Key Words
cochlear implants - hearing aids - hearing loss - pediatric - SII - verification
INTRODUCTION
It is well known that children with hearing loss (HL) demonstrate deficits in speech
and language skills compared with similarly aged children with normal hearing sensitivity
([Nicholas, 2000]; [Geers et al, 2009]; [Pittman and Schuett, 2013]; [Davidson et al, 2014]; [Tomblin et al, 2014]). The major reason for these deficits is the reduced ability to hear spoken language
([Tomblin et al, 2014]). That is, infants and children with HL have reduced audibility of speech and sounds
in their environment compared with those without HL and, hence, much of spoken language
is inaudible. When audibility is reduced, the ability to learn a spoken language is
affected. For children with HL, hearing aids (HA) are often used as a (re)habilitation
approach to improve the audibility of speech and sounds. For infants and children
who do not adequately benefit from appropriately fit amplification, additional approaches,
such as cochlear implant(s) (CI), may be considered.
Even though many children with severe to profound HL will eventually receive CI(s)
([Sininger et al, 2010]), a trial with well-fit HAs is typically recommended before CI surgery (often as
part of the CI candidacy process). Before implant surgery, the everyday audibility
of speech for these children will depend on both the child’s degree of HL (e.g., unaided
pure-tone average [PTA]) and on how well his/her HA(s) is/are fit. Many studies of
pediatric CI users focus on the first item and ignore the second; they report the
amount of preimplant hearing and its relation to various eventual outcomes, such as
speech perception and vocabulary development. For example, several studies of pediatric
CI recipients found that greater amounts of preimplant residual hearing were associated
with better speech perception and language outcomes ([Cowan et al, 1997]; [Nicholas and Geers, 2006]; [2007]). Likewise, [Phan et al (2016)] found that more preimplant residual hearing (i.e., lower unaided better-ear PTA;
0.5, 1, and 2 kHz) was associated with better speech discrimination performance at
two to four weeks post-CI stimulation for infants (N = 17) with a mean age of 16.4
months (standard deviation [SD = 3.6]). Notably, this result was found even though
the average “best unaided” PTA was 108.7 dB HL and ranged from 87 to 120 dB HL. Thus,
[Phan et al (2016)] concluded that even very-limited residual preimplant hearing may facilitate speech
discrimination abilities in the time period immediately following implantation.
In contrast to the many studies that report degree of preimplant hearing for eventual
pediatric CI recipients, there are few reports on the second factor affecting audibility,
namely, HA fittings before CI surgery. Although some clinicians and researchers may
report the age at which HA use was initiated ([Geers et al, 2013]; [Davidson et al, 2014]), very few report on the fittings of the HAs themselves or on the amount of speech
that is audible during HA trials. For children with less profound degrees of HL, there
are some studies of HA fittings. In studies of children with mild to severe HL, [McCreery et al (2013)] and [McCreery et al (2015)] report that HA fittings were variable. In these two studies, the proximity of the
HA output to prescribed pediatric targets (Desired Sensation Level-[DSL]) was examined
for each child’s HAs (defined as “deviation from target”). Then, aided Speech Intelligibility
Indices (SII; [ANSI S3.5, 1997]) were calculated and compared with published normative ranges for children ([Bagatto et al, 2011]; [2016]; [Moodie et al, 2017]). Approximately 55% of the children’s HAs deviated significantly from their prescribed
targets in both ears ([McCreery et al, 2013]). “Deviations from target” tended in the direction of underfitting, such that children’s
HAs with greater deviations tended to have SIIs that were lower (poorer) than the
normative range for a given degree of HL; 35% of the HAs had SIIs that were below
the published normative range for their severity of HL ([McCreery et al, 2015]).
For children with profound levels of HL, however, there is only one report on HA fittings
([Strauss and van Dijk, 2008]) and none that report the amount of speech that is audible when aided (SII data).
In several reports of an association between better preimplant residual hearing and
better postimplant outcomes (e.g., [Cowan et al, 1997]; [Nicholas and Geers, 2006]; [2007]), there is an assumption that speech was optimally audible during the preimplant
period. Yet, these studies provided no verification or documentation of the aided
HA fit (i.e., HA verification) and/or quantification of the audibility of speech.
One study, by [Davidson and Skinner (2006)], examined relations among PTAs, SIIs, and speech recognition scores for 26 children
with severe to profound HL. Although HA fits were not reported explicitly, SIIs for
two input speech levels (soft: 55 dB SPL [at the time of this study, the lowest input
level available on the Audioscan system was 55 dB SPL] and raised: 70 dB SPL) were
calculated using the Audioscan Verifit System (Etymonic Design Inc., Ontario, Canada).
The mean better-ear unaided PTA for the group was 79 dB HL (range: 60–98 dB HL), and
mean SIIs were 27% and 42% for the soft and raised input levels, respectively. Although
the calculated SIIs were relatively low, they were significantly correlated with speech
recognition scores on the Lexical Neighborhood Test (LNT; [Kirk et al, 1995]) test at both levels. That is, soft-input SIIs were significantly correlated with
speech recognition scores for soft speech (50 dB SPL) and raised-input SIIs were significantly
correlated with speech recognition scores for loud speech (70 dB SPL).
Because many children with profound HL will eventually receive one or two CIs, some
may question the value of ensuring good, or best-possible, HA fits and speech audibility
during the period when these children are using HAs. Before implant surgery, the everyday
audibility of speech will depend, as stated previously, on both a child’s degree of
HL (e.g., preimplant unaided PTA) and on how well his/her HA is fit. Clinically, although
the audiologist has no influence on a child’s degree of HL (i.e., unaided PTA), the
audiologist does have some influence on how well fit a child’s HAs are such that speech
is as audible as possible. Presumably, any improvement in the audibility of speech,
albeit even small increases in aided SII, might help the child while likely not causing
any harm. Thus, a careful and systematic examination of HA fittings is warranted.
In addition, because children with greater degrees of residual hearing are now potential
CI candidates, the need for careful HA fitting and adjustment, and documentation thereof,
are critical ([Gifford et al, 2010]; [Sampaio et al, 2011]; [Mowry et al, 2012]). Such information may assist in determining CI candidacy and with decisions about
continued use of an HA at a nonimplanted ear (i.e., bimodal use). Thus, the primary
aim of this project is to characterize HA fittings and aided audibility of speech
for pediatric HA users with severe to profound HL who later received CI(s). A secondary
aim is to examine the relation between preimplant aided audibility and postimplant
speech perception.
METHODS
Participants
Participants were drawn from a larger study examining the effects of early acoustic
hearing for pediatric CI recipients, and were recruited from audiology centers and
oral schools for the deaf across the United States. Participants met the following
criteria: severe to profound sensorineural HL in at least the poorer ear (a few exceptions
were children with etiologies of auditory neuropathy spectrum disorder, enlarged vestibular
aqueduct, and/or cytomegalovirus), HL that was congenital or acquired before 15 months
of age, educated in an oral communication setting, and at least one CI surgery before
4.5 years of age. For this present study (HA fits & SII), data were collected retrospectively
using pre-CI audiologic records.
Although there are 117 children in the larger study, only those participants with
HA verification information in their pre-CI audiological records were included in this present study. In addition, for some participants,
more than one ear fit the inclusion criteria (e.g., child has bilateral CIs and HA
verification data for both ears). When available, data collected from the participants’
records included gender, age at HA fitting(s), age at CI(s), unaided air conduction
thresholds, aided thresholds, etiology of HL, device history (i.e., type of device,
manufacturer, and date fit), present device use (i.e., simultaneous bilateral CIs,
sequential bilateral CIs, and bimodal devices), and manufacturer of devices (HA manufacturers:
Phonak, Widex, Oticon, and Unitron; CI manufacturers: Cochlear Americas [Englewood,
CO], Advanced Bionics [Valencia, CA], Med-EL [Innsbruck, Austria]). A total of 45
pediatric participants (23 female and 22 male) consisting of 71 ears were included
in this present study (see [Table 1]). These participants include bilateral CI users, simultaneous (n = 9) and sequentially
(n = 21) implanted, and bimodal CI users (n = 15), that is, those who use an HA at
the nonimplanted ear. For these 45 participants, the average age at the time of the
test (testing performed for the larger study) was 6.8 years (SD = 1.3), and the average
age at first CI was 2.3 years (SD = 1.1) with a mean preimplant unaided PTA of 99
dB HL (SD = 16). For the other 72 participants from the larger study, that is, those
who did not have preimplant HA-fitting data available in their audiological records,
the average age at the time of test was 7.2 years (SD = 1.3), average age at first
CI was 2.0 years (SD = 1.1), and mean preimplant unaided PTA was 98 dB HL (SD = 17).
Table 1
Participants and Ears with Verification Data; HAs with Active Frequency Lowering
|
HA Verification Data
|
Number of Participants
|
|
|
|
With Data for One Ear
|
With Data for Two Ears
|
Total
|
Number of Aided Ears
|
Number of HAs with Active Frequency Lowering
|
|
Any input level
|
19
|
26
|
45
|
71
|
35
|
|
Soft level
|
16
|
14
|
30
|
44
|
26
|
|
Conversational level
|
16
|
20
|
36
|
56
|
30
|
|
Loud level
|
13
|
17
|
30
|
47
|
23
|
Note: Number of participants and aided ears with available HA verification data, and number
of HAs with active frequency lowering, in total (any input level) and for three verification
speech input levels.
HA Verification
HA-fitting data from Audioscan Verifit™ records were collected and included the following (when available): measured HA output
in dB SPL for running speech (carrot passage) at different input levels ranging from
soft to loud and Audioscan software–computed SIIs ([ANSI S3.5, 1997]) for different input levels (when available). For most participants, age-related
average real ear to coupler difference values were used to simulate in situ measurements
of HA output in the coupler of the Audioscan system. From audiology records and HA-fitting
reports, participants who used HAs with frequency-lowering capability, specifically
nonlinear frequency compression, were identified. Centers differed in their definition
of input levels, with 5 dB differences noted at each level: soft = 50 or 55 dB SPL,
conversational = 60 or 65 dB SPL, and loud = 70 or 75 dB SPL. From Audioscan Verifit™ records, deviations of HA output from DSL targets (defined as HA output minus DSL target) were examined for octave frequencies from 250 to 4000 Hz, when available.
Only the deviations for the conversational speech input level (60 or 65 dB SPL) are
reported here (N = 56 HAs; 56 ears) because that level was most frequently used with
the Verifit™ (see [Table 1]). In some audiologic records, graphical data were provided in lieu of numerical
data; in these instances, a MATLAB (Mathwords, Natick, MA) script named Graph Picker
was used to estimate the numerical values of prescriptive targets and HA output at
each frequency. Finally, aided SIIs as a function of unaided preimplant PTA for these
participants’ ears were compared with normative data from the University of Western
Ontario Pediatric Audiological Monitoring Protocol (UWO-PedAMP) ([Bagatto et al, 2011]; [2016]; [Moodie et al, 2017]).
Speech Recognition Performance
As part of the larger study, participants, at ages five to eight years, were administered
the LNT while using their bilateral devices (2 CIs or CI+HA). This open-set word recognition
test consists of monosyllabic, 50-word lists drawn from the vocabulary of three- to
five-year-old typically developing children. Each child heard one 50-word list in
the presence of four-talker babble presented with an 8-dB signal to noise ratio, with
the speech level at 60 dB(A). The speech tests were administered in a sound booth,
with the prerecorded signals presented via an audiometer to a speaker located at 0°
azimuth about 3 feet from the child. The child was instructed to repeat what he/she
heard. A percent-correct word score was computed that represents the percent of words
(responses) that were recognizable as the target word.
STATISTICAL ANALYSES
To determine whether the deviations from target (HA output – DSL target) were significantly
greater than zero, 95% confidence intervals of the mean deviation (upper and lower
limits [LL, UL]) for each frequency (250, 500, 1000, 2000, and 4000 Hz) were calculated.
In addition, t-tests were used to compare preimplant-aided SIIs of HAs of listeners in each of the
three present device groups (simultaneous CIs, sequential CIs, and bimodal) and to
compare aided SIIs across ears for two groups of listeners: those who were sequentially
implanted with CIs and those who used bimodal devices. Correlational analyses were
used to examine the relationship between preimplant-aided SIIs (for the ear receiving
the first CI) and postimplant speech perception in noise. In addition, correlational
analyses were used to explore relationships between preimplant-aided SIIs and demographic
data.
RESULTS
For the ears in this study, that is, those with HA verification data, mean unaided
preimplant thresholds at octave frequencies from 250 to 4000 Hz are shown in [Figure 1]. Mean preimplant thresholds are severe to profound at all octave frequencies.
Figure 1 Mean unaided preimplant thresholds are plotted in dB HL with +1/−1 SD at octave frequencies
from 250 to 4000 Hz, for the pediatric participants in this study (N = 45). The mean
unaided thresholds are 78, 83, 89, 91, and 88, at octave frequencies 250, 500, 1000,
2000, and 4000 Hz, respectively. (This figure appears in color in the online version
of this article.)
Deviation from (Prescriptive DSL) Target
Deviations from prescriptive target, defined as “HA output minus DSL target,” at octave frequencies from 250 to 4000 Hz for conversational speech
input levels (60 or 65 dB SPL) are shown in [Figure 2]. Because HA outputs were not available at every frequency for every ear, there are
different Ns (different numbers of ears) at each frequency. [Table 2] shows, for each octave frequency, the mean, SD, and LL and UL of the 95% confidence
interval of the mean for the deviations from prescriptive target for 56 HAs (or, equivalently,
for 56 ears). Overall, for conversational speech levels, HA outputs were very close
to prescriptive targets, that is, the deviations from target are close to zero. At
4000 Hz, however, the average HA output was significantly below the prescriptive (DSL)
target, with a mean deviation of −14.2 dB. At all other frequencies, deviations from
target were not significantly different from zero. Because this analysis included
different numbers of HAs (ears) at each frequency, the analysis was repeated using
only those HAs (participants’ ears) for which conversational-level data were available
at all octave frequencies (N = 28). The pattern of results remained unchanged, namely,
only at 4000 Hz did deviations from target differ significantly from 0 (at 4000 Hz,
mean deviation = −17 dB).
Figure 2 Box plots of “deviation from prescriptive target” for conversational input levels
(60 or 65 dB SPL); box plots include HAs with active frequency lowering, from a total
of 56 HAs. In each box plot, median and interquartile range (IQR) are indicated by
the thick line in the middle, and the distance between the upper and lower extent
of the box. Whiskers indicate the minimum and maximum values, except for outliers
(circles indicate outliers >1.5 IQR’s but <3 IQR’s from the median, whereas asterisks
represent outliers >3 IQR’s from the median). Numbers of HAs included in the box plots
are listed below the graph at each octave frequency. The activation of frequency lowering
in some HAs (30 out of 56) is reflected in larger deviations from prescriptive target
at the high frequencies, particularly 4000 Hz. Use of frequency lowering is presumably
related to poor audibility at high frequencies for some participants and suggests
that clinicians were likely seeking to improve high-frequency audibility when it was
not achieved with conventional amplification. (This figure appears in color in the
online version of this article.)
Table 2
Deviations from Target
|
Frequency (Hz)
|
N
|
Mean (dB)
|
SD (dB)
|
95% Confidence Interval [LL, UL] (dB)
|
|
250
|
33
|
1.4
|
7.8
|
[−1.4, 4.1]
|
|
500
|
56
|
2.0
|
7.8
|
[−0.1, 4.1]
|
|
1000
|
54
|
1.5
|
7.2
|
[−0.5, 3.5]
|
|
2000
|
52
|
−2.1
|
8.6
|
[−4.5, 0.3]
|
|
4000
|
46
|
−14.2
|
17.3
|
[−19.3, −9.1][*]
|
* Denotes significant.
Notes: Deviations from (prescriptive DSL) target (“HA output minus target”) for conversational
speech input levels and for all HAs. The number of ears (N), mean, SD, and the LL
and UL of the 95% confidence intervals of the mean are listed. If the 95% confidence
interval does not include zero, then the mean deviation from prescriptive target is
considered significantly different from zero; any such instances are indicated with
an asterisk.
[Figure 3] and [Table 3] show analogous results, deviations from prescriptive DSL target, but excluding HAs
using frequency compression. Again, deviations from target are shown and provided
at octave frequencies from 250 to 4000 Hz for conversational speech input levels (60
or 65 dB SPL), and again, because HA outputs were not available at every frequency
for every ear, there are different Ns at each frequency. At 1000 Hz, the average deviation
was 3 dB (SD = 7.2) above target and was significantly different from zero. At all
other frequencies, the deviations from target were not significantly different from
zero.
Figure 3 Box plots of “deviation from prescriptive target” for conversational input levels
(60 or 65 dB SPL); box plots exclude HAs with active frequency lowering, from a total
of 26 HAs. For details on the box plots, see the caption for [Figure 2]. Numbers of HAs included in the box plots are listed below the graph at each octave
frequency. (This figure appears in color in the online version of this article.)
Table 3
Deviations from Target Excluding Frequency Lowering
|
Frequency (Hz)
|
N
|
Mean (dB)
|
SD (dB)
|
95% Confidence Interval [LL, UL] (dB)
|
|
250
|
11
|
3.3
|
7.3
|
[−1.6, 8.2]
|
|
500
|
26
|
2.7
|
8.0
|
[−0.5, 5.9]
|
|
1000
|
26
|
3.0
|
7.2
|
[0.1, 5.9][*]
|
|
2000
|
26
|
−0.7
|
8.0
|
[−3.9, 2.6]
|
|
4000
|
23
|
−1.6
|
9.4
|
[−5.7, 2.4]
|
* Denotes significance.
Notes: Deviations from (prescriptive DSL) target (“HA output minus target”) for conversational
speech input levels, excluding HAs that use frequency compression. The number of ears
(N), mean, SD, and the LL and UL of the 95% confidence intervals of the mean are listed.
If the 95% confidence interval does not include zero, then the mean deviation from
prescriptive target is considered significantly different from zero; any such instances
are indicated with an asterisk.
SII
Audioscan software–computed aided SIIs for soft, conversational, and loud speech input
levels are shown, using box plots, in [Figure 4] (N = 52 aided ears; only those aided ears with SII values at all three speech-input
levels are shown in this figure). As expected, SIIs are lowest for soft levels (mean
= 22; SD = 19), slightly higher for conversational levels (mean = 30; SD = 20), and
then, again slightly higher for loud levels (mean = 36; SD = 20). SIIs were also examined
for various subgroups of participants. For the simultaneously implanted bilateral
CI participants (N = 9), the aided SII associated with the better ear’s preimplant
HA records are used. For the sequentially implanted bilateral CI participants (N =
21), the aided SII associated with the first-CI ear’s preimplant HA records are used,
and for the participants with bimodal devices (N = 15), the aided SII associated with
the (only) CI ear’s preimplant HA records are used. For aided ears of listeners in
these present “device” groups, mean preimplant SII values for conversational speech
input levels were 35 (SD = 24) for the simultaneous bilateral CI group, 27 (SD = 19)
for the sequential bilateral CI group, and 30 (SD = 22) for the bimodal group. Mean
SII was not significantly different across these three groups. Finally, SIIs for conversational
speech input levels were also compared, across ears, for the participants in the sequential
CI and bimodal device groups. (A similar comparison was not done for participants
who received bilateral CIs simultaneously because presumably, they would have had
similar unaided preimplant PTAs in both ears.) For the sequential bilateral CI participants,
preimplant-aided SIIs for the first CI ear (mean = 28; SD = 28) were generally lower
than preimplant-aided SIIs for the second CI ear (mean = 37; SD = 21), although this
difference was not statistically significant. For the bimodal device users, aided
SIIs for the preimplant ears (mean = 30; SD = 22) were significantly lower [t
(14) = −3.34, p < 0.01] than the aided SIIs for the ears currently wearing HAs (mean = 52; SD = 18).
Figure 4 Box plots of aided SIIs (percentage of audible speech), computed from Audioscan software,
for three speech input levels. Only those aided ears with SII values at all three
speech-input levels are shown in this figure (N = 52). (This figure appears in color
in the online version of this article.)
In [Figure 5], Audioscan software–computed aided SIIs, for these participants’ ears for conversational
speech input levels, are plotted as a function of unaided, preimplant PTA. Also shown
are the normative data from the UWO-PedAMP ([Bagatto et al, 2011]; [2016]; [Moodie et al, 2017]). Shown in red is the line of best fit for the 67 ears in our sample; these are
all the aided ears with SIIs available for a conversational-level speech input. (For
four aided ears, SIIs were available only for soft- and/or loud-level speech inputs.)
The red line departs from the UWO-PedAMP data, especially for unaided PTAs > 75 dB
HL. Note that there are a few ears with PTAs in the moderate loss range; these ears
belong to children who had a diagnosis of progressive HL (i.e., enlarged vestibular
aqueduct or cytomegalovirus) or auditory neuropathy spectrum disorder.
Figure 5 Aided SII vs. unaided preimplant PTA for these participants’ ears (N = 67) are shown
individually by diamonds, and the line of best fit is shown in red. Normative data
are from the UWO-PedAMP. For the normative data, the light gray line represents average
aided SII values and the dark gray lines represent SDs (2 SD above and 1 SD below)
about the mean normative data. (This figure appears in color in the online version
of this article.)
Postimplant Speech Recognition Scores
The mean score for the LNT word list presented in noise (+8 dB signal to noise ratio)
was 60% correct (SD: 22.7; range 0–92%).
Correlations
Aided SIIs (at conversational speech input levels) were not significantly correlated
with age at first HA nor were they correlated with age at first CI. As expected, unaided
PTA was significantly correlated with preimplant-aided conversational-level SIIs (r = −0.936, p < 0.01) (see [Figure 5]). Preimplant-aided conversational SIIs for the first CI ear were also significantly
correlated with postimplant LNT in noise scores (r = 0.35, p < 0.05), when tested in these same children at ages five to eight years. Interestingly,
however, age at first CI was not correlated with postimplant LNT in noise scores.
DISCUSSION
The pediatric CI candidacy process typically includes a trial period with an HA, or
HAs, before any possible CI surgery. Not only does this practice assist in determining
CI candidacy but it also allows the child to receive some audibility of speech, although
perhaps quite limited. Pediatric Amplification Guidelines ([AAA, 2013]) provide specific protocols for achieving and verifying audibility targets with
HAs. Studies from children with mild to severe loss have demonstrated positive effects
of well-fit HAs on spoken language outcomes ([McCreery et al, 2015]). Thus, for pediatric recipients, it is reasonable to attribute the positive effects
of preimplant residual hearing on post-CI outcomes in-part to the audibility of speech provided by HAs.
This study sought to characterize, for pediatric CI recipients, preimplant HA fittings
and aided audibility (which is, of course, affected by PTAs and HA fittings). Of a
total of 117 participants in a larger study (NIH R01 DC012778), preimplant HA fitting
records were retrievable for 45 participants (approximately 38%). Many CI and audiology
centers reported that such data were not routinely saved in their patient records.
Although there was some variation in the speech input levels (soft, conversational,
or loud) used for matching prescriptive targets, most centers used a 60–65-dB SPL
level to represent average conversational speech. Overall, in this sample of children
with severe to profound HLs, who later received at least one CI, HAs were generally
closely matched to prescriptive targets. Because HA fitting data were unavailable
for most of the participants (i.e., the other 62%), it is uncertain whether these
data are representative of pediatric CI recipients in general. These 45 participants,
however, are seemingly no different from the remaining 72 participants with respect
to preimplant PTA, age at first CI, and age at the time of speech perception tests
(see description of participants in “Methods”).
For children with mild to profound losses, studies evaluating proximity to HA fitting
targets have found that a substantial number of children have HA outputs that deviate
significantly from target ([Strauss and van Dijk, 2008]; [McCreery et al, 2013]; [Ching et al, 2015]; [McCreery et al, 2015]). Moreover, some researchers suggest that deviations from target increase with more
severe HLs, especially deviations at high frequencies such as 4–6 kHz ([Ching et al, 2015]). Yet, for speech recognition and ultimate development of spoken language skills,
high-frequency audibility is critical for children ([Stelmachowicz et al, 2004]; [Stelmachowicz et al, 2008]). Thus, clinicians do their best, using any method possible—including frequency-lowering
methods—to provide at least limited high-frequency audibility before receiving CIs.
For the present study, children with HA outputs that deviated significantly from target
at 4000 Hz were those who used HAs with active frequency lowering. This type of processing
shifts (or lowers) high-frequency information to lower frequency regions, where listeners
generally have better residual hearing ([Simpson et al, 2005]; [Glista et al, 2009]). This type of processing aims overcoming limitations in high-frequency audibility
resulting from a combination of poor residual hearing and limited HA bandwidths. When
the data in this present study were analyzed with frequency-lowering HAs excluded,
deviations from target were not significantly different from zero. This result makes
sense because HAs with frequency-lowering do not amplify high-frequency energy (beyond
a specified cut-off frequency), but instead transmit high-frequency information at
lower frequencies.
The aided SIIs, for these participants, estimate the audibility of speech, preimplant,
in specific circumstances. When the distance between a talker and a child was approximately
1 m, then approximately 22% of softly spoken speech would have been audible (mean
preimplant SII = 22). Similarly, ∼30% and ∼36% of conversationally spoken and loudly
produced speech would have been audible, respectively, to an infant (preimplant) at
this same 1-m distance. All these values are very low. [Stiles et al (2012)] suggest that for children with mild to moderately severe HL, those with aided SII
values < 65 exhibit greater delays in vocabulary development. For the children in
the present study, with more severe to profound HL, these low SIIs represent inadequate
audibility of speech (even of “loud” speech) such that CI surgery appears clearly
warranted for all participants (recall, there was no significant difference in preimplant-aided
SII across the three device groups). For the children who currently use bimodal devices,
the significant difference in aided SII of the preimplant ear versus that of the present
HA ear indicates that the ear with poorer hearing was chosen for implantation.
Calculation of the aided SII depends on both unaided PTA and on the fit of the HA
to DSL target ([McCreery et al, 2015]). Consequently, it is unsurprising that for this study, SII and unaided PTA are
highly correlated with each other, a result consistent with other studies ([Sininger et al, 2010]; [McCreery et al, 2015]). The correlation (r = −0.936, p < 0.01) is especially strong in this study because HA fits were all very close to
DSL target. That is, overall, deviations from target were approximately zero. The
exact relation, between PTA and SII, for children with HL that range from mild to
severe, is shown by the UWO-PedAMP normative data ([Bagatto et al, 2011]; [2016]; [Moodie et al, 2017]). The regression line for SII versus PTA in the present sample, however, departs
from the normative data for PTAs > 75 dB HL. This seems likely due to the fact that
the normative values were obtained from children with greater residual hearing (i.e.,
with less severe HL) than those in the present sample. Despite these reasonably well-fit
HAs, these children were deemed ultimately and quite appropriately as candidates for
CIs.
Interestingly, despite limited residual hearing (PTA) and limited aided audibility
(SII), there was a positive and significant, albeit somewhat weak, correlation between
aided SII and post-CI speech perception in noise (and, equivalently, between unaided
PTA and post-CI speech perception in noise). These results along with those from other
studies support the notion that even limited audibility before receiving a CI may
be beneficial for speech recognition ([Cowan et al, 1997]; [Nicholas and Geers, 2006]; [2007]; [Sininger et al, 2010]). Unlike other studies ([Dowell et al, 2002]; [Geers et al, 2013]), age at first CI was not correlated with speech perception in noise for this sample.
The potential effects of age at first CI on eventual speech perception ability may
have been mitigated by the consistent use of HAs, before receiving CIs, for the participants
in this study. Regardless, this weak positive correlation provides further support
that early audibility of speech, presumably provided via HAs, could have positive
effects for children who subsequently receive CIs.
There are several considerations that may limit generalizing the results of the present
study to other pediatric CI recipients. Although most of the participants in the present
study had HAs that were fit to prescriptive target, these participants represent only
about one-third of the total population of participants in the larger study. It is
possible that centers that keep carefully documented HA-fitting records may be more
likely to place a high priority on preimplant fitting procedures and thus bias the
results toward more optimal HA fittings. In addition, all of the preimplant data were
obtained retrospectively from centers across the United States several years after
the measurements were obtained. Thus, differences in equipment, procedures, and personnel,
across centers and across time, may contribute to variations in these measurements.
There are, of course, many demographic and child factors that are very likely to affect
post-CI outcomes (i.e., maternal education, enrollment in early intervention programs,
etc.). The sample size (N = 71 ears) in the present study, however, is not adequate
for sophisticated regression analyses that would allow quantitative determination
of the contributions of child, family, educational, and audiological variables to
post-CI outcomes. Rather, the primary scope of the present study was to characterize
preimplant-aided audibility profiles of children who receive CIs and determine whether
preimplant audibility, as measured by HA fitting parameters, is related to postimplant
outcomes. An examination of aided audibility (SII), using the HA fit to target, may
guide clinicians in determining CI candidacy and should be encouraged. As children
with more residual hearing are considered for CIs, such an examination may be more
critically important than in the past.
Abbreviations
CI:
cochlear implant(s)
HA:
hearing aid(s)
HL:
hearing loss
DSL:
Desired Sensation Level
IQR:
interquartile range
LL:
lower limits
LNT:
Lexical Neighborhood Test
PTA:
pure-tone average
SD:
standard deviation
SII:
Speech Intelligibility Index
SPL:
sound pressure level
UL:
upper limits
UWO-PedAMP:
University of Western Ontario Pediatric Audiological Monitoring Protocol
