user preference and reliability of bilateral hearing aid ......settings. performance differences...

158

J Am Acad Audiol 19:158–170 (2008)

*Department of Hearing and Speech Science, Vanderbilt Bill Wilkerson Center

Benjamin Hornsby, Dan Maddox Hearing Aid Research Laboratory, Vanderbilt Bill Wilkerson Center, Room 8310 Medical CenterEast, South Tower, 1215 21st Ave. South, Nashville, TN 37232-8242; Phone: 615-936-5132; Fax: 615-936-5013; E-mail:[email protected]

This research was supported in part by a grant from Siemens Hearing Instruments, Inc., and the Dan Maddox Hearing AidResearch Foundation.

Portions of this work were presented at the American Academy of Audiology’s 17th Annual Convention, Washington, DC, April2005.

Disclosure: H. Gustav Mueller is a consultant for Siemens Hearing Instruments.

User Preference and Reliability of BilateralHearing Aid Gain Adjustments

Benjamin W.Y. Hornsby*H. Gustav Mueller*†

Abstract

The purpose of the current study was to evaluate the consistency and reliabilityof user adjustments to hearing aid gain and the resulting effects on speechunderstanding. Sixteen bilaterally aided individuals with hearing loss adjustedtheir hearing aid gain to optimize listening comfort and speech clarity whilelistening to speech in quiet and noisy backgrounds. Following these adjustments,participants readjusted their aids to optimize clarity and comfort while listeningto speech in quiet. These final gain settings were recorded and compared tothose provided by NAL-NL1 prescriptive targets. In addition, speechunderstanding was tested with the hearing aids set at target and user gainsettings. Performance differences between the gain settings were thenassessed.

Study results revealed that although some listeners preferred more or lessgain than prescribed, on average, user and prescribed gain settings were similarin both ears. Some individuals, however, made gain adjustments between earsresulting in “gain mismatches.” These “mismatches” were often inconsistentacross trials suggesting that these adjustments were unreliable. Speechtesting results, however, showed no significant difference across the differentgain settings suggesting that the gain deviations introduced in this study werenot large enough to significantly affect speech understanding.

Key Words: Bilateral, gain, hearing aids, hearing loss, prescriptive formula,speech understanding, volume control

Abbreviations: HINT = Hearing in Noise Test; NAL-NL1 = National AcousticLaboratories—Nonlinear; REIG = real ear insertion gain; SNR = signal-to-noiseratio; VC = volume control

Sumario

El propósito del actual estudio fue evaluar la consistencia y la confiabilidadde los ajustes ganancia del auxiliar auditivo por parte del usuario, en cuantoa mejorar la comprensión del lenguaje. Dieciséis individuos hipoacúsicos conamplificación bilateral ajustaron la ganancia de sus auxiliares auditivos paraoptimizar la comodidad al escuchar y la claridad del lenguaje mientrasescuchaban lenguaje en ambientes silenciosos y ruidosos. Después de estosajustes, los participantes reajustaron sus auxiliares para optimizar la claridad

DOI: 10.3766/jaaa.19.2.6

The use of wide dynamic range com-pression can, in theory, reduce theneed for user adjustments of hearing

aid gain via a volume control (VC) mecha-nism. For example, Kochkin (1995) reportedthat a large percentage of completely-in-the-canal (CIC) hearing aid users were satisfiedwith their hearing aid’s volume adjustment,even though the CIC hearing aids that theywere using did not have a volume control. Inreality, however, many individuals with hear-ing loss may prefer access to a VC to allowgain adjustments in various environments(Surr et al, 2001; Kochkin, 2003). This is notsurprising given that the most comfortablelevel (MCL) for listening is more accuratelydescribed as a most comfortable range(MCR). When listening to speech stimuli, theMCR can be as large as 20–30 dB for personswith hearing loss, depending on the meas-urement method and degree of hearing loss(Ventry and Johnson, 1978). In addition, pre-ferred listening levels can vary based on thelistening conditions, signal-to-noise ratio(SNR), and type of noise (Kuk et al, 1994;Keidser et al, 2005; Smeds et al, 2006a,2000b). Therefore, although low-thresholdcompression systems may reduce the needfor a VC, compared to a linear aid, in manyinstances user access to VC adjustments

remains quite desirable (Surr et al, 2001;Kochkin, 2003).

When hearing aids are fit bilaterally, asthey are in approximately 80% of fittings inthe United States (Strom, 2006), an impor-tant goal of the bilateral fitting is to adjustthe gain of each hearing aid to create a “bal-anced” perception of loudness between earsand to place amplified conversational speecharound the midpoint of the patient’s “com-fortable” range. To achieve this goal the fre-quency response and overall gain of the hear-ing aids are initially set based on some pre-scriptive rule that takes into account thepatient’s audiometric thresholds (e.g.,National Acoustic Laboratories—Nonlinear 2(NAL-NL2; Dillon, 2006), Desired SensationLevel Version 5 (DSL 5.0; Scollie et al, 2005),or a manufacturer’s first fit algorithm).Overall loudness and loudness balance arethen typically verified using subjective meth-ods. Although there is no consensus on howto best verify aided loudness perception forspeech, a reasonable approach is to presentcontinuous discourse in the sound field andsystematically raise and lower the overallgain until optimal loudness, based on subjec-tive report, is achieved (e.g., Cox and Gray,2001). Loudness balance can be verifiedusing a similar technique by systematically

Bilateral Hearing Aid Gain Adjustments/Hornsby and Mueller

159

y la comodidad al escuchar lenguaje en silencio. Estos ajustes finales deganancia fueron registrados y comparados a aquellos indicados por las metasde prescripción del NAL-NL1. Además, se evaluó la comprensión del lenguajecon los auxiliares auditivos graduados en el nivel meta y en el nivel escogidopor el usuario. Entonces se evaluaron las diferencias de desempeño entre dichosajustes de ganancia.

El estudio reveló que aunque algunos sujetos prefirieron más o menosganancia de la prescrita, en promedio, los ajustes de ganancia del usuario ylos prescritos fueron similares en ambos oídos. Algunos individuos, sinembargo, realizaron ajustes de ganancia entre sus dos oídos que resultaronen “desajustes de ganancia”. Estos “desajustes” fueron a menudo inconsistentesen los diferentes ensayos sugiriendo que eran no confiables. Los resultadosde las pruebas de lenguaje, sin embargo, no mostraron diferencias significativasentre los diferentes ajustes de ganancia, sugiriendo que las desviaciones deganancia introducidas en este estudio no fueron lo suficientemente grandespara afectar significativamente la comprensión del lenguaje.

Palabras Clave: Ganancia bilateral, auxiliares auditivos, hipoacusia, fórmulade prescripción, comprensión del lenguaje, control de volumen

Abreviaturas: HINT = Prueba de Audición en Ruido; NAL-NL1 = LaboratoriosNacionales de Acústica – No lineal; REIG = ganancia de inserción de oído real;SNR = tasa de señal/ruido; VC = control de volumen

adjusting the gain of the “louder” or “softer”side until a subjective report of “balance”between ears is obtained.

Given the relatively large range of “com-fortable” and the fact that MCL varies withSNR, it is expected that listeners will makegain adjustments as they move from oneenvironment to another. Research has shownthat many elderly hearing aid users have dif-ficulty manipulating VC mechanisms onhearing aids (Ward et al, 1979), and for thisreason, their adjustments might not be pre-cise. A potential negative consequence ofthese user gain adjustments is creating a“mismatch” between ears. In other words thecarefully balanced gain settings providedduring the initial fit may be significantlyaltered due to user gain adjustments.Depending on the magnitude of this mis-match, there is a potential for speech under-standing to be affected in certain environ-ments.

We know that, in many conditions, bilat-eral listening can result in improved speechunderstanding in background noise. Themagnitude of the bilateral benefit will varybased on several factors including the config-uration of the speech and noise and the typeof masking noise (Bronkhorst and Plomp,1989; Gallun et al, 2005). Ricketts (2000)compared bilateral and unilateral aided per-formance and showed an approximately 2.5dB improvement in the SNR needed for 50%sentence understanding when listening in abackground noise of cafeteria babble thatsurrounded the listener. Although reducingthe gain at one ear does not automaticallyresult in a unilateral condition, at somepoint, if the mismatch becomes severeenough, the speech understanding wouldonly be representative of the hearing aid(ear) with the most gain, and bilateral bene-fit would be reduced or absent.

The issue of user gain adjustments also isimportant when considering hearing aidsthat have a “linked-gain” adjustment. Thatis, when gain is turned up or down in one, thesame adjustment is applied to the other hear-ing aid. Moreover, hearing aids are nowavailable that automatically readjust gainbased on the users preferred settings.Knowledge about hearing aids users’ abilityto adjust gain reliably would help determineif this “gain learning” should occur with thehearing aids in a linked or unlinked mode.

Unfortunately, a review of current litera-

ture reveals no systematic research evaluat-ing the magnitude of hearing aid gain mis-match or its potential effects on speechunderstanding. The purpose of this experi-ment was to (1) quantify the consistency andmagnitude of gain adjustments made byhearing aid users and compare these gainsettings to their National AcousticLaboratories—Nonlinear (NAL-NL1) pre-scribed settings and (2) determine if differ-ences from clinician-adjusted gain, resultingfrom user adjustments of the hearing aidgain, affect speech understanding.

PROCEDURES

Participants

Participants consisted of sixteen (16)adults (eight male and eight female) withsymmetrical, mild-severe sensorineuralhearing loss (SNHL). Symmetrical wasdefined as ≤20 dB difference between ears atany audiometric test frequency. Ninety-sixpercent (96%) of participant thresholds (138out of 144 thresholds) were ≤10 dB betweenears. All participants’ hearing thresholdswere between 15 and 75 dB HL at 500 Hzand between 50 and 85 dB HL at 3000 Hz(see Figure 1). Participants with hearing lossranged in age from 23 to 82 years (medianage 75.5 years). Fifteen of the 16 participantswere experienced hearing aid users (averageof 12.7 years of usage). In addition 14 of 15users had manual volume controls on theircurrent aids while one participant used anexternal remote control to adjust the gain ofhis hearing aids. Additional demographicinformation regarding the study participantsis provided in Table 1.

Hearing Aids

All participants were fitted with bilateralTriano 3P BTE aids. The devices were pro-grammed for omnidirectional mode only. Inaddition, given that the focus of the projectwas on user adjustments of gain, all noisereduction (termed “Speech ComfortSystem™” processing) and feedback suppres-sion algorithms were disabled during fittingand testing. Disabling these features limitedthe influence of these potentially confoundingvariables and ensured that overall gain pref-erence, as opposed to potential interactionsbetween gain and feature specific processing,

Journal of the American Academy of Audiology/Volume 19, Number 2, 2008

160

was the primary factor responsible for ourresults. To allow for substantial gain changes,the VC range, which can be controlled via themanufacturers fitting software, was set to 16dB (±8 dB from the programmed gain set-ting). The test hearing aids used a toggleswitch, rather than a volume control wheel, tocontrol the aid volume. Holding the toggleswitch in a fixed position (up or down) sys-tematically increased or decreased the vol-ume until the toggle switch was released orthe maximum change in volume (8 dB) was

reached. Pushing the toggle switch up ordown and then releasing it increased ordecreased the volume in 1 dB steps. An audi-ble beep was heard with each 1 dB change inlevel. In addition, opening and closing thehearing aid battery door reset the devices tothe programmed gain setting.

Participants were fit bilaterally usingNAL-NL1 prescriptive real ear insertion gain(REIG) targets. The REIG targets werederived based on the following software choic-es from NAL-NL1’s selection screen options:BTE for hearing aid type, bilateral fitting,and four compression channels. Vent sizesranged from 3 mm to unvented and werebased on degree of hearing loss.

All fittings were verified using the steady-state composite noise test signal, presentedfrom a 0° azimuth, on the Frye Systems Fonix6500 real ear analyzer. The hearing aids wereadjusted until a best match to target wasobtained. Other than adjustments made tolimit feedback, no further “fine-tuning” was per-formed. For a 65 dB input, measured REIGvalues were, on average, within 3 dB of pre-scribed gain values through test frequenciesof 2000 Hz and within about 7 dB of targetout to 4000 Hz (see Figure 2). While the focusof this research was gain for the 65 dB SPLinput signal, the hearing aids used in thestudy employed WDRC, and compressionkneepoints and ratios across the 16 channelswere adjusted according to the NAL-NL1 pre-scription. Hence, hearing aid gain for inputsbelow 65 dB SPL were also close to NAL-NL1targets. REIG measures made with a 50 dB


161

Figure 1. Mean and range of participant audio-grams. Filled and open symbols represent the rightand left ear, respectively. The dashed lines representthe best and poorest thresholds as a function of fre-quency across all subjects.

Table 1. Demographic Data of Study Participants

Years Wear Monaural/ Years of Years of Volume Control Participant Age of HL HAs? Bilateral HA use Education on Current Aid

1 73 12 Yes Bilateral 5 14.5 No2 82 28 Yes Bilateral 18 16 Yes3 72 12 Yes Bilateral 8 8 Yes, remote4 81 7 Yes Monaural 6 13.5 Yes5 76 10 No None 0 17 No6 23 23 Yes Bilateral 15 17 Yes7 71 40 Yes Bilateral 25 18 Yes8 79 3 Yes Monaural 0.5 10 Yes9 79 14 Yes Bilateral 12 14 Yes

10 75 8 Yes Bilateral 6 13 Yes11 80 3 Yes Bilateral 1 12 Yes12 51 51 Yes Bilateral 48 12 Yes13 82 8 Yes Bilateral 4 14 Yes14 61 11 Yes Bilateral 11 13 Yes15 72 11 Yes Bilateral 10 14 Yes16 82 40 Yes Bilateral 34 14 Yes

Average 71.2 17.6 12.7 13.75

SPL input signal were, on average, within3–4 dB of target values for test frequenciesthrough 3000 Hz (data not shown).

Following real ear verification and adjust-ment, the hearing aids were removed fromthe listener, and 2-cc coupler gain measures,using a 65 dB composite noise as an inputsignal, were made for the right and left hear-ing aids. A “midfrequency” average gain(MFA; gain at 1, 2, and 3 kHz) measure wasused for comparison purposes. This providedus with a “baseline” measure of the clinicianfit gain for both hearing aids.

Volume Control AdjustmentProcedures

Prior to testing, all participants receivedtraining in the use of the hearing aid VC. Thetraining was informal but involved requiringthe participants to adjust the gain on thehearing aids to full on (or just prior to feed-back) and then reduce gain on the hearingaids to a very low level. Participants wereallowed as long as necessary to work with the

devices until they felt comfortable makingadjustments. This generally required nomore than five to ten minutes. If a potentialstudy participant was unable to consistentlymake VC adjustments they were excludedfrom the study.

Following the training procedure, the aidswere reset to their programmed gain settingsby opening and closing the battery doors.Participants were then asked to adjust theirhearing aid gain, if required, to optimize clar-ity and comfort while listening to speech inquiet (65 dBA). Participants were instructedto listen first and only make adjustments ifrequired. Following this procedure, the hear-ing aids were removed, and hearing aid gainwas remeasured in a 2-cc coupler, againusing a 65 dB composite noise as the input.This allowed us to compare participants’ pre-ferred gain with the clinician-prescribedgain.

During the adjustment process, speechmaterials were presented through a TannoySystem 600™ loudspeaker located approxi-mately 1.25 m from the listener at a 0°azimuth. Practice passages from theConnected Speech Test (CST; Cox et al, 1987)were used as the speech materials. Each pas-sage consists of eight to ten sentences on aspecific topic read from a children’s encyclo-pedia by a female talker. Quiet speech waschosen as the test material given NAL-NL1targets are based on optimizing understand-ing, clarity, and comfort of speech in quiet(Ching et al, 2001).

After this initial check of each individual’spreferred gain settings, participants listened,in a random order, to speech in three simu-lated environments that were designed toencourage user adjustments of hearing aidgain. These environments (described below)were created by varying the location of thespeech and noise loudspeakers, and the typeof background noise, in a 3.2 x 3.2 x 3.2 msound-treated room (Rt60 ~450 msec). Speechunderstanding testing (described later) wasalso conducted in this same room. In all sim-ulated environments the speech and noiseloudspeakers were located 1.25 m from thecenter of the listener’s head.

Simulated Environment 1 (SE1)

This condition was designed to simulate aloud bar/restaurant setting and encouragelisteners to decrease gain compared to the


162

Figure 2. Measured real ear insertion gain using a65 dB input signal for the right and left ears. NAL-NL1 prescriptive targets are shown for reference.Error bars show 1 standard deviation of the differ-ences between measured and prescribed gain.

prescribed settings. Continuous discourse,using the CST practice passages, was pre-sented at a +5 dB SNR from a 0° azimuth.The speech materials were presented at anoverall level of 80 dBA. The backgroundnoise consisted of five uncorrelated segmentsof cafeteria babble presented from loud-speakers located at equal intervals aroundthe listener (36°, 108°, 180°, 252°, and 324°).The segments of cafeteria babble were pre-sented through Definitive Technologies loud-speakers (Model BP-2X). These loudspeakershave a bipolar pattern and use two matcheddriver pairs, which are mounted 150° toeither side of the central axis of the loud-speaker resulting in increased sound disper-sion within the test environment. The centerof both speech and noise loudspeakers werepositioned at approximately ear level of thelistener (36”). Noise levels from individualspeakers were equated, and their overall lev-els were adjusted to create a combined back-ground noise level of 75 dBA (Larson DavisModel 814 SLM, slow averaging).


This condition was designed to simulatelistening to soft speech or speech at a dis-tance and thus encourage listeners toincrease gain compared to the prescribed set-tings. In this condition, continuous discourse,again CST practice passages, was presentedin quiet at an overall level of 45 dBA from thespeech speaker (0° azimuth).


This condition was designed to simulatean asymmetric listening environment inwhich the noise was presented predominate-ly to one ear and the speech predominately tothe other. This setting was included toencourage participants to adjust their VCs tocreate a “mismatch” between ears.Specifically, this setting was designed in anattempt to encourage the listener to increasegain on the ear near the speech (right ear)and decrease gain on the ear near the noise(left ear). The background noise consisted oftraffic noise recorded from an intersectionnear the Vanderbilt Bill Wilkerson Center.The traffic noise was presented at an overalllevel of 65 dBA using three BP-2X loud-speakers placed on the left side of listener atazimuths of -45°, -90°, and -135°. The speech

materials were presented from a loudspeak-er located to the right of the listener (90°azimuth) at an overall level of 65 dBA, result-ing in a 0 dB SNR listening condition.

While listening in each environment, sub-jects were instructed to adjust the gain oftheir hearing aids to achieve optimal clarityand comfort. No speech testing was conduct-ed in any simulated environment. In addi-tion, we did not record the final VC settingsmade by the user in these simulated envi-ronments. These conditions were includedsolely to encourage listeners to make sub-stantial (but realistic) gain changes.

After listening in each simulated environ-ment, each participant was asked to readjustthe gain of their hearing aids for optimalclarity and comfort while listening to speechin quiet presented at a level of 65 dBA (asthey did after the initial fit). With the gainleft at this setting (i.e., for speech in quiet)the hearing aids were removed from the lis-tener by the experimenter, and 2-cc couplergain measures were again obtained andrecorded.

To obtain an estimate of gain adjustmentconsistency, all participants repeated theprocess of listening and adjusting their hear-ing aids in the three simulated environments(presented again in a random order) a secondtime during the study. Following this repeti-tion, participants again listened to speech inquiet (65 dBA) and readjusted hearing aidgain for clarity and comfort a third time (onceprior and twice after listening in the simulat-ed environments).

Sentence Understanding Procedures

Speech understanding was assessed usinga modified version of the Hearing in NoiseTest (HINT; Nilsson et al, 1994). The modifi-cations related to the type and locations ofthe noise used during testing are describedbelow. Speech test materials were presentedthrough a single Tannoy System 600 loud-speaker positioned at a 0° azimuth. Testingwas performed at two noise levels (50 and 65dBA) and in two noise source configurations.Condition 1 used five-speaker uncorrelated,cafeteria babble with speakers located atequal intervals around the listener (36, 108,180, 252, and 324°). Condition 2 used a three-speaker noise configuration with all noisesources located on the right side of the listen-er at 45°, 90°, and 135° (speech still at a 0°


163

azimuth). This is similar, but on the oppositeside, to the noise configuration used in thethird simulated environment. The motiva-tion for switching the side of the noise loud-speakers was to create a test condition thatwould be more likely to be negatively affectedby user adjustments of volume control gain.During testing, the levels of each noisespeaker were equated prior to adjusting theircombined overall levels to either 50 or 65dBA. A lower noise level of 50 dBA wasincluded under the assumption that theeffects of “mismatches” in gain adjustmentsmay be more observable at low speech levels.

As per the HINT protocol, speech levelswere adaptively varied to determine theSNR required for 50% correct sentencerecognition. Performance was assessedusing 20 sentences (two 10-sentence lists)per condition. Speech understanding wasassessed in a total of eight experimentalconditions (two different noise source con-figurations x two noise levels x two volumecontrol settings [user adjusted and clinicianadjusted]). Participants were counterbal-anced into groups based on whether speechtesting would be performed first with thegain set based on the NAL-NL1 prescriptivefit or the user-adjusted fit. Noise configura-tion and speech presentation level duringspeech testing were randomized within eachVC setting.

If speech testing was to be done with theuser-adjusted gain setting first, then partic-ipants listened in the simulated environ-ments (in a random order), made VC adjust-ments, listened to speech in quiet, readjust-ed their gain (and settings were then record-ed), and then completed speech testing usingthe HINT materials as described above (twonoise configurations x two levels). If speechtesting was to be done first with the gain setat the prescribed gain setting, participantslistened to speech in quiet (without first lis-tening in the simulated environments) andadjusted their gain as needed. These valueswere recorded then the gain was reset to theclinician-programmed gain (by opening andclosing the battery door), and speech testingwas performed using the HINT materials asdescribed above (two noise configurations xtwo levels). For these participants, afterspeech testing was completed they listenedin the three simulated environments, in arandom order, and adjusted their hearingaid gain settings as needed. After this they

listened again in quiet and readjusted thehearing aid gain as needed. These final gainvalues were then recorded. Figure 3 shows ablock diagram of the study procedures.

RESULTS

Consistency of User Gain Adjustments

Our first comparison of interest examinedthe relationship between user adjusted gainsettings and the settings programmed by cli-nicians when fitting to NAL-NL1 targets.Recall that the VCs were programmed toallow a 16 dB range (±8 dB) around the pre-scribed gain. Although no participantsreported an inability to achieve desired gainsettings during the study procedures, it ispossible that the presence of feedback mayhave limited upper levels of gain adjust-ments for some of our listeners. The datashown in Figure 4 were derived by subtract-ing individual 2-cc coupler midfrequency


164

Figure 3. Block diagram of study protocol.

average gain values (MFA; average at 1, 2,and 3 kHz) for the right and left ears, basedon user-adjusted gain settings, from the cor-responding right or left ear coupler gain val-ues based on our fit to NAL-NL1 targets. Forexample, Subject 1’s MFA 2-cc NAL-NL1 cou-pler gain was 38 and 37 dB for the right andleft ears, respectively. Following user gainadjustments, after listening to speech inquiet, the hearing aid gain was 39 and 32 dBfor the right and left ears, respectively. Thusthe deviation from NAL best fit was 1.0 and-5.0 dB for the right and left ears, respective-ly. Described in this fashion, negative valuesrepresent an adjustment resulting in lessgain than that given by the clinicians fit toNAL-NL1 targets. These data provide, foreach ear, a "deviation from NAL-NL1 bestfit," essentially showing how the user adjust-ed the gain, relative to the programmedNAL-NL1 gain, when listening to speech inquiet.

Figure 4 shows the user gain adjustmentsimmediately following the real ear fittingprocedure. Participants’ gain settings werepreset at the clinician’s fit to NAL-NL1 (byturning the hearing aid off and then back on)after which they were instructed to makegain adjustments, if needed, to optimize thecomfort and clarity of speech in quiet (65dBA). The purpose of this was to observe ifuser-adjusted gain was significantly different

than our best fit to NAL-NL1 targets.Results in Figure 4 show that, on average,

subjects adjusted the hearing aid gain toslightly less (1–2 dB) than our best fit to theNAL-NL1 targets. Results of a T-test com-paring these deviations showed that the dif-ferences were not statistically significant (t =-1.075, p = 0.30 and t = -1.962, p = 0.069, forthe right and left ears, respectively). Therewas, however, substantial variability acrossparticipants. Six of the 16 participantsadjusted the hearing aid gain in at least oneear so that it was at least 4 dB different thanthe programmed gain. Three of the four par-ticipants adjusted the gain in both ears thanto less than programmed gain while one par-ticipant adjusted the gain in both ears tomore than programmed gain settings. Twoadditional participants, discussed below,chose gain values at least 4 dB different thanprescribed in one ear only resulting in a “gainmismatch.” Gain mismatches due to useradjustments (i.e., differences in the deviationfrom NAL-NL1 targets between ears) were,on average, quite small (~1 dB). Only two ofthe 16 participants showed gain mismatches>2 dB. These participants showed gain mis-matches of ~8 and 11 dB, respectively. Basedon further adjustment trials, however, itappears that these mismatches were notreflective of a true subjective preference forthese VC adjustments. These same partici-


165

Figure 4. Ear specific deviations, due to user adjustment of the hearing aid gain, from the bestfit to prescribed gain settings. These data are from the first trial in which participants did notlisten in simulated noises prior to making adjustments. The large symbol represents the aver-age results while the small symbols represent individual deviations. Positive values representuser gain adjustments that resulted in more gain than prescribed while negative values reflectadjustments leading to less gain than prescribed.

pants did not maintain a consistent prefer-ence for reduced gain in the left ear com-pared to the right during latter adjustmenttrials. In fact for both participants the gainmismatch was reversed on at least one of thesubsequent trials, suggesting that these indi-viduals were simply unreliable in their abili-ty to make gain adjustments.

A correlation analysis of the data in Figure4 (right and left ear deviations) showed astrong and significant (r = 0.70, p < 0.01) cor-relation. This suggests that when startingwith the gain set at baseline (i.e., the clini-cian’s fit), individuals may adjust overall gainup or down, compared to the clinician fit; how-ever, the relative gain between ears remainedcomparable to that provided by the clinician’sfit to NAL-NL1 prescriptive formula.

Figure 5 shows the gain deviations afterlistening in our three divergent simulatedenvironments during two separate trials.Recall that in these conditions participantswere encouraged to make gain adjustments.After listening in the divergent settings allparticipants again listened to speech in quiet(65 dBA) and then reset the hearing aid gainto their preferred settings.

Results shown in Figure 5 show again that,on average, deviations in adjusted gain fromour clinical fit to NAL-NL1 targets are smalland uniform between ears. The “overall aver-

age” deviations (i.e., the deviations averagedacross both trials) in Figure 5 are near 0 dBfor both the right and left ears; however, therewas again substantial variability across sub-jects. In addition, the magnitude of ear “mis-matches” was somewhat larger than observedduring the first trial (Figure 4) with mis-matches up to 16 dB. Baumfield and Dillon(2001) reported that rms differences of 6 dB ormore between average prescribed target gain(averaged across 1000, 2000, and 4000 Hz)could affect benefit and satisfaction with anaid. In the 32 trials (16 subjects x two trials),there were only five instances where listeners(four listeners) generated mismatches >6 dB.Only one listener created a mismatch of >6dB on both trials, and in this case the mis-matches were in opposite directions; that is,there was not a consistent preference for amismatch in one direction.

A correlation analysis continued to show asignificant correlation between adjustmentsmade to the right and left ears (r = 0.534, p <0.01) for the data combined across the two tri-als and for each trial individually (r = 0.536, p <0.05 and r = 0.570, p < 0.05 for trials two andthree, respectively). This finding also suggeststhat our study participants were relatively goodat readjusting their hearing aids to achieve“balanced” gain between ears even after listen-ing in divergent simulated environments.


166

Figure 5. Same as Figure 4 except participants first made gain adjustments while listening inthree noise conditions designed to encourage gain adjustments and then listened to speech inquiet (as in Figure 4) and made their final gain adjustments. Deviations reflect final gain adjust-ments after listening to speech in quiet. Participants were evaluated on two separate trials. Largesymbols represent average results while small symbols represent individuals on a given trial.

Effect of User Gain Adjustments onSpeech Understanding

Of primary interest in this study was thepotential effects of user gain adjustments onspeech understanding. To examine this wemeasured HINT sentence thresholds in twonoise configurations and at two noise levelswith (1) the hearing aid gain set at the NAL-NL1 prescribed setting and (2) the gain set by

the user. To obtain the user setting, subjectsfirst listened to speech and made gain adjust-ments as needed to optimize clarity and com-fort in the simulated environments describedabove. Following this, subjects listened tospeech in quiet (at 65 dBA) and readjustedtheir hearing aid gain to optimize clarity andcomfort in this setting. Speech understandingwas assessed at this gain setting (e.g., withthe gain set, by the user, to result in possiblemismatches or substantial deviations fromthe clinician gain settings) and in a separatetrial with the aids set at the prescribed gainsettings. Mean HINT scores as a function ofnoise level, noise configuration, and gain con-dition are shown in Figure 6.

Results show that, on average, user adjust-ment of aid gain had no significant impact onspeech understanding. A repeated measuresANOVA was used to examine hearing aidgain effects on speech understanding. Thewithin subjects independent variables weregain condition (NAL-NL1 or user adjusted),noise configuration (five noises surroundingthe listener or three noises on the right side),and noise level (65 or 50 dBA). The dependentvariable was the HINT score in eachgain/noise configuration/noise level condition.Of specific interest in this study, there was nosignificant effect of gain condition and no sig-nificant interaction between gain condition


167

Figure 6. HINT thresholds as a function of Noise Con-figuration (C1 = Noise Condition 1; Five noise speakers sur-rounding the listener and C2 = Noise Condition 2; Threenoise speakers on the right side of the listener) and Noiselevel (65 or 50 dBA). Solid and striped bars show data forprescribed and user gain settings, respectively.

Figure 7. Ear “mismatch” (filled circles) and HINT difference scores (filled bars) plotted sepa-rately for each speech in noise test condition and as a function of subject (see text for more details).C1: Noise Condition 1 (Speech at 0° and five noise loudspeakers surrounding the listener). C2:Noise Condition 2 (Speech at 0° and three noise loudspeakers on the right side of the listener).65 and 50 refer to the overall noise levels of 65 and 50 dBA used during speech testing.

and any other test condition.Given the substantial variability observed

in gain adjustments across users, it was alsoof interest to see if individual cases of sub-stantial gain mismatches affected speechunderstanding. Figure 7 shows differences inthe HINT scores as a function of magnitudeof gain mismatch for each of our 16 subjects(each bar represents a HINT difference scorefor a single subject in a given test condition).Ear mismatch (or gain mismatch betweenears) refers to the “difference in the devia-tions” from the clinician fit for the right andleft ear that was present during the portionof the speech testing in which the user-adjusted gain settings were used. These earmismatches were derived by subtracting thedeviation from prescribed target in the leftear from the deviation from prescribed targetin the right ear. For example, the participantwith the largest mismatch (16, mismatch of12 dB) achieved this by adjusting the gain inthe right aid up (+6 dB) compared to the pre-scribed gain and adjusting the gain in the leftaid down (-6 dB) compared to the prescribedgain. This 6 dB increase in gain in the rightear coupled with a 6 dB decrease in gain inthe left ear resulted in a 12 dB “mismatch.”The mismatches are shown by the filled sym-bols on each graph. Subjects are orderedbased on the magnitude of their mismatch.The HINT difference scores reflect thechange in HINT scores, in each condition, fol-lowing user adjustments of the gain. A posi-tive value means the HINT score was betterwhen tested with aid gain set in each ear bythe user (rather than at prescribed gain lev-els). The solid horizontal lines in the figureshow a range of ±4 dB around a change scoreof 0 dB.

There are no normative data for the criti-cal difference in HINT scores between twotest conditions, based on the noise type andconfigurations used in this study. Pastresearch using the HINT and single, steady-state noise sources (placed at 0°, 90°, or 270°)reported a 95% confidence interval of ~2 dBwhen two 10-sentence lists are used (as wasdone in our study). Given the variability inthe temporal structure of our noise (cafeteriababble), an estimate of a 4 dB 95% confidenceinterval for our test environment wasassumed.

Figure 7 shows that the difference inHINT scores exceeded the ±4 dB criteria in~15.6% of the cases (10 of 64 instances). Of

those cases where the difference exceeded 4dB, half (5 of the 10) had poorer HINT scoresand half had better HINT scores followinguser adjustment of gain. In addition, none ofthe subjects showed a consistent trendtoward better or poorer HINT performancefollowing user adjustments. A correlationanalysis of user mismatch and change inHINT score also showed no significant rela-tionship between performance on the HINTand user adjustments.

DISCUSSION

The current study had two primary purpos-es. First we were interested in whether

hearing aid users would adjust their hearingaid gain to match prescribed gain settings,particularly after being in situations thatencouraged substantial gain adjustments.Our results showed that, on average, individu-als adjusted their gain to match NAL-NL1prescribed gain settings in each ear. Althoughsome individuals preferred more or less gainthan prescribed, this preference was generallysimilar between ears (see Figures 4 and 5).There was, however, substantial variabilityacross participants with some listeners intro-ducing relatively large mismatches in gainlevels between ears. This was particularlytrue following listening in our various simu-lated environments. The introduction of rela-tively small mismatches by listeners is notunexpected. Although the binaural system inpersons with normal hearing is quite sensitiveto level differences between ears, the perform-ance of persons with hearing loss is generallypoorer and quite variable (Koehnke et al,1995). Thus some of the mismatches intro-duced by our listeners may be due to largerthan normal interaural level difference (ILD)thresholds. In addition, the task of balancingthe level between ears while listening to freefield presentation of speech is substantiallydifferent than the same/different paradigmused to quantify ILD thresholds under head-phones. It is possible that the interaural leveldifference needed to detect a perceptual “mis-match” while listening to aided speech in thefree field is somewhat larger than the thresh-old for a headphone ILD task.

A second purpose of this study was to deter-mine if gain mismatches affected aided speechperception. The study findings in this areawere clear and consistent. For the conditionstested in the study, speech at 0° and noise sur-


168

rounding the listener or to one side, useradjustment of gain did not significantly affect(negatively or positively) speech understand-ing compared to performance with the hearingaids set at the prescribed gain settings (seeFigure 6). This suggests that the magnitude ofthe deviations from prescribed gain intro-duced by users was not large enough to signif-icantly affect speech understanding.

There are several factors that may beresponsible for this outcome. A general inter-pretation of these results is that adult hearingaid users are sensitive to gain changes thatcould significantly affect speech understand-ing and are able to prevent changes of thatmagnitude from occurring. In addition, thelack of a significant effect on speech under-standing may be due to the fact that gain devi-ations were limited to ±8 dB. Allowing largerdeviations would have introduced the possibil-ity of larger effects on audibility and thus per-formance. Another factor that may have limit-ed the effect of user gain adjustments was thespeech recognition task we chose. We testedour participants in two fixed levels of back-ground noise (65 and 50 dBA). It is possiblethat the fixed level of the noise dictated audi-bility (rather than auditory thresholds) andthus speech recognition performance. That is,with noise levels fixed, adjusting the hearingaid gain changed both the speech and noiselevels, leaving the SNR largely unchanged(except in bands where audibility may havebeen affected by the gain adjustments). We diduse a lower noise level (50 dBA) in an attemptto address this issue, but no effect wasobserved at this level either. It is possible,however, that at even lower noise levels (orwith larger VC adjustments) user deviationswould affect speech understanding.

Binaural squelch (the improvement indetection and recognition of a signal with bin-aural versus the better ear monaural listen-ing) is another factor that could have beennegatively affected by user gain adjustments.Bronkhorst and Plomp (1992) reported anapproximately 3 dB improvement in the SNRneeded for 50% sentence recognition, when lis-tening binaurally compared to monaural lis-tening, in a background noise that surroundedthe listener, similar to our noise condition 1.The introduction of a large gain mismatchbetween ears has the potential to change abilateral fitting into essentially a unilateralone (e.g., turn down the gain such that the sig-nal to one ear is largely inaudible) thus reduc-

ing any potential benefit provided by binauralsquelch effects. Although there is the potentialfor user adjustments in gain to affect the ben-efits of binaural squelch, past research andthe experimental conditions used suggest thiseffect, if present, would be small.

Specifically, the largest gain mismatchintroduced during the speech testing portionof the study was approximately 12 dB (one aidturned up 6 dB and the other down 6 dB fromtarget). The binaural system, however,appears to function quite efficiently in thepresence of interaural disparities in intensity(level differences between ears) of this magni-tude, as long as the SNR between earsremains unchanged. For example, Bronkhorstand Plomp (1989) presented speech and noiseto a KEMAR manikin (in an anechoic cham-ber) and routed the sounds from the KEMAR’sears to listeners. This gave them independentcontrol of the levels at each ear. In one condi-tion a single noise source was used and waspresented from a 90° azimuth while thespeech was presented from a 0° azimuth. Inthis condition binaural speech recognition waslargely unaffected (only a 0.5 dB decrease inSRT was observed) by a 20 dB decrease inspeech and noise levels at the ear near thenoise. A larger decrease (approximately 3.5dB) was observed when the output from theear opposite the noise was attenuated by 20dB. In this condition, however, the ear withthe better SNR was being attenuated.

Further evidence for the limited effect ofuser gain adjustments on binaural benefitcomes from past research evaluating theeffects of interaural differences in level on themasking level difference (MLD). The MLDrefers to the improvement in detection orrecognition of a signal (speech or nonspeech)when listening in a dichotic (different betweenears) versus a monotic or diotic (same to bothears) condition. McFadden (1968) measuredMLDs in the NmSm (monaural noise andmonaural signal) and N0Sπ (binaural noise inphase between ears and binaural signal 180°out of phase between ears) conditions. In a setof the N0Sπ conditions, the noise level wasfixed at one ear while the signal and noisewere systematically reduced at the other ear(thus maintaining the SNR at each ear). Thisis somewhat analogous to the current study inwhich users were able to adjust the gain of onehearing aid down (decreasing both the signaland noise at one ear). The MLD was quiteresistant to changes in interaural level. No


169

effect on the MLD was observed until the levelat one ear was reduced more than 10 dB.These findings are consistent with our studyresults showing no effect of the limited adjust-ments in gain made by our listeners on speechunderstanding (Figure 6).

Finally, an important caveat to the currentstudy is that our results represent perform-ance in a laboratory setting with the listenersfocused specifically on making consistent andreliable gain adjustments. These results sug-gest that under these optimal conditions olderadults with hearing loss can make reliableand reasonable gain adjustments. Theseresults imply that the potential for the intro-duction, by hearing aid users, of gain mis-matches that will negatively affect speechunderstanding are minimal, and concern overpotential negative effects should not be a moti-vation for not providing a volume control forusers. However, recent work by Chalupperand Heuermann (2006) suggests that in real-world settings, other factors, such as time con-straints, more rapidly changing environ-ments, and physical constraints (e.g., some-thing in one hand limiting their ability toadjust a given hearing aid) may reduce thereliability and magnitude of user gain adjust-ments thus increasing the potential for nega-tive effects on speech understanding.

Acknowledgments. The authors would like tothank Erin Picou and Katherine Quinlan for theirassistance in data collection.

REFERENCES

Baumfield A, Dillon H. (2001) Factors affecting the useand perceived benefit of ITE and BTE hearing aids. BrJ Audiol 35(4):247–258.

Bronkhorst AW, Plomp R. (1989) Binaural speech intel-ligibility in noise for hearing-impaired listeners. J AcoustSoc Am 86(4):1374–1383.

Bronkhorst AW, Plomp R. (1992) Effect of multiplespeechlike maskers on binaural speech recognition innormal and impaired hearing. J Acoust Soc Am92(6):3132–3139.

Chalupper S, Heuermann H. (2006) Learning optimalgain in real world environments. Paper presented atthe International Hearing Aid Conference, Lake Tahoe,CA.

Ching TY, Dillon H, Katsch R, Byrne D. (2001)Maximizing effective audibility in hearing aid fitting.Ear Hear 22(3):212–224.

Cox RM, Alexander GC, Gilmore C. (1987) Developmentof the Connected Speech Test (CST). Ear Hear 8(5,Suppl):119S–126S.

Cox RM, Gray GA. (2001) Verifying loudness percep-tion after hearing aid fitting. Am J Audiol 10(2):91–98.

Dillon H. (2006) What’s new from NAL in hearing aidprescriptions? Hear J 59(10):10, 12, 14–16.

Gallun FJ, Mason CR, Kidd G. (2005) Binaural releasefrom informational masking in a speech identificationtask. J Acoust Soc Am 118(3):1614–1625.

Keidser G, Brew C, Brewer S, Dillon H, Grant F, StoreyL. (2005) The preferred response slopes and two-chan-nel compression ratios in twenty listening conditionsby hearing-impaired and normal-hearing listeners andtheir relationship to the acoustic input. Int J Audiol44(11):656–670.

Kochkin S. (1995) Consumer satisfaction and benefitwith CIC hearing instruments. Hear Rev 2(4):7–12.

Kochkin S. (2003) Isolating the impact of the volumecontrol on customer satisfaction. Hear Rev 10(1):26, 28,30–32, 34–35.

Koehnke J, Culotta C, Hawley M, Colburn S. (1995)Effects of reference interaural time and intensity dif-ferences on binaural performance in listeners withnormal and impaired hearing. Ear Hear 16:331–353.

Kuk FK, Harper T, Doubek K. (1994) Preferred real-ear insertion gain on a commercial hearing aid atdifferent speech and noise levels. J Am Acad Audiol5(2):99–109.

Nilsson M, Soli S, Sullivan J. (1994) Development ofthe Hearing In Noise Test for the measurement of speechreception thresholds in quiet and in noise. J Acoust SocAm 95(2):1085–1099.

McFadden D. (1968) Masking-level differences deter-mined with and without interaural disparities in maskerintensity. J Acoust Soc Am 44(1):212–223.

Ricketts T. (2000) The impact of head angle on mon-aural and binaural performance with directional andomnidirectional hearing aids. Ear Hear 21(4):318–328.

Scollie S, Seewald R, Cornelisse L, Moodie S, BagattoM, Laurnagaray D, Beaulac S, Pumford J. (2005) TheDesired Sensation Level multistage input/output algo-rithm. Trends Amplif 9(4):159–197.

Smeds K, Keidser G, Zakis J, Dillon H, Leijon A, GrantF, et al. (2006a) Preferred overall loudness. I: Soundfield presentation in the laboratory. Int J Audiol45(1):2–11.

Smeds K, Keidser G, Zakis J, Dillon H, Leijon A, GrantF, et al. (2006b) Preferred overall loudness. II: Listeningthrough hearing aids in field and laboratory tests. IntJ Audiol 45(1):12–25.

Strom K. (2006) The HR 2006 dispenser survey. HearRev 13(6):16–39.

Surr RK, Cord M, Walden B. (2001) Response of hear-ing aid wearers to the absence of a user-operated volumecontrol. Hear J 54(4):32–34, 36.

Ventry IM, Johnson JI. (1978) Evaluation of a clinicalmethod for measuring comfortable loudness for speech.J Speech Hear Disord 43(2):149–159.


170

user preference and reliability of bilateral hearing aid ......settings. performance differences...

Documents