From detecting cancer and remote patient monitoring to reduced physician burnout and improved CMS star ratings, here are 27 proven clinical uses of AI — with results:
- Rochester, Minn.-based Mayo Clinic developed an AI model using ECG data that outperforms the model for end-stage liver disease score in predicting severe liver disease and liver-related death among transplant patients. Trained on data from more than 75,000 patients, the model detects complications MELD misses and could be combined with it to optimize organ allocation and refine transplant decisions.
- Atlanta-based Emory Healthcare utilized agentic AI to improve its CMS star rating for blood pressure control from one to four stars. The AI agents telephoned older adults with hypertension, who provided recent blood pressure readings or conducted a live blood pressure measurement during the call.
Use of AI agents reduced the system’s average cost per reading by 88.7% and improved the system’s controlled blood pressure gaps by 17%. - New York City-based Mount Sinai Health System researchers developed AEquity, a tool designed to detect and mitigate biases in health datasets used to train machine-learning models. Tested on medical images, patient records and national survey data, the tool uncovered both well-known and previously overlooked biases, helping prevent AI models from perpetuating disparities in care.
- Mount Sinai also developed an AI model that helps determine which atrial fibrillation patients benefit from blood thinners to prevent stroke. Trained on data from 1.8 million patients, the tool accurately assessed stroke and bleeding risks and recommended against anticoagulation in up to 50% of patients who would have received it under current guidelines, signaling a potential shift toward more personalized stroke prevention.
- Mayo Clinic researchers also developed an AI model that predicts ALS and anticipates patient survival using data from F-wave nerve conduction studies. The model outperformed clinical annotations and identified factors linked to decreased survival, such as older age at onset and family history of ALS.
- Somerville, Mass.-based Mass General Brigham and Emory Healthcare saw major reductions in clinician burnout after adopting ambient AI scribes, with a 21% decline in burnout at Mass General Brigham and a 30.7% rise in documentation-related well-being at Emory.
- Yale New Haven (Conn.) Health implemented an AI-powered clinical deterioration tool eCART across its seven hospitals, leading to a 13% relative decrease in mortality. The tool monitors 97 clinical variables used to detect patient decline about four hours before critical events, helping clinicians intervene earlier and focus on patients who need attention most.
- Sacramento, Calif.-based Sutter Health doubled its rate of early-stage lung cancer diagnoses after implementing an AI-powered tracking tool within its Epic EHR. The tool has led to more than 70% of all lung cancer diagnoses to occur during stage 1 or stage 2, when survival is more likely.
- Providers at Renton, Wash.-based Providence saved 2.5 hours of after-hours work per week using Microsoft’s DAX Copilot ambient AI tool, according to a six-month study. Clinicians reported reduced documentation burden, frustration and burnout, and researchers concluded that ambient clinical intelligence could help address physician burnout.
- An AI tool developed by researchers from Columbia University and NewYork-Presbyterian, both based in New York City, successfully identified patients at risk of having undiagnosed structural heart disease.
The tool, called EchoNext, was trained on more than 1.2 million electrocardiogram and echocardiogram data pairs from 230,000 patients. The tool accurately identified 77% of structural heart problems in 3,200 electrocardiograms compared to 64% accuracy of 13 cardiologists analyzing the same data. - New Brunswick, N.J.-based St. Peter’s University Hospital cut emergency department visits among high-risk patients by 7% using an AI tool that predicts avoidable ED visits within 90 days by analyzing social determinants of health. The tool helped clinicians identify patients’ food and transportation needs, enabling earlier interventions and reducing ED visits from 16.7% to 9.5%.
- New York City-based Mount Sinai Hospital’s cardiac catheterization lab developed and utilized an AI agent called Sofiya to call patients prior to stenting procedures to answer logistical questions and provide preprocedural instructions.
The AI agent saved over 200 nursing hours in five months, making a full day of calls in two to three hours.
“Nurses now have time to support us in other areas where we need their clinical expertise,” Annapoorna Kini, MD, director of the cath lab and of the hospital’s interventional structural heart disease program, told Becker’s. - Nearly 2,000 physicians and advanced practice providers at Altamonte Springs, Fla.-based AdventHealth are using Nuance’s DAX Copilot ambient AI scribe.
Since implementing the technology in April 2024, 86% of users report less burnout, 83% report improved work-life balance and job satisfaction, 80% report better patient experience, and 3 in 4 say they are less likely to leave clinical practice. - Yale New Haven (Conn.) Health System researchers employed an AI model called PanEcho to perform echocardiogram interpretation.
The model accurately estimated left ventricular ejection fraction, and detected moderate or worse left ventricular systolic dysfunction, right ventricular systolic dysfunction and severe aortic stenosis with a median normalized mean absolute error of 0.13. - Charleston, S.C.-based MUSC Health has been leveraging agentic AI for administrative functions. The technology has been associated with a 7.6% reduction in no-show rates and a 30% increased digital intake for Spanish-speaking patients.
The health system has also deployed the AI agents to execute 40% of prior authorizations without human involvement, cutting 30 minutes of manual work to about one minute. - Since integrating Ambience’s generative AI for clinical documentation in 2024, Boise, Idaho-based St. Luke’s Health System has generated an extra $13,049 per clinician.
The health system reported a 41% reduction in active documentation time, a 41% reduction in time to chart closure, a 39% decrease in time spent documenting after hours, a 36% reduction in the number of clinicians reporting near-daily clinician burnout, a 22% increase in patient face time and a 17% decrease in time spent documenting after appointments. - Anthony Law, MD, PhD, an assistant professor in the department of otolaryngology at Atlanta-based Emory University School of Medicine, developed an AI model to detect throat cancer by listening to a patient’s voice.
The model, designed for in-clinic use through an app, has about an 93% success rate for identifying patients who have a mass in their larynx. - The Permanente Medical Group, part of Oakland, Calif.-based Kaiser Permanente, saved nearly 16,000 hours in documentation time over a 15-month period through the use of ambient AI.
- Chicago-based Northwestern Medicine radiologists used a generative AI system — designed by Northwestern engineers — to boost documentation efficiency by 15.5%.
- Cleveland-based MetroHealth improved its medication prior authorization process with an AI tool, cutting submission times from 18 hours to around five minutes and increasing approval rates by 15%.
In one case, AI submitted prior authorization for a hepatitis C patient’s prescription in five minutes, leading the health system to receive insurance approval within three hours and provide the medication within seven. - Houston-based Texas Children’s Hospital developed an AI model to assess bone age in pediatric patients, reducing the time radiologists spend on image interpretation by 30% to 50%.
- Sacramento, Calif-based UC Davis Health developed an AI model to identify patients who may be at-risk for hospitalization or an ED visit within 12 months. Care teams then conduct proactive outreach to the at-risk patients. The initiative reduced hospitalization rates by 5% to 10% among at-risk patients.
- Somerville, Mass.-based Mass General Brigham employed ambient documentation using generative AI to alleviate administrative burden among its clinicians.
Around 60% of providers reported they were more likely to extend their clinical careers because of the technology, 20% reported reduced burnout symptoms and 80% said they were spending more time looking at their patients. - Kettering (Ohio) Health leveraged an AI-powered risk assessment tool to cut length of stay in half and improve care quality among patients undergoing angioplasty.
Use of the AI tool was associated with a decline in contrast-induced acute kidney injury from 10% to an average of 2.18%.
The tool was also associated with a decrease in bleeding complications from 2.15 events per month to an average of 1.54 events per month, and the average patient length of stay decreased from 3.44 days to 1.79 days.
- Sioux Falls, S.D.-based Sanford Health found the use of an ambient AI documentation tool significantly improved workforce satisfaction and retention.
After using the technology for less than a year, 88% of clinicians reported a reduction in burnout or fatigue, 90% reported higher job satisfaction and an improvement in their work-life balance, and 95% of clinicians stating it has helped reduce mental strain.
Additionally, 76% of clinicians said they were more likely to remain with Sanford Health and 80% reported were more likely to continue practicing in the medical field. - Aurora, Colo.-based UCHealth uses AI to monitor about 22,000 hospital beds through a virtual care center. With the help of in-room cameras and AI embedded in the Epic EHR, a team of nurses virtually monitors medical-surgical and step-down beds across 14 hospitals for signs of sepsis.
The technology has enabled care teams to identify sepsis two to four hours earlier than if they had not been using AI, reducing mortality risk by 30% or more.
The technology also helps virtual care center teams remotely monitor patients at home, reducing at-home diabetes patients’ hemoglobin A1c levels by about 25%. - The use of ambient AI-scribe technology at Philadelphia-based Penn Medicine reduced clinical note time by about two minutes per appointment and after-hours “pajama time” from 50.6 minutes to 35.4 minutes per workday.
