High-Pass and Low-Pass Filtered Digits-in-Noise Tests for Estimating Frequency-Specific Hearing Loss

This study aimed to evaluate the test characteristics of low-pass filtered speech (LPS) and high-pass filtered speech (HPS) digits-in-noise (DIN) tests. The method used filtered digit triplets presented in broadband (BB) noise to limit speech information above or below the filter cutoff frequency. The relationship between these filtered DIN results and pure-tone average (PTA) was explored.

A total of 125 participants with normal hearing or sensorineural hearing loss were recruited. Each participant completed BB, LPS, and HPS DINs for both ears, with 32 normal-hearing participants completing a retest of the LPS and HPS DINs in each ear. Analysis was restricted to the better-hearing ear.

Test-retest reliability was strong for both filtered DINs, with nonsignificant mean speech recognition threshold (SRT) differences (0.2 dB SNR) and measurement errors of 1.3 dB for LPS and 1.6 dB for HPS DINs. The HPS DIN demonstrated the highest diagnostic accuracy for detecting hearing loss, with areas under the curve up to 0.94. In contrast, the LPS and BB DINs showed lower accuracy across frequency ranges. Differences between HPS and LPS DIN SRTs were weakly related to audiometric slope (R² = .19-.24), suggesting limited correspondence with hearing loss configuration.

HPS DIN demonstrated superior sensitivity and specificity for detecting hearing loss compared to both BB and LPS DINs across conventional, low-frequency, and high-frequency PTAs. Findings reflect the dominant effect of hearing in the high frequencies on the DIN SRT. Estimating the slope of the audiogram from filtered DIN variants is only possible to a limited extent.