Prediction of an unfavourable course of low back pain in general practice: comparison of four instruments

Several instruments can be used to identify patients with an unfavourable course of low back pain in general practice. However, it is unclear which instrument is the predictor of outcome. To compare the predictive performance (that is, calibration and discrimination) of risk estimation by GPs with assessments using the Orebro Musculoskeletal Pain Screening Questionnaire, the Low Back Pain Perception Scale (LBPPS), and a prediction rule developed for this purpose. Design of study: A prospective cohort study with 1-year follow-up. General practice in The Netherlands. The outcome 'unfavourable course of low back pain' was defined as having no clinically important improvement at minimally 50% of the measurements at 6, 13, 26, and 52 weeks. Logistic regression analyses were used to study associations between potential predictors and outcome. In total, 60 GPs recruited 314 patients to the study (16 patients were excluded from analysis due to missing data on the course of low back pain). Over a third of patients (112/298) showed an unfavourable course of low back pain on follow-up. Risk estimation by GPs, the Orebro questionnaire, the LBPPS, and the prediction rule had discriminative ability (area under the curve) of 0.59 (95% CI [confidence intervals] = 0.52 to 0.66); 0.61 (95% CI = 0.54 to 0.67); 0.59 (95% CI = 0.52 to 0.66); and 0.75 (95% CI = 0.69 to 0.81) respectively. The prediction rule included history of low back pain, self-perceived risk to develop chronic low back pain, no solicitous responses of the patient's partner (as reported by the patient), frequent walking at work, and 'pain catastrophising'. Although the prediction rule performed best with regard to calibration and discrimination, it needs to be externally validated. Risk estimation by GPs performs as well as other instruments and, at present, seems to be the best available option