Assessment of High-Variability Speech Recognition in Cochlear Implant Candidates

Abstract

Background: Current clinical speech recognition measures to evaluate cochlear implant (CI) candidacy may not reflect the challenging listening conditions encountered in everyday life, including talker variability. Consequently, these measures may offer limited insight into CI candidates’ functional communication abilities. Purpose: We aimed to compare performance on high-variability sentences with conventional clinical measures (Consonant-Nucleus-Consonant [CNC] words, AzBio sentences), identify the contribution of working memory to performance, and explore the relationship between performance and hearing-related quality of life in adult CI candidates. Research Design: This was a prospective, cross-sectional study. Study Sample: The study included 32 postlingually deafened adult CI candidates (age 42.0–85.0 years). Data Collection and Analysis: Participants were tested in their everyday listening configuration (i.e., using their own personal devices). Speech recognition tests included CNC words in quiet, AzBio sentences in quiet and multitalker babble (MTB), and Perceptually Robust English Sentence Test Open-set (PRESTO) sentences in quiet. Participants also completed measures of auditory sensitivity (Spectral-Temporally Modulated Ripple Test), working memory (Reading Span Test), and hearing-related quality of life (Cochlear Implant Quality of Life-35 Profile questionnaire, CIQOL-35). Statistical analyses included a one-way repeated measures analysis of variance (RM ANOVA), hierarchical regression, and Pearson correlations. Results: The RM ANOVA revealed a significant main effect of test on speech recognition accuracy; PRESTO was significantly more challenging than AzBio sentences in quiet but was comparable to CNC words and AzBio sentences in MTB. In hierarchical regression models, working memory significantly improved the model fit for PRESTO but was not a significant predictor for the other tests. Finally, correlational analyses showed that PRESTO scores showed a strong relationship with the CIQOL-35 Global score and was more consistently related to CIQOL-35 domains than the other tests. Conclusions: PRESTO is challenging for adult CI candidates and captures individual differences in cognitive-perceptual processing. High-variability materials may better represent real-world listening demands and hearing-related quality of life before implantation. Clinical Relevance Statement: Including high-variability sentence materials like PRESTO in the pre-CI battery may be useful for assessing CI candidates’ speech recognition abilities in everyday, challenging listening conditions. This may help clinicians set realistic expectations for post-CI outcomes and inform individualized rehabilitation protocols.