Related Experiment Video
Updated: Jun 11, 2026
07:44
Evaluation of Changes in Hydration and Body Cell Mass with Bioelectrical Impedance Analysis after Exercise Program for Rheumatoid Arthritis Patients
Published on: July 14, 2023
Electronic Health Record-Based Machine Learning Model for Predicting Disease Activity in Patients with Rheumatoid
Xiaoying Zhang1,2, Chun Li1,3, Zelin Yun1,3
1Department of Rheumatology & Immunology, Peking University People's Hospital, Beijing, China.
Health Data Science
|June 10, 2026
Summary
Machine learning models accurately predict rheumatoid arthritis (RA) disease activity using longitudinal electronic health records. This enables personalized treatment strategies for better patient management.
Area of Science:
- Rheumatology
- Medical Informatics
- Artificial Intelligence
Background:
- Precision medicine increasingly utilizes machine learning (ML) for predicting therapeutic outcomes.
- Initial clinical assessments are crucial for forecasting treatment success in chronic diseases.
- Rheumatoid arthritis (RA) management can benefit from predictive models to optimize clinical decision-making.
Purpose of the Study:
- To develop and validate ML models for predicting disease activity in RA patients.
- To leverage longitudinal electronic health records (EHRs) for prognostic modeling.
- To enhance personalized treatment selection and patient management in RA.
Main Methods:
- A multicenter retrospective study analyzed EHRs from 1,864 RA patients across 5 Chinese tertiary hospitals (2017-2022).
- Longitudinal data (demographics, labs, medications) at baseline, 3, and 6 months were used.
- Four ML models were trained to predict clinical remission (DAS28-ESR ≤ 2.6) at 6 months post-treatment.
Main Results:
- The optimal internal validation model achieved 95.3% accuracy and 0.971 AUROC for predicting remission.
- External validation demonstrated model generalizability with 87.3% accuracy and 0.922 AUROC.
- A deep neural network model accurately stratified disease activity (remission to high) with 68.6% accuracy and 0.860 AUROC.
Conclusions:
- Longitudinal clinical data from EHRs are effective for developing prognostic models in RA.
- Deep learning approaches trained on large, multicenter cohorts can accurately predict RA disease trajectories.
- These predictive models offer a valuable tool for personalized RA patient management.
