Prediction model underestimates CV risk in disadvantaged neighborhoods

A widely accepted model systematically underpredicts major atherosclerotic cardiovascular disease (ASCVD) risk in disadvantaged communities, according to a new study in Annals of Internal Medicine.

Cleveland Clinic researchers created a neighborhood disadvantage index (NDI) for northeastern Ohio, which accounted for more than three times the amount of geographic variability in ASCVD event rates when compared to the Pooled Cohort Equations Risk Model (PCERM) of the American College of Cardiology and American Heart Association.

The PCERM accounted for 10 percent of variation in ASCVD event rates at the census-tract level, while the NDI accounted for 32 percent.

“Efforts are needed to enhance risk prediction by incorporating aspects of neighborhood SEP (socioeconomic position) and discerning its systemic effects on individuals,” wrote lead researcher Jarrod E. Dalton, PhD, and colleagues. “Such efforts are particularly important in the context of health disparities in ASCVD, whereby the mechanisms involved in ASCVD progression may differ qualitatively among subpopulations defined according to social strata.”

According to the authors, “although the goal of the PCERM was to establish more demographically representative models for ASCVD events, it did not incorporate variation in risk directly related to SEP.”

This has financial implications, the authors noted, because reimbursement rates for some incentive programs are tied to the PCERM and may unfairly penalize providers who treat socioeconomically challenged populations.

To create the NDI, researchers used the following variables from the 2010 U.S. Census: percentage white, non-Hispanic; percentage with a high school degree; percentage with Medicaid, aged 18 to 64 years; percentage uninsured, aged 18 to 64 years; median income; percentage of households below the federal poverty level; percentage of children living in households receiving supplemental security income, cash public assistance income, food stamps, or Supplemental Nutrition Assistance Program benefits; and percentage of households headed by an unmarried mother.

Dalton et al. analyzed patients from 21 northeastern Ohio counties who had at least one outpatient lipid panel performed from 2007 through 2010, stated they were of white or black race and were at least 35 years old. The study was limited to black and white participants because the PCERM is only applicable to those groups.

After establishing a sample of 109,793 patients, the authors considered the time from baseline to the first occurrence of a major ASCVD event—defined as MI, stroke or cardiovascular death—in relation to the PCERM prediction. The follow-up period was five years.

In addition to the personal challenges of poverty, the researchers pointed out, residents in disadvantaged neighborhoods typically have less access to healthy food, exercise options and preventative cardiovascular care.

In an accompanying editorial, Sandro Galea, MD, Boston University; and Katherine M. Keyes, PhD, Columbia University; wrote the study is “a good first step” in reconceptualizing predictive health models.

“Although ultimately all health is ‘under the skin,’ mediated by individual behavior and biology, both behavior and biology are shaped by a person's social networks, neighborhoods, policies that influence access to healthy resources, and social norms that guide behaviors,” they wrote. “Dalton and colleagues illustrate this concept nicely through the predictive power of the NDI. Clearly, neighborhood disadvantage produces a set of conditions that increase the risk faced by those living in marginalized neighborhoods in ways that an individual risk score simply cannot capture.”