The course of rheumatoid arthritis was analyzed in 50 newly-diagnosed adults followed prospectively for an average of over five years. Eighteen patients had no swelling and no erosion (outcome 1), 22 patients had swelling without erosion (outcome 2) and 10 patients had erosion (outcome 3). Computer-assisted data screening and multivariate analytic technics were employed to derive a simple index from entry data which correlated closely with the patients' subsequent articular course. The index included six entry items: positive rhematoid factor; two or more swollen upper extremity joints on examination (2 points each); history of Raynaud's-like symptoms; malaise or weakness at the onset of arthritis; white race and female sex (1 point each). This index was 80 per cent accurate over-all in classifying patients into the three mutually exclusive articular outcome categories. Zero to 2 points on entry predicted no swelling, 3 to 5 points swelling without erosion and 6 to 8 points the development of bony erosion. Although results must be qualified to this patient population in the early stages of disease, the analytic approaches developed promise to assist with a broad range of clinical research and medical care delivery problems.