We prospectively studied outcomes for 179 patients with low back pain. Predictors of future function, employment, and medical utilization were drawn from 21 clinical, demographic, and psychosocial variables using multivariate techniques. Education, previous episodes, and whether the patient "always feels sick" were independently associated with most outcome measures, but prescribed therapy and physical findings were not. These 3 items created a scale defining subgroups with 3-fold differences in outcomes (e.g., 35% functionally improved in the worst group vs 93% in the best, p less than 0.001). Data from a national survey supported the importance of education and self-rated health as correlates of back related disability.