Evaluating epistemic uncertainty in a hybrid bayesian network model of spinal injury during lifting

Traditional occupational biomechanics models of lifting use body posture and hand load to predict L5/S1 compression force estimates. A novel hybrid Bayesian network implementation of a static lifting model has been developed that incorporates disc injury status as well. However, it is unclear how sensitive results are to assumed probability distributions used to model intervertebral disc compressive strength. A simulation study was conducted to investigate the effect of assumed probability distributions (normal, Gamma, and Weibull). The measure of interest was the probability that L5/S1 compression force exceeded the National Institute for Occupational Safety and Health (NIOSH) recommended design limit of 3,400 N. Simulations indicated that this measure was insensitive to the probability distribution used to model disc load tolerance.