A Large-Scale Study of the Relationship Between Degree and Type of Hearing Loss and Recognition of Speech in Quiet and Noise

Objectives: Understanding speech in noise (SIN) is the dominant complaint of individuals with hearing loss. For decades, the default test of speech perception in routine audiologic assessment has been monosyllabic word recognition in quiet (WRQ), which does not directly address patient concerns, leading some to advocate that measures of SIN should be integrated into routine practice. However, very little is known with regard to how SIN abilities are affected by different types of hearing loss. Here, we examine performance on clinical measures of WRQ and SIN in a large patient base consisting of a variety of hearing loss types, including conductive (CHL), mixed (MHL), and sensorineural (SNHL) losses. Design: In a retrospective study, we examined data from 5593 patients (51% female) who underwent audiometric assessment at the Stanford Ear Institute. All individuals completed pure-tone audiometry, and speech perception testing of monaural WRQ, and monaural QuickSIN. Patient ages ranged from 18 to 104 years (average = 57). The average age in years for the different classifications of hearing loss was 51.1 (NH), 48.5 (CHL), 64.2 (MHL), and 68.5 (SNHL), respectively. Generalized linear mixed-effect models and quartile regression were used to determine the relationship between hearing loss type and severity for the different speech-recognition outcome measures. Results: Patients with CHL had similar performance to patients with normal hearing on both WRQ and QuickSIN, regardless of the hearing loss severity. In patients with MHL or SNHL, WRQ scores remained largely excellent with increasing hearing loss until the loss was moderately severe or worse. In contrast, QuickSIN signal to noise ratio (SNR) losses showed an orderly systematic decrease as the degree of hearing loss became more severe. This effect scaled with the data, with threshold-QuickSIN relationships absent for CHL, and becoming increasingly stronger for MHL and strongest in patients with SNHL. However, the variability in these data suggests that only 57% of the variance in WRQ scores, and 50% of the variance in QuickSIN SNR losses, could be accounted for by the audiometric thresholds. Patients who would not be differentiated by WRQ scores are shown to be potentially differentiable by SIN scores. Conclusions: In this data set, conductive hearing loss had little effect on WRQ scores or QuickSIN SNR losses. However, for patients with MHL or SNHL, speech perception abilities decreased as the severity of the hearing loss increased. In these data, QuickSIN SNR losses showed deficits in performance with degrees of hearing loss that yielded largely excellent WRQ scores. However, the considerable variability in the data suggests that even after classifying patients according to their type of hearing loss, hearing thresholds only account for a portion of the variance in speech perception abilities, particularly in noise. These results are consistent with the idea that variables such as cochlear health and aging add explanatory power over audibility alone.