SRF301 - Repeatability of Clinical Measures of Speech Recognition in Quiet and Noise

Abstract:
For more than 70 years, the standard test of speech recognition in the audiogram has been word recognition in quiet, despite the fact that individuals with hearing loss routinely struggle with speech recognition in noise. To facilitate the transition of speech-in-noise measures into routine practice, we examined the between-session repeatability of tests of word recognition in quiet (NU-6) and speech in noise (QuickSIN). Preliminary results suggest that after accounting for ceiling effects routinely observed with word-recognition tests, both measures had comparable between-session reliability. This further supports the idea that speech in noise can be routinely measured in clinical practice.

Summary:
Rationale
The most common complaint of individuals with hearing loss is difficulty understanding speech in noise. Despite this concern, for more than 70 years, the default test of speech recognition in routine audiologic practice has been word recognition in quiet. This discrepancy has led many researchers and clinicians to advocate for speech-in-noise measures to be routinely provided in clinical practice or to even replace word recognition in quiet as the primary measure of speech recognition. Implementing such fundamental changes requires key information for clinicians and researchers. A critical factor is the between-session repeatability of any speech recognition measure in clinical populations, as this consistency is vital for the medical and rehabilitative management of the patient. Unfortunately, large-scale studies in clinical populations are lacking in this regard. Here, we address this issue by examining between-session repeatability of speech recognition results as measured by the NU-6 word lists (word recognition in quiet) and the QuickSIN (speech in noise) in thousands of patients seen in our facility.

Design
We first identified over 100,000 audiograms from patients seen in our facility. All of these patients underwent pure-tone audiometry, and most of these also completed monaural word recognition in quiet as measured by the NU-6 word lists. A subset of approximately 12,000 audiograms also contained monaural speech in noise performance as measured by the QuickSIN. We then identified patients with minimal changes in hearing between sessions as defined by ≤ 5 dB difference in the high-frequency pure tone average (1, 2, and 4 kHz). 6,187 patients had similar hearing thresholds with monaural word-recognition scores, while 440 of these patients also completed monaural QuickSIN. We then determined the repeatability of each measure by comparing both word-recognition scores and QuickSIN SNR losses between each session. We were particularly interested in whether each measure fell within a critical difference determined by confidence intervals. For word recognition scores, we also examined pre-specified values used to determine referral for medical management (e.g., 10-15 percent).

Results
While preliminary and in progress, our results suggest that approximately 10 percent of word-recognition scores showed between-session differences in performance that might trigger referral to a physician, while 19 percent of patients revealed a difference in QuickSIN SNR loss > 2.7 dB. However, these results are complicated because most patients showed word-recognition scores at ceiling (e.g., at or near 100 percent). In contrast, QuickSIN SNR losses are rarely at ceiling levels. Our preliminary results indicate that, once ceiling effects are accounted for, the between-session reliability of the two measures becomes more comparable. We are currently exploring these effects in greater detail, along with examining the influence of hearing loss on the repeatability of these measures.

Conclusions
While preliminary and in progress, our results suggest that measures of speech recognition in quiet and noise have comparable levels of between-session repeatability after accounting for ceiling effects routinely observed when measuring word recognition in quiet. We plan to expand this work to examine these effects in pathologies commonly associated with repeat audiograms, such as sudden hearing losses, vestibular schwannomas, or Meniere’s disease.

Brief Summary of Clinical Takeaways:
This investigation provides necessary information to further integrate measures of speech in noise into routine practice.