A PASS/REFER CRITERION FOR SCREENING NEWBORNS USING DPOAEs

When screening infants for hearing loss using otoacoustic emissions (OAEs), a single pass/refer criterion must be established that will minimize referrals of babies with normal hearing (false positives), yet maximize the likelihood that babies with actual hearing loss will be identified (test sensitivity). How to establish this criterion, and the test specificity and sensitivity corresponding to a particular pass/refer criterion, is sometimes not well understood. It is not possible to establish one standard pass/refer criterion across OAE devices because measurement characteristics differ across devices. While some OAE manufacturers specify a pass/refer criterion for their devices, many do not. The manufacturers of those units leave this decision up to the user. Those units that do recommend pass/refer criteria base these criteria upon the measurement choices included in the device (i.e., presentation levels of f1 and f2, averaging times, and f1 and f2 frequencies), and thus criteria will differ across devices. This project reports the derivation of the pass/refer criterion for newborn hearing screening using a hand-held DPOAE screener. It is hoped that presenting information on the derivation of a pass/refer criterion will help users of all OAE devices understand how setting a pass/refer criterion will affect test specificity and sensitivity. INTRODUCTION In 1993, the National Institutes of Health (NIH) recommended in their consensus statement, “Early Identification of Hearing Impairment In Infants and Young Children,” that all babies born in the U.S. be screened for hearing loss within the first 3 months of life (NIH, 1993). In addition, the European Consensus Statement on Neonatal Hearing Screening also concluded that implementing newborn hearing screening programs is desirable and “should not be delayed” (Grandori, 1998). These consensus statements have fostered the development of many OAE measurement devices. The devices recently introduced to the market measure distortion product otoacoustic emissions (DPOAEs). Although many devices have been introduced, several do not specify a pass/refer criterion to apply to newborn testing. In addition, because each device’s measurement characteristics differ, no one criterion can be applied to every device. This leads to arbitrary criteria from the literature being applied to DPOAE units. Criteria such as 3 or 6 dB SNR at all or some test frequencies are commonly in use. According to Gorga (1999), application of these arbitrary criteria can lead to low sensitivity rates especially for infants with mild hearing losses. Recently a hand-held DPOAE unit was introduced to the market (See Figure 1). After developing this device, a single pass/refer criterion for neonatal screening was established using probability distributions and the binomial probability theory. No pass/refer criteria will result in perfect performance, but criteria can be established that will minimize both referrals of normal hearing children and misses of children with hearing loss. METHOD PASS/REFER CRITERIA DECISIONS The decision that a DPOAE exists is based on detecting a signal whose level is significantly above the background noise level. Using OAEs, we look at the signal-to-noise ratio (SNR) to make a pass/refer decision. To determine if the SNR is an actual emission instead of noise requires a statistical decision because the random noise level in the DPOAE filter channel can be expected to exceed the average of the random noise levels in the four adjacent filter channels used as the reference for comparison roughly half the time. For the hand-held device used in this study, the primaries at f1 and f2, DPOAE at 2f1f2, and noise floor sound pressure levels are estimated via a digital signal processor and the discrete Fourier transform (bin resolution =31 Hz). This device estimates the noise floor based on the four closest frequency bins to the emission bin. The noise floor is limited at -18 dB SPL. Therefore, in this study, we made the statistical decisions by making extended measurements of the noise distributions in both the DPOAE filter channel (DP level) and the rms average of the four adjacent channels (N level). From these measurements, cumulative frequency distributions and probability theory were used to estimate the “miss rate.” The goal of the present study was to determine the “miss rate” of the hand-held device using the default test settings and different criteria. The “miss rate” can be defined as the percentage of ears with hearing loss that will be missed using the pass/refer criterion. The default characteristics for this unit were: Number of Frequencies Tested = 3 Frequency Range = 2000 4000 Hz P1 (Intensity of F1 in dB SPL) = 65 P2 (Intensity of F2 in dB SPL) = 55 Averaging Time = 2 seconds N um be r of O cc ur re nc es SNR (dB) 100 90 80 70 60 50 40 30 20 10 0 0 1 2 3 4 5 6 7 8 9 10 SNR Cumulative Probabilty Distribution 2000 Hz 1 in 10 probability of SNR>7 dB at one frequency N um be r of O cc ur re nc es "Signal" Re: Noise in Adjacent Bands 100