“Would you be surprised if this person were not alive 12 months from now?”
Known as “the surprise question,” this question along with clinical risk calculators help doctors make decisions around care for patients near the end of life. Increasingly, clinicians are also using AI models to predict remaining life expectancy and better inform palliative care decisions.
In a July study out in the Journal of Palliative Medicine, SPH professor Mika Hamer sought to understand what palliative care teams think about AI prognostication tools – which give an estimated life expectancy or prognosis – in end-of-life care. What they had to say was mixed.
Hamer and a team at the University of Colorado interviewed a variety of palliative care practitioners, including doctors, nurses, spiritual and religious guides and social workers about their perspectives on AI prognostication tools.
“There was consensus around the utility of AI prognostication for more accurately predicting how long someone has to live, which could inform what kinds of care they need or when they might qualify for hospice,” Hamer said. “But there was also deep concern about the sensitivity of this kind of data, and the potential for it to be used against the patient’s best interests. For example, is there potential for health insurance companies to use this information to deny care?”
Medicare covers hospice care in the final six months of life.
“Most palliative care practitioners said they saw AI as one tool in their toolkit that informs their clinical decisions – a “screening tool” of sorts. But they said that using AI-based prognostication alone would pose ethical issues.”
AI tools can be highly accurate and can improve the information that’s used for decision-making. The palliative care teams see AI models as companions to other prognostic and diagnostic tools, not as a replacement for end-of-life care.