Our analysis found that different metrics for occupational hearing loss currently employed in the aluminum industry lead to markedly different estimates of the burden of disease when applied to the same set of industrial audiograms. At the same time, a number of metrics, including an age-corrected 10 dB threshold shift criteria and a 15 dB shift criteria, showed good correlation with the observed–expected rate of high-frequency hearing loss occurring at each study location over the 5-year period of observation. These metrics demonstrate potential for providing a reasonable “signal” of hearing loss occurrence that could serve as an international standard for comparison of occupational hearing loss across countries and industries.
The variation in absolute rates between different metrics implies differences in sensitivity of the measures. Certain metrics appeared to be “highly sensitive,” as exemplified by metrics #4 and 6. Approximately 10% of the population or more met these metrics’ criteria for hearing loss each year of the study period. Our previous analysis of metric 6 as a candidate “early flag” for detecting cases of early hearing loss in individuals found that the high number of individuals flagged by such sensitive metrics on annual screening may limit its clinical usefulness due to the requirement for greater resources for follow-up of many individuals who may not go on to develop significant hearing loss during employment.
At the same time, such sensitive measures may be useful when assessing overall program quality, and the lack of age correction may limit the usefulness of such measures, as explained below.
Other metrics appeared to be relatively insensitive, with lower annual rates, as demonstrated by metrics 2 and 3. For metric 3, the criteria requiring a significant shift from baseline (25 dB age-corrected shift in the average of 2, 3 and 4 KHz) means that only a few individuals who work at a location for many years will be identified. By that time, it will be too late to salvage much of their hearing, and the noise problem of the company will also have gone undetected for a prolonged period. Such measures, because they produce low annual rates of loss, are likely, in a smaller worker population, to spuriously produce rates of 0 due to their inability to detect small hearing changes in the population.
Whether or not to adjust a baseline following the occurrence of a hearing shift is another important consideration in developing an international standard. If no baseline revision takes place, sensitive metrics such as #4 would produce results several-fold higher (data not shown) and be less likely to detect recent trends, as much of the annual hearing loss reported would be cumulative from previous years.
For valid comparison of rates of health outcomes across different at-risk populations, the methods of analysis need to reduce possible biases that could lead to erroneous conclusions about where unacceptable rates of occupational hearing loss are occurring. A particular problem arises in comparing rates of diseases that are more common in different age groups across populations with varying age structures. To account for this, methods of age standardization have been developed. Hearing loss risk (due to presbycusis) increases markedly with age. The age correction of surveillance audiograms is a form of age standardization, whereby each individual audiogram is “adjusted” for the expected amount of hearing loss due to normal aging (using standard tables such as appendix F of the OSHA hearing conservation standard).
The hearing loss metric is then applied in the setting of this age correction. While age correction may have some disadvantages when applied to an individual audiogram,
age correction does allow the hearing loss experience in an industry where the workers are older, to be compared with the hearing loss occurring in an industry where the workers are relatively younger, without as much bias due to a differential effect of age. One key potential problem with this type of age standardization of hearing loss as an international standard is that rates of hearing loss with normal aging may vary across different populations and countries. Little is known about this, and if subsequent studies confirm such variability, development of country- or population-specific age adjustment tables would be indicated. At the same time that age correction appears to be intuitively important, our finding that the 15 dB shift metric (with no age correction) demonstrated a similar degree of correlation with the observed–expected hearing loss as the age-corrected 10 dB shift suggests that age correction may not be critical for the purposes of comparing occupational hearing loss across locations or countries. In fact, the two measures (10 dB age-corrected shift and 15 dB nonage-corrected shift) showed a high degree of correlation between the measures. Further study of these metrics in populations that diverge in their age structure more than the locations studied here is necessary to determine whether age correction is necessary for comparing hearing loss rates between occupational populations.
One finding of this study was that at approximately half of the study locations, we observed less hearing loss over the study period than would be expected by aging alone (observed–expected <1). The prevalence of this finding suggests that the aging formulae in the ANSI 3.44 standard may not accurately reflect the aging rates of current populations, as the U.S. population today may be losing less hearing as they age compared with previous generations. This finding is consistent with other recent reports that suggest that Americans hear as well or better than 40 years ago, and that the age correction tables in current use may need to be updated.
Whether this same trend for improving hearing holds true in other countries remains unclear.
While this study reports promising results for at least two of the candidate metrics, we recommend further studies before setting an international standard. The population of workers used to validate these metrics was based in the US, and further development of an international standard should involve analysis of databases from other countries as well to ensure that the metrics perform well in a number of settings.
The study was also limited by the small number of locations used for the comparison (10). As a result, although the correlations between certain metrics and the difference of observed and expected hearing loss were statistically significant, and the R2 values for some metrics appeared much larger than others, a larger sample should be used to assess the statistical significance of the differences between the performance of individual metrics.
This study used the observed–expected rate of high-frequency hearing loss (either at 2, 3 and 4 KHz or 3, 4 and 6 KHz) as a comparison standard to approximate the actual degree of high-frequency hearing loss occurring in each location. One could argue that this type of measure would be even better than the other proposed metrics that are based on either shifts from baseline or absolute values of hearing loss at particular frequencies. Calculating observed and expected rates of loss, however, requires more sophisticated computing that is beyond the capabilities of many industrial settings, and an approach that is unfamiliar to many hearing conservation professionals. The good correlation between metrics 1 and 6 and the observed–expected hearing loss argues that existing methods such as age-corrected shift rates or the 15 dB shift criteria that we used could provide useful comparison metrics for assessing the relative burden of NIHL in different industrial settings across countries.