Artificial intelligence technologies are advancing the ability to conduct proactive screening to catch the warning signs of serious health conditions, like heart disease. With earlier detection of cardiovascular risk, clinical teams can intervene sooner to improve patients’ long-term health.
Often, key details about a patient's health risks are buried in complex, disparate data sources.
The University of Texas Medical Branch, a major academic medical center based in Galveston serving the Gulf Coast region, has been working with health tech company Bunkerhill Health to use AI to identify and ensure appropriate treatment for patients with cardiovascular risk. The use of a coronary artery calcium (CAC) detection algorithm has helped flag patients with incidental CAC and initiate referrals to cardiology. Early results showed that the AI-enabled cardiovascular analysis identified that 11.7% of patients getting routine chest CTs had unaddressed, clinically significant CAC and triggered follow-up, according to the health system.
These AI detection tools have identified patients who would have otherwise been missed and ensured they received life-saving interventions, said Peter McCaffrey, M.D., Chief AI Officer at UTMB.
For one UTMB patient, a police officer, the AI solution detected warning signs of a heart attack, including a blockage in the so-called “widow-maker” artery of his heart, before the patient experienced permanent heart damage. The patient underwent a successful revascularization procedure and has since recovered.
"We think it has definitely saved the lives of patients," McCaffrey told Fierce Healthcare. These AI-powered screening programs use technology to find and surface information while still relying on physicians' expertise for adjudication, management and reasoning, McCaffrey noted.
"You leave the AI aside, you leave the dollars aside, this is such a wonderful application of new ideas that are just helping patients live healthier lives," Nishith Khandwala, co-founder and CEO of Bunkerhill Health, told Fierce Healthcare.
Bunkerhill Health holds Food and Drug Administration (FDA) clearances for nine clinical algorithms, most focused on cardiovascular health. It developed algorithms that detect coronary artery and aortic valve calcium on both non-contrast and contrast-enhanced chest computed tomography (CT) scans. The company currently has AI-enabled cardiovascular screening programs in place at more than 10 health systems.
But there are challenges to scaling these AI tools in practice at health systems to enable broader population health impacts.
There are incremental costs for health systems to roll out these screening programs, Khandwala noted. And, until recently, there was no Medicare payment model that supported opportunistic screening.
Bunkerhill Health saw an opportunity to drive that forward.
In April, the company revealed it has secured a reimbursement pathway for AI-enabled cardiovascular analysis. The Centers for Medicare and Medicaid Services (CMS) established a national billing code (G0680) and associated payment under the Hospital Outpatient Prospective Payment System (OPPS) for algorithmic analysis of coronary artery calcium (CAC) and aortic valve calcium (AVC) on chest CT scans.
Establishing a reimbursement pathway is an important step toward broader adoption of AI in clinical care, Khandwala said.
“With this pathway in place, health systems have greater clarity on how to incorporate AI-driven insights into clinical workflows in appropriate clinical contexts to help identify and manage patients with meaningful cardiovascular findings,” he said.
“This is one of many steps along the way of just methodically and systematically addressing every potential barrier to adoption for cardiovascular analysis programs powered by AI,” noted Sean Bennett, vice president of operations at Bunkerhill Health.
Health tech companies that use AI to extract opportunistic screening information available in cardiac or lung CT scans can take advantage of the new billing codes and associated payments.
Jonathan Govette, CEO of Oatmeal Health, an AI lung cancer screening startup, touted that the new CMS payment pathway creates a business model to unlock the value from medical images.
“CMS created an OPPS billing pathway specifically for this opportunistic screening. For the first time, hospitals can get reimbursed for using AI to extract additional diagnostic value from scans already being performed,” Govette wrote in a LinkedIn post. “This changes everything about preventive cardiology economics. Every chest CT for pneumonia, lung nodules, or trauma now becomes a cardiac risk assessment tool. No extra radiation. No extra appointment. No extra scan time. Just AI analyzing data that was already there.”
In an interview with Fierce Healthcare, Khandwala and Bennett offered a deeper look at the process for securing the CMS reimbursement pathway.
Making the case for a billing pathway for AI-enabled cardio analysis
Bunkerhill Health was founded by Khandwala and David Eng in 2018, growing out of the Stanford University Center for Artificial Intelligence in Medicine and Imaging (AIMI). The company initially started by working with a consortium of academic medical centers, including UCSF, Emory University and MedStar Health, to validate AI medical imaging algorithms and then commercialize them and deploy the algorithms at health systems.
Millions of patients undergo chest CT scans each year, including both contrast-enhanced and non-contrast studies performed for a wide range of clinical indications. But reviewing all these scans manually to screen for incidental coronary artery calcium would be extremely labor-intensive and time-consuming.
Bunkerhill Health’s AI algorithms for coronary artery calcium and aortic valve calcium on chest CT scans can analyze existing imaging for previously undetected heart disease. The algorithms enable automated identification and quantification of these findings on scans patients already receive as part of routine care, company executives said.
The ambition was to use existing imaging to help hospitals find at-risk patients and close critical gaps in care. For cardiologists, the algorithms can help flag at-risk patients who might otherwise fall through the cracks, Khandwala noted.
As part of a randomized trial at Stanford Health Care, called the NOTIFY-1 study, clinical teams used the AI algorithm to identify coronary artery calcium on routine CT scans and then notified clinicians and patients about the incidental findings. The company's algorithm was shown to drive a 44-percentage-point increase in statin prescription rates, with the potential to reduce future cardiovascular risk.
Even with clinical evidence supporting the AI-enabled screening, Bunkerhill Health ran into challenges with scaling adoption at health systems.
These AI detection tools enable health systems to do opportunistic screening for cardiovascular risk, which represents a new use case to improve outcomes in high-stakes cardiac monitoring.
“This isn’t something that was being done before and now is being automated. It was not done before, and now we are starting to do it,” Khandwala said.
“At a very large health system, let’s say they are running on over 100,000 scans and there are 9,000 patients that are identified as at-risk. Now imagine handing over a list of 9,000 patients to the health system and saying, good luck with this,” he noted.
Healthcare AI solutions that only flag patients with findings help to surface problems, but fall short of solving them. It still leaves care teams to manually review charts and determine next steps. Care teams are already stretched thin, so identifying patients for follow-up and scheduling those appointments only adds to the workload.
Bunkerhill Health’s goal was to improve outcomes for at-risk patients, without adding work for clinicians by using AI to automate operational workflows. From end to end, the company’s platform scans medical images and electronic health records, flags patients for follow-up and then automates care coordination.
Health systems can use the company’s iCAC algorithm to detect elevated CAC on routine chest CT scans, and other AI tools can then automate patient chart review, identify patients not on statins, confirm other risk factors and trigger outreach and appointment scheduling.
There is a business case for opportunistic cardiac screening, but the downstream financial return-on-investment isn't clear at this point.
“For a health system to adopt a new technologies, you have to invest so much effort. You need to invest so much political capital. You need to invest so much in change management,” Khandwala said.
He added, “These are use cases that are only enabled because of AI. With anything completely new, there are known unknowns and unknown unknowns, and in order for those to be de-risked, we think that some kind of reimbursement is very helpful."
The company decided to initiate and lead a submission to CMS for a reimbursement pathway to financially “de-risk” AI-driven opportunistic screening for providers, executives said.
Last June, working in consultation with law firm McDermott Will & Schulte, Bunkerhill Health submitted an application through the CMS New Technology Ambulatory Payment Classification (APC) assignment process. The new technology APC pathway provides payment for innovative outpatient services not adequately covered by existing codes.
For several months, the company worked directly with CMS to review the technology, provide the clinical justification and NOTIFY-1 study results and help define how the analysis can be evaluated and supported in practice.
CMS has now established Healthcare Common Procedure Coding System (HCPCS) code G0680 for the detection and quantification of coronary artery calcium and/or aortic valve calcification using AI-based algorithmic analysis of chest CT scans for insurance claims. CMS set the payment rate at $15.50.
The reimbursement pathway does not set up “blanket approval” for broad, opportunistic screening programs, Bennett asserted.
“I think the real value is just providing a layer of financial de-risking for a subset of patients who might be part of these programs,” he said.
He added, “We’re proud that we're able to support all of the great clinical outcomes that this program can bring at scale across health systems, but do so in a very responsible way with a lean cost structure and really accurately represent the modest additional costs health systems incur by running programs like this, but also make sure that there is pathway, at least for a subset of patients, that they recoup the costs at the actual rate that they are incurring.”
CMS has advanced reimbursement for several AI-powered cardiac screening and analysis tools. Last year, Viz.ai announced that the agency established a national Medicare payment rate for AI-powered electrocardiogram (ECG) analysis. Late last year, CMS finalized a national payment rate for Eko Health's AI-assisted cardiac exam, performed with its digital stethoscope. The agency also set a Medicare reimbursement rate for Caristo Diagnostics' AI-powered coronary plaque analysis performed on coronary CT angiography (CCTA) scans.
As of January 2026, there are 26 CPT codes for clinical AI solutions, according to the Bipartisan Policy Center.
The new CMS reimbursement pathway for algorithmic analysis of CAC supports the use of AI-powered screening tools to identify risk earlier and prevent more advanced disease, McCaffrey said.
"I think it's a very good thing in the name of public health. If we really want to bend the curve down, we've got to be farther up in the funnel, which necessitates new kinds of things and necessitates screening, in general, as a much more active thing. For health systems broadly to adopt that, it helps that there is some kind of reimbursement mechanism aligned to that," he said.
In the past few years, Bunkerhill Health has evolved, developing a clinical reasoning platform, called Carebricks, that enables health systems to design and deploy AI agents across clinical and operational workflows.
The company now has a library of AI-powered, ready-to-use clinical workflows for a range of clinical, administrative and operational use cases, from prior auth automation to escalating high-risk referrals to high-risk patient outreach triage.
“The company looks quite different than when I joined back in 2021, but the clear through-line is that we're trying to bridge the gap between the potential of AI in healthcare and its reality in clinical practice,” Bennett said.
Rather than delivering predefined solutions or narrow point tools, the company works directly with health systems to identify their most pressing challenges and rapidly build AI-powered workflows tailored to those needs, executives said.
UTMB Health has rolled out Bunkerhill Health's Carebricks system-wide with more than 20 AI agents now live to support cardiovascular risk identification, aortic aneurysm identification and follow-up and incidental pulmonary nodule management. UTMB also uses AI agents for administrative workflows like referral queue triage, prior authorization automation and case mix optimization.
"One of our big mandates is access to care. We have a lot of patients in our area who don't have a lot of resources, who are notoriously challenged with access in general, and we want to do the best that we can to make sure that they have access. One way this looks like is being more efficient in our referrals and prioritizing who is the most critical, who has a finding that really can't wait and who should be at the top of that list to be navigated," McCaffrey said.
More broadly, Bunkerhill Health's Carebricks platform aligns with UTMB's AI strategy to focus on agentic AI that can be used to orchestrate workflows rather than focus on narrow solutions, McCaffrey said.
"When we plot the course of frontier intelligence and look at where is GPT going? Where is Claude going? Where is Gemini going? If they become more flexible and more capable, then the bottleneck really becomes how quickly you can iterate and how quickly you can deploy. We have seen a lot of benefit in getting good at a way of doing things in a platform, and focusing on being iterable and responsive. Our technical philosophy has been to go with platforms, double down on platforms, focus on workflow iteration and really trying things, responsibly, yes, but trying things and being disciplined about incremental improvements all the time," he said.