The 2 AM phone call is one aspect of heart failure treatment that is not often discussed. A patient at home discovers a problem. fluid construction. breathing in different ways. They give the hospital a call. An on-call doctor, probably not a cardiologist, answers the phone, fumbles to retrieve records they have never seen before, and attempts to make a sound clinical decision while exhausted and lacking the complete picture. It is a system that has never been ideal and is only maintained by institutional habit and goodwill. A Cincinnati research team made the decision to take action.
Located in Cincinnati’s Mt. Auburn neighborhood, the Christ Hospital Health Network has 136 years of history. As of early 2026, a study that was published in the Journal of the American College of Cardiology has been the subject of quiet discussion in the cardiology community. The focus of the project is an AI virtual assistant designed especially for heart failure with reduced ejection fraction, one of the most deadly and resource-intensive diagnoses in American medicine.
The Cincinnati Project — AI Heart Failure Assistant: Key Facts
| Institution | The Christ Hospital Health Network, Cincinnati, Ohio — ranked #1 hospital in the Cincinnati region by U.S. News & World Report; named to Newsweek’s World’s Best Hospitals list in 2026 |
| Research Center | The Carl & Edyth Lindner Center for Research and Education — one of the top cardiovascular research centers in the US; introduced more than 25 first-in-the-world medical devices or medications |
| Lead Researcher | Dr. Eliano Navarese, MD, PhD — Senior Scholar, Lindner Center; previously practiced clinical-academic medicine in Canada, Italy, Germany, the Netherlands, Poland, and the United States |
| What the AI Does | Generative, interactive AI virtual assistant — not a chatbot or algorithm; provides real-time diagnosis and optimal treatment recommendations for heart failure with reduced ejection fraction, then routes decision to cardiologist for approval |
| Study Type & Result | Pilot randomized controlled trial — AI assistant outperformed standard of care; reduced hospitalizations and improved patient outcomes; published in the Journal of the American College of Cardiology (JACC) |
| Heart Failure Scale (US) | Heart disease is the No. 1 killer of both men and women in America; heart failure is the No. 1 hospital discharge diagnosis for Medicare patients; annual cost exceeds $35 billion |
| Key Differentiator | Unlike rule-based algorithms, the system is generative and interactive — incorporating the accumulated knowledge of expert physicians directly into real-time patient care decisions |
| Parallel Research: Cornell/Columbia | Weill Cornell Medicine and Columbia University researchers developed an AI model using cardiac ultrasound data achieving ~85% accuracy in identifying advanced heart failure patients — published March 2026 in npj Digital Medicine |
| Advanced Heart Failure Gap | An estimated 200,000 Americans have advanced heart failure; current diagnosis requires specialized cardiopulmonary exercise testing available only at large medical centers — fewer than a few thousand receive appropriate care annually |
| Rural Application Potential | Dr. Navarese specifically cited rural settings — where cardiologists are scarce — as a primary target for deployment; tool enables primary care providers to access specialist-level guidance in real time |
| Next Steps | Larger clinical trials underway; FDA approval pathway being mapped; Cornell/Columbia team planning clinical studies of ultrasound-based AI model |
The cardiologist is not replaced by the assistant. More specifically, it provides real-time treatment recommendations that are reviewed and approved by the on-call cardiologist after incorporating the expert physicians’ accumulated knowledge into a generative and interactive system. quicker choices. more knowledgeable choices. while a human is still signing off.
The system’s creator, Dr. Eliano Navarese, is an Italian-born physician-researcher who has worked in Canada, Germany, Italy, the Netherlands, Poland, and the United States. This career path has given him an exceptionally broad perspective on how various health systems handle the same clinical problems in different and costly ways. One thing he has made clear is that this is not a chatbot. It’s not a decision tree disguised in contemporary terminology. Because the system is interactive and generative, it makes decisions based on clinical data rather than by comparing patterns to a predetermined set of rules. In practice, where patient presentations are rarely flawless and the incorrect recommendation at the wrong time can have disastrous consequences, this distinction is crucial.
The American healthcare system has never quite managed to control heart failure, a disease of enormous scope and expense. For Medicare patients, it is the most common diagnosis for hospital discharge. Over $35 billion is spent annually. Despite decades of advancements in medicine and surgery, heart disease continues to be the nation’s leading cause of death for both men and women. Over the course of its existence, the Lindner Center at Christ Hospital has introduced over 25 first-in-the-world medical devices. It is evident that the next frontier will not be a new drug or stent, but rather a different kind of intelligence applied to clinical decision-making. It’s a significant institutional wager.
The pilot trial’s conclusions are worth carefully considering. The current standard of care was outperformed by the AI assistant. Not just a little bit. Not in limited technical metrics that don’t correspond to actual results. Better in terms of lowering hospitalizations, which are costly, inconvenient, and frequently avoidable incidents that characterize the experience of heart failure for patients and their families. It’s still a pilot study, and Dr. Navarese has made it clear that more extensive trials are required before generalizations are made. However, a pilot randomized controlled trial that outperforms standard care is not insignificant. For a first try, that’s actually a pretty high bar to meet.
In March 2026, Weill Cornell Medicine and Columbia University published their own research on AI heart failure, which is an interesting parallel thread from New York. In order to identify patients with advanced heart failure, a condition that affects an estimated 200,000 Americans, their system applies machine learning to cardiac ultrasound data—the standard echocardiograms already included in routine care. The majority of these patients never receive the proper treatment because the standard diagnostic test requires specialized equipment only found at large academic medical centers. Using pre-existing data sources, the Cornell model identified high-risk patients with about 85% accuracy. In other words, the infrastructure needed to locate patients is the bottleneck, not the patients themselves. The bottleneck is eliminated by the AI.
There is a sense that cardiology may be getting close to a true turning point as these two projects—one centered on diagnosis and the other on treatment choices—develop concurrently. The quieter kind, where the day-to-day operations of patient care begin to change in ways that add up to something significant over years, is not the kind that is made public with a press release and a stock price spike. In particular, the Christ Hospital project has an almost understated quality, coming from a mid-sized Ohio city rather than a coastal academic behemoth, and it was created by a team that seems more concerned with the patients than the attention. Whether the larger trials will support the pilot’s recommendations is still unknown. However, it’s difficult not to think the entire thing is worth closely observing.
