Patient questionnaires and formal education as more significant prognostic markers than radiographs or laboratory tests for rheumatoid arthritis mortality--limitations of a biomedical model to predict long-term outcomes.