A predictive risk model of low-back pain (LBP) disability was developed by a panel of six experts in the fields of chronic pain and disability. It comprised 28 factors organized into eight categories: job, psychosocial, injury, diagnostic, demographic, medical history, health behaviors, and anthropometric characteristics and was administered as a 15-minute written questionnaire. The model was tested prospectively on 250 patients (age range, 18–65 years) attending two secondary-care low-back clinics. Disability, as predicted by the model, was compared with 1) actual disability assessed 3 and 6 months later; 2) predictions of disability made by the attending physicians; and 3) predictions obtained from an empirically derived model. These results showed that 1) the expert-generated risk model had a predictive accuracy of 89% and did better in predicting disability than the physicians across all samples and 2) the empirically weighted model did best of all (91% predictive accuracy), suggesting that the expert model used appropriate factors but that the weights assigned to these factors by the panel of experts could be improved.