[Computer analysis of audiograms in epidemiologic surveillance of cohorts exposed to excessive noise]

We have developed a method for the assessment of auditive loss using a sample of 1794 Bekesy audiograms recorded in young military students. A rectangular digital filter was used to smooth rough audiogram signals so as to detect pathological patterns such as scotoms and recruitments. Three factors were extracted from principal component analysis. They were correlated with the usual auditory indices and explained 70% of the total observed variance. The first factor is a general indicator of deafness, while the second and third describe the shapes of the hearing threshold level (asymmetry and convexity). This method can be used for rapid identification of suspect audiograms and is thus of value for epidemiological surveillance of populations exposed to impulsive noise.