Cincinnati’s heart specialists are testing out a new kind of teammate: a generative AI “virtual assistant” that helps manage newly diagnosed heart failure patients, speeds up medication tweaks and, in a small pilot, appeared to cut short term hospitalizations. The idea is to put high level specialist know how into the hands of busy primary care teams, while using far less clinician time than traditional follow up.
Study Shows Faster Med Changes and Fewer Returns to the Hospital
In the ASSIST HF SIRIO randomized pilot, 60 patients with new heart failure with reduced ejection fraction were assigned to virtual assistant guided care or standard care and followed for 12 weeks. As reported in JACC, those in the virtual assistant group had better optimization of guideline directed medical therapy across drug classes, lower NT proBNP levels and fewer heart failure hospitalizations over the short follow up.
How the AI Sidekick Actually Worked
The research team built a generative AI virtual assistant that used retrieval augmented generation and expert prompt engineering to pull key data from patients’ records and generate treatment recommendations. According to the methods paper in JACC Advances, nonmedical staff handled virtual assistant guided contacts every two weeks, while supervising cardiologists reviewed every AI output before any changes were made as a safety backstop.
Homegrown Project, Big Expectations
The work was led by investigators at The Christ Hospital and the Lindner Research Center. Dean Kereiakes, chair of the hospital’s Heart & Vascular Institute and a co author, told the local paper that “advanced technology like this is essential to address issues facing healthcare delivery today.” Hospital officials described the assistant as among the first of its kind for routine outpatient heart failure follow up, according to The Cincinnati Enquirer.
Why This Matters for a Costly, Common Disease
Heart failure is both widespread and expensive. Nearly 6.7 million adults in the United States are living with it, and heart failure was mentioned on 452,573 death certificates in 2023. Treatment and hospital care have racked up more than $30 billion in direct costs in past estimates, according to the CDC. That kind of price tag helps explain why health systems are kicking the tires on scalable tools that might keep patients stable and out of the hospital.
Privacy, Safety and What Comes Next
The authors stress that this was a single center pilot meant to test feasibility and safety, not a definitive proof of long term benefit. They call for larger, multi center trials to see whether the early gains hold up in different health systems. The methods and supplemental appendix detail how the model was trained, how red flag symptoms were detected and how oversight was structured to protect patient safety and privacy, but any broader rollout would still need local validation and clear data access agreements, according to JACC Advances.
If the approach scales, the hope is that primary care teams could start patients on recommended heart failure medications faster while freeing cardiologists to focus on the sickest cases. Christ Hospital and the study authors say they plan to test a wider rollout and longer follow up to see whether those 12 week improvements stick around, as reported by The Cincinnati Enquirer.
