The aim of this study was to develop a predictive model for treatment outcome in patients with low back pain (LBP) receiving chiropractic treatment.This multicenter practice-based predictive validity study was conducted in private chiropractic practices in Sweden. Of 64 previously compliant chiropractors, 58 recruited a maximum of 30 consecutive patients with LBP each. Information was provided on 1061 patients, of which 1057 questionnaires were valid. Chiropractic treatment was decided by the treating chiropractor. The outcome variable was the self-reported “definite improvement” at the fourth visit. The predictor variables included model 1, 3 hypothesized prognostic groups (best, intermediate, and least favorable) based on clinical information collected at baseline and at the second visit; and 4 additional models based on the following variables: age, sex, pain intensity during past 24 hours, description of disability, duration and pattern of pain during present attack, total duration of pain, and pain pattern during the past 12 months.Three of our factors were best at predicting the absence of improvement by the fourth visit and were able to correctly classify 79% of patients and to cover 74% of the receiver operated characteristics curve. These were (1) no definite overall improvement by the second treatment, (2) presence of leg pain, and (3) the minimum total duration of pain over the last 12 months being 30 days.In this study, patients with LBP who also had leg pain and LBP occurring sufficiently frequently or having lasted sufficiently long to add up to at least 30 days in the past year, and who did not report definite general improvement by the second treatment were not good candidates for short-term recovery. It is suggested that patients who fit the criteria of potential nonresponders should be carefully monitored to allow a selective approach of care.