Mayo Clinic and health-tech startup Bayesian Health announced in May 2026 that a jointly developed AI platform designed to flag hospitalized patients who could benefit from palliative care has produced striking results in a validated randomized clinical trial: a 44% increase in timely palliative care referrals, a 25% reduction in 60-day readmissions, and a 28% reduction in 90-day readmissions. The findings, detailed in releases from Mayo Clinic News Network and BioSpace, mark one of the more rigorously tested AI deployments in U.S. hospital medicine to date and offer a rare glimpse of an AI health tool validated not just for accuracy, but for measurable changes in patient outcomes.
A quiet crisis in palliative care access
Palliative care — specialized medical support focused on relieving pain and stress for patients with serious illness — is chronically underused in American hospitals, not because clinicians doubt its value but because identifying which patients need it often depends on a physician noticing subtle, easy-to-miss signs: a declining health trajectory, a caregiver who seems overwhelmed, a symptom burden that has quietly become unmanageable. Those signals can get lost in the chaos of a busy inpatient unit, and by the time a palliative consult is ordered, patients may have already endured avoidable suffering or a preventable readmission. Mayo Clinic’s Department of Medicine set out to see whether a real-time AI system, built under the health system’s Practice Transformation Ventures framework, could catch those signals earlier than clinicians reliably do on their own.
How the AI platform actually works
The system continuously analyzes data already flowing into the electronic health record — vital signs, lab trends, medication changes, prior admissions, documented symptoms — looking for patterns associated with unmet palliative care needs. Rather than operating as a standalone alert that clinicians must remember to check, the tool is embedded directly into existing EHR workflows, surfacing guidance to frontline physicians and nurses at the point of care. Bayesian Health, which specializes in integrating predictive models into hospital record systems, handled that embedding, while Mayo Clinic’s clinical teams led development and validation of the underlying model. A hospital-wide operational dashboard also gives palliative consultation teams a real-time view of which patients across the system may need their attention, rather than waiting for a referral to land in an inbox.
Why a 28% readmission drop is a big deal
Hospital readmissions within 30, 60, and 90 days are among the most closely watched metrics in American healthcare, both because they signal gaps in care quality and because they carry direct financial consequences under Medicare’s Hospital Readmissions Reduction Program. A double-digit reduction in 90-day readmissions tied to earlier palliative care referral suggests that getting seriously ill patients into supportive care sooner does not just ease suffering — it may prevent the kind of medical crises that send patients back to the emergency department weeks after discharge. For a field that has seen plenty of AI pilots produce modest efficiency gains without moving core outcome metrics, results validated through a randomized trial design carry unusual weight.
The case for cautious optimism
Supporters of the rollout point out that this is not a chatbot summarizing a chart or a triage tool flagging sepsis risk — categories where AI has had a mixed track record — but a system built around a specific, well-understood clinical workflow gap: palliative care referrals that come too late. Because the tool was tested in a stepped-wedge randomized design rather than a retrospective chart review, Mayo Clinic can point to a level of evidence that many hospital AI tools never reach before commercial rollout. If the results hold up as the tool scales beyond Mayo’s own hospitals, health systems nationally could see both better end-of-life and serious-illness care and meaningful cost savings from fewer avoidable readmissions.
Where skepticism remains
Palliative care specialists have long cautioned that predictive algorithms flagging “declining health trajectories” risk encoding bias if the underlying data reflects unequal access to care in the first place — patients from under-resourced communities may show different documentation patterns not because they are healthier, but because they saw a doctor less often. Independent replication outside Mayo Clinic’s own health system, which has more resources and clinical infrastructure than most U.S. hospitals, will be an important test of whether the 44% referral increase and readmission reductions generalize to community hospitals with leaner staffing and different patient populations. Critics also note that a real-time alert is only as good as the clinical capacity to act on it — flagging more patients for palliative consults does little good if a hospital lacks enough palliative care specialists to see them.
What comes next
Mayo Clinic and Bayesian Health have not yet detailed a timeline for licensing the platform to other health systems, though the companies have signaled interest in expanding beyond Mayo’s own network given the strength of the trial data. The next milestones to watch will be peer-reviewed publication of the full trial results, any external validation studies at hospitals with different patient demographics, and whether Medicare or other payers begin referencing readmission-reduction tools like this one in future value-based care contracts. For a healthcare AI market still searching for proof that predictive tools change outcomes and not just workflows, Mayo’s palliative care trial offers one of the more concrete answers yet.
