Mayo Clinic, Bayesian Health Develop Solution to Expand Palliative Care Access

In 2021, Healthcare Innovation interviewed Suchi Saria, Ph.D., a professor of machine learning and healthcare at Johns Hopkins University in Baltimore, about a company she founded called Bayesian Health to develop artificial intelligence-based clinical decision support to make the EHR more dynamic and predictive. Fast-forward five years, and Bayesian has partnered with Mayo Clinic to co-develop an AI solution to identify hospitalized patients who may benefit from palliative care earlier in their stay. 


The solution is designed to support timely consultations, with the objective of improving goal-concordant care for patients with serious illness and reducing non-beneficial readmissions.

Roughly one-third of readmissions involve patients with serious illness, many of whom experience repeated hospitalizations. However, fewer than half of these patients receive palliative care consultations. To address this challenge, Mayo Clinic and Bayesian Health built a solution that identifies patients with unmet palliative care needs earlier and equips clinicians with the context they need, within their workflow, to navigate the complex conversations and care coordination that follows.

(Healthcare Innovation recently wrote about how, in a similar effort, St. Louis-based BJC HealthCare created an algorithm to help prioritize which patients to have goals-of-care or advanced care planning conversations with, which led to an increase in palliative care utilization.)

In a randomized trial conducted at Mayo Clinic, validated findings from an earlier version of the program demonstrated that use of the tool was associated with a 44% increase in   timely palliative care referrals, a 25% reduction in 60-day readmissions and a 28% reduction in 90-day readmissions, along with improved patient quality of life.

"The challenge in palliative care is not just identifying unmet needs but doing so early enough to change the course of care," said Jacob J. Strand, M.D., chair of Palliative Care at Mayo Clinic, in a statement. "What makes the difference is tailoring workflows to local culture, based on patient acuity, and across central and bedside teams across the organization. When high-quality, patient-specific signals reach frontline clinicians in the moments that matter, it cuts through the complexity of inpatient care, drives more consistent decision-making and supports teams in delivering the best possible care to every patient."

Mayo Clinic’s Department of Medicine led the clinical development and validation of the AI solution. Bayesian Health supported integration of the model into the EHR, allowing care teams to access this information within existing clinical workflows.

This is the first collaboration of its kind at Mayo Clinic to use AI across the entire care process in a complex hospital setting, helping its care teams spot unmet needs earlier, connecting the patient with the right specialists at the right time, while keeping the patient’s health information coordinated but confidential for an overall improvement in care.

Bayesian Health provides a clinical foundation of real-time clinical intelligence that shifts care from reactive to proactive. The newest module on the Bayesian platform brings this approach to palliative care, identifying unmet needs such as pain or caregiver support so clinicians can reach patients earlier in their care journey. Palliative care teams get a real-time, hospital-wide view of patients who may benefit from a consult, while bedside clinicians get clear, interpretable guidance and a streamlined path to action, so the moment of insight becomes a moment of care rather than another notification, the company said. The clinical AI continuously learns from clinician feedback and local patient populations, improving identification accuracy over time.

"Palliative care is exactly the kind of problem our platform is built for: reaching patients earlier, when clinicians still have time to change the course of their care," said Saria in a statement. "It takes purpose-built infrastructure, rigorous validation and a thoughtful partnership between AI and clinical experts. That's how trustworthy AI gets built, and it's how care actually improves for both patients and caregivers."