The validation of biodynamic models

Griffin, Michael J. (2001) The validation of biodynamic models. Clinical Biomechanics, 16 (1), S81-S92. (doi:10.1016/S0268-0033(00)00101-7).

Abstract

Biodynamic models may: (i) represent understanding of how the body moves (i.e., ‘mechanistic models'), (ii) summarise biodynamic measurements (i.e., ‘quantitative models'), and (iii) provide predictions of the effects of motion on human health, comfort or performance (i.e., ‘effects models').
Model validation may involve consideration of evidence used to derive a model, comparison of the model with alternatives, and a comparison between model predictions and independent observations of the predicted qualities or quantities. Models should be associated with a specified range of independent and dependent variables and indicate how intra-subject variability and inter-subject variability are accommodated. Models of the mechanisms of body movement may be validated by demonstrations that the mechanisms are well represented. Models giving numerical predictions (‘quantitative models' and ‘effects models') should specify the expected accuracy of predictions. ‘Effects models' advocated for predicting health, comfort or performance require that: (i) vibration or shock is a proven cause of the specified effect, (ii) within all reasonable ranges of model inputs, there must be reason to expect a positive correlation and acceptable error between the model predictions and the effect, (iii) other variables having a large influence on the effect must be taken into consideration. It is more useful to report the accuracy of ‘quantitative models' and ‘effects models' models than to state that they are ‘validated' or ‘un-validated'.
Checklists for assessing the quality of a biodynamic model are proposed, taking into account the type of model and the model assertions, the evidence, the assumptions, the accuracy, and the appropriateness of the model.
Relevance - Biodynamic models can be used to predict risks of injury or disease. Models can be used to optimise designs in order to minimise predicted risks. However, models can be promulgated and used without knowledge of their accuracy or usefulness.

This record has no associated files available for download.

Contributors

Author: Michael J. Griffin

Download statistics