Comparison of task-based estimates with full-shift measurements of noise exposure

Abstract

Using a large data set of noise exposure measurements on construction workers, task-based (TB) and full-shift (FS) exposure levels were compared and analyzed for the sources and magnitudes of the error associated with this methodology. Data-logging dosimeters recorded A-weighted sound pressure levels in decibels using Occupational Safety and Health Administration criteria for every minute of monitoring and were combined with information from task cards completed by subjects. Task-related information included trade, construction site type, location, activity, and tool. A total of 502 FS measurements were made, including 248,677 min of exposure on five construction trades. Six TB models of varying degrees of specificity were fit to the minute-level data and the results used to obtain TB estimates of the daily FS exposure levels. The TB estimates were derived using the predictions alone and also including subject and shift-specific residual means and variances. The predictions alone, which ignore within-task variability, produced a significant negative bias that was corrected by incorporation of the residual variance. This bias is only an issue in this setting in which the exposure of interest is noise, which follows a nonlinear averaging relationship. These estimates explained 10 to 60% of the variability in FS measures; adding the residual mean produced estimates that explained about 90% of the variability. In summary, TB estimates are important for exposure estimation when task time varies substantially. However, TB estimates include a substantial degree of error when there is large interindividual or intershift variability in exposure levels for a given task. Methods to improve the prediction of task-associated exposure, or adjusting for individual and shift differences, are needed.