What is the potential measurement error in occupational exposure studies?

This paper demonstrates statistical methods that estimate measurement error from available industrial hygiene data. Errors in measuring a continuous e xposure variable may arise when all individuals in a work area are assigned the same exposure. An example is when the mean of exposure measurements ob tained on a sample of individuals is assigned to all workers with similar j obs. This may lead to inaccurate point and interval estimates in exposure-r esponse modeling. A method of simulating the distribution of true (i.e., un observed) individual exposures is described in order to estimate the mean a nd variance of measurement error. The minimum variance unbiased estimator a pproximates the mean of lognormally distributed exposure measurements. The distribution of true individual exposures is approximated by the distributi on of simulated estimates of mean exposure. The methodology is illustrated by exposure data from work areas manufacturing refractory ceramic fiber (RC F) and RCF products. Results show that exposure is slightly underestimated in work areas with between 25 and 113 exposure measurements; measurement er ror variance averages about 1.3% of the total variance.