AI and Palliative Care
I recently had a great discussion about palliative care with a grad student. She’s working on her PhD in a health administration field, and she was looking for help on her thesis.
Having worked in the acute care setting for several years, she recognized the utility and benefit that a strong palliative care program could bring to that setting, decreasing costs and improving patient outcomes / experiences. So, she wanted to focus on that topic in her dissertation.
Her professors insisted that all candidates include AI (“artificial” / adaptive intelligence) in their theses, so that’s what she was trying to do: Would an AI predictive tool assist physicians and providers in identifying patients appropriate for palliative care or Hospice, increasing conversations and palliative care utilization?
For those of us who are proponents of Palliative Care and who have been using analytics (and AI) tools to identify patients appropriate for a Palliative Care conversation, the answer is an overwhelming “YES!” It can help to identify patients who might benefit from a palliative care or end-of-life conversation.
But simply identifying the people isn’t the solution, nor will aiding identification solve the larger problems hobbling the appropriate use of Palliative Care or Hospice. The median length of stay in Hospice in the US is 17 days – that too late in the game to glean the most benefit of comfort and palliation.
Even without an AI tool to identify patients for potential conversations about realistic treatment goals, we fail, as a profession, to initiate those conversations with OBVIOUS patients/families until very late in the game, if at all. For physicians and providers, there’s a comfort level issue at play, as well as an “intervene and cure” bias, and both serve as obstacles to this conversation. (Those are topics of a couple of chapters in a book, so I’ll just leave them glossed-over here.)
Add to those problems that acute care is not the best setting for this conversation. There are at least four problems with waiting until a patient is admitted into a hospital to have the sensitive and thoughtful conversations around palliation and end-of-life.
1) Usually, a patient is in a state of emotional duress when they’re sick enough to be in the hospital. Such is not a good time to be making big decisions, and that’s if the patient is even capable to do so, which leads to #2.
2) Family presence and interest in the conversation may not be available. If a patient is unable to speak or decide for themselves, someone needs to be available to step up, preferably someone who’s already the DPoA for Healthcare – unlikely if the conversation has never been initiated before.
3) Since the conversation will be occurring in the throes of acute care of a condition, will the efforts be misinterpreted as “We’re just done with you” or worse? This conversation is usually even more difficult than usual if there’s no relationship or context to the caregiving.
Which leads to probably the biggest challenge of a first-time palliative care discussion in an acute care setting:
4) When a patient is identified as appropriate for discussions around palliation and end of life, WHO will have that conversation?
Let’s agree that using an AI tool to identify patients that would most benefit from palliation and then would automatically trigger patients’ orders – DNR, DNI, Comfort Care Only, etc. – would be unethical (and downright wrong).
So, with that said, if/when a patient is identified as a good candidate for a palliation discussion, who’s going to have that conversation in today’s hospitals?
These difficult topics are best addressed in the context of a trusting relationship. Who is that in today’s acute care setting?
The Hospitalist just met the patient, who has no contact or relationship with the patient’s PCP, and who will go off rotation in 2 days?
That same Hospitalist that only has time to move from one copy-and-paste note to another due to the pressures of corporatized medicine? (My apologies to the conscientious Hospitalists out there, but in my experience lately, you are in the minority.)
A Palliative Care or Hospice Team? They may have more training in how to have these conversations, but they don’t have a foundation of a trusting relationship. They constitute a better option but still fall short of ideal.
For the patient with late-stage COPD, HF, or Cirrhosis, perhaps the specialist managing them outside the hospital would be best for the conversation – if they see patients at that facility. (Not touching cancer here, as that’s a whole travel trunk of baggage to unpack.) Again, these docs could possibly be better than a “cold” Hospitalist if they’re available and they, themselves, are prepared for this type of conversation.
Interestingly, the grad student told me that this was consistent with the responses she’d been getting from every physician and provider she had interviewed. Identifying the patients without the cognitive and comfort infrastructure won’t have any great impact on decision-making, resource utilization, or moves to improve end-of-life care.
We need an education or re-education process for physicians and providers to get them on board with conversations of advanced care planning and end-of-life.
Perhaps we need to frame palliation and comfort care in terms of an intervention in and of itself, rather than abandonment of treatment or giving up. That would fit better with our “do something” mentality in medicine. Making someone comfortable IS doing something, and we’re not abandoning them.
Biggest need: We need to have these conversations BEFORE someone lands in the hospital or has a late stage or terminal diagnosis.
Advanced care planning conversations need to happen early and often (annually) in the PCP or specialist office – where a trusting relationship should exist. (Oh, and through Advanced Care Planning, Medicare will pay for that conversation.)
Unfortunately, outpatient physicians and providers are generally not any more comfortable with this than their acute care counterparts. Therefore, the same education and re-education needs to occur with the outpatient teams. Doing so and having these talks earlier will lead to the improved outcomes and experiences we’re looking for, while decreasing utilization costs.
Additionally, the more often these conversations occur, the more comfortable we all become with the topic of end-of-life, death and dying – physicians, providers, patients, and families.
Once again, in healthcare, AI can serve to augment tools that help us do what needs to be done. But it can’t do it all for us. Can we be better at identifying patients that might benefit from a palliative care or end-of-life conversation? Of course! But that only leads us to a finer-tuned cohort of patients who we choose to neglect because we lack the comfort, skills, experience, and expertise to have a conversation.
Let’s work on how to build relationships and better prepare those who are best suited to have that conversation.