Researchers leverage EHR data to identify at-risk heart patients earlier
A team of researchers at the University of North Carolina School of Medicine and UNC Gillings School of Global Public Health have developed an algorithm using electronic health record (EHR) data to identify tens of thousands of patients across the state who face an increased risk of developing cardiovascular disease.
Principal investigator Sam Cykert, MD, a professor of medicine at UNC-Chapel Hill, said the university’s Heart Health Now! program took advantage of the fact that nearly all patients in the U.S. health system now have some kind of electronic medical record, since that wasn’t always the case.
“Over the years, I’ve cared for many people who suffered the debilitating side effects of a heart attack or stroke much too early,” he said in a release. “Because of the lack of sophisticated information systems and processes, doctors could not quickly identify patient risk and prioritize new evidence for care. So many of these folks missed opportunities that could have prevented severe outcomes resulting from premature disease.”
For their research, Cykert and his colleagues crafted a network of 219 primary care clinics in North Carolina, more than half of which were situated in rural areas. Those clinics treated a total of 345,440 primary care patients aged 40 to 79, but more than 30 percent of those patients lacked cholesterol scores in their EHR records. Without those numbers, physicians can’t formally determine a person’s cardiovascular risk.
Cykert et al. leveraged data from the 70 percent of patients who did have cholesterol data to devise an equation that would estimate the cholesterol levels of patients with missing information. The researchers said the formal—and admittedly impractical—method of data imputation allowed them to identify 43,205 individuals in North Carolina who were at an increased risk of heart disease but had never been identified before.
In a second attempt, the team substituted conservative cholesterol numbers (170 mg/dL for total cholesterol and 50 mg/dL for HDL) for the patients who lacked information. Using that approach and accounting for other risk factors, like smoking status, weight, exercise habits and gender, they identified 40,565 patients at a high risk for heart disease.
Cykert said around half of patients with missing data got their cholesterol checked during the course of the study, allowing the researchers to test the sensitivity and specificity of their equations. They found the formal method, which took far longer, yielded a lower specificity and a higher false-positive rate than the estimation.
“Whether doctors are part of a large health system or in a small rural practice, the fact that all these patients have digital data now means we can identify patients who are at high risk of developing a very serious condition without waiting six months for them to make an appointment,” Cykert said. “Doctors can engage with these patients immediately and re-engage with them as needed to decrease risk, which is so crucial when it comes to decreasing the number of heart attacks and strokes.”
Cykert’s team was one of seven to receive a $15 million federal grant from the Agency for Healthcare Research and Quality in 2015. Their work was published in the Journal of the American Medical Informatics Association in late November 2018.
After graduating from Indiana University-Bloomington with a bachelor’s in journalism, Anicka joined TriMed’s Chicago team in 2017 covering cardiology. Close to her heart is long-form journalism, Pilot G-2 pens, dark chocolate and her dog Harper Lee.