Abstract
Rheumatoid arthritis (RA) affects approximately 1% of the global population and causes progressive, irreversible joint destruction if not treated early. The challenge is that early RA can be difficult to diagnose — symptoms are nonspecific, serological markers are not always present, and patients may cycle through primary care visits for months or years before receiving a definitive diagnosis. By the time treatment begins, significant joint damage may have already occurred.
An EHR-based predictive model for RA onset would address this problem by identifying high-risk individuals from the pattern of their clinical encounters — laboratory values, visit frequencies, prescribed medications, diagnostic codes — before a rheumatologist makes the diagnosis. Such a model could function as a scalable, low-cost screening tool integrated into existing clinical workflows, flagging patients for targeted evaluation and potentially accelerating the time to treatment.
[Additional details on study design, dataset, features, and modeling approach to be added.]