Mayo Clinic taps AI to flag patients needing palliative care (Courtesy of the Jacksonville Business Journal) — Mayo Clinic is deploying artificial intelligence across its hospital network to help doctors identify patients who might benefit from palliative care, a practice to relieve pain or bring comfort to patients facing serious or terminal illness.
The Rochester, Minnesota-based health system said Tuesday it is deploying the tool across its network after co-developing it with New York-based Bayesian Health. The platform analyzes patient data such as diagnosis history, admission frequency and pain symptoms. It flags cases where additional supportive care may be appropriate.
“What we’re trying to do is identify if there are needs that aren’t being met by our current medical care,” said Dr. Jacob Strand, chair of palliative care at Mayo Clinic in an interview.
Palliative care focuses on relieving pain and other symptoms for patients with serious illness, as well as providing emotional and psychological support. It is often used for patients requiring end-of-life care.
Mayo has been working since 2018 to develop predictive models to identify patients whose needs may be overlooked, Strand said. Earlier versions were tested at Mayo Clinic Hospital in Rochester, Mayo’s Mankato campus and its intensive care units in Jacksonville, Florida. Mayo says this marks the first time it has deployed this type of AI across its entire hospital network.
Mayo launched the latest version of the platform last month with plans to expand to additional sites, Strand said. As the system flags patients for potential palliative care consults, clinicians can accept or ignore the recommendation, and no action is required, Strand said.
The AI model was designed to improve accuracy over time based on clinician feedback and local patient populations, according to Mayo.
“If certain patient characteristics are leading to more decisions being deferred, that allows us to make improvements to the AI model,” Strand said.
In a randomized clinical trial conducted at Mayo with more than 3,000 patients, use of the tool increased palliative care referrals by 44%.
Mayo said expanded use of palliative care also could help it avoid costly hospitalizations, reducing 60-day readmissions by 25% and 90-day readmissions by 28%.
Strand said the project reflects Mayo’s focus that AI be used to identify unmet patient needs rather than predict recovery or mortality. Predicting a patient’s prognosis is inherently challenging and often comes too late to meaningfully help patients.
Many eligible patients are not receiving palliative care, Mayo said. Fewer than half of seriously ill patients who are hospitalized multiple times receive palliative consultations, according to Mayo.
“It’s hard to do this at scale and get visibility for patients with unmet care needs, particularly when other priorities might take precedence, like infection or complex heart failure,” Strand said. “That’s not an algorithm problem. That’s a workflow problem that we sought to address. Because of that, conversations about unmet care needs are often recognized too late.”
By automating part of the screening process, Mayo hopes to expand access to palliative care while reducing the burden on clinicians.
“Hopefully this not only gets us to see more patients, but more time at their bedside instead of in front of computers,” he added.
Photo courtesy of Wikimedia Commons