|| Checking for direct PDF access through Ovid
Objectively identifying patients at baseline who may not respond well to a generic muscle strengthening intervention could improve clinical practice by optimizing treatment strategies. The purpose of this study was to determine whether pelvic acceleration measures during running, and clinical and demographic variables could classify patellofemoral pain patients according to their response to a 6-week hip/core and knee exercise-based rehabilitation protocol.Forty-one individuals with patellofemoral pain participated in a 6-week exercise intervention program and were sub-grouped into treatment Responders (n=28) and Non-responders (n=13) based on self-reported pain and function measures. Baseline pelvic acceleration measures were reduced using a principal component analysis and combined with patient reported outcome measures and demographic variables in a support vector machine to retrospectively classify patient treatment response.The final classification model had 85.4% classification accuracy, which was significantly better than treatment success rate, with excellent detection rates for Responders (recall: 96.4%), but 23.1% of misclassifications among Non-responders (precision: 90.0%). Thus, it resulted in an F1-score of 0.93 and a Matthews correlation coefficient of 0.69.Overall, the classifier successfully separated patellofemoral pain patients into exercise-based treatment Responders and Non-responders based on a combination of three components of the pelvic accelerations. While this model requires independent validation, it has the potential for further development and to be applied in clinical practice and improve treatment strategies for patellofemoral pain.Pelvic acceleration can be used to classify sub-groups of patellofemoral pain treatment outcome.The proposed classifier is specific to hip/core and knee muscle strengthening protocols.This classification model better identifies Responders to exercise treatment.This is a preliminary evidence-informed method for treatment response prediction.