Reliability and validity of industrial audiometry: implications for hearing conservation program design.

Hearing conservation in industry relies heavily on monitoring audiometry to detect early noise-induced hearing loss in workers who are exposed to potentially damaging noise, with or without hearing protectors. The "real-world" reliability and validity of these measurements, as well as otoscopic observations in industry, have not been extensively investigated. In addition, there is considerable controversy over the selection of a definition of "significant threshold shift" in industrial audiometry. These and related issues were considered in a series of three studies utilizing data from an active hearing conservation program. Test-retest variability in industry is much higher than has been reported for clinical settings; this variability is reduced by pure-tone averaging. Workers referred for otologic evaluation were found to have hearing levels which were, on the average, about 5 dB better than indicated by plant audiometry, even without excluding 4% of referred workers who had unilateral deafness and showed "shadow curves" on the plant audiograms. Otoscopic data obtained by the plant audiometrists were uncorrelated with the results of otoscopy by consultant otologists. Techniques borrowed from decision theory and signal detection theory were used to evaluate possible criteria for significant threshold shift. Criteria based on pure-tone averaging were superior to those based on a certain amount of threshold shift for any frequency tested. It is proposed that a significant threshold shift be defined as a 10 dB or greater change for the worse for either the 0.5, 1, 2 kHz pure-tone average or the 3, 4, 6 kHz pure-tone average, in either ear, and that such shifts be validated by prompt retesting. Even with this criterion, a substantial number of shifts (most shifts, in some situations) will be either spurious or attributable to disorders other than noise-induced hearing loss, such as presbycusis. Otologic referral in cases of large or repeated shifts may prevent unjustified administrative actions, to the advantage of both workers and management. A practical consequence of the use of monitoring audiometry may be a de facto lowering of the permissible exposure level to 85 dBA TWA.