By Gregory Dorn, MD, MPH, President of Hearst Health
As advanced technologies such as artificial intelligence (AI) gain popularity, the fundamental mission of healthcare—to improve health—should be kept front and center. Recently an AI program from Dr. Alexander Sandhu of Stanford Medicine won national recognition for doing just that. Its AI solution helps find patients at risk for heart attack by identifying coronary calcium in existing chest CTs.
Part of the elegance of Stanford’s program is that it shows the patient what is actually happening in their own body, which is much more compelling than a risk factor statistic. “When you pull up an image of their heart and you circle the calcium in their heart and say, ‘That's not normal,’ all of a sudden their motivation for behavior change changes substantially,” said Sandhu.
Stanford’s Incidental Coronary Calcium team aimed to use the detection of coronary artery calcium from computed tomography (CT) of the chest to improve the primary prevention of atherosclerotic cardiovascular disease. Coronary artery calcium—an established predictor of heart attack and stroke—can be identified on chest CTs.
About 15 million chest CTs are performed for various reasons in the US each year, while chest CTs specifically to detect coronary artery calcium are conducted only about 60,000 times per year. The Stanford team developed an AI algorithm that searches existing chest CTs in the patient record to identify calcium deposits and present this information to primary care physicians.
A multi-center study of the program showed that across the patients identified by Stanford’s algorithm, the following results were achieved compared with usual care:
- Statin therapy: 51.2% received a statin prescription versus 6.9% with usual care (p<0.001)
- Shared decision making: 77.9% had a statin discussion or prescription compared with 12.0% with usual care (p<0.001)
- LDL cholesterol reduction: 97.2 mg/dL versus 115.3 mg/dL with usual care (p=0.005)
The program was named winner of the 2023 Hearst Health Prize, a $100,000 award for excellence in data science in healthcare we offer annually in partnership with the UCLA Center for SMART Health. Dr. Sandhu presented his program at UCLA Health Data Day on May 11.
Using AI to Move the Needle on Health Outcomes
The boon in health IT activity over the past decade has yielded myriad innovations in data science applications. But to what extent are these applications actually helping improve health outcomes? Even with the volume and pace of innovation and the increasing investment in data science, very few applications demonstrate a measurable and reproduceable improvement in patient health outcomes. Programs that demonstrate a real health impact should be recognized and promoted such that best practices proliferate.
AI tools that support clinical judgment will help clinicians be more effective clinically. Arash Naeim, MD, PhD, co-director of UCLA Center for SMART Health, commented “Gaining clinically actionable insights from existing electronic health record data makes our healthcare system better for patients and clinicians alike.”
In the Stanford example, AI does the heavy lifting of evaluating a large volume of records to isolate those with significant calcifications that present a real risk to patients’ cardiovascular health. By AI automating analysis that would take many clinicians hours of review, AI enables clinicians to be more productive and effective with the care they deliver.
“Deploying AI to identify patient attributes across a vast dataset can help alleviate the workload of clinicians while empowering them to deliver timely care,” said Alex Bui, PhD, co-director of the UCLA Center for SMART Health.