Sharing the load: modeling loads in OpenSim to simulate two-handed lifting

Static Optimization (SO) procedures are commonly used to estimate muscle forces and joint loads from kinematics and external force data. The method of modeling hand-mass interaction during lifting tasks may affect the kinematics and/or external forces applied to the model, yet the extent to which different modeling decisions affect the estimated spinal joint loads is unknown. The present work compares five hand-mass interaction modeling approaches that differ in the complexity of implementation and runtime for the kinematic and SO analyses during two-handed lifting tasks. Intraclass correlation coefficients demonstrated strong agreement among the modeling approaches for the prediction of both maximum and average L5S1 resultant forces across all tasks. However, the five modeling approaches resulted in maximum relative differences in the L5S1 resultant force of up to 35% (2.6 kN). To compare the accuracy of each modeling approach, the resulting dynamic inconsistencies (i.e., residual forces and moments) were evaluated. The approach that resulted in the overall lowest residuals and incurred the least computational expense is recommended in the present study. The present work illustrates how different external-load modeling approaches can result in substantial differences in predicted spinal loads, especially as the movement speed increases, and how some models may perform better in terms of residual forces.