265 The benefit of subgroup analysis when predicting rehabilitation outcomes in athletic groin pain patients. (3rd March 2020)

• Record Type: Journal Article
• Authors: Gore, Shane; Richter, Chris; Franklyn-Miller, Andrew; Falvey, Eanna; King, Enda; Moran, Kieran
• Is Part Of: British journal of sports medicine. Volume 54(2020)Supplement 1
• Page Start: A109
• Page End: A110
• Publication Date: 2020-03-03
• Subjects: Sports medicine -- Periodicals; 617.1027
• DOI: 10.1136/bjsports-2020-IOCAbstracts.265
• Languages: English
• ISSNs: 0306-3674

Abstract

Abstract : Background: Response to rehabilitation might be different across athletes due to different movement strategies utilised pre-injury, and/or individual changes following injury (e.g. effect of athletic groin pain [AGP] to an athlete). While there is evidence that AGP athletes utilize different movement strategies, it is unknow if these movement strategy groups respond differently to rehabilitation. Objective: The aim was to assess the ability to predict changes in the HAGOS-subscale score of physical function in sport and recreation (HAGOS-subscale) using movement adaptations following rehabilitation on a group and subgroup-level. Design: Cross-sectional Setting: Clinical Patients: Eighty-six subjects with AGP undertook motion capture screening of a 110° change of direction task pre- and post-rehabilitation. Interventions (or assessment of risk factors): Each athlete was classified to one of three predefined movement clusters using their pre-rehabilitation data. Subsequently, four elastic net models (1 whole group using all subjects; 3 subgroup, one model for each identified movement cluster) were built to predict the change in the HAGOS-subscale using changes in biomechanical features and cross-validated. Main outcome measurements: To assess the ability of each elastic net model to predict changes in the HAGOS-subscale, the average R-squared (cvR2) and root-mean-square (cvRMSE) during the cross-validations were used. Results: Using the group-level model, a cvR2 of 0.46 (cvRMSE=12) was observed. Using the subgroup-level models resulted in cvR2 of 0.74, 0.85 and 0.91 (cvRMSE=7, 4, 5), while using these models on athletes outside their subgroup reduced their cvR2 to <0.42. Conclusions: The ability to predict changes in the examined HAGOS-subscale doubled for most subgroup-level models compared to a group-level model, while the cvRMSE reduced from 12 to 5.3 on average. This highlights the benefit of considering subgroup specific movement characteristics over group-level characteristics and illustrates that the examined subgroups respond differently to rehabilitation.