OBJECTIVES : Five databases were mined to examine distributions of airbone
gaps obtained by automated and manual audiometry. Differences
in distribution characteristics were examined for evidence of influences
unrelated to the audibility of test signals.
DESIGN : The databases provided air- and bone-conduction thresholds
that permitted examination of air-bone gap distributions that were free
of ceiling and floor effects. Cases with conductive hearing loss were
eliminated based on air-bone gaps, tympanometry, and otoscopy, when
available. The analysis is based on 2,378,921 threshold determinations
from 721,831 subjects from five databases.
RESULTS : Automated audiometry produced air-bone gaps that were normally
distributed suggesting that air- and bone-conduction thresholds
are normally distributed. Manual audiometry produced air-bone gaps
that were not normally distributed and show evidence of biasing effects
of assumptions of expected results. In one database, the form of the
distributions showed evidence of inclusion of conductive hearing losses.
CONCLUSIONS : Thresholds obtained by manual audiometry show tester
bias effects from assumptions of the patient’s hearing loss characteristics.
Tester bias artificially reduces the variance of bone-conduction
thresholds and the resulting air-bone gaps. Because the automated
method is free of bias from assumptions of expected results, these distributions
are hypothesized to reflect the true variability of air- and boneconduction
thresholds and the resulting air-bone gaps.