Are Speech Perception Scores in Cochlear Implant Recipients Consistent Across Different Tests?

Abstract

Objective: Cochlear implant (CI) candidacy and postoperative outcomes are assessed using sets of speech perception tests that vary from center to center, limiting comparisons across institutions and time periods. The objective of this study was to determine if scores on one speech perception test could be reliably predicted from scores on another test.

Study design: Arizona Biomedical (AzBio) Sentence Test, Consonant-Nucleus-Consonant word (CNCw), and Hearing in Noise Test (HINT) scores in quiet for the implanted ear were collected for individuals who received a CI between 1985 and 2019. Scores collected during the same testing session were analyzed using Bland-Altman plots to assess agreement between testing methods. Simple linear regression with logit transformation was used to generate predictive functions and 95% confidence intervals for expected mean and individual scores.

Setting: Single academic medical center.

Patients: A total of 1,437 individuals with a median age of 59.9 years (range, 18-95 yr) and 46% (654 of 1,437) male.

Interventions: N.A.

Main outcome measures: Agreement as a function of test score, mean, variance, and correlation coefficients.

Results: A total of 2,052 AzBio/CNCw, 525 AzBio/HINT, and 7,187 CNCw/HINT same-session score pairings were identified. Pairwise test comparisons demonstrated limited agreement between different tests performed in the same session, and a score correlation between different speech tests revealed large variances.

Conclusion: Transformation functions between test batteries were predictive of mean scores but performed poorly for prediction of individual scores. Point-wise comparisons of scores across CI test batteries should be used with caution in clinical and research settings.

Figure 1.

Bland-Altman plots for each pair of speech recognition scores (A: AzBio vs CNCw; B: AzBio vs HINT; C: CNCw vs HINT). Mean values are represented by the blue lines and limits of agreement for each test pair are represented by red dashed lines (95% confidence interval or ±2 SD). Dots are ‘jittered’ to make all observations visible.

Figure 2.

Bland-Altman plots of AzBio/CNCw (A), AzBio/HINT (B), and CNCw/HINT (C) scores stratified by age at implantation (<60 yo vs ≥60 yo). Mean values are represented by the solid lines and limits of agreement for each test pair are represented by dashed lines (95% confidence interval or ±2 SD).

Figure 3.

Bland-Altman plots of AzBio/CNCw (A), AzBio/HINT (B), and CNCw/HINT (C) scores stratified by acoustic (pre-operative) versus CI hearing (post-operative). Mean values are represented by the solid lines and limits of agreement for each test pair are represented by dashed lines (95% confidence interval or ±2 SD).

Figure 4.

Bland-Altman plots of post-operative AzBio/CNCw (A), AzBio/HINT (B), and CNCw/HINT (C) scores stratified by time since implantation (<1 year vs ≥1 year). Mean values are represented by the solid lines and limits of agreement for each test pair are represented by dashed lines (95% confidence interval or ±2 SD).

Figure 5.

Pair-wise comparison of individual speech perception scores obtained on same-day testing between differential modalities of speech perception tests (A: AzBio vs CNCw; B: AzBio vs HINT; C: CNCw vs HINT). The blue shaded area represents the 95% confidence interval for the correlation of best fit for the group mean. The red dashed lines represent the expected ranges (95% confidence interval or ±2 SD) for the individual expected scores.