Health economic simulation modeling of an AI-enabled clinical decision support system for coronary revascularization

Smarter Choices for Heart Care

People with serious heart artery blockages often face a tough choice between medicine, minimally invasive procedures, or open‑heart surgery. This study asks a simple but important question: if doctors had help from a powerful artificial intelligence (AI) tool that forecasts future health and costs, could patients live better lives while the health system saves money? Using real data from tens of thousands of Canadians with heart disease, the researchers show that an AI guide could often recommend a different treatment than what was actually given — and that these changes could both improve outcomes and reduce spending.

Three Different Paths for a Sick Heart

When arteries that feed the heart become narrowed, doctors usually consider three main options. One is medicine only, using drugs to control symptoms and risk factors. Another is a catheter‑based procedure that opens blocked vessels from inside. The third is bypass surgery, which reroutes blood around clogged arteries using blood vessels taken from elsewhere in the body. Large clinical trials have compared these options in selected groups of patients, but real‑world cases are often more complicated than the people who enroll in research studies. Age, other illnesses, anatomy, and personal preferences can all make the choice unclear, leaving doctors and patients to navigate uncertainty.

What the AI System Actually Does

The team evaluated an AI‑powered decision support system called Revaz AI. It was trained on data from more than 42,000 patients who underwent imaging of their heart arteries in Alberta, Canada. For each new patient, and for each of the three treatment options, Revaz AI predicts the chances of dying or suffering major heart‑related problems over the next three and five years. The researchers then fed these predictions into a detailed computer model that follows each patient over the rest of their life, tallying two things: the costs of care and the amount of time spent in good health, measured as "quality‑adjusted life years" (QALYs), a standard way to combine length and quality of life into a single number.

Testing Different Ways Doctors Might Use AI

Using records from 25,942 real patients, the authors replayed history on a virtual stage. For each person, the model simulated what would have happened under all three treatments, guided by Revaz AI’s risk forecasts. They then compared these simulated futures to what actually occurred, under three imaginary rules for how doctors might use the AI tool. In the first rule, doctors chose the option that delivered the best mix of cost and health, assuming the health system was willing to pay a standard amount for each extra year of good‑quality life. In the second, doctors ignored cost and picked the option that produced the most healthy life years. In the third, more cautious rule, doctors stuck to their original plan unless the AI suggested a different treatment that clearly improved the patient’s expected quality of life.

Big Shifts Toward Surgery and Better Value

Across these scenarios, the AI‑guided choices often differed from the original decisions, most commonly shifting patients toward bypass surgery rather than catheter procedures or medicine alone. Under the cost‑conscious rule, nearly three quarters of patients in the main analysis would have received a different treatment, leading on average to both lower lifetime costs and more time in good health. Even under the cautious rule, where doctors changed course only when the AI showed a clear benefit, more than half of patients would have been treated differently, again with modest cost savings and meaningful health gains. When the model focused purely on maximizing healthy life years, some strategies became more expensive overall but still offered good value for money by common health‑policy standards.

Why These Results Matter — and Their Limits

The study suggests that AI‑based decision support could help health systems get more benefit from every dollar spent on complex heart care. By tailoring choices to each patient’s predicted future — not just today’s anatomy — the tool often favors treatments that cost more upfront but prevent repeat procedures and hospitalizations later, especially bypass surgery. However, the findings represent a best‑case estimate. The model assumes that the AI’s risk predictions are accurate, that doctors can always follow its recommendations, and that future patients resemble those in the Alberta data and older clinical trials. In real life, surgical capacity, patient frailty, personal preferences, and evolving medical techniques will all constrain how closely practice can match the simulated ideal.

What This Means for Patients and Health Systems

For patients, the takeaway is not that a computer should decide their treatment, but that better use of data can sharpen the conversation with their doctors. For health systems, the work shows that AI is not just a high‑tech gadget but a tool that could meaningfully improve both outcomes and budgets when thoughtfully integrated into care. Before such systems can be widely adopted, randomized clinical trials will need to confirm that using AI in real clinics truly changes decisions in the ways this simulation predicts — and that those changes translate into longer, healthier lives at acceptable cost.

Citation: Mullie, T., Puri, A., Bogner, E. et al. Health economic simulation modeling of an AI-enabled clinical decision support system for coronary revascularization. npj Digit. Med. 9, 249 (2026). https://doi.org/10.1038/s41746-026-02430-x

Keywords: coronary artery disease, artificial intelligence, clinical decision support, health economics, bypass surgery