Rochester — Mayo Clinic is bringing artificial intelligence into some of the most delicate moments in hospital care, rolling out a new tool that flags inpatients who may need palliative care and surfacing those cases earlier in a hospital stay. Co-developed with clinical-AI firm Bayesian Health, the system is built to call out red flags like uncontrolled pain, overwhelmed caregivers, and other signals that can quietly fall through the cracks. Hospital leaders say the hope is straightforward: fewer avoidable readmissions and more timely, consistent conversations about serious illness.
According to a release from the Mayo Clinic News Network, Mayo’s Department of Medicine led the clinical development of the tool under the organization’s Practice Transformation Ventures framework. The model now sits inside clinicians’ electronic health record workflows, so teams see specific patient signals in the same moments they are making care decisions. "The challenge in palliative care is not just identifying unmet needs but doing so early enough to change the course of care," Jacob J. Strand, M.D., chair of Palliative Care at Mayo Clinic, said in the release. Mayo Clinic reported that the approach was validated through randomized testing inside its hospitals.
Local business coverage by the Minneapolis / St. Paul Business Journal reported that a randomized clinical trial at Mayo Clinic showed a 44% increase in timely palliative care referrals for patients flagged by the tool, along with roughly a 25% reduction in 60-day readmissions and a 28% reduction in 90-day readmissions. As reported by the Minneapolis / St. Paul Business Journal, those gains showed up after the model was wired directly into bedside workflows and clinicians acted on its guidance.
How the technology fits into care
Bayesian Health says its platform continuously reads the full patient record and then surfaces interpretable alerts and recommended next steps directly in the electronic health record. The goal is to spare clinicians from hopping between screens and systems to figure out what the AI is trying to say. The company describes the palliative care module as a learning system that fine-tunes its identification accuracy over time as clinicians accept, modify, or reject its recommendations.
Why this could matter for patients and hospitals
Palliative care remains underused even when it can clearly improve symptom control and discharge planning. Mayo Clinic notes that fewer than half of hospitalized patients who could benefit actually receive specialty palliative consultations. The trial’s reductions in readmissions hint at one very practical payoff for hospitals that spot unmet needs earlier and link patients to the right teams, according to the Mayo Clinic News Network.
Open questions and limits
Researchers are quick to point out that strong performance in a trial does not guarantee smooth sailing once a tool is scaled up. A recent systematic review in npj Digital Medicine found mixed evidence on automated systems that try to identify palliative needs and flagged implementation headaches such as alert fatigue, staffing constraints, and the need to track equity impacts. On the flip side, large prospective deployments like the TREWS sepsis system described in Nature Medicine show that careful design, committed clinician buy-in, and tight production monitoring can translate into measurable patient benefits. The lesson is that the workflow can be just as important as the model’s math.
Mayo Clinic and Bayesian Health say they will keep evaluating the tool as it moves into routine use, and clinicians in Rochester and beyond will be watching to see whether the early trial gains hold up across more patients and services. For hospital leaders, the pitch is tempting: if the model scales, it could help make palliative care conversations more consistent while cutting down on readmissions. Whether that promise holds will depend on the unglamorous details of workflow design, staffing, and transparent oversight.
