Health status measures in rheumatoid arthritis that have been extensively validated for use in clinical trials are generally used also in correlative studies, e.g. to predict future health status. This application requires stability (repeatability of measurements). The purpose of our study was to determine the stability of commonly used health status measures. Two measurements at an interval of 6 months were then in 99 patients. High stability (alpha = 0.78 to 0.94) was observed for five biomedical measures (grip strength, walking time, platelet count, haemoglobin and erythrocyte sedimentation rate) and five self-report measures (mobility, self care, impact daily activities, anxiety and cheerful mood). Moderate stability (alpha = 0.65 to 0.72) was observed for joint scores, pain, C-reactive protein and depressive mood. The highly stable measures most adequately reflect individual differences, may be applied most reliably in correlative studies and appear to have the largest clinical utility with regard to long-term prediction of health status.