Using machine learning to uncover joint involvement patterns linked to disease activity and disability in rheumatoid...

Rheumatoid arthritis (RA) is a chronic inflammatory disease marked by painful, progressive joint damage that leads to disability and a significant socioeconomic burden. Given the growing interest in data-driven healthcare, this study applied machine learning (ML) techniques to identify patterns of joint involvement associated with greater disease activity and functional impairment in patients with RA.

In this cross-sectional study, 936 individuals with RA were assessed using demographic and clinical data, along with detailed joint examinations, including tenderness, swelling, and limitation of motion. Multiple ML models, including random forest, extreme gradient boosting (XGB), k-nearest neighbors, CatBoost (CAT), and LightGBM, were evaluated to predict disease severity and disability. Disease activity and disability were measured using Disease Activity Score based on 28 joints (DAS28) and Health Assessment Questionnaire Disability Index (HAQ-DI) scores, respectively.

For disability prediction, the XGB model performed best (F1 score, 0.60; AUC, 0.85), while the CAT model showed the highest performance for disease activity classification (F1 score, 0.63; AUC, 0.81). Key joint features associated with higher disability included the knee, wrist, acromioclavicular joint, and several hand joints. In contrast, high disease activity was primarily linked to tenderness and swelling in the knee and small hand joints.

Overall, involvement of the knee, wrist, shoulder, and small hand joints was associated with functional disability, whereas disease activity was mainly reflected in the knee and hand joint involvement. These findings emphasize the importance of recognizing specific joint involvement patterns to improve clinical assessment and guide treatment strategies in RA. Key Points • This study is the first to apply machine learning to comprehensively evaluate joint-specific patterns of involvement and their association with both functional disability and disease activity in rheumatoid arthritis. • Large joints, particularly the knee, wrist, and shoulder, were identified as the strongest determinants of significant functional disability, alongside key small hand joints involved in grip and fine motor function. • High disease activity was primarily driven by small hand joints (especially MCP and PIP joints), with additional contributions from selected large joints such as the knee, wrist, and acromioclavicular joint. • These findings highlight the clinical value of detailed joint-level assessment beyond traditional composite scores, supporting more targeted and personalized management strategies in rheumatoid arthritis.