Evaluation of the predictive abilities of a qualitative exposure assessment model

Various qualitative exposure assessment models based on different underlying assumptions requiring distinct inputs and providing diverse outputs are used by occupational hygienists and risk managers to evaluate the magnitude of occupational exposures. Although a wide variety of exposure assessment models are available, most models have not been validated. This study compared the inhalation risk factor of a qualitative exposure assessment model with quantitative exposure data collected at a manufacturing facility for more than 9 years for 12 worker groups involving 24 chemical agents. A Spearman's rho correlation found no significant correlation between the model's risk factor and the maximum measured exposure (rs=0.119, p=0.496). A Fisher exact test found that the maximum measured exposure was independent of the model's inhalation risk factor (chi2=0.203, p=0.653). The model accurately classified measured exposures in 18 out of 35 cases (51%), 53% of the measured exposures classified as acceptable were correctly classified by the model (sensitivity), whereas 33% of the measured exposures classified as uncertain/unacceptable were correctly classified by the model (specificity). There was a 67% probability that the model would result in a false low classification and a 47% probability that the model would result in a false high classification. Although the model was simple to apply, the overall predictive ability for the low level exposures seen in the study were poor and inconsistent among worker groups compared.