater than 50% of the time and that time is increasing,” she said, “you can roughly estimate their prognosis at three months or less.”
A 2017 study published in the Journal of the American College of Cardiology that looked at five-year mortality among heart failure patients regardless of their ejection fraction status was, said Dr. Gundersen, “a wake-up call for me.” That study demonstrated that five years after being admitted for heart failure, patients had an almost 80% mortality rate.
“Even more striking,” she noted, “two years after an admission, patients had around a 50% mortality rate. That’s why we should take the opportunity—sooner rather than later—to have these discussions with patients who have chronic diseases.”
You’re not crushing their hope
Part of the reason why physicians shy away from discussing prognosis is they don’t want to take away patients’ sense of hope. Dr. Gundersen pointed to a 2024 study published in Palliative Medicine that looked at data on close to 700 patients with advanced solid cancers.
The outcome: “There was not a statistically significant difference in patients’ hope, regardless of them having a more or less realistic expectation of prognosis,” she said. “Will patients be upset when you share a poor prognosis? Yes, they will. But ultimately, they will still maintain hope, so you don’t have to worry about crushing their spirit when you do share serious news.”
Dr. Gundersen also steered physicians toward another resource: the “Serious Illness Conversation Guide,” which is now being widely used.
“Institutions adopting it have reported that once you’re well-practiced in the technique, you can have these conversations in 15 minutes, particularly when they’re done earlier in patients’ illness,” she said. As Dr. Gundersen explained, the guide focuses on having patients first share their understanding of their health, followed by the clinician’s sharing—with the patient’s permission—their understanding of that patient’s health, then exploring goals and values. Clinicians should then summarize what they’ve learned about the patient and make a recommendation for their care.
Another recommended site: vitaltalk.org. “This is chockful of communication tips,” Dr. Gundersen said. “Many of those help you attend to patient and family emotions so you don’t go into robot mode having these conversations.”
Literature shows, she added, that patients feel most abandoned not when they get serious news, but when they feel their physician has checked out. “Making statements like, ‘Our team is here to support you’ or ‘We will figure out next steps together’ helps align you with patients and families and won’t engender a sense of abandonment.”
She also listed five questions based on the work of palliative care expert Susan Block, MD, that, Dr. Gundersen said, “are as much of a shortcut that we can use with patients that will still give us a lot of information.” Those questions are:
• What have you been told about your prognosis?
• What are you hoping for?
• What worries you most?
• What matters most to you when time gets short?
• What trade-offs are you willing to make for the possibility of more time?
Words count
Dr. Gundersen next steered her listeners to an article published in October 2024 in Mayo Clinic Proceedings by authors including Rana Awdish, MD, an intensivist who almost died from multiple organ failure.
That article, titled “Never-Words: What Not to Say to Patients with Serious Illness,” urges clinicians to steer clear of words and phrases that can frighten patients and sabotage efforts to share decision-making. Instead of saying, “She will not get better,” for instance, a better alternative is, “I’m worried she won ‘t get better.”
Dr. Gundersen noted that she’s training herself to remove the word “want” from these conversations.
“Naturally, patients ‘want’ to get better, so I don’t ask families what a patient might want,” she noted. “If you are considering ventilating a patient or putting in a trach and PEG, I instead say, ‘What would your loved one say or think about that plan?’ I often hear family members say, ‘They’d kill us for doing all that stuff.’ ”
A 2020 article in the Annals of Emergency Medicine on code status conversations with seriously ill older adults with respiratory failure also offered some tips for making care recommendations going forward. When family members ask, “What would you do if he were your dad?,” Dr. Gundersen suggested responding based on what you’ve learned about the patient’s goals and values. An example: “From what you’re telling me about him, I think your dad would find invasive treatments to be unacceptable. Is that right?”
If family members agree with you, that allows clinicians to make “we would recommend” statements about providing either intensive treatments to help the patient recover or intensive comfort care.
Another tip: Don’t just say someone is “sick.”
“We often say, “Your loved one is really sick,” and we expect people to know what we mean by that,” Dr. Gundersen pointed out. The author of a 2019 article in The BMJ coined the phrase “sick enough to die,” which she has been using ever since.
“That really seems to hammer home exactly how sick someone is, and it helps put things in perspective,” she said.
And be careful how you use the word “need,” an insight Dr. Gundersen said she gained from a 2023 JAMA viewpoint. When physicians say, “If your mother’s breathing gets any worse, she will need to be intubated,” that creates the expectation that patients need life-sustaining therapy.
Instead, she reframes such a statement as, ” ‘Your mother’s breathing is getting worse. Can we talk about what this means and what to do next?’ That leaves the conversation much more open-ended.”
Symptoms at end of life
Dr. Graves began her discussion introducing the Palliative Performance Score, a scale that palliative physicians use to determine if the prognosis they feel as a kind of gestalt is accurate or not. That score considers functional status factors like ambulation, activity level, self-care, intake and consciousness level to estimate a patient’s median number of days of survival.
“This is one way you can get a sense of what’s going on,” said Dr. Graves. Recognizing end-of-life symptoms allows doctors to stop treatments and tests that are more burden than benefit. It also helps reduce the unexpected, not only for patients and their families, but also for staff.
“We’ve talked about how most people want to spend their final days at home in their community, but most don’t,” she noted. “If we can identify when someone is declining and we can have that conversation, can we make the hospital more like home if home isn’t an option?”
In terms of ending burdensome treatments at the end of life, Dr. Graves said an article published in 2024 in the Journal of Pain and Symptom Management provides great guidance on when to deprescribe medications. (The authors give recommendations for patients who have six or fewer months to live and for those with three or fewer months.) Nearly 50 experts conferred to reach those recommendations.
“If you have a patient on beta-blockers, for instance, and it’s for mild or moderate hypertension, 90% of the experts in this article agree you should deprescribe it when patients have a prognosis of less than six months to live,” Dr. Graves said. Another table in the guidance tells you what order to stop various medications for patients on multiple prescriptions.
“Decreasing patients’ pill burden is actually a big piece of quality of life for a lot of patients when they have only months to live,” she pointed out.
When a patient has only days to weeks
For patients on comfort care, a host of close to 20 end-of-life signs—including apnea, peripheral cyanosis and inability to close their eyelids—signal that the patient has roughly three days left to live.
As for the prevalence of end-of-life symptoms, Dr. Graves noted that a 2013 systematic review that looked at impending death symptoms within the last two weeks of life found that 57% of the dying patients studied had dyspnea, 52% had pain, 51% had secretions and 50% had confusion.
“Be prepared,” she noted “to deal with these four symptoms.”
Giving opioid infusions at the end of life for dyspnea can be “one of the most emotionally laden things that we do in end-of-life care,” said Dr. Graves. “Are we giving it too soon or too late? And how do we adjust it?”
Fortunately, a 2022 retrospective study in the Journal of Pain and Symptom Management looked at opioid infusions at end of life in close to 200 patients in one academic medical center. It found that nearly half—44%—of those infusions were potentially inappropriate.
“Inappropriate” infusions were defined as those started in patients who were on less than 50 mg oral morphine equivalents in less than 24 hours or those started in patients with fewer than 3 PRN boluses who had a frequency of less than every two hours. Also inappropriate: increasing patients’ infusion rate more than three times in 24 hours.
“Inappropriate opiate infusions,” Dr. Graves said, “can lead to more patient and staff distress.” The bottom line: Avoid continuous opiate infusions unless patients are receiving more than 50 mg morphine equivalents consistently over a few days. Also, utilize frequent PRN boluses (every two hours or less) and take your time assessing the need for an infusion.
What about patients on comfort care who have delirium? A study in NEJM in 2022 looked at the use of haloperidol in ICU patients with delirium, which is “a bit out of this scope,” Dr. Graves pointed out. The trial, which randomized about 1,000 patients to IV haloperidol or placebo, found no evidence that haloperidol increased mortality or length of stay.
“It’s probably OK to continue to use although, in this study specifically, very few patients had underlying cognitive disorders,” Dr. Graves said. “While it’s not exactly what’re we’re using it for, there’s still some evidence that it’s probably OK.”
As for nausea and vomiting at end of life, a 2023 study in the Journal of Pain and Symptom Management that looked at a broad group of patients—including those who did not have cancer—found that olanzapine was helpful.
“That’s some evidence that we can use this in palliative patients as an adjunct, starting at 2.5 mg at bedtime,” said Dr. Graves. She cautioned physicians to be mindful of QTc prolongation as well as the possibility that patients might end up taking two dopamine antagonists, such as olanzapine and prochlorperazine.
Stopping nutrition
One of the most emotional things for families is withdrawing nutrition. “As a society, we see feeding as an act of love,” she pointed out. “I tell families about how it’s really normal for people at the end of life to not want a lot to eat or drink—and let them know that food can actually make symptoms worse.”
As for difficult family dynamics, Dr. Graves noted that no hospitalist or palliative physician can fix families who have 25 years of dysfunction.
“We can make sure that everyone is safe and respectful, and we can set the tone in meetings,” she noted. Sometimes, just communicating with the patient if possible or the person holding that patient’s medical power of attorney is the way to go.
Another tough conversation can crop up between the hospitalist and a patient’s specialists. “Often, patients have really strong relationships with a specialist who has difficulty seeing a patient’s decline,” Dr. Gundersen pointed out. “It’s important to respect their relationship with the patient, so I ask specialists for their perspective on what’s going on. I then offer mine and share with them what I think the next steps should be.”
Phyllis Maguire has been Executive Editor of Today’s Hospitalist since 2006. Based in Bucks County, Pa., her health care interests are hospital medicine and long-term care options. She also likes zydeco, hiking, and reading memoirs and romances.
