Algorithm uses EHR data to identify undiagnosed hypertension patients

A computerized algorithm used to analyze electronic health record data from 10 health centers was able to improve the identification and diagnosis of patients with hypertension in underserved communities.

The 10 health centers in Arkansas, California, Kentucky and Missouri were selected for the EHR analysis because they had a high prevalence of hypertension in their respective patient populations, compared with the national average, according to Margaret Meador, director of clinical integration and education at the National Association of Community Health Centers.

Meador, who led the study funded by the Centers for Disease Control and Prevention, contends that researchers developed an algorithm specifically for use in safety-net health centers to confirm elevated blood pressure readings and to diagnose patients with hypertension. Ultimately, Meador says what they discovered by mining EHR data were cases of undiagnosed hypertension.

“After implementation of algorithm-based interventions, diagnosed hypertension prevalence increased significantly from 34.5 percent to 36.7 percent,” according to the study, published in the March issue of The Joint Commission Journal on Quality and Patient Safety. “A cohort of patients was tracked from eight of the 10 health centers to assess follow-up evaluation and diagnosis rates; 65.2 percent completed a follow-up evaluation, of which 31.9 percent received a hypertension diagnosis.”

As a result, researchers concluded that undiagnosed hypertension is “hiding in plain sight” because of “provider inertia” characterized by a “conservative wait-and-see approach” in which clinicians order further follow-up “rather than diagnosing hypertension for those with repeated elevated blood pressure readings.”

Evaluating the pattern of blood pressure readings over time and making sure that these measurements were accurate was critical to the study, according to Meador, the lead author. “What you don’t want is a scenario where providers don’t believe the readings,” she notes. “We really had a lot of emphasis on ensuring accuracy of measurement.”

Blood-pressure-CROP.jpg
Medical assistant Astrid Garcia, center, takes a patient's blood pressure at a Community Clinic Inc. health center in Takoma Park, Maryland, U.S., on Wednesday, April 8, 2015. Led by the American Medical Association, three of the top five spenders on congressional lobbying have waged a campaign to urge Congress to revamp the way Medicare pays physicians and end the cycle of "doc fix" patches. Senate leaders predict quick action on House-passed legislation when Congress returns April 13 from its two-week recess. Photographer: Andrew Harrer/Bloomberg *** Local Caption *** Astrid Garcia

Also See: EHR analysis shows blood pressure can be too low

“It was a surprising finding that one in three patients were actually walking around with hypertension,” observes Meador. “I think all the health centers in the project were shocked at how many patients they had hiding in plain sight. However, this isn’t a health center issue. This is a national issue.”

Writing in an editorial accompanying the study in The Joint Commission Journal on Quality and Patient Safety, Stephen Persell, MD, associate professor at Northwestern Medicine, points out that about 32 percent of the U.S. adult population has hypertension, but 16 percent are not aware they have the condition. However, he asserts that there are system-level approaches that can be put into place to help make sure that opportunities to properly diagnose hypertension are not wasted.

“Meador and colleagues demonstrate the practical application of a population health strategy driven by electronic health record data to improve the diagnosis of hypertension in health centers that provide care to otherwise underserved populations and communities,” states Persell. “Their approach builds on prior work that systematically analyzed EHR data to find individuals with high blood pressure measurements and no hypertension diagnosis.”

While the CDC-funded study focused on health centers, Meador believes that the algorithm and supporting care processes could be applied to other healthcare settings, although it would require testing. “It’s not a highly sophisticated algorithm, but it can be translated into more sophisticated tools,” she adds. “It’s possible to do this in any EHR.”

For reprint and licensing requests for this article, click here.