Every year 735,000 Americans suffer a heart attack. To help prevent heart attacks, health care providers use risk calculators to predict who is likely to develop cardiovascular disease. A potential limitation with the calculators is they were developed using data from people in population-based studies who might be different from patients seeing their doctor for a checkup. A new study by researchers at the University of Minnesota School of Public Health tested two popular cardiovascular risk calculators using patient electronic health data (EHD) and found that they maintain their accuracy at predicting cardiovascular risk when they are used in a clinical setting.
The study was published in the Journal of the American Heart Association.
“This is the first study to look how well the models are working for people who are actually showing up in the clinic,” said lead author and assistant professor Dr. Julian Wolfson.
The study tested the performance of the two most popular calculators — the Framingham Risk Score and the ACC/AHA Pooled Cohort Equations — using EHD from patients in a large metro-area health system.
“We found that both models performed reasonably well,” said Dr. Wolfson. “On top of that, the performance of the models could be improved by tailoring it to work with a health care provider’s particular electronic health data system.”
The study also found that the Framingham Risk Score model outperformed the Pooled Cohort Equations in their test with real-world data despite being an older and simpler calculator.
“The people who should take notice of the results are those who work with decision support systems and put software on clinic computers or advise health care providers on assessing cardiovascular disease risk,” said Dr. Wolfson. “This study suggests that people should use or continue using the Framingham Risk Score. And we found that tailoring the risk calculators to work with a health system’s own data will improve the ability to predict risk.”
Dr. Wolfson is conducting additional research focusing on creating new risk models that are specifically designed to work with EHD as well as examining how best to measure the accuracy of these models.