Good generalizability of a prediction rule for prediction of persistent shoulder pain in the short term

Abstract Objective To evaluate the generalizability of recently developed clinical prediction rules for the prognosis of shoulder pain in general practice. Study Design and Setting A large research program, consisting of a prognostic cohort study and three randomized controlled trials with 6 months follow-up, was carried out in The Netherlands. The clinical prediction rules were derived from the results of the prognostic cohort study ( n = 587). The main outcome measure was persistent symptoms at 6 weeks or 6 months. The control groups of the trials who received usual care were merged ( n = 212), and used to validate the prediction rules by studying calibration and discrimination. Results The prediction rule for short-term outcome showed reasonable calibration and discriminative ability in this validation cohort. The area under the receiver operating characteristic curve (AUC) was 0.72 compared to 0.74 in the derivation cohort. The prediction rule for long-term outcome performed less well. Discriminative ability (AUC) decreased to 0.56 in the validation cohort compared to 0.67 in the derivation cohort. Conclusion The prediction rule for the short-term (6 weeks) prognosis showed good generalizability. The prediction rule for the long-term prognosis showed poor generalizability.