AI Can Identify People at Risk of Cardiovascular Complications
Researchers at the University of Utah Health have for the first time demonstrated that artificial intelligence could lead to better ways of predicting the onset and course of cardiovascular diseases.
Working with physicians at Intermountain Primary Children's Hospital, the researchers developed unique computational tools to precisely measure the synergistic effects of existing conditions on the heart and blood vessels.
The researchers said this comprehensive approach could help doctors anticipate, prevent or treat serious heart problems, possibly even before a patient learns of the underlying condition.
Although the study focused only on cardiovascular diseases, the scientists believe it could have much broader significance. They suggest these findings could ultimately usher in a new era of personalized preventive medicine. Doctors will actively contact patients to warn them about possible diseases and what can be done to ease the problem.
“We can turn to AI to clarify risk for practically every medical diagnosis. Cancer risk, thyroid surgery risk, diabetes risk — any medical term you can imagine”
According to Mark Yandell, PhD, senior author of the study, and Edne Benning, president of U of U Health and co-founder of Backdrop Health, existing methods for calculating the combined impact of various risk factors, such as demographics and medical history, on cardiovascular disease are often inaccurate and subjective. As a result, these methods fail to identify certain interactions that can have serious consequences for heart and blood vessel health.
To more accurately measure how these interactions, also known as comorbidities, affect health, Tristani-Firouzi, Yandell and their colleagues at U of U Health and Intermountain Primary Children’s Hospital used machine learning software to sort through more than 1.6 million electronic health records (EHRs) after names and other identifying information were removed.
These electronic records, which document everything that happens to a patient, including lab tests, diagnoses, medication use and medical procedures, helped researchers identify comorbidities most likely to worsen a specific condition, such as cardiovascular disease.
In their current study, they used a form of artificial intelligence called probabilistic graphical networks (PGM) to calculate how any combination of these comorbidities may affect risks associated with heart transplantation, congenital heart defects or sinus node dysfunction (SND, a disruption or failure of the heart's natural rhythm driver).
Among adults, the researchers found that:
- People previously diagnosed with cardiomyopathy syndrome (a heart muscle disease) were 86 times more likely to need a heart transplant than those without it.
- Those with viral myocarditis had roughly 60 times higher risk of needing a heart transplant.
- Use of milrinone, a vasodilator drug used to treat heart failure, increased transplant risk 175-fold. It was the strongest single predictor of heart transplantation.
In some cases, the combined risk was even higher. For example, among patients with cardiomyopathy who took milrinone, the risk of needing a heart transplant was 405 times higher than for those with healthier hearts.
According to Tristani-Firouzi, comorbidities had markedly different effects on transplant risk in children. Overall, the risk of heart transplantation in children ranges from 17 to 102 times higher than in children with no prior heart conditions, depending on the primary diagnosis.
The researchers also examined the impact of maternal health during pregnancy on their children. Women who had high blood pressure during pregnancy were roughly twice as likely to give birth to children with congenital heart and circulatory problems. In children with Down syndrome, the risk of a heart anomaly is about three times higher.
In infants who underwent a Fontan operation to correct a congenital blood flow defect in the heart, the likelihood of developing SND was about 20 times higher than in those who did not need the surgery.
The researchers also found important demographic differences. For example, in a Hispanic patient with atrial fibrillation (rapid heartbeat), the risk of SND was twice as high as in Black and White patients with similar medical histories.
Josh Bonkowski, MD, director of the Primary Children's Center for Personalized Medicine, who was not an author of the study, believes the research could lead to a practical clinical tool for patient care.
“This new technology shows that we can accurately estimate the risk of medical complications and even identify medications better suited to individual patients”
Looking ahead, Tristani-Firouzi and Yandell hope their research will also help doctors untangle the growing web of disorienting medical information that surrounds them every day.
“No matter how knowledgeable you are, there is no way to keep in your head all the knowledge you need as a medical professional these days to treat patients in the best way. The computational machines we are developing will help doctors make the best possible patient care decisions, using all relevant information available in our electronic age. These machines are vital to the future of medicine”
