Predictive Modeling in RA may Start with EHR Data


Electronic health records data can be used to create forecast models of disease outcomes in rheumatoid arthritis patients, according to new research.

(©Wladimir1804, AdobeStock)

(©Wladimir1804, AdobeStock)

Electronic health records (EHRs) data can be used to create forecast models of disease outcomes in rheumatoid arthritis (RA) patients, according to new research.

Results published in the Journal of the American Medical Association revealed artificial intelligence systems can accurately predict a patient’s disease activity state at their next clinical visit. American College of Rheumatology data indicates 42 percent of RA patients had moderate-to-high disease activity at their most recent visit, suggesting more tools can personalize disease management.

“We aimed to used structured data from the EHR to build a model that would most accurately predict RA disease activity,” researchers said. “The ability to forecast disease activity could be clinically used to inform the aggressiveness of treatment on an individualized basis at each clinical visit.”

They discovered a patients’ clinical disease activity index (CDAI) from most recent visits was a poor future indicator. Instead, combining one-year disease activity history, laboratory values, and all medications was the strongest indicator.

To determine forecast model accuracy, researchers pulled EHR data on 578 university hospital (UH) and 242 public safety-net hospital (SNH) patients, including medications, patient demographics, laboratories, and prior disease activity measures. Researchers measured disease activity with composite index scores and quantified model performance using the area under the receiver operating characteristic curve (AUROC).

According to results, the UH-trained model reached an AUROC of 0.91 (95% CI, 0.86-0.96), accurately predicting next-visit uncontrolled disease activity in 115 of 117 test samples. The same model reached an AUROC of 0.74 (95% CI, 0.65-0.83) in the SNH population. Using the UH-trained model is both settings demonstrated larger model training size efficacy.

Subsequent predictive models will include data from multiple large and small institutions, investigators said. Using EHRs from more diverse patient populations will create the best system to help physicians understand and predict disease trajectories.

“In the future, models built from large pooled patient populations may be the most accurate, giving everyone access to the most robust models trained on the largest and most diverse patient populations possible,” they said. “The methods used to develop models for predicting RA disease activity may be informative for other heath conditions with quantifiable outcomes.”



Norgeot B, Glicksberg B, Turpin L, Lituiev D, Gianfrancesco M, Oskotsky B, Schmajuk G, Yazdany J, Butte A, Assessment of a Deep Learning Model Based on Electronic Health Record Data to Forecast Clinical Outcomes in Patients With Rheumatoid Arthritis. JAMA Network Open (2019), doi: 10.1001/jamanetworkopen.2019.0606.

Related Videos
Signs and Symptoms of Connective Tissue Disease
Connective Tissue Disease Brings Dermatology & Rheumatology Together
© 2024 MJH Life Sciences

All rights reserved.