Influence of interval censoring and bias on injury risk curve development

Injury risk research has compared parametric and non-parametric methods of developing risk curves to express bio-mechanical data, accounting for injury mechanism complexity and censorship assumptions. However, risk curves may over-predict the injury risk under these assumptions. This may occur because inherent biases in bio-mechanical data may influence the risk curve. Interval censoring can be applied when an interval is known within a normally doubly censored data set. Parametric and non-parametric models were applied to several data sets, evaluating their response to bias and interval censorship. Accounting for biases may or may not improve the fit to the data set depending on the risk curve. Interval censoring caused a decrease in the goodness of fit to the experimental data. While not normally a desired outcome, interval-censoring may actually produce risk curves that better predict the injury risk of the population of interest instead of the experimental data.