Census tract-level socioeconomic data, while invariably linked to health outcomes, contributes little to risk prediction models beyond what’s already available through electronic health records (EHRs), Duke researchers report in JAMA Network Open this month.
Social determinants of health—things like a patient’s ancestry, language proficiency, income and employment characteristics—have always factored into medicine, corresponding author Benjamin A. Goldstein, PhD, and colleagues wrote in JAMA. But it wasn’t until more recently that clinicians began using EHR data to predict disease risk in individual patients.
“A key limitation of EHRs, when used for research purposes, is that they do not reliably collect sociodemographic and neighborhood information, which has long been recognized to be strongly associated with health,” Goldstein, of the Department of Biostatistics and Bioinformatics at the Duke University School of Medicine in Durham, North Carolina, and co-authors said. “Social and behavior measures linked to clinical variables within EHRs may improve clinical care and population health while also helping to inform population-level risk reduction strategies.”
A handful of clinical trials have suggested combining neighborhood socioeconomic status (nSES) indicators with disease risk factors can improve the accuracy of prediction models, including work that’s shown nSES variables improve the accuracy of the Framingham risk score in estimating coronary heart disease risk. To establish whether the approach is applicable across other specialties, Goldstein et al. studied data from 90,097 patients logged in Duke University Health System’s EHR.
The researchers linked patients’ medical records with census tract data to quantify the association between nSES and the risk of adverse outcomes, they said. Machine learning was applied to develop risk models, and Goldstein’s team assessed how those models changed with the input of various nSES values.
Of the patient pool, the authors said those who lived in neighborhoods of a lower socioeconomic status saw shorter times to the use of emergency services, inpatient encounters and hospitalizations due to accidents, asthma, influenza, myocardial infarction and stroke. When added to existing EHR variables, nSES didn’t seem to improve predictive performance for any health outcome.
“This work reaffirms that the social environment is associated with health outcomes,” Goldstein and colleagues said. “However, these results suggest that information about the environment in which a person lives may not contribute much more to population risk assessment than is already provided by EHR data.”
Still, the authors were careful not to rule out nSES data entirely.
“Although this result does not mean that integrating social determinants of health into the EHR has no benefit, researchers may be able to use EHR data alone for population risk assessment,” they said.