Audiometric Reliability in Industry

Fast, Continuous Audiogram Estimation Using Machine Learning

“…For the 6 standard audiogram frequencies, the mean estimated threshold difference was −0.011 ± 5.61 dB HL, the mean absolute estimated threshold difference was 4.16 ± 3.76 dB HL, the median absolute estimated threshold difference was 3.00 dB HL with an interquartile range of 5.00 dB HL, and the percent 5-dB difference in threshold estimates was 66.25. These values compare favorably with historical differences in audiogram estimation methodologies (Gosztonyi Jr. et al 1971; Schmuziger et al 2004; Ishak et al 2011; Mahomed et al 2013). Judging from the relatively low percent 5-dB difference yet comparable mean absolute difference, the ML procedure appears to produce somewhat more outlier estimates at individual frequencies than methods that estimate directly at those frequencies.…”

Validity of Automated Threshold Audiometry

“…Three of these studies used children (Hartly & Siengenthalar 1964;Delany et al 1966;McPherson et al 2010) and one used an adult population (Gosztonyi et al 1971). Three of these studies used children (Hartly & Siengenthalar 1964;Delany et al 1966;McPherson et al 2010) and one used an adult population (Gosztonyi et al 1971).…”

“…Delany et al (1966) indicated that automated audiometry for participants provided results substantially in agreement with manual audiometry, however, as observed with adults, automated audiometry tends to produce thresholds that are slightly lower (−0.8 to −3.3 dB) than manual testing. Gosztonyi et al (1971) reported on industrial screening conducted on salaried and hourly workers (N = 38 ears). As age decreases, however, a greater proportion of children are either unable to 7 perform the test at all or frequently lose concentration so that portions of the test need to be repeated at a later stage to obtain a full audiogram.…”

“…Eleven reports in this systematic review included results on test-retest reliability, of which four used the method of limits and seven the method of adjustment for threshold audiometry. Several reports indicated that the second test session produced slightly lower (i.e., better) thresholds than the first session when manual and automated audiometry were used (Burns & Hinchcliffe 1957;Gosztonyi et al 1971;Robinson & Whittle 1973;Erlandsson et al 1979a;Lutman et al 1989;Fautsi et al 1990;Fromby et al 1996;Ho et al 2009;Ishak et al 2011, Swanepoel et al 2011. Several reports indicated that the second test session produced slightly lower (i.e., better) thresholds than the first session when manual and automated audiometry were used (Burns & Hinchcliffe 1957;Gosztonyi et al 1971;Robinson & Whittle 1973;Erlandsson et al 1979a;Lutman et al 1989;Fautsi et al 1990;Fromby et al 1996;Ho et al 2009;Ishak et al 2011, Swanepoel et al 2011.…”

Validity of Automated Threshold Audiometry

“…Eleven reports in this systematic review included results on test-retest reliability, of which four used the method of limits and seven the method of adjustment for threshold audiometry. In each case, reported test-retest reliability for automated audiometry was indicated to be within typical variability when compared with the test-retest reliability of manual audiometry (Burns & Hinchcliffe 1957;Gosztonyi et al 1971;Formby et al 1996;Robinson & Whittle 1973;Erlandsson et al 1979a, b;Lutman et al 1989;Fautsi et al 1990;Ho et al 2009;Ishak et al 2011;Swanepoel et al 2011). Only Ishak et al ( 2011) reported higher test-retest variability with Bèkèsy sweep-frequency audiometry, but reported that using a slower sweep rate of 20 seconds per octave would improve the acquired test-retest reliability.…”