Artificial Intelligence to Predict Rheumatoid Disease Activity Save

A recent study in JAMA Network Open shows that artificial intelligence models can use electronic health record data to prognosticate future patient outcomes in rheumatoid arthritis (RA).

The hypothesis was whether artificial intelligence (AI) models were capable of forecasting future patient outcomes, in a complex disorder like RA.  Model performance was quantified using the area under the receiver operating characteristic curve (AUROC). Disease activity in RA was measured using the CDAI - a composite index score of RA activity.

A prospective study included 820 RA patients from 2 separate and dissimilar medical centers.  The electronic health record (EHR) served as the source for patient data, medication exposures, demographics, laboratories, and prior measures of disease activity. A longitudinal deep learning model was used to predict disease activity for patients with RA at their next rheumatology clinic visit and was compared to last disease activity for predictive value.

A total of 820 RA patients were included; they had a mean age of 57-60 yrs, and nearly 80% were female.

At one site, the model reached an AUROC of 0.91 (95% CI, 0.86-0.96) in a test cohort of 116 patients. The second site training model had an AUROC of 0.74 (95% CI, 0.65-0.83) in 117. 

By comparisoins prediction of clinical status using each patients’ most recent disease activity score had statistically random performance and inferior to the AI models.

Thus, longitudinal deep learning model had strong performance in a test cohort of 116 patients, whereas baselines that used each patient’s most recent disease activity score had statistically random performance.