Benefits of a Hearing Registry: Cochlear Implant Candidacy in Quiet Versus Noise in 1,611 Patients

Abstract

Purpose

This retrospective study used a cochlear implant registry to determine how performing speech recognition candidacy testing in quiet versus noise influenced patient selection, speech recognition, and self-report outcomes.

Method

Database queries identified 1,611 cochlear implant recipients who were divided into three implant candidacy qualifying groups based on preoperative speech perception scores (≤ 40% correct) on the AzBio sentence test: quiet qualifying group, +10 dB SNR qualifying group, and +5 dB SNR qualifying group. These groups were evaluated for demographic and preoperative hearing characteristics. Repeated-measures analysis of variance was used to compare pre- and postoperative performance on the AzBio in quiet and noise with qualifying group as a between-subjects factor. For a subset of recipients, pre- to postoperative changes on the Speech, Spatial and Qualities of Hearing Scale were also evaluated.

Results

Of the 1,611 patients identified as cochlear implant candidates, 63% of recipients qualified in quiet, 10% qualified in a +10 dB SNR, and 27% qualified in a +5 dB SNR. Postoperative speech perception scores in quiet and noise significantly improved for all qualifying groups. Across qualifying groups, the greatest speech perception improvements were observed when tested in the same qualifying listening condition. For a subset of patients, the total Speech, Spatial and Qualities of Hearing Scale ratings improved significantly as well.

Conclusion

Patients who qualified for cochlear implantation in quiet or background noise test conditions showed significant improvement in speech perception and quality of life scores, especially when the qualifying noise condition was used to track performance.

Cochlear implants (CIs) are an efficacious and cost-effective intervention for many adults with moderate-to-profound sensorineural hearing loss who do not benefit from traditional amplification (Bond et al., 2009; Francis et al., 2002; Saunders et al., 2016). However, despite the reported benefits of implantation, only an estimated 5%–7% of adults who are candidates obtain a CI (Sorkin, 2013; Sorkin & Buchman, 2016). Low utilization of CIs may relate to the absence of routine hearing screenings and limited referrals from primary care physicians as well as unestablished referral pathways to CI centers (Sorkin, 2013; Sorkin & Buchman, 2016). Additional roadblocks to implantation may occur due to differences in preoperative candidacy assessment protocols across CI centers.

The current National Coverage Determination for cochlear implantation, outlined by the Centers for Medicare and Medicaid Services (CMS, 2005), provides CI coverage for adults with bilateral moderate-to-profound sensorineural hearing loss who score ≤ 40% correct in the best-aided listening condition on recorded tests of open-set sentence recognition. The CMS-aided speech recognition performance criterion is more stringent than the labeled indications approved by the Food and Drug Administration for most of the currently available CI systems. However, CMS does not outline specific testing procedures, allowing clinicians to qualify patients using different speech recognition protocols while still adhering to the CMS National Coverage Determination standard. This flexibility produces variability in candidacy determination procedures across centers. For example, a center could test and qualify listeners for a CI based on speech recognition performance in quiet or background noise. Two recent studies reported preoperative candidacy decisions based on speech recognition in noise (Holder et al., 2018; Mudery et al., 2017), and results from one study indicated adults who qualified in noise obtained significant postoperative benefit (Mudery et al., 2017). Given the limited research, though, many clinicians might be reluctant to qualify candidates in noise due to concerns about potential postoperative CI benefit or insurance coverage when qualifying in noise.

Documenting average preoperative to postoperative improvement on speech-in-noise tests across studies can be problematic because of the inherent variability in small clinical samples. Using a meta-analytic approach to examine CI benefit in noise is a reasonable approach but is also difficult due to center-to-center variability in testing, lack of studies performing candidacy evaluations in noise, and inconsistent reporting standards in the literature (Adunka et al., 2018; Gifford et al., 2008; Holder et al., 2018; Sladen et al., 2017). Another approach is to compile CI patient data from multiple implant centers in a clinical registry to address questions regarding candidacy and postoperative benefits in background (Chen et al., 2017; Miller et al., 2020; Schafer et al., 2016). Given the significant positive impact of CIs as well as the lack of standardization in audiological candidacy assessments, this study aims to use a large CI registry to (a) explore the proportion of candidates in the database who qualify for a CI (i.e. ≤ 40% preoperative speech recognition in the best-aided condition) in quiet or background noise, (b) define the demographic characteristics of the cohorts who qualify for a CI in quiet and/or noise conditions, (c) examine postoperative sentence recognition outcomes in patients who qualify for a CI based on speech scores in quiet or noise, and (d) assess subjective satisfaction with CIs for those who qualify for the devices in a quiet versus noise condition.

Method

Database Procedures

Data were mined from a national, multicenter CI database composed of clinical data from 7,275 patients from 12 clinical sites across nine states (Colorado, Iowa, Maryland, New York, Ohio, Oklahoma, Tennessee, Texas, and Wisconsin). The development of the database has been described previously by Schafer et al. (2016). The 7,275 patients represent both CI candidates and/or recipients ranging in age from 0 to 102 years who received implants from 1978 to 2019. Because not all 7,275 patients in the registry were implanted (some were not implant candidates or chose not to be implanted), the data set was initially narrowed to a total of 6,599 implantations in the database.

To address Aim 1 and explore the proportion of implanted candidates in the United States who qualified for a CI in quiet and/or a noise test condition, the database of 6,599 unique implantations was queried to include only those cases with preoperative AzBio sentence recognition scores (Spahr et al., 2012). This query resulted in a 2,979 patient sample described in Table 1. After excluding bilateral recipients, the data set was narrowed to include only those with an AzBio sentence recognition score ≤ 40% in the best-aided condition in quiet or noise. This query resulted in a 1,611 patient sample with implantation dates ranging from 2008 to 2019. The best-aided condition was defined as the listening condition with the highest score whether this represented the aided score of the ear to be implanted or the bilateral hearing aid score. Although we are unable to confirm each center's fitting procedures, the authors can attest that the contributing clinical sites followed clinical practice guidelines for fitting and verification of hearing aids using developmentally appropriate prescriptive targets. Presentation levels ranged from 48 to 65 dB SPL (A weighted) in greater than 99% of tests. To determine the proportion of patients who qualified for a CI with sentence scores administered in quiet versus noise, the 1,611 patients were divided into three implant qualifying groups based on the test condition that resulted in ≤ 40% correct for AzBio sentences. Groups included those who qualified in (a) quiet (N = 1,020), (b) at a +10 signal-to-noise ratio (SNR; N = 155), or (c) at a +5 dB SNR (N = 436). Most patients had scores in more than one test condition (e.g., +10 dB SNR and in quiet), but patient data were assigned to the least adverse listening condition in which they scored ≤ 40% correct. For example, if a patient scored 29% correct on AzBio sentences in quiet and 15% correct on AzBio sentences at +10 dB SNR, he/she was assigned to the AzBio quiet group. Figure 1 summarizes the process of creating these separate cohorts.

Table 1.

Demographic and preoperative hearing characteristics by qualifying group.

Demographics and characteristics	Qualifying group
	AzBio Quiet	AzBio +10 dB SNR	AzBio +5 dB SNR
Age at implant: Years (SD)	65.7 (17.3) N = 1,005	66.1 (17.1) N = 151	66.5 (16.3) N = 432
Manufacturer distribution (%)	17, 75, 8	8, 85, 6	16, 77, 8
Gender	50.7% Female N = 1,020	49.5% Female N = 155	39.0% Female N = 436
Race and ethnicity a	N = 1,020	N = 155	N = 436
African American	2.1%	0.6%	1.1%
American Indian	0.5%	0.6%	0.0%
Asian	0.1%	0.6%	0.4%
Hispanic	2.1%	0.0%	0.9%
White	50.1%	72.3%	30.0%
Other	10.8%	3.9%	13.8%
Unknown	34.3%	22.0%	53.8%
Duration hearing loss: Years (SD)	25.2 (18.6) N = 809	24.5 (16.0) N = 121	25.7 (16.6) N = 363
PTA dB HL: Years (SD)	88.3 (17.1) N = 883	80.4 (16.2) N = 149	74.8 (14.6) N = 390
LFPTA dB HL: Years (SD)	69.3 (19.3) N = 451	56.2 (23.5) N = 79	50.2 (19.6) N = 277

Note. PTA = pure-tone average at 500, 1000, and 2000 Hz; LFPTA = low-frequency pure-tone average (250 and 500 Hz); HL = hearing level.

^a Racial and ethnic categories chosen as outlined by the National Institutes of Health; Manufacturer distribution are listed in percentages as follows: Advanced Bionics, Cochlear, MED-EL.

Figure 1.

Flow diagram demonstrating creation of the Quiet, +10 dB SNR, and +5 dB SNR qualifying groups.

The demographic and hearing characteristics of the candidates were collected, when available, for each group of qualifying CI candidates including the following: age at implant, gender, manufacturer type, race and ethnicity, and preoperative hearing characteristics including duration of hearing loss, pure-tone average (PTA) hearing thresholds (500 Hz, 1 kHz, 2 kHz), and low-frequency pure-tone average (LFPTA; 250 and 500 Hz). These data are provided in Table 1.

Speech Recognition Outcomes

To address our third aim and examine postoperative improvement in speech recognition performance by qualifying group (i.e., qualified in quiet, +10 dB SNR, or +5 dB SNR), the three qualifying groups of CI candidates were refined to include patients with at least one preoperative AzBio score and one best postoperative AzBio score in the same listening condition (quiet, +10 dB SNR, +5 dB SNR) collected between 3 and 24 months postactivation. Patients who qualified in quiet typically had pre–post AzBio data in quiet and a +10 dB SNR or +5 dB SNR condition. Likewise, patients who qualified in one of the noise conditions often had pre–post data on the AzBio in quiet. Thus, speech recognition was assessed in quiet and the two noise conditions for patients in the three qualifying groups. The AzBio sentence test contains 20 recorded sentences. Each sentence was scored for number of words correct, and performance was expressed as an overall percentage correct across all sentences.

Speech recognition outcomes on the AzBio were analyzed separately for the three test conditions (quiet, +10 dB SNR, and +5 dB SNR) using repeated-measures analysis of variance (RM ANOVA) and Number Cruncher Statistical Software. In each separate analysis, CI qualifying group (qualified in quiet, +10 dB SNR, and +5 dB SNR) was included as a between-subjects factor and test session (AzBio score at pretest and posttest) was included as a within-subject factor. In addition, effect sizes were calculated using methods appropriate for an RM design (Morris & DeShon, 2002).

Patient-Reported Outcomes

To assess patient-reported outcomes associated with implantation, subsets of patients from the quiet qualifying group (N = 68) and the +5 dB SNR qualifying group (N = 37), who also had preoperative and postoperative data on the Speech, Spatial and Qualities of Hearing Scale (SSQ; Gatehouse & Noble, 2004), were queried. Subjects from the +10 dB SNR qualifying group were not included due to limited pre- and postoperative SSQ data. The SSQ is a quantitative patient-reported outcome tool given to patients either in an interview format or in a patient self-filled format. This questionnaire assesses self-reported hearing disabilities on a 0–10 scale (0 = great difficulty) across speech-hearing, spatial, and qualities of hearing domains. These data were analyzed with paired t tests and effect sizes (Morris & DeShon, 2002). This study qualified for an exemption from the University of North Texas Institutional Review Board given the use of only de-identified data.

Results

Group Demographic Analyses

Known preoperative characteristics of age at implantation, duration of deafness, PTA, LFPTA, and manufacturer breakdown for the three CI qualifying groups are provided in Table 1. Because some of these characteristics were not reported for everyone in the three groups, the N reported for each characteristic is reported. According to a one-way ANOVA, there was no significant difference among the three qualifying groups for age at implant or duration of deafness, F(2, 1591) = 0.68, p = .51; F(2, 1301) = 0.54, p = .59. The implanted ear PTA and LFPTA were significantly different among the three groups, F(2, 1422) = 95.3, p < .0001; F(2, 807) = 81.8, p < .0001. Specifically, post hoc analyses with the Tukey–Kramer Multiple Comparisons test suggested that preoperative audiometric thresholds were poorest for the quiet qualifying group and best for the +5 dB SNR qualifying group, indicating patients who qualified in noise had better hearing thresholds. Gender, race, and ethnicity data are also summarized in Table 1.

Sentence Recognition Outcomes in Quiet

Figure 2 illustrates average preoperative and postoperative sentence recognition performance for all patients in the three qualifying groups who had preoperative and postoperative AzBio sentence recognition scores in quiet. Table 2 provides additional descriptive statistics and displays the medium-to-large effect sizes between pre- and postoperative scores in quiet.

Figure 2.

Mean pre–postoperative AzBio sentence recognition in quiet. Bars represent 1 SD.

Table 2.

Descriptive statistics for AzBio sentence scores by qualifying group.

		Qualifying group
Test condition	Descriptive statistics	AzBio Quiet		AzBio +10 dB SNR		AzBio +5 dB SNR
		Preop %	Postop %	Preop %	Postop %	Preop %	Postop %
Quiet	Average (SD)	15 (14)	67 (27)	60 (12)	74 (26)	71 (16)	83 (19)
	Mdn	13	75	58	83	72	89
	Range	0–40	0–100	41–90	0–100	16–100	0–100
	Improvement	52 (29)		15 (25)		12 (22)
	Effect size	2.8		0.99		0.59
	Confidence interval	[2.7, 3.0]		[0.62 ,1.4]		[0.41, 0.78]
+10 SNR	Average (SD)	16 (15)	59 (24)	26 (9)	61 (22)	57 (12)	72 (19)
	Mdn	9	59	27	59	55	73
	Range	0–60	5–100	3–39	8–95	12–93	19–100
	Improvement (SD)	43 (29)		35 (24)		15 (18)
	Effect size	2.0		2.8		1.1
	Confidence interval	[1.4, 2.6]		[2.3, 3.3]		[0.42, 1.9]
+5 SNR	Average (SD)	16 (23)	48 (29)	22 (15)	54 (22)	22 (11)	54 (27)
	Mdn	8	45	18	43	22	55
	Range	0–97	0–99	0–39	25–83	0–40	0–100
	Improvement	31 (29)		32 (17)		32 (27)
	Effect size	1.2		1.8		2.3
	Confidence interval	[0.82, 1.6]		[0.71, 2.9]		[2.1, 2.5]

Note. Improvement represents postoperative average minus preoperative average. Effect sizes represent the effect size d between preoperative and postoperative scores for repeated measures as well as the 95% confidence interval for d. Preop = preoperative data; Postop = postoperative data; SNR = signal-to-noise ratio.

A two-factor RM ANOVA was used to examine the effects of CI qualifying group (quiet, +5 dB SNR, or +10 dB SNR) and test session (pre- vs. postoperative AzBio score). This analysis indicated significant main effects of CI qualifying group, F(2, 2048) = 521.5, p < .0001, and test session, F(1, 2048) = 402.0, p < .0001. The interaction between CI qualifying group and test session was also highly significant, F(2, 2048) = 228.1, p < .0001.

Post hoc analyses were conducted with the Tukey–Kramer Multiple Comparisons test. For the main effects of CI qualifying group and test session, all group comparisons were significantly different (p < .0001), and across the three groups, there was a significant improvement between preoperative and postoperative scores. Post hoc analysis of the two-way interaction between CI qualifying group and test session yielded several notable findings. The quiet qualifying group experienced significant mean improvement on the AzBio in quiet between preoperative and postoperative tests (p < .05). However, average preoperative and postoperative scores for the quiet qualifying group were significantly poorer at both test sessions compared to the two groups qualified in noise (p < .05; see Figure 2). When comparing the +5 dB SNR and +10 dB SNR qualifying groups, both groups showed significant average improvements after implantation in the quiet test condition (p < .05). However, the +5 dB SNR qualifying group had significantly higher average preoperative (p < .05) and postoperative (p < .05) scores in quiet than the +10 dB SNR qualifying group.

Sentence Recognition Outcomes in Noise

Figures 3 and 4 show average preoperative and postoperative AzBio sentence recognition in noise performance across the three qualifying groups for the +10 dB SNR condition and the +5 dB SNR condition, respectively. Table 2 provides descriptive statistics and displays the large effect sizes between pre- and postoperative scores. Two separate, three-factor RM ANOVAs were conducted to examine the effects of CI qualifying group (quiet, +5 dB SNR, or +10 dB SNR) and test session (pre- vs. postoperative score) for the +5 dB SNR AzBio test condition and the +10 dB SNR AzBio test condition.

Figure 3.

Mean pre–postoperative AzBio sentence recognition at a +10 dB SNR. Bars represent 1 SD.

Figure 4.

Mean pre–postoperative AzBio sentence recognition at a +5 dB SNR. Bars represent 1 SD.

Sentence Recognition Outcomes in Noise: +10 dB SNR

The RM ANOVA for the +10 dB SNR test condition suggested a significant main effect of CI qualifying group, F(2, 212) = 25.7, p < .0001, and of test session, F(1, 212) = 133.4, p < .0001. The interaction between CI qualifying group and test session was also significant, F(2, 212) = 7.6, p < .01. The post hoc analysis on the significant main effect of qualifying group found no significant difference between the +5 dB SNR and +10 dB SNR qualifying groups (p < .05); however, the quiet qualifying group had significantly poorer scores than the two qualified-in-noise groups (p < .05; see Figures 3 and 4). The post hoc analysis on the significant interaction effect showed that each CI qualifying group improved significantly from preoperative to postoperative sessions (p < .05). Preoperative test scores were significantly higher for the +5 dB SNR qualifying group when compared to preoperative scores for the remaining two groups (p < .05); however, there were no significant differences in postoperative scores across the three qualifying groups (p > .05; see Figures 3 and 4).

Sentence Recognition Outcomes in Noise: +5 dB SNR

The analysis of the +5 dB SNR condition yielded no significant main effect of CI qualifying group, F(2, 602) = 2.3, p = .10, but a significant main effect of test session, F(1, 602) = 91.2, p < .0001. The interaction between CI qualifying group and test session was not significant, F(2, 602) = 0.00, p = .10. The post hoc analysis on the significant main effect of test session suggested significantly higher postoperative scores (p < .05).

Degree of Change in Speech Recognition Outcomes

In addition to the primary analyses, an additional two-factor RM ANOVA was conducted to compared the degree of change (i.e., postoperative–preoperative score) or percentage-point improvement experienced by each CI qualifying group in each of the test conditions (see Table 2). This analysis suggested a significant main effect of CI qualifying group, F(2, 1429) = 23.9, p < .0001, a significant main effect of test condition, F(2, 1429) = 7.1, p < .1, and a significant interaction effect between CI qualifying group and test condition, F(4, 1427) = 87.0, p < .0001.

Post hoc analyses on the main effect of CI qualifying group found significant differences across all groups with the greatest improvements achieved by the quiet qualifying group followed by the +10 and +5 dB SNR qualifying groups. When examining the main effect of test condition, the largest improvements were obtained in the +10 and +5 dB SNR test conditions when compared to the quiet test condition. Post hoc analyses of the interaction effects yielded multiple findings that are summarized in Table 3. Most of the post hoc findings were expected with the exception of the similar levels of improvements in all three qualifying groups for the +5 dB SNR test condition. Even the poorest preoperative performers (i.e., quiet qualifying group) experienced an average 31% gain in speech recognition after implantation at a +5 dB SNR.

Table 3.

Summary of post hoc analyses on the degree of change in preoperative to postoperative speech recognition scores.

Comparison	Variable	Summary of results
Group comparison within each test condition	Quiet condition	• Quiet group greater improvements than +10– and +5 dB SNR groups • Similar improvements for +10– and +5 dB SNR groups
	+10 dB SNR condition	• Quiet and +10 groups had similar improvements that were greater than the improvements in the +5 group
	+5 dB SNR condition	• All groups achieved similar improvements
Condition comparison within each qualifying group	Quiet group	• All conditions showed significantly different improvements • Quiet condition yielded largest improvements followed by +10– and +5 dB SNR conditions
	+10 dB SNR group	• +10– and +5 dB SNR conditions yielded greater improvements than the quiet condition
	+5 dB SNR group	• +5 dB SNR condition yielded greater improvements than the +10 dB SNR and quiet conditions • Similar improvements in +10 dB SNR and quiet conditions

Note. Analyses conducted with Tukey–Kramer Multiple Comparisons Test. p < .05; SNR = signal-to-noise ratio.

Individual Speech Recognition Outcomes

In addition to group level analyses, individual patient data were analyzed to determine the number of patients in each qualifying group who experienced significant postoperative benefit in quiet, +10 dB SNR, and +5 dB SNR test conditions. Figure 5 plots individual preoperative and postoperative AzBio scores for the (a) quiet, (b) +10 dB SNR, and (c) +5 dB SNR test conditions for all patients who qualified based on their performance in quiet (open squares), +10 dB SNR (gray circles), or +5 dB SNR (black triangles). The solid lines on each plot represent the critical difference scores for the AzBio sentence test (Spahr et al., 2012), which were used to compute whether a patient experienced a significant postoperative benefit relative to the preoperative score. Table 4 summarizes the number and percentage of patients in each qualifying group who experienced significant improvements in the three test conditions.

Figure 5.

Pre–postoperative AzBio sentence scores in (A) quiet, (B) +10 dB SNR, and (C) +5 dB SNR test conditions for individuals who qualified for implantation in quiet (open squares), +10 dB SNR (filled circles), or +5 dB SNR (filled triangles). Ninety-five percent critical difference scores are represented by the solid lines.

Table 4.

Number of patients (percentage in parenthesis) in each qualifying group who experienced significant postoperative improvement relative to preoperative performance (p < .05) in the three different test conditions.

	Qualifying condition
Test condition	AzBio Quiet	AzBio +10 dB SNR	AzBio +5 dB SNR
Quiet	641/721 (88.9%)	34/65 (52.3%)	115/238 (48.3%)
+10 dB SNR	29/34 (85.3%)	44/55 (80%)	8/17 (47.1%)
+5 dB SNR	36/55 (65.5%)	6/9 (66.7%)	174/237 (73.4%)

Note. Significant improvement based on 95% critical difference scores (Spahr et al., 2012); SNR = signal-to-noise ratio.

Patient-Reported Outcome Measures

Table 5 displays the descriptive SSQ statistics at preoperative and postoperative test sessions in the quiet qualifying and +5 dB SNR qualifying groups who also had AzBio pre- and postoperative scores. Figure 6 illustrates the total ratings averaged across the three SSQ domains. Ratings are on a 0-point (not at all) to a 10-point (all of the time) scale. The quiet qualifying group showed an average total SSQ improvement of 2.7 points after implantation, and the +5 dB SNR qualifying group showed an average SSQ improvement of 1.7 points after implantation. For the quiet qualifying group, paired t tests indicated significant improvements across the speech-hearing, t(67) = −8.7, p < .0001; spatial, t(67) = −7.7, p < .0001; and qualities of hearing, t(67) = −8.71, p < .0001, domains, as well as total SSQ, t(67) = −9.6, p < .0001. Patients in the +5 dB SNR qualifying group also reported significantly improved performance on the speech-hearing, t(36) = −8.3, p < .0001; spatial, t(36) = −4.5, p < .0001; and qualities of hearing, t(36) = −4.4, p < .0001, domains as well as total SSQ, t(36) = −6.3, p < .0001.

Table 5.

Descriptive SSQ statistics by qualifying group for patients with both SSQ and speech recognition results.

SSQ domain	Descriptive statistics	Qualified AzBio Quiet		Qualified AzBio +5 dB SNR
		Preop (0–10)	Postop (0–10)	Preop (0–10)	Postop (0–10)
Speech–hearing	Average (SD)	1.5 (1.5)	4.1 (2.1)	2.6 (1.2)	4.9 (1.9)
	Mdn	1.2	4.5	2.6	5.1
	Range	0–8	0–9.5	0–5.6	1.4–8.6
Spatial	Average (SD)	1.9 (1.9)	4.3 (2.6)	4.4 (2.1)	5.7(2.2)
	Mdn	1.65	54.7	4.1	6.1
	Range	0–8.3	0–9.5	0–8.7	0.4–8.8
Qualities of hearing	Average (SD)	2.42 (2.0)	5.1 (1.9)	4.8 (1.5)	6.3 (1.7)
	Mdn	2.1	5.2	5.2	6.4
	Range	0–7.8	0–9	0–7.2	1.8–8.9
Overall SSQ	Average (SD)	1.93 (1.6)	4.5 (1.9)	3.9 (1.3)	5.6 (1.6)
	Mdn	1.8	4.6	4.2	5.8
	Range	0–6.9	0–9.1	0–6.33	1.3–7.9

Note. SSQ = Speech, Spatial and Qualities of Hearing Scale; Preop = preoperative data; Postop = postoperative data; SNR = signal-to-noise ratio.

Figure 6.

Mean pre–postoperative overall Speech, Spatial and Qualities of Hearing Scale ratings averaged across the three domains. Bars represent 1 SD.

Discussion

To qualify for a CI, CMS stipulates aided sentence recognition scores must be 40% or poorer in the best-aided condition, but whether testing takes place in quiet or background noise is not specified. This study used a large CI database to determine how many patients qualified for a CI based on preoperative sentence recognition performance in quiet and noise. Postoperative self-report and speech perception benefit for patients who qualified in quiet and noise listening conditions were also examined. The results suggest that of the 1,611 patients identified as CI candidates in the registry, 63% of recipients qualified in quiet, 10% qualified in a +10 dB SNR, and 27% qualified in a +5 dB SNR. Patients who qualified in quiet obtained significant speech perception benefit in all listening conditions, but experienced the greatest improvements in quiet. For patients who qualified in noise, the greatest postoperative speech perception benefits occurred for the noise test conditions. The implications of these findings will be discussed.

Postoperative Speech Perception Benefit in Quiet

For all CI qualifying groups, mean postoperative speech recognition performance significantly improved in quiet. In the quiet listening condition, those who qualified in quiet experienced an average improvement of 52% points compared to the +10 dB SNR or +5 dB SNR qualifying groups who experienced 14% and 12% point improvements, respectively. The smaller observed improvements in postoperative performance in quiet for the two noise qualifying groups can be attributed to their higher average preoperative performance. Mean preoperative performance on the AzBio in quiet was 71% for the +5 dB SNR group, 60% for the +10 dB SNR group, and 15% correct for the quiet qualifying group. As a result, there was less room for improvement for patients who qualified in noise. This finding is consistent with Mudery et al. (2017) who found older adults with preoperative speech scores > 40% in quiet but ≤ 40% in noise showed smaller postoperative improvements in quiet.

Postoperative Speech Perception Benefit in Noise

In this study, patients in all three qualifying groups experienced significant postoperative improvement in noise. However, patients who qualified in noise experienced greater postoperative benefit in noise than in quiet. Moreover, when patients who qualified in noise were tested postoperatively using the same noise level as used to determine candidacy, the derived benefit from CI for these patients was evident. For example, the +10 dB SNR qualifying group experienced an average improvement of 35% when postoperative testing was performed in the same +10 dB SNR condition compared to a 14% improvement in quiet and 32% improvement in a +5 dB SNR condition. Likewise, the +5 dB SNR qualifying group experienced the most benefit in the same postoperative +5 dB SNR condition (32% improvement) compared to the smaller improvements experienced in the +10 dB SNR (15% improvement) or quiet (12% improvement) conditions. Of note, when tested in the most difficult noise condition (+5 dB SNR), patients in the quiet qualifying group obtained similar average postoperative improvement as observed for the two noise qualifying groups.

Patient-Reported Benefits

While measures of quantitative sentence recognition are an important postoperative metric of CI benefit, an equally important measure is the subject's perceived change in disability. In this study, a subset of patients in the quiet and +5 dB SNR qualifying groups also had pre- and postoperative SSQ data, and rating changes for the speech-hearing, spatial, and qualities of hearing domains were assessed. Overall, all patients showed significantly improved mean ratings (increases) across all SSQ domains (see Table 5). Similar to the postoperative speech perception results, the patients in the +5 dB SNR group had higher preoperative scores on all SSQ domains compared to patients in the quiet qualifying group. Both qualifying groups, though, showed similar overall postoperative quality of life improvements relative to preoperative ratings.

Clinical Implications

Despite the use of speech-in-noise testing to determine adult CI candidacy, few studies have evaluated how patient selection and clinical outcomes are affected when qualifying in noise. For example, in a recent survey of 81 neurotologists who routinely engage in CI care, 68% reported routinely using speech-in-noise testing to determine adult CI candidacy, 28% reported selective use of when scores in quiet were borderline, and only 4% reported never using speech-in-noise testing to determine adult CI candidacy (Carlson et al., 2018).

In the current study, 63% of a 1,611 patient sample met CMS criteria for implantation using a quiet AzBio sentence paradigm, whereas 10% met criteria in a +10 dB SNR and 27% met criteria in +5 dB SNR. Because these results suggest that nearly one third of adult CI recipients qualify in noise, it is important to understand how these patients respond to cochlear implantation both for policy considerations as well as patient counseling. The results of this study suggest that patients qualified in noise obtain significant speech perception and quality of life benefits. However, it is important to note that the centers in this data sample used a +5 dB SNR test condition more often than a +10 dB SNR test condition (see Figures 3 and 4), and this may not represent a national trend.

The results of this study highlight two important findings for CI practitioners. First, regardless of qualifying condition, individual CI patients derived the greatest postoperative speech perception benefit in their qualifying condition (see Table 4). For example, 90% of patients who qualified in quiet showed significant postoperative speech recognition benefit when tested in quiet. In contrast, only 66% of patients who qualified in quiet experienced significant improvement in the most adverse noise condition. Likewise, for patients in the +5 dB SNR qualifying group, 73% of individuals had significantly higher postoperative scores on the AzBio when tested at a +5 dB SNR. In contrast, if patients who qualified in +5 dB SNR were only tested postoperatively in quiet, our results indicate only 50% of patients would show significant improvement with their implant. Thus, these results indicate if the patient qualified in noise, he or she will need postoperative testing in noise given that CI benefit in quiet will be smaller due to ceiling performance effects (see Figure 2). Second, patients should be counseled preoperatively that expectations and room for postoperative improvement differ for patients based on their qualifying test condition and self-reported hearing abilities. The results from this study are encouraging, though, as 65% of patients who qualified in quiet experienced significant speech perception benefit in the most difficult +5 dB SNR test condition.

Benefits and Limitations of a CI Registry

This study utilized a multicenter, clinical registry for CI recipients to examine how candidacy determination in quiet versus noise conditions impacts patient selection and speech recognition. The data set compiled outcomes from 12 different clinical sites in a variety of practice settings that were located in geographically different regions of the United States. The registry employed in this study has been used in other studies (Chen et al., 2017; Grisel et al., 2018; Miller et al., 2020; Schafer et al., 2016), but no other large-scale CI registries exist in the United States. The American Academy of Otolaryngology—Head & Neck Surgery has a clinical data registry for otolaryngology (Denneny, 2016); however, that registry has yet to yield clinical outcome studies for persons with CIs. In addition, although the requirements for developing a successful national registry for auditory implants has been published (Mandavia et al., 2018), an actual registry has yet to be created. The results of this study demonstrate the value of focused, clinical registries in establishing benchmarks for postoperative improvements and understanding variability in practice patterns. A potential weakness of this study is that clinical sites are not rigorously monitored for data quality or completeness as participation in the registry is voluntary. While clinical sites have no reason to enter inaccurate information, site visits confirming data accuracy have not been performed. In addition, as all data entry is voluntary, the completeness of data elements for each patient is variable. Finally, the data collection procedures slightly differed across clinics, resulting in small differences in presentation level for speech recognition testing. However, it should be noted that the large CI centers that contribute to the registry adhere to well-established and validated CI evaluation protocols.

Conclusions

Using a clinical CI registry, the results of this study suggest CI candidates who qualify for implantation in quiet or background noise test conditions show significant postoperative improvement in speech perception and quality of life scores. For patients qualifying in noise, the greatest speech perception gains were seen when the qualifying noise condition was used to track performance.

Author Contributions

Christopher Silva: Data curation (Lead), Software (Lead).

Funding Statement

This research was supported in part by Research Grant 2P50DC000242 from the National Institutes on Deafness and Other Communication Disorders, National Institutes of Health (University of Iowa) and National Institutes on Deafness and Other Communication Disorders Grant RO1 DC13117 (Vanderbilt University Medical Center), the Lions Clubs International Foundation, and the Iowa Lions Foundation (University of Iowa).