Enhancement of a musculoskeletal shoulder model in OpenSim and validation with in vivo force data: The influence of OpenSim workflow configurations

Musculoskeletal modeling is widely used to investigate biomechanical mechanisms that are difficult or impossible to measure experimentally, such as glenohumeral joint reaction force (GH-JRF), used to characterize shoulder stability. However, objective quality criteria for OpenSim workflow configurations remain limited. This study evaluated the influence of different modeling choices on GH-JRF estimation accuracy using in vivo reference data from an instrumented shoulder prosthesis (Orthoload). In addition, commonly used quality metrics, scaling RMSE and inverse kinematic (IK) RMSE, were assessed as indicators of dynamic simulation accuracy. Experimental kinematic data were reprocessed through multiple workflow configurations, combining two musculoskeletal models, five scaling methods, two IK setups, and three force calculation approaches across unloaded abduction, forward flexion and backward flexion. GH-JRF accuracy was defined using RMSE and waveform similarity. Model selection and force calculation approach showed the strongest influence on GH-JRF accuracy, whereas scaling methods and IK setup selection had little systematic effect. No consistent association was found between scaling or IK accuracy and GHJRF estimation. These findings indicate, that improvements in scaling or kinematic accuracy do not necessarily translate into enhanced dynamic validity and highlight the need for outcome specific validation in musculoskeletal shoulder modeling.