Stay on target
Aside from the almost-instantaneous warning signs—chest pain, shortness of breath, nausea—heart attacks are notoriously hard to predict.
Scientists at the University of Nottingham, however, may have a new way of anticipating heart failure among patients.
According to Stephen Weng, an epidemiologist at the UK university, established risk assessment techniques—those recommended by the American Heart Association (AHA) and American College of Cardiology (ACC)—no longer cut it.
Standard models, he explained, tend to “oversimplify” cardiovascular disease (CVD); eight core baseline variables—gender, age, smoking status, systolic blood pressure, blood pressure treatment, total cholesterol, HDL cholesterol, diabetes—don’t automatically foretell a heart attack.
“There remain a large number of individuals at risk of CVD who fail to be identified” by current tools, “while some individuals not at risk are given preventive treatment unnecessarily,” Weng and his colleagues explained in a recent research paper.
“Approaches that better incorporate multiple risk factors, and determine more nuanced relationships between risk factors and outcomes, need to be explored,” the study said.
Machine learning offers that alternative approach, exploiting big data to minimize human error.
Using four computer-learning algorithms, Weng and his team input data for 378,256 CVD-free patients, aged 30 to 84, from nationwide medical practices. More than 295,000 served to train the machines, while the remaining 82,989 subjects provided validation.
Nearly 25,000 fatal or non-fatal cardiovascular events were documented over the study’s 10-year period (Jan. 1, 2005 to Jan. 1, 2015)—about 75 percent of which were accurately predicted by Weng’s algorithms.
The neural network formula tested highest, beating existing guidelines by 7.6 percent. In fact, the MVP algorithm correctly identified 355 more patients than human doctors, proving its ability to help save lives.
“Unlike established approaches to risk prediction, the machine learning methods used were not limited to a small set of risk factors, and incorporated more pre-existing medical conditions,” the research paper said.
“Machine learning approaches offer the exciting prospect of achieving improved and more individualized CVD risk assessment,” the team continued. “This may assist the drive towards personalized medicine, by better tailoring risk management to individual patients.”