0 I96/0~02/9O/l101/0040$02.00/0 E A R AN11 H E A R I N G

Copyright 0 1990 by The Williams & Wilkins Co.

~~~

~

AMPLIFICATION AND AURAL REHABILITATION

Hearing Aid Gain and Frequency Response Requirements for the Severely/Profoundly Hearing Impaired” Denis Byrne, Aaron Parkinson, and Philip Newall

National Acoustic Laboratories, Chatswood NSW 2067 Australia [D. 6.1,and Speech, Language and Hearing Centre, Macquarie University, North Ryde NSW 21 13 Australia [A. P., P. N.]

ABSTRACT

The optimal frequency response slope, from the low frequencies (250 or 500 Hz) to 2000 Hz, was estimated for each of 46 severely or profoundly hearing-impaired adults. The estimates were derived from paired comparison judgments of speech filtered to simulate different frequency response conditions, from home trials and ratings of different tone settings of high-powered, behind-the-ear hearing aids, and for 28 subjects, from speech recognition testing. The estimated optimal response, expressed as the slope from 250 to 2000 Hz and as the slope trom 500 to 2000 Hz, was compared with the response prescribed by the National Acoustic Laboratories (NAL) procedure and its relationship to audiometric variables was analyzed. Insertion gain was measured for the preferred volume setting with the best frequency response. Preferred gain was typically about 10 dB higher than the NAL prescribed gain. Considering these results in relation to other data, it appears that the “half-gain’’ rule ceases to apply when HTL exceeds about 70 dB. The estimated optimal frequency response agreed with the NAL response for some subjects but relatively more low frequencies were required for between a third and half of the subjects, depending upon how frequency response is expressed. Generally, more low frequencies were required if HTL at 2000 Hz exceeded 95 dB, whereas the NAL response was usually appropriate for other cases.

Hearing aid gain and frequency response selection procedures have been based, primarily, on research with the mildly and moderately hearing impaired (for review and discussion see Braida et al, 1979; Byrne, 1983; Skinner, 1988). There are reasons, however, to believe that the standard prescriptive procedures may not be optimally effective when applied to the severely and profoundly hearing impaired. First, when gain is prescribed according to the “half-gain rule” [i.e., gain to equal approximately ‘This study was supported by a National Health & Medical Research Council grant (No. 87/0492).

40

half of hearing threshold level (HTL)], which is the basis of many prescriptive formulae, it would appear to be insufficient to amplify normal speech to an adequate level for the severely hearing impaired. This may be more so for some frequencies than for others and, therefore, the appropriateness of the prescribed frequency response, as well as overall gain, could be questioned (Newall, Byrne, & Plant, 1986; Seewald, Ross, & Stelmachovicz, 1987, Fig. 4, p. 30). In theory, the problem of gain selection should not arise in procedures based on most comfortable loudness level (MCL) measurements although, because of signal output limitations, it might not be possible to measure MCL, at all the desired frequencies, for some of the severely hearing impaired. Implicit in most procedures for selecting the frequency response of a hearing aid, are the assumptions that useful aided hearing can be provided over a wide frequency range and that the wider the range, the better. These assumptions are very doubtful when there is no measurable hearing at some frequencies and may be invalid in some other cases where the hearing is much worse at some frequencies than at others. In some cases it might even be detrimental to provide audible signal at some frequencies and, in fact, this appears to be so for some of the moderately hearing impaired who have steeply sloping high-frequency hearing losses (Murray & Byrne, 1986). For the moderately hearing impaired, the rationale of amplifying all frequency bands of speech to the preferred listening level, or something similar, is well supported by research (Byrne, 1986; Lippmann, Braida, & Durlach, I98 1 ; Skinner, 1980) but it may not necessarily be effective for all severely hearingimpaired persons. The need to modify hearing aid selection procedures, to make them suitable for the severely hearing impaired, has been recognized by their proposers (Byrne & Dillon, 1986; McCandless & Lyregaard, 1983) and modifications have been suggested (Byrne, 1978; Cox, 1985; Schwartz, Lyregaard, & Lundh, 1988). In addition, some procedures have been devised specifically for the severely hearing impaired.

Amplification for the Severely Hearing-Impaired

These are based on the principle of amplifying the various frequency bands of speech, of a typical level, to specified target sensation levels (SL). Gengel, Pascoe, and Shore (1 97 1) proposed that the widest spectrum of sound (rms band levels of speech of 70 dB SPL) be amplified to at least, 10 to 20 dB SL. Huntington (1975) suggested amplifying the whole speech area, for speech of 65 dB SPL, to be above threshold. This is equivalent to amplifying the rms band levels, at all frequencies, to about 20 dB SL. These procedures are “constant SL” procedures in that the target SL is the same regardless of frequency or HTL. After examining the aided thresholds of 100 severely to profoundly hearing-impaired children, Macrae ( 1986) concluded that the National Acoustic Laboratories’ (NAL) audiologists, who had fitted these children, seemed to be trying to amplify speech bands to a constant SL. Recently, Matkin ( 1989) has recommended an SL procedure (expressed as optimal aided thresholds) in which the target SL is constant regardless of HTL but varies with frequency. A theoretical objection to constant SL procedures is that they do not account for the fact that, for sloping audiograms, a graph of MCL as a function of frequency will usually be less sloping than the threshold curve (Byrne & Murray, 1985). We might expect, therefore, that amplifying to a constant SL across the frequency range will tend to produce the loudest signal at frequencies where hearing is most impaired. In principle, this difficulty is avoided in the procedure of Seewald, Ross, and Spiro ( 1985) in which the desired SL decreases as HTL increases, until a minimum of 5 to 10 dB SL, depending on frequency, is reached. Another type of desired SL procedure is that of Ling (1 976). He recommended that gain equal to half the dynamic range, at each octave frequency from 250 to 4000 Hz, be added to the minimum gain required to make audible five speech sounds (three vowels and two consonants) which, altogether, span the frequency range. A focal point of debate in the literature on hearing aid selection for the severely hearing impaired, has been the extent to which the lower frequencies should be amplified, relative to the higher frequencies (Byrne, 1978). It can be argued that for severely hearing-impaired clients who have little high-frequency hearing, the hearing aid’s frequency response should have relatively more low-frequency amplification than would be suitable for the moderately hearing impaired. A contrary argument is that too much low-frequency amplification may prevent the use of the residual high-frequency hearing, however little that may be. This could occur because of upward spread of masking or because the additional low frequencies would increase the loudness of sound with the result that, to maintain comfortable listening, the hearing aid wearer would need to reduce the volume setting to the extent that the higher frequencies would be at an inadequate, and perhaps even inaudible, level. For severe hearing losses that are considerably greater at the high frequencies, some procedures (Schwartz et al, 1988) will actually prescribe more highfrequency emphasis than would be prescribed for a moderately hearing-impaired client with the same audiometric

41

configuration. Sensation level procedures also tend to prescribe more high-frequency emphasis than most other procedures such as NAL (Seewald et al, 1987, Fig. 4). In summary, a very wide range of frequency response prescriptions can be obtained depending upon which procedure is used. Although logical argument can be put in favor of one or another prescription, there is little evidence to support or refute the effectiveness of any particular procedure. The present study was aimed at determining the optimal frequency response characteristics for 46 severely or profoundly hearing-impaired adults. These data were used to ascertain whether any generally applicable prescription rules could be formulated or conversely, to indicate to what extent detailed evaluation is required for every individual. More specifically the main research questions were: 1. What is the optimal frequency response characteristic for each subject? 2. Can the optimal frequency response be predicted from the audiogram and, if so, what would be an appropriate prescription rule? 3. Does the gain used by the severely hearing impaired differ from that indicated by the half-gain rule and, if so, what would be an appropriate gain prescription rule? METHOD

Subjects The subjects were 46 severely to profoundly hearing-impaired adults (15 male, 31 female), aged 20 to 88 years (mean, 56.3 years). Their mean hearing threshold levels (HTL), standard deviations, and ranges of HTLs, are shown in Table 1. The three frequency average (500, 1000, 2000 Hz) HTLs ranged from 73 to 1 13 dB. In calculating the averages, nonmeasurable thresholds at 1000 or 2000 Hz have been counted as 125 dB (audiometer limit plus 5 dB). All thresholds were measurable at 500 Hz. All subjects had predominantly sensonneural hearing losses but a conductive component was present in 13 cases. Hearing Aids and Fitting Procedures Two high powered behind-the-ear hearing aid models were used (Phonak Superfront PPC2 and Phonak Superfront PPCLA). They were chosen because the tone controls, which have five discrete switch positions, provided a wide variation in the frequency response slope which, for the PPC2, was in approximately equal steps of about 4 dB/octave from 500 to 2000 Hz. The frequency response for each tone setting, for the two hearing aids, is shown in Figure 1. Twenty-three subjects were wearing PPC2 or PPCLA aids when the study was commenced. The remainder were loaned such aids which they agreed to use instead of their own aids for the duration of the study. All subjects had worn hearing aids regularly for a number of years. Three subjects with symmetrical hearing losses were fitted binaurally and the tone settings were adjusted to be the same at all times. For these subjects, data for only one ear, selected randomly, were used in all analyses. New earmolds were made, when necessary, and all subjects were able to wear their hearing aids on the preferred gain setting without incumng acoustic feedback, Ear and Hearing, Vol. 11, No. 1, 1990

42

Byrne et al.

Table 1. Mean HTL, standard deviations, and range of HTLs for the 46 subjects (n = 46 at all frequencies). HTLs exceeding the audiometer limit were assigned the value of audiometer limit plus 5 dB. No subjects exceeded limits at 0.25, 0.5, 0.75 kHz. At other frequencies, the audiometer limits and the number of subjects exceeding them were: 0.125 kHz = 85 dB (n = 12), 1 kHz = 120 dB (n = l), 1.5 kHz = 120 dB (n = 3),2 kHz = 120 dB (n = 5),3 kHz = 120 dB (n = 7). 4 kHz = 120 dB (n = 15).

Frequency (kHz) HTL

0.125

0.25

0.5

0.75

1

1.5

2

3

4

Mean SD Range

71.5 17.8 25-90

77.9 18.0 20-1 05

85.4 15.2 45-1 10

92.4 11.3 65-1 15

97.3 12.2 75-1 25

99.5 14.2 70-1 25

99.6 15.6 70-1 25

99.7 17.6 65-1 25

104.8 19.5 60-1 25

I

I

I

I

I

I

I

1

2

4

.25

.5

I

I

I

1

2

4

- I/’. 73

60

25

.5

FREQUENCY

(kHz 1

Figure 1. Gain (frequency response) curves, for all tone settings, of the two behind-the-ear hearing aids that were used in this study.

The hearing aid saturation sound pressure level (SSPL) was not selected by any specific formula. However, at the beginning of the study we enquired closely about possible problems related to the SSPL being used and, in general, proceeded from the principle that SSPL should permit the greatest amount of signal, averaged over a wide frequency range, that was possible without incurring any tolerance problems. Unless contraindicated, SSPL was left on the setting that the subject had been using with his or her own hearing aid or was adjusted to an equivalent setting if a different hearing aid was to be used. However, changes were made in a number of cases, usually ones where the SSPL setting was less than maximum and the SSPL curve had a peak around 1000 Hz. For these subjects the peak was reduced or removed by placing a suitable damping element in the friction hook, and the SSPL setting was increased. This change was reported to be beneficial by all these subjects. Throughout the study SSPL remained constant for each subject. Tone settings were systematically varied by the experimenter, in the manner to be described later, and at all times the subject was free to adjust the hearing aid volume control. Overview of Procedures The study was designed to determine the best tone control setting, and the corresponding real-ear frequency response, for each subject. The real-ear gain that was used, for the best tone setting, was also measured. Regression analyses were used to examine how the gain and frequency response requirements of our subjects were related-to their pure tone audiograms and to prescriptions derived from the NAL hearing aid selection procedure (Byme & Dillon, 1986). Three methods were used to estimate the best tone setting for each subject. An initial estimate was made by using paired comparison judgments of the relative intelligibility of speech samples filtered to approximate the effects of different tone Ear and Hearing, Vol. 11, No. 1, 1990

settings. For the remainder of the experiment, the tone setting initially estimated to be best, was compared with a tone setting providing one degree (about 4 dB/octave) less frequency response slope and with a tone setting providing one degree more slope. (For the PPCLA aid, the tone settings were sometimes vaned by two positions from the initial estimate, to provide the desired variation in frequency response.) The hearing aid was worn on each of the three tone settings over a series of home trials. After each trial, the subjects completed a questionnaire rating how well they could hear in various situations with the setting that had been on trial. The first 28 subjects were given a speech recognition test after each trial. A final estimate of the best tone setting was made by considering jointly the paired comparison judgments, the questionnaire ratings and, if obtained, the speech recognition results. The following sections give details of the assessment methods, the testing and trial arrangements, and the determination of the best tone setting. Basic Measurements Pure-tone air conduction thresholds were measured, using a Madsen OB 822 audiometer, at the octave frequencies from 125 to 8000 Hz and at 750, 1500, and 3000 Hz. Real-ear insertion gain was measured, using a Rastronics CCI 10/3 system and a swept warble tone, for the hearing aid tone setting determined to be best, at the preferred volume setting. Gain was measured in a 2 cc coupler for all of the tone settings on trial at the preferred volume for each setting. Speech Intelligibility Judgments The procedure used to obtain an initial estimate of the best hearing aid tone setting is described in the following sections. Briefly, samples of differently filtered speech were presented in pairs and the subject judged which member of the pair was more intelligible. The effect of changing from one hearing aid tone setting to another was simulated by switching back and forth from one to the other of the speech samples that were being compared. The speech was presented through a loudspeaker to the subjects who were wearing their hearing aids. The use of the filtered speech recordings, instead of changing the settings on the hearing aids, was to permit comparisons to be conducted without any delay in switching from one condition to the other. This is not only convenient but, in fact, is essential because it has been found that the delay incurred in changing hearing aid settings greatly reduces the reliability of the paired comparison procedure (Schum & Collins, 1985). The hearing aid was used as the final stage of the sound delivery system to ensure that the effects of the earmold and “plumbing” were taken into account. A further advantage was that the use of different tone settings, for different subjects or stages of the procedure, minimized the number of filtered speech recordings

Amplification for the Severely Hearing-Impaired

that were required. The procedure is similar to one used in previous research (Byrne & Cotton, 1988) and to one recommended for clinical use (Appendix of Byrne & Cotton, 1988). Materials The base speech material was a descriptive passage read by a male talker whose speech approximated the average, long-term spectrum. This speech sample was filtered, by making small adjustments in some one-third octave bands, to match precisely the average spectrum reported in Byrne and Dillon ( 1 986). This sample (a 30 sec passage recorded repeatedly) is designated “unfiltered” speech. This was then filtered in the following ways: ( 1 ) 6 dB/octave cut from 2000 to 250 Hz, (2) 6 dB/octave boost from 2000 to 250 Hz, (3) 12 dB/octave boost from 2000 to 250 Hz. The four speech samples (unfiltered and the three filtered) were recorded (without any competing noise) side by side on four tracks of a 4-channel tape recorder. The filtering was designed to approximate the effects of differences between hearing aid tone settings. However, to improve the prospects of getting a clear preference for a particular sample, the differences between samples were made slightly greater than the difference between adjacent tone settings. Equipment The test equipment consisted of a 4-channel tape recorder (Ampex AG-440C), a subject switch box, an experimenter control box, a custom-made amplifier including an attenuator with 1 dB steps, and a loudspeaker (Foster). Procedure The subjects were seated in a sound-treated test room facing the loudspeaker and 1 m from it. Their hearing aids were set on the preferred volume setting. Each of the four speech samples was presented and was adjusted to the most comfortable level (MCL). MCL was defined as the level that would be preferred for listening over an extended period (half-hour or more). Throughout the testing all speech samples were presented at MCL. For each test trial, the subject was instructed to listen to two samples of speech and to judge which sample was more intelligible, defined as “easier to understand.” He or she was asked to listen to sample A for a few seconds, then switch to sample B for a few seconds, then switch back and forth as desired until ready to make a judgment. Eighteen subjects had such poor speech discrimination that they could not judge intelligibility. Instead, they were asked which sample corresponded to the quality they would prefer when using a hearing aid. The initial estimate of the best tone setting was derived as follows. First, the hearing aid was set on the tone setting corresponding most closely to what would be prescribed by the NAL procedure. Then all six possible pairs of speech samples were judged 10 times. The trials were randomised and each member of a pair was presented as sample A half the time. Each condition (i.e., speech sample) was scored as the total number of times (out of 30) that it was judged to be more intelligible. If the unfiltered condition received the highest score, then the initial estimate of the best tone setting was the one that had been used for the test. If the 6 dB/octave boost condition received the highest score, then the best setting was estimated to be one position removed from the setting being worn, in the direction providing more low frequencies. If either the 12 dB/octave boost or the 6 dB/octave cut (i.e., the conditions providing the most or least low frequencies, respectively) received the highest score, then the procedure was repeated after adjusting the tone control by two steps in the direction indicated (i.e., toward more low frequencies if the 12 dB/octave boost had been best, toward less low frequencies if the 6 dB/octave cut had been best). In this way, the best tone setting could be identified because it had been shown to be superior to at least one condition having more low frequencies and at least one condition having less low frequencies.

43

Home Trials and Ratings Most (22) of the first 28 subjects had three home trials, of about 2 weeks, with each of three tone settings. These were the ones estimated to be best, one position in the direction of less low frequencies, and one position in the direction of more low frequencies. The other 6, of the first 28 subjects, had only two or just one, trial with some settings. After each trial, a questionnaire was administered to obtain ratings, on a five point scale, of how well the subject could hear, with the aid setting on trial, in each of eight listening situations. The word “hear” was interpreted liberally to mean auditory visual functioning in situations where lipreading was used. Ratings were also obtained of the quality of speech and music, hearing of environmental sounds, and difficulty (if any) tolerating background noise. In addition, there was a rating of the overall effectiveness of the setting. The questionnaires were scored by assigning points to the responses to each question where “1” indicated the best response ( e g , “very good”) and “5” indicated the poorest response (e.g., “very poor”). Points were totalled to obtain an overall rating. The last 18 subjects were given only four trials consisting of a practice trial and one trial with each tone setting. This new protocol was adopted after reviewing the results for the first 28 subjects. After concluding all the trials, the subjects were asked which tone setting they had found best and would like to continue using. Speech Recognition Testing The first 28 subjects were given a speech recognition test after each home trial. Materials The materials were a vowel test and a consonant test from the COMMTRAM manual (Plant, 1984). Both are presented as 6-alternative, forced choice items. The vowel test (55 items) consists of five presentations of each of the eleven vowels which occur in stressed monosyllables in Australian English. In the consonant test (1 10 items), the 22 consonants that appear initially in English are presented five times with most of the other consonants acting as response foils at least once. The tests were recorded by a male talker and each vowel was amplitude normalized with respect to Leq. The recording also contained a continuous passage of speech at the same level. Equipment The test equipment consisted of a tape recorder (Ampex AG-440C), a custom-made amplifier and 1 dB step attenuator, a loudspeaker (Foster), a desk top computer (HPSS), a custom-made psychoacoustic test unit (used for switching functions), a video monitor screen, and a set of six subject response buttons. Procedure The subject was seated 1 m in front of the loudspeaker and wore the aid (or aids if fitted binaurally) on the tone setting being tested and o n the volume usually used. The continuous speech was presented through the loudspeaker and was adjusted to MCL. The test words were then presented and the subject responded by pressing one of six response buttons positioned beside the response foils which were displayed on the monitor screen. The administration was controlled by the computer which changed the response foils between each item, collected the response, and paused the tape recording if the subject had not responded within 6 seconds. The vowel test was administered first, followed by the consonant test. After testing each tone setting once, the scores for the vowel test and for the consonant test were examined for “floor” or “ceiling” effects. If the scores for either test exceeded 96% or were below 24% for all tone settings, then the test concerned was not repeated. Ear and Hearing, Vol. 11, No. 1, 1990

44

Byrne et at.

Scoring Scoring was in terms of percent correct and according to various types of distinctions. Only the percent c’orrectwas used for estimating the best tone setting. A detailed analysis of the subjects’ speech test performance will be presented in a separate article. Final Estimate of Best Tone Setting The final estimate of the best tone setting was derived by considering the intelligibility judgments, the speech recognition scores, the questionnaire ratings and, in a few marginal cases, the subjects preferences. The first step in reaching this estimate was to calculate which pairs of tone settings were significantly different according to each of the three tests. For the intelligibility judgments, two settings (actually the filtered speech conditions most closely resembling them) were considered significantly different if one had been chosen 8 or more ofthe 10 times that they had been compared. The probability of this occurring by chance is 5.47% as calculated from the binomial distribution. For the speech recognition tests, the mean and standard error of the repeated tests, for each setting, were calculated. The significance ( 5 % level) of the difference between settings was then calculated using the means and standard errors in the manner described by Byrne (1986) which is essentially a t-test. The same procedure was applied to the questionnaire ratings of the first 22 subjects. Typically, the difference between the overall scores, for two different tone settings, needed to be 4 or more points to be significant at the 5% level. This criterion was subsequently adopted for the last 24 subjects to decide whether differences between questionnaire ratings were “significant.”

68

3 F A . HTL (dB) 90 112 133

Figure 2. Three frequency (500, 1000, and 2000 Hz) average insertion gain for preferred volume control and best tone settings, as a function of three frequency average (3FA) gain prescribed by NAL procedure and as a function of 3FA hearing threshold level (HTL).

60

RESULTS

Gain Figure 2 shows the three frequency average (3FA) realear gain for the volume setting normally used, for the estimated best tone setting. This is compared with the 3FA gain prescribed by the NAL procedure. The 3FA HTL is also shown on the horizontal axis as this vanes in proportion to 3FA NAL gain. It can be seen that the used gain was never significantly less than the NAL gain and was typically about 10 dB more. If we compare the gain values with HTL, we find that used gain was somewhat more than half of HTL. For example, for 90 dB HTL the used gain averaged about 50 dB but the NAL gain was only 40 dB. Figure 3 shows the gain values averaged over 250, 500, and 1000 Hz (low FA gain). Again, the used gain was more than the NAL gain and, in fact, the difference tended to be greater than for the 3FA gain. A further analysis was conducted to assist in interpreting the significance of using different tone settings. This was necessary because in everyday usage the relative amounts of signal provided in different frequency regions, will depend not only on the tone setting but also on the volume setting that is used. We, therefore, compared the LFA gain for the three tone settings that were on trial for each subject, after the hearing aids’ volume control had been adjusted to the level usually used for each tone setting. Specifically, we calculated the difference in LFA gain Ear and Hearing, Vol. 11, No. 1, 1990

a;

20 20

30

40

50

60

LOW F.R. (-25, . 5 , 1 kHZ) NRL GRIN (dB) Figure 3. Low frequency (250, 500, and 1000 Hz) average insertion gain for preferred volume control and best tone settings, as a function of low FA gain prescribed by NAL procedure.

between the settings providing the most and the least LFA gain, for each subject. For one subject, the difference was less than 3 dB, which means that he adjusted the hearing aid volume control to receive virtually the same amount of low-frequency signal for all tone settings. The other subjects, however, did receive different amounts of lowfrequency gain, mostly ranging from 3 to 18 dB, with different tone settings. For 22 subjects, the tone settings differed only in low-frequency gain in that the 2000 Hz gain was the same for all settings. The other subjects adjusted their volume controls so that some tone settings provided more low-frequency gain at the expense of less high-frequency gain and vice versa.

Amplification for the Severely Hearing-Impaired

45

Frequency Response 0 In the following analyses, frequency response has been expressed in terms of the average slope, in dB/octave, N between 250 and 2000 Hz, and in terms of the average I (0 2i slope between 500 and 2000 Hz. The relationship of the cv estimated best slope, for each subject, to the slope prescribed by the NAL procedure is shown in Figures 4 and 1 y . J 5. Consider first the data for slopes from 250 to 2000 Hz mi, (Fig. 4). For about three quarters of the subjects (35), the 6 O , NI best slope agrees with the NAL slope within +6 dB/octave. This tolerance seems a reasonable allowance to make for inaccuracies in estimating the best response (see “Discussion”). It corresponds to slightly more than one tone setting. For the remaining subjects (except one), the best I0 m slope was less steep than the NAL slope; that is, relatively W I SINGLE POINT 0 m more low frequencies were required. If we consider the DOUBLE POINT 0 slopes from 500 to 2000 Hz (Fig. 5 ) , the proportion of 0 subjects requiring relatively more low frequencies than the NAL prescription, is 5 5 % . In view of the discrepant esti-2 2 6 10 14 18 mates of the proportion of subjects requiring more low frequencies, it is relevant to consider how well each of the NAL SLOPE 0.5 - 2 kHz slope measures (250-2000 Hz, and 500-2000 Hz) repre(dB 1 OCT) sents the relative amounts of low-frequency versus highfrequency gain. If we refer back to Figure 1 we see that Figure 5. Estimated best frequency response slope as a function of the the response curves of the hearing aids tend to roll off slope prescribed by the NAL procedure, when slopes are expressed as from a point somewhat above 250 Hz. Consequently, the the average in dB/octave over the range from 500 to 2000 Hz (see also difference in gain between 250 and 2000 Hz indicates a Fig. 4 legend). slope which is usually steeper than would be representative of the response curve. On the other hand, the slope from 500 Hz to 2000 Hz tends to be insufficiently steep to be vicinity of 500 Hz for some tone settings. We suggest, representative of the frequency response curve because therefore, that the proportion of subjects requiring more there tends to be a slight “hump” in the curve in the low frequencies than the NAL response, is underestimated by the data of Figure 4 but overestimated by the data of Figure 5. As a first step toward ascertaining whether there 16I i n were any differences between those subjects who were suited by the NAL response and those who required more N r low frequencies, the average audiograms of the two groups 12 ru were calculated. These are shown in Figure 6. I On the average, the subjects of the “Low” group have more sloping audiograms and poorer high-frequency thresholds than the “NAL” subjects. Despite their more sloping audiograms, the Low subjects required less sloping frequency responses, namely 1.3 dB/octave, on the average, compared with 6.7 dB/octave for the NAL subjects, for the 250 to 2000 Hz slopes, or -3.1 dB/octave versus 2.6 dB/octave, for the 500 to 2000 Hz slopes. This figure gives only a general indication of group trends. Within -4 each group, there were individuals who differed from the I I L * ’ l I I I I I I general pattern that is shown. -4 0 4 8 12 16 Multiple linear regression analyses were performed usNRL SLOPE 0.25 - 2 kHz ing, as the dependent variables, the estimated best slope from 250 to 2000 Hz, and from 500 to 2000 Hz. The (dB / OCT) independent variables were HTL at 250, 500, 1000, 2000 Figure 4. Estimated best frequency response slope as a function of the Hz, and the differences 2000 to 500 Hz, and 2000 to 250 slope prescribed by the NAL procedure, when slopes are expressed as Hz. The correlation of the dependent variables with each the average in dB/octave over the range from 250 to 2000 Hz. Best of the independent variables is shown in Table 2. slope is calculated from insertion gain measurements for the tone setting estimated to be best. NAL slope is calculated from prescribed real ear It can be seen that there is a moderate correlation gain. between the best slope and HTL at 1000 Hz or HTL at .-I

-

4t

--

t

Ear and Hearing, Vol. 11, No. 1, 1990

46

Byrne et al.

Table 2. Correlations between the best frequency response slope from 250 to 2000 Hz, or from 500 to 2000 Hz, and audiometric variables. Best Slope 500-2000 HZ

250-2000 HZ

Variable

-0.44’ -0.43 -0.24

HTL 1000 Hz HTL 2000 Hz HTL 500 Hz HTL 250 Hz HTL 2000-250 HZ HTL 2000-500 HZ

-0.51 -0.44 -0.32 -0.23 -0.13 -0.1 1

-0.17 -0.17 -0.18

* Correlations that were significant, at the 5% level, after partialling out the variance accounted for by variables having higher correlations.

15

i

I

n

I

I

I-

2000 Hz. (The reason why the correlations, at the two frequencies, are of similar magnitude and why either becomes nonsignificant after partialling out the variance explained by the other, is that HTL at the two frequencies is highly correlated; Y = 0.68.) Furthermore, the correlation is negative; the subjects with the largest hearing losses tended to require the least steep response slopes. The difference between the NAL slope and the best slope was calculated and Figure 7 shows how this is related to HTL at 2000 Hz. (This difference was more predictable from the 2000 Hz HTL than from the 1000 Hz HTL.) There appears to be a “break-point’’ about 95 dB HTL in that for most higher HTL values the best slope is less than the NAL slope, whereas for HTLs below 95 dB the differences are about equal in both directions. DISCUSSION

This study has confirmed the two general predictions that could be made on logical grounds, namely, that most of the severely and profoundly hearing impaired require more gain than would be prescribed by the half-gain rule and that some of these clients require relatively more low frequencies than would be appropriate for less impaired clients with the same audiogram configuration. Our data regarding gain raise two questions: (1) above what hearing level does the half-gain rule, which has been verified for the moderately hearing impaired by many studies (for review see Byrne, 1983; Byrne &Cotton, 1988; Lyregaard, 1986) become inapplicable, and (2) what is the appropriate rule for the severely hearing impaired? To examine these questions we have compared the present data with those of a previous study as shown in Figure 8. Ear and Hearing, Vol. 11, No. 1, 1990

2

m I

w

15 11

I

I

0 I

I

I

-v

I

7

I 0

s 3 [L

v,

_1

U

z

-1

-5 i

0

-9 HEARING L E V E L A T 2kHz (dB) Figure 7. Difference between NAL slope and best slope as a function of HTL at 2000 Hz. Top panel shows data for response slopes described by average slope from 250 to 2000 Hz. Lower panel shows data for response slopes described by average slope from 500 to 2000 Hz. Lines indicate: (a) agreement between NAL and best slopes (horizontal line), (b) regression (sloping line), and (c) division between subjects with HTLs greater than versus less than 95 dB at 2000 Hz (vertical broken line).

The data points show the average NAL gain for subjects grouped according to hearing levels. The diagonal line shows the gain prescribed by the NAL procedure, that is, increasing at 46% the rate that HTL increases. The figure suggests that the need for more gain than that prescribed by the NAL procedure, occurs for HTLs from about 70 dB upward. This agrees closely with data represented by Pascoe (1988) and Lyregaard (1988). Both those data sets show that MCL increases at half the rate that HTL increases over the range from 0 to 60 dB HTL but that for greater hearing levels, MCL increases at a rate exceeding half of HTL. It appears that, for HTLs exceeding 60 dB, required gain increases at about two-thirds of the rate that

Amplification for the Severely Hearing-Impaired

24

46

68

90

47

112

much divergence from the NAL response is accepted before a difference is considered to be significant. Previously we have suggested that use of the 250 to 2000 Hz m slope probably underestimates the proportion requiring TJ 50 more low frequencies, whereas that proportion is probably overestimated by using the 500 to 2000 Hz slope. We feel z that a realistic conclusion would be that between a third 40 and half of our subjects required relatively more low a frequencies than the NAL response, &6 dB/octave. In c3 discussing Figures 4 and 5 we assumed that a tolerance of &6 dB/octave would be a reasonable allowance for inacI 3 30 W curacies in estimating the best response. A possible indiv, cation of whether this tolerance is appropriate might be 3 obtained by considering the proportions of clients found 20 in a previous study to require less low frequencies or more a low frequencies than the NAL response. In that study (Byrne & Cotton, 1988) about 8% of mildly to moderately LL 10 hearing-impaired listeners would require less low frequencies than the NAL response, if both speech intelligibility m and pleasantness are considered. If a similar percentage were assumed to apply to the severely hearing impaired 10 20 30 40 SO this would suggest a tolerance of +-5 dB/octave would apply to Figure 4. However, in the present study there was 3 F.R. ( . 5 i l i 2 kHz1 a substantial proportion of subjects requiring more low frequencies, whereas there was only one (1.5%) in the NRL GRIN ( d B ) Byrne and Cotton study. We might expect, therefore, that, Figure 8. Three frequency average used gain versus 3FA NAL prescribed in the present study, there would be a smaller proportion gain or 3FA HTL, from two studies. (Asterisks refer to data from Byrne of subjects requiring less low frequencies. Therefore, a & Cotton, 1988), solid circles refer to present study. Each point shows tolerance of more than 5 dB/octave, say 6 dB/octave, average used versus average prescribed gain for subjects grouped seems appropriate. according to prescribed gain of 0-10 dB, 10.1-20 dB, 20.1-30 dB, 30.1The prediction of which clients would require relatively 40 dB, 40.1-50 dB, 50.1-60 dB. Numbers near each point indicate more low frequencies, seems possible to a moderate extent. number of subjects in each group. As shown in Table 2, the best response showed correlations of -0.43 to -0.51 with the audiometric variables 1000 HTL increases if we allow for the fact that the required and 2000 Hz HTL. Inspection of Figure 7 suggests that a simple rule for gain/HTL, function should be slightly less steep than the MCL/HTL function because typical hearing aid input varying the NAL procedure would be to decrease the slope levels tend to increase with increasing hearing loss (Byrne (i.e., provide relatively more low frequencies) by 4 dB/ & Cotton, 1987; Parkinson, Byrne, & Newall, 1989). The octave if the 2000 Hz HTL is between 95 and 1 10 dB and present data (Fig. 8) are consistent with such a rule, to decrease the slope by 8 dB/octave if the 2000 Hz HTL although individual variability in preferred gain would exceeds 110 dB. This suggestion is based on the observapreclude the formulation of any precise rule from a sample tion that for HTLs of 95 dB or less, the data points are of only 46 subjects. [Also, the data point of Fig. 8 for the about equally distributed above and below the horizontal most severe group may underestimate preferred gain as 0 dB/octave line (which indicates agreement between the four of these subjects were using the maximum volume NAL and best responses) whereas most points are above control setting (Parkinson et al, 1989)l. It might be ex- the line for HTLs of 95 to 110 dB. A similar argument pected that subjects with mixed hearing losses would prefer would apply if the data for HTLs above 110 dB were more gain than subjects with purely sensorineural hearing considered in relation to a line drawn at 4 dB/octave. To help explain the present data, especially with regard losses. We could not find any evidence of such a trend in our data (Figs. 2 and 3 were originally constructed to show to the best frequency responses, we may briefly refer to the “mixed” and “sensorineural” subjects separately) but three types of analysis which are being reported separately. this is not conclusive as our mixed subjects were predom- First, a detailed analysis of the speech recognition performance of the first 28 subjects indicated that the perception inantly sensorineural. . Our second major finding is that a significant proportion of the vowels and consonants was reasonably predictable of subjects required a frequency response with relatively from the 2000 Hz HTL. The correlations were 0.83 and more low frequencies than the NAL response. Any esti- 0.7 6, respectively. This predictability was best for HTLs below 100 dB mate of what the proportion is depends on what metric is used to represent the frequency response and on how (perception was uniformly good) and for HTLs above 1 15

v

Ear and Hearing, Vol. 11, No. 1, 1990

48

Byrne et al.

dB (perception was uniformly poor). This conclusion ap- inappropriate for some of the severely and profoundly plies to the recognition of most speech features, namely, hearing impaired. consonant place and manner (but not voicing) and vowel Earlier we suggested some simple rules for varying the height, place, and length, as well as to the percentage NAL procedure, with respect to frequency response, when correct for vowels and consonants. In the present context applying it to the severely and profoundly hearing imit is interesting to note that the point at which aided speech paired. We could add that the overall gain (i.e., gain at all perception deteriorates (2000 Hz HTLs exceeding 95 dB) frequencies) should be increased by about 10 dB for all coincides approximately with the point at which the best 3FA HTLs exceeding 70 dB. Clearly, however, the present frequency response has more low frequencies than the data could be used in more sophisticated ways to calculate NAL response. Of the subjects who were suited by the a revised formula. This has not been done at present as NAL response, about 75% scored above the group mean we wish to also consider the results of some further invesscores for vowels and consonants, whereas only 25% of tigations. In particular, we are conducting a similar study the subjects who required more low frequencies, scored to the present one but using severely hearing-impaired above the group mean score. It may be that the desirability children as subjects. It is possible, although unproven, that of providing more low frequencies is related to whether the amplification requirements of such children may differ the subject can recognize speech satisfactorily by audition from those of adults. alone. This is in line with the suggestion of Lindblad, CONCLUSIONS Haggard, and Foster (1983) that visual information can 1. Most of the severely and profoundly hearing imsubstitute for the mid- and high-frequency auditory information used in speech perception and, therefore, the need paired require more gain than would be prescribed by the for auditory information should move toward the lower NAL procedure or any other procedure using a half-gain rule. frequencies when lipreading is being used. 2. Clients with little high-frequency hearing (typically In another analysis (Parkinson et al, 1989) we calculated the sensation levels (SL) of one-third octave bands of HTLs exceeding 95 dB at 2000 Hz), usually require a speech at various frequencies for the best frequency re- frequency response with relatively more low frequencies sponse and the preferred volume setting. The calculations than would be prescribed by the NAL procedure. 3. Gain and frequency response requirements are sufincluded adjustments for the difference between coupler and insertion gain (measured for each subject), the speech ficiently predictable that it is feasible to use an HTL-based presentation level preferred for the volume setting used, prescription formula for an initial selection. It would, and any limitation on SL imposed by the hearing aid’s however, be highly desirable to supplement this approach gain or (more frequently) SSPL. This analysis is relevant with the use of evaluation procedures whenever possible. when examining possible explanations for the present References finding that subjects with 2000 Hz HTLs exceeding 95 Braida LD, Durlach HI, Lippmann RP, Hicks BL. Rabinowitr WM. and Reed CM. Hearing aids-a review of past research on linear amplification, amplitude dB, required frequency responses with relatively large compression and frequency lowering. ASHA Monographs, No. 19. Rockville. degrees of low-frequency emphasis. This did not occur (or MD: American Speech-Language Hearing Association, I 979. at least not usually) because such subjects were not receiv- Byrne D. Effects of band width and stimulus type on most comfortable loudness levels of hearing-impaired listeners. J Acoust Soc Am 1986:80:484-493. ing any audible signal around 2000 Hz. Nearly all subjects D. Theoretical prescriptive approaches to selecting the gain and frequency with 2000 Hz HTLs from 100 to I10 dB received audible Byrne response of a hearing aid. Monographs in Conrempary Audiology, 4/l. Upper signal at 2000 Hz when using the best tone setting and Darby PA: Instrumentation Associates, 1983. preferred volume setting. Furthermore, most such subjects Byrne D. Selectionofhearingaidsforseverelydeafchildren. BrJ Audiol 1978:12:922. could have, if they had so desired, adjusted the hearing Byrne D and Cotton S . Preferred listening levels ofsensonneurally hearing-impaired aid’s volume control (by turning it up) and tone control listeners. J Audiol I987;9:7- 14. (by selecting a position with more low cut) so that they Byme D and Cotton S. Evaluation of the National Acoustic Laboratories new hearing aid selection procedure. J Speech Hear Res 1988;3 1: 178- 186. would have received higher signal levels around 2000 Hz Byrne D and Dillon H. The National Acoustic Laboratories (NAL) new procedure for selecting the gain and frequency response of a hearing aid. Ear Hear with little change in signal levels at the lower frequencies. 1986;7:257-265. It is, however, true that most of the more severely hearing- Byrne D and Murray H. Relationships of HTLs, MCLs, LDLs and psychoacoustlc impaired subjects received little, if any, audible signal at tuning curves to the optimal frequency response characteristics of hearing aids. Aust J Audiol 1985;7:7-16. 3000 or 4000 Hz because of the limitations of the hearing Cox PM. A structured approach to hearing aid selection. Ear Hear 1985;6:226aids. 239. Another analysis, that was undertaken to clarify inter- Gengel RW, Pascoe D, and Shore I. A frequency-response procedure for evaluating and selecting hearing aids for severely hearing-impaired children. J Speech Hear pretation of our results, involved comparing the NAL Disord 1971 ;36:34 1-353. frequency responses with ones calculated from speech- Huntington A. Selecting hearing-aids for young children. Br J Audiol l975;9:7580. band MCLs. This showed good agreement between the A-C, Haggard MP, and Foster JR. Audio versus audio-visual testsNAL and MCL responses and that both failed to prescribe Lindblad implications for hearing aid frequency responses. Report TA No. 108, Stockholm: sufficient low frequencies for a proportion of subjects. Karolinska Institute, 1983. This indicates that the NAL procedure did fulfill its aim Ling D. Speech and the hearing-impaired child. Washington, DC: AG Bell, 1976. RP, Braida LD, and Durlach HI. Study of multichannel amplitude of amplifying all speech bands to MCL but that this aim, Lippmann compression and linear amplification for persons with sensorineural hearing loss. which is similar to that of many other procedures, is J Acoust SOCAm 1981:69:524-534. Ear and Hearing, Vol. 11, No. 1, 1990

Amplification for the Severely Hearing-Impaired Lyregaard PE. On the practical validity of POGO. Hear Instrum 1986;37(5):1316.

Lyregaard FE. POGO and the theory behind. In Jenson JH, Ed. Hearing Aid Fitting: Theoretical and Practical Views. Copenhagen: Stougaard Jensen, 1988. Macrae J. Relationships between hearing threshold levels, aided threshold levels and aided speech discrimination of severely and profoundly deaf children. NAL Rpt. No. 107. Canberra; Aust. Govt. Publishing Service, 1986. McCandlessGA and Lyregaard PE. Prescription of gain/output (POGO) for hearing aids. Hear lnstrum 1983;34(1):16-21. Matkin HD. Hearing instruments for children: premises for selecting and fitting. Hear lnstrum 1989:38(9):14-16, Murray H and Byrne D. Performance of hearing-impaired and normal hearing listeners with various high frequency cut-off in hearing aids. Aust J Audiol 1986:8:21-28. Newall P. Byrne D, and Plant G. Amplification for the severely and profoundly hearing-impaired:a pilot study. Aust J Audiol 1986;8:42-55. Parkinson A. Byrne D, and Newall P. (in press) Aided speech sensation levels received by severely hearing-impaired listeners. Aust J Audiol 1989. Pascoe DP. Clinical measurements of the auditory dynamic range and their relation to formulas for hearing aid gain. In Jensen JH, Ed. Hearing Aid Figging: Theoretical and Practical Views. Copenhagen: Stougaard Jensen, 1988. Plant G. COMMTRAM communication training program for profoundly deaf adults. Sydney: National Acoustic Laboratories, 1984.

49

Schum DJ and Collins MJ. Test-retest reliability of two paired-comparison hearing aid evaluations. ASHLA Annual Convention, Washington, DC, October 1985. Schwartz D, Lyregaard PE, and Lundh P. Hearing aid selection for severe-toprofound hearing loss. Hear J 1988:41(2):13-17. Seewald RC, Ross M, and Spiro MK. Selecting amplification characteristics for young hearing-impaired children. Ear Hear 1985:6:48-53. Seewald RC, Ross M, and Stelmachovicz PG. Selecting and verifying hearing aid performance characteristics for young children. J Acad Rehabil Audiol 1987;20:25-37. Skinner MW. Speech intelligibility in noise-induced hearing loss: effect of highfrequency compensation. J Acoust SOCAm 1980:67:306-3 17. Skinner MW. Hearing Aid Evaluation. Englewood Cliffs, NJ: Prentice Hall, 1988.

Acknowledgments: We wish to thank the following for suggestions and assistance at various stages of this project: Richard Tyler (University of Iowa); William Noble (University of New England, NSW); Gary Walker, Greg Upfold, and Geoff Plant (NAL). Address reprint requests to Denis Byrne, National Acoustic Laboratories, 126 Greville St., Chatswood, NSW 2067, Australia. Received June 26, 1989; accepted September 10, 1989.

Ear and Hearing, Vol. 11, No. 1, 1990