Socioeconomic status may influence a patient's health, but researchers don't need this information to develop accurate predictive models, according to a study published in JAMA.
Researchers from Durham, N.C.-based Duke University used machine learning — a type of artificial intelligence — to develop risk models to predict health events through the use of EHR data, neighborhood socioeconomic status data or both.
The researchers trained the models on data from more than 90,000 patients, including EHR data from adults who lived in North Carolina's Durham County and visited the Duke University Health System and the Lincoln Community Health Center, both in Durham, N.C., between 2009 and 2015. They linked this information with neighborhood socioeconomic status data based on census tracts.
They tested the two models in more than 122,000 patients, and found how well neighborhood socioeconomic status data predicted health events depended on the outcome in question. However, using neighborhood socioeconomic status data did not improve predictions for any health events compared to models that only used EHR data.
"The results of this study suggest that information on [neighborhood socioeconomic status] may not contribute much more to risk prediction above and beyond what is already provided by EHR data," the study authors concluded. "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."