Data processing options and response scoring for OAE-based newborn hearing screening

Scoring of click-evoked otoacoustic emissions (CEOAEs) is typically achieved by the evaluation of the reproducibility of the whole emission and/or within narrow bands. Screening outcomes are influenced not only by the specific combination of the subdivision scheme (i.e., the number, position, and bandwidth of the narrow bands) and the threshold used to determine pass and refer, but also by the accuracy with which the reproducibility is estimated. This study was designed to examine what factors affect the accuracy of the reproducibility estimate and how the accuracy of the reproducibility estimate together with the choice of the subdivision scheme/thresholds affect CEOAE scoring. Simulations with real CEOAEs corrupted with synthesized noise indicated that the reproducibility estimate is influenced by time-windowing and band-pass filtering: the longer the time-window or the broader the bandwidth of the filter, the more accurate the estimate. Quantitative figures on numerical scoring were given in terms of the referral rate and were derived from CEOAEs recorded in a clinical environment from more than 3400 newborns. The narrow bands were extracted according to 12 different subdivision schemes covering the 1.5-4-kHz range. The referral rate was found to depend on the subdivision scheme being used: (i) the worst results were obtained considering four narrow bands at 1.6-2.4-3.2-4 kHz; (ii) the best results were obtained considering two narrow bands at 2.25 and 3.75 kHz; (iii) bandwidths greater than 1 kHz resulted in the lowest referral rates. Also, scoring based on the extraction of four narrow bands produced the most unstable results, i.e., a small change in the threshold might cause even a great change in the referral rate.