Good eating habits, regular exercise, and conducting speech testing when we fit hearing aids. Those are three good ideas that we all talk a lot about, but the day-to-day application never quite equals the rhetoric. Dieting and exercise are topics best reserved for the beginning of the new year, so, this month, let's talk about speech audiometry.
Speech audiometry can be applied in a variety of ways in the typical hearing aid fitting protocol. It can be used for loudness scaling, intelligibility judgments, quality assessment, and even to evaluate the occlusion effect. Traditionally, however, when we think of speech audiometry, we think of objective measures of speech understanding—testing that results in a signal-to-noise ratio or percent correct score. It's these objective measures of speech understanding that will be the focus of this article.
A patient walks in your door with problems understanding speech. Doesn't it seem that, before he leaves with his new hearing aids, he should know if his understanding has improved? Don't you want to know? Objective measures of speech understanding can answer that question.
From the very beginning of audiology, the use of speech audiometry to differentiate one hearing aid from another has been the basis of most speech testing in the hearing aid evaluation. For example, in 1946, long before the current digital versus analog issue arose, Wiener and Miller published a fitting protocol using speech testing that was designed to compare carbon versus vacuum tube hearing aids. (This article was published in the monograph Combat Instrumentation, so you know they were serious about their procedure!)
The publications of Raymond Carhart and others in the late 1940s helped promote speech testing in the evaluation and differentiation of hearing aids, and speech audiometry quickly became the preferred method to evaluate real-ear hearing aid performance—a position it would hold for over 30 years.
Yet, even in these early years, not everyone believed that speech audiometry was a worthwhile tool for comparative hearing aid evaluation. For example, in 1960, researchers Shore, Bilger, and Hirsh, who conducted comparative testing of hearing aids using SRTs and monosyllabic word recognition in quiet and in noise, concluded:
“The reliability of these measures is not good enough to warrant the investment of a large amount of clinical time with them in selecting hearing aids.”
Dave Resnick and Marshall Becker reached the same conclusions 3 years later, as they announced in ASHA magazine the discontinuation of comparative speech testing at Washington Hospital Center. In general, these negative reports were not well received by the audiologic community and, in fact, were considered by some to be Carhart bashing. Speech audiometry, therefore, continued to be the cornerstone of hearing aid selection, as surveys conducted in the 1960s and 1970s revealed that about 85% of audiologists used speech testing for hearing aid differentiation. But then things changed rapidly.
The decline of aided speech testing
In the early 1980s, clinic after clinic began to drop the use of speech testing when fitting hearing aids. This was due to a combination of four separate factors, all occurring about this time:
* Articles, monographs, and book chapters by opinion leaders were accumulating that showed that speech testing was not a reliable method to differentiate between hearing aids. That is, inter-aid differences tended to be no greater than test-retest differences of the same hearing aid. Many of these reports were influenced by the 1970s work of Aaron Thornton and Mike Raffin, who modeled speech-discrimination scores as a binomial variable.
* Prescriptive fitting methods such as the Berger approach and the original NAL were also gaining popularity. These fitting approaches were verified with functional gain, not speech audiometry.
* The custom ITE hearing aid was now a widely available option. Unlike with BTEs, when the patient was fitted with ITEs, there were not several different products available for clinical comparative testing.
* Audiologists were moving into private practice and hearing aid dispensing. Office managers quickly pointed out that the 2-hour hearing aid evaluation (common at the time at university clinics, often referred to as the “hang three” approach) was not a cost-effective method of fitting hearing aids.
As we moved into the mid-1980s, probe-microphone measurements became a popular method for verifying hearing aid performance. This added yet another reason not to do speech testing, as hearing aids were now being fitted outside the test booth in rooms that did not have the equipment to deliver sound field speech stimuli. By the 1990s, speech testing was rarely used for hearing aid differentiation, and many dispensers did not conduct any aided speech testing.
SPEECH AUDIOMETRY TODAY
So much for history. What's happening today? To find out, our annual Hearing Journal survey of dispensers included a section asking readers several questions related to aided speech audiometry.
Hearing Journal survey
Overall, 52% of dispensers who responded to the survey reported that they “usually” or “always” do some type of aided speech testing. This testing is most commonly conducted binaurally, although 38% of those who do speech testing said that they also test each ear separately. Regarding the frequency of aided speech testing, there was no significant difference between hearing instrument specialists and audiologists or among audiologists in different work settings.
There also was no difference in the frequency of speech testing for dispensers who price their products differently, or for dispensers fitting mostly high-end products versus those who typically fit conventional products. In other words, when we looked at aided speech testing, about half of dispensers said they do it, the other half do not. And we found no specific characteristics that differentiate these two groups.
Not surprisingly, monosyllabic word recognition in quiet was the most common speech task that was routinely (“usually” or “often”) conducted (92%), followed by monosyllabic word recognition in noise (35%). Only 6% of the dispensers who conduct aided speech testing reported that they routinely use sentence tests (e.g., SIN, HINT, SPIN). By an overwhelming margin, the survey showed that the leading reason that dispensers conduct aided speech testing is to demonstrate aided benefit, which may explain why monosyllables in quiet were the speech material of choice.
Our love affair with monosyllables
Our survey showed that monosyllables are nearly always the choice when dispensers select speech material for fitting hearing aids. It is somewhat surprising that this fascination with monosyllables continues. Over the past 10 to 15 years, numerous new speech tests have been developed, and nearly all of them use sentence-length material. We'll review some of these tests shortly.
Why would we expect that the recognition of monosyllables, without any competing noise, in a non-reverberant audiometric test booth, presented at a level that often fails to represent average speech in the real world, would have anything to do with long-term benefit and satisfaction with hearing aids? Unless the scores are at or near 0%, it usually doesn't! James Jerger probably was questioning the same thing in 1980 when he penned this oft-cited statement:
“We are, at the moment, becalmed in a windless sea of monosyllables. We can sail further only on the fresh winds of imagination.”
Our imagination hasn't taken us too far in 20 years. The reliance on monosyllable word scores is probably a carry-over from the diagnostic audiologic battery (of course, the use of monosyllabic word scores for diagnostic purposes has its own problems).
Given the long-term impact that the fitting of hearing aids has on the patient, conducting a couple of extra speech tests that are specifically related to the hearing aid fitting seems reasonable. And, in many cases, we're talking about an investment of only 5 to 10 minutes. True, the busy dispenser might say that monosyllables offer more flexibility since they can be easily delivered live voice. But with CD players selling for $89, is this really a major issue?
Todd Ricketts, PhD, of Vanderbilt University, who has used several different types of speech materials both in the clinic and in hearing aid research, states:
“Given the lack of reliability and real-world face validity of monosyllabic word-recognition tests, I certainly do not believe they are the best choice, or, frankly, have much utility when applied to hearing aid fittings. My personal belief is that if speech-recognition tests are selected in order to enhance hearing aid selection and fitting, they should be selected in an attempt to assess specific difficulties expressed by the patient (e.g., ‘I really wish I could understand women's voices better in noise’).”
The move to sentence-length material
As mentioned earlier, several speech tests have been developed that use sentence material, most of them geared toward the evaluation of hearing aids. They are all designed to be used with some type of competing stimulus, and all are available on CD. If you follow the research literature, you'll notice that one or more of these tests have been used in almost every hearing aid study conducted in the past decade.
To add to the face validity of these measures, some researchers have conducted the testing outside the audiometric test booth (gasp), in typical reverberant room environments—a concept that could be easily worked into the fitting protocol of a typical dispensing office.
SENTENCE-LENGTH SPEECH TESTS
The following is a review of some common sentence-length speech tests that have been used in the evaluation of hearing aids. One has been around since the 1960s, and one was released just this year.
* Connected Speech Test (CST): Each test item of the CST is a passage of speech, 9 or 10 sentences in length, about a familiar topic. It is presented with a multi-talker babble that can be adjusted to the desired signal-to-babble ratio. The passage score is based on the percent correct of 25 scoring words in each passage. Scores typically are averaged across several test passages. The CST can also be conducted using an audiovisual format.
* Dichotic Sentence Identification (DSI): Six sentences from the SSI test are paired in different combinations and presented dichotically, with a total of 30 sentences presented to each ear. The patient only has to identify the number of the sentence presented. The DSI test is scored in percent correct for the right and left ears. If right or left scores from the “free-recall” presentation are depressed, the test can be repeated using a “directed right” or “directed left” mode (the patient is required only to identify the sentence presented to the directed ear), which is helpful in distinguishing reduced cognitive function from APD.
* Hearing In Noise Test (HINT): Sentences are delivered in groups of 10, with speech-shaped noise as the competition (several researchers have substituted multi-talker babble as the background noise). The patient must repeat back all key words of a sentence for a correct response; each sentence is scored as correct or incorrect. The background noise is typically presented at 65 dB SPL, and the presentation level of the sentences is varied adaptively, in 2-dB steps, louder for an incorrect response, softer for a correct response. The HINT is scored by calculating the Reception Threshold for Sentences (RTS), the signal-to-noise ratio where 50% of sentences were repeated correctly.
* QuickSIN: The QuickSIN consists of 18 lists of sentences, six sentences in each list, five key words in each sentence, presented with multi-talker babble. The six sentences are presented at pre-recorded signal-to-noise ratios which decrease in 5-dB steps from +25 dB to 0 dB. The test can be scored in total percent correct, SNR-50 (signal-to-noise ratio for 50% correct), or SNR loss (the difference between the patient's SNR-50 and the SNR-50 of normal-hearing persons). The QuickSIN test also has high-frequency emphasis (HFE) lists, and high-frequency emphasis/low-pass filtered (HFE-LP) lists (filter set at 2500 Hz).
* Speech In Noise (SIN): Replaced by the QuickSIN test.
* Speech Perception In Noise (SPIN): The SPIN test consists of sentences five to eight words long, presented with a multi-talker babble. Only the last word in the sentence, always a monosyllabic noun, is a test item. Half of the test items are identified as having high predictability (syntactic and semantic cues) and the other half of the words have low predictability (minimal contextual cues). The test is scored in percent correct, often with separate scores calculated for the low- and high-predictability items.
* Synthetic Sentence Identification (SSI): The SSI is presented in groups of 10 sentences with a single-talker competing message at 0-dB message-to-competition ratio. The patient only has to identify the number of the sentences presented; performance is scored in percent correct for each list. This test is usually presented at various hearing levels to construct a performance/intensity function.
These are some of the most popular sentences tests, all of which result in either a percent correct score or a signal-to-noise ratio for 50% correct. There are other sentence tests, such as the Speech Intelligibility Rating (SIR), that use subjective estimations of intelligibility. The SIR test has also been used in the comparative evaluation of hearing aid performance.
While sentence tests may not be any better than monosyllables for differentiating between similar hearing aids, they do add face validity to the evaluation procedure. Moreover, by manipulating the test material, the presentation level, and the signal-to-noise ratio, dispensers can personalize the tests for their individual patients, creating the listening conditions that they experience in the real world. These results are often helpful during the counseling process.
HEARING AID DIFFERENTIATION— THEN AND NOW
As mentioned earlier, one reason that speech testing was abandoned by many dispensers in the 1980s was the realization that traditional monosyllabic word lists were not very efficient for detecting small differences in hearing aid performance. A major reason for this, of course, was that we tended to “pre-select” hearing aids that were very similar (and then conducted testing for 2 hours trying to make them look different!).
This would probably still happen today if we conducted comparative speech testing on digital hearing aids from two different manufacturers, or even if we compared a digital product to an analog one. The instruments are going to be programmed pretty much the same, regardless of whether they are digital or analog or of which manufacturer built them.
One of those who was publishing articles on the hearing aid differentiation issue in the early 1980s is Brian Walden, PhD, director of research at Walter Reed Army Medical Center. I asked him if he had changed any of his thinking in the past 20 years. He replied:
“It seemed then—and for the most part I think it still appears—that what hearing aids must do fundamentally is to restore as much audibility as possible to the patient. Further, although my thinking on this varies slightly from time to time, it seems that subtle differences in the specific amplification parameters that are provided are lost on most patients. So, what use is speech-recognition testing in hearing aid fitting? Fitting hearing aids to hearing-impaired patients is still primarily about improving the ability to understand speech. But, understanding speech seems to be primarily about audibility, and subtle suprathreshold spectral differences in audibility don't seem to make a whole lot of difference to the garden-variety patient.
“I'm intentionally painting a pretty simple picture here. There are other issues, such as broadened auditory filters and potential ‘dead hair cell regions’ in the cochlea, and possible changes in frequency and temporal resolution, that may make the reality more complex, at least for some patients. When these phenomena are better understood and hearing aid signal processing algorithms are developed that effectively alter the input signal in these domains, the picture may look very different. And, the importance of speech-recognition testing in hearing aid fitting may also be quite different from what it is today.”
WHAT TESTS ARE CLINICS USING?
Given the variety of sentence-length speech tests that are being used in hearing aid research, I thought it might be interesting to see if any of these tests are finding their way into clinical protocols. We'll start by looking at two major hearing healthcare networks, HEARx and Sonus, which both have standard fitting protocols used in their clinics across the U.S.
Cindy Beyer, vice-president of quality assurance at HEARx, says that her company's offices conduct monosyllabic word-recognition testing, both monaurally and bilaterally, prior to the fitting, but that little speech testing is conducted at the time of the fitting, as verification consists primarily of probe-mic results and subjective reports.
Similarly, Michelle Fusco, director of operations and training at Sonus, reports that, in addition to the traditional word-recognition testing, Sonus dispensers also conduct a “Binaural Preference Check” using running conversational speech presented both binaurally and monaurally to illustrate the advantage of binaural listening. She adds that the QuickSIN has recently been added to both the pre-fitting and fitting protocol.
The general fitting protocols of HEARx and Sonus appear to be similar to those of large audiology clinics. For example, Therese Walden, AuD, director of the audiology clinical section at Walter Reed Medical Center where 2000 or more hearing aids are fitted each year, notes that the hearing aid selection and verification protocol has changed a little since the Army days of Captain Carhart, and even since the time of Captain Don Worthington and Captain Jerry Northern. She relates that hearing aid verification at Walter Reed is accomplished by obtaining a “family of curves” with probe-mic measures, and that speech-recognition tests are routinely used only in the fitting of directional-microphone hearing aids in order to illustrate the potential benefit of this technology to the patient.
Michael Valente, PhD, director of another leading audiology clinic, that at Washington University in Saint Louis, states:
“Clinically, we do not use speech audiometry due to the well-known issue of the need to demonstrate large differences in scores between conditions in order to say with any degree of confidence that these differences are in fact significant. Also, there is the issue of testing in a sound-treated room (with or without the presence of noise) and how that relates to performance in the real-world (i.e., external validity). Perhaps the use of the recently introduced ‘virtual reality’ systems may address this concern.”
Things aren't too much different in Kalamazoo, MI, where Dennis Burrows, PhD, is the director of the Constance Brown Hearing and Speech Center. Burrows does add, however, that his clinic recently began using the QuickSIN as part of its selection process to assess SNR loss. Audiologists there also use the separated version of the QuickSIN for assessing directional hearing aid fittings and the high-frequency version to determine if the patient might benefit from an extended high-frequency emphasis.
These comments from clinic directors seem to reflect the secondary role of speech audiometry today in the fitting of hearing aids. It's interesting to note that several people mentioned the QuickSIN test—a test that has been commercially available for only the past 6 months. Maybe the word “quick” has captured people's attention.
WHERE ARE WE GOING?
So what is the role of speech testing in the fitting of hearing aids? We seem to have learned that it's difficult for our clinical speech tests to detect small variations in hearing aid performance, although we keep wishing they would. Certainly, our newer speech tests, most of them using sentence material, have greater face validity, which is helpful for patient counseling. But, in general, we seem to be using speech audiometry on only a limited basis when we fit hearing aids, and things haven't changed too much in the past 20 years.
However, if there is a trend, albeit a slender thread of a trend, speech audiometry does seem to be gradually creeping back into the hearing aid fitting process. I suspect that this trend will grow as new digital hearing aid processing algorithms are unveiled. So, are speech audiometry and hearing aid fittings going steady or just casual acquaintances? Maybe neither, but I do think there's a little bit of dating going on.
SPEECH TESTING—NOT FOR FITTING ONLY
The focus of our cover story is on the use of speech testing in the fitting of hearing aids. However, speech testing has many other uses in both the pre-fitting and the hearing aid verification process. Let's talk about some of them.
PRE-FITTING REASONS TO CONDUCT UNAIDED SPEECH TESTING
• To determine hearing aid candidacy: Monosyllabic word-recognition results can be used to help determine hearing aid candidacy. Unusually high word-recognition scores do not mean that the person is a poor candidate, since PB testing is typically conducted at a higher input level than average speech in the real world. However, very low scores can be significant for determining hearing aid candidacy. If the testing was conducted at a level that provided a good approximation of PB-Max (e.g., usually around 75 DB HL or 80 dB HL) and the PB scores are very low in both ears, the patient might be a candidate for other intervention strategies, such as a cochlear implant (assuming a cochlear pathology has been documented).
Speech-in-noise testing also can be conducted to help determine candidacy for border-line patients. For example, Robert Sweetow, PhD, director of audiology at the University of California at San Francisco, states:
“I have found patients whose hearing loss was so mild, I might not have recommended amplification, despite their subjective reports. In a few cases, HINT and QuickSIN testing provided evidence that I used in counseling to show patients that their personal demands upon their hearing may be unrealistic. Their HINT and QuickSIN scores were at or near the performance of people with normal hearing. But, more often than not, their subjective comments were confirmed, sometimes to my surprise.”
• To determine when a binaural fitting might not be appropriate: In most cases, binaural is the standard fitting, but there are exceptions. When PB-Max scores differ greatly (e.g., 80% in the right, 30% in the left), a cautious approach to binaural amplification is warranted, especially if the patient is an experienced user of amplification for the better ear. Many dispensers conduct a binaural word-recognition test for such patients, although I'm not sure this would provide much useful information. I suspect that the binaural score would simply reflect the better monaural score.
There's also evidence to suggest that in cases where the PB-Max scores might be similar for both ears, if the DSI score is exceptionally low in one ear, binaural amplification may be contraindicated.
• To determine the existence of a cognitive or central processing disorder: Tests such as the SSI or the DSI can be used to detect an auditory processing disorder (APD). While it seems important to know if your patient has a significant APD before fitting hearing aids, these tests are used even less frequently than the other sentence tests we've mentioned.
Catherine Palmer, PhD, of the University of Pittsburgh, explains her clinic's approach:
“We don't conduct these tests routinely because the results would not change the hearing solution that we recommend based on the other tests that we've conducted and our interaction with the patient and family. In other words, I think the difficulties these types of problems will cause often become obvious without the testing.”
• To predict the aided outcome with amplification: It is helpful to obtain a general idea of how well a person will do with amplification, and many of the speech tests we've mentioned will provide that information.
Brian Walden, PhD, of Walter Reed Army Medical Center offers these comments on this topic:
“I think it's reasonable to want to know the upper limit of a patient's speech-recognition ability; that is, what's the best that the patient can be expected to do with amplification? For the most part, we can't expect a hearing aid to provide significantly better speech recognition than is obtained under earphones. And, given the excellent quality of today's hearing aids, we shouldn't expect that the speech-recognition ability provided by the hearing aid would be much poorer than that obtained under ‘ideal’ audiometric test conditions. So, establishing the upper limits of speech understanding for the patient and seeing how close we can come to that with amplification appears useful.”
It's important to point out that these comments make the assumption that maximum performance was obtained during the earphone testing. If the earphone speech testing was conducted at a presentation level that didn't maximize audibility (which often is the case), then the aided results would likely be better than the earphone results (assuming that effective audibility was maximized with the hearing aid).
• To demonstrate to patients that their understanding of speech is impaired: It is sometimes helpful in counseling to have objective information to show the patient that his speech-understanding performance is significantly poorer than that of normal-hearing individuals. In this case, we're not interested in the patient's maximum performance, but rather his speech-understanding ability when the signal is presented at an “average” or “slightly-below-average” intensity. If we assume that soft speech is around 45 dB SPL to 55 dB SPL, then a presentation level around 35 dB HL or 40 dB HL would be a fair challenge. Several different speech tests could be used for this demonstration. It's important, of course, to know the results for normal-hearing individuals for these same presentation levels so that appropriate comparisons can be made to the patient.
• To assist in determining amplification characteristics and features: This is an area of interest to Catherine Palmer, and she comments on the current protocol at the University of Pittsburgh:
“We are implementing speech-in-noise testing in the hearing aid evaluation using either the QuickSIN or the HINT, in order to assist us in providing technology recommendations. I believe it will help us determine who absolutely needs a special noise solution such as directional-microphones technology, ALDs, or some type of noise reduction, and also be of assistance in creating realistic expectations for the patient and family. I think this type of testing also can be very helpful in separating out the level of technology that is most beneficial (cost/benefit) to the patient.”
The QuickSIN also has some filtered lists (HFE-LP) that might be helpful in determining hearing aid characteristics. These lists have a slight high-frequency gain boost (around 10 dB at 1500 Hz, 20 dB at 2000 Hz) and then are sharply filtered above 2500 Hz. In theory, if the patient's performance on the standard lists (recommended presentation level of 70 dB HL) is no better than on the HFE-LP lists, then it should be questioned if hearing aid gain above 3000 Hz is beneficial.
REASONS TO CONDUCT AIDED SPEECH TESTING
We pointed out that speech audiometry probably lacks the sensitivity to detect small differences in hearing aid performance on an individual basis. Similarly, it's unlikely that the advantages provided by a binaural versus a monaural fitting will be reflected in typical clinical tests. However, there are several reasons for conducting aided speech testing, the results of which can be helpful for both fitting decisions and patient counseling.
• To demonstrate that aided performance is better than unaided: Patients walk into the clinic with trouble understanding speech in background noise. When they are fitted with their new hearing aids, the first thing they ask is, “Can I now understand speech better in background noise?” It doesn't matter how many times you show them that the probe-mic results are precisely on target for six different input levels, they still want the answer to that question. The answer can be easily provided with speech testing, as Robert Sweetow explains:
“I am consistently performing the QuickSIN in the sound field at relatively soft levels, around 45 dB HL, for border-line and unmotivated hearing aid candidates. I routinely do at least four lists, and the procedure still only takes a couple of minutes. It definitely demonstrates unaided versus aided changes.”
For this type of demonstration, either the QuickSIN test or the CST should work quite well, using a presentation level around 40 dB HL or 45 dB HL. It may not even be necessary to record the results if the only purpose is to illustrate the benefits of audibility. As may be obvious, the HINT would not work well for this purpose. Even if the patient's RTS improves his perception will be that the hearing aids aren't helping, because he will be missing the same percent of sentences for both the unaided and aided condition.
• To demonstrate the advantage of special features: One feature that can be easily demonstrated with speech audiometry is directional-microphone technology. Since we know that some directional-microphone hearing aids are not “directional” when they come out of the box, this demonstration should probably be routine whenever directional hearing aids are fitted.
Speech tests such as the HINT, QuickSIN test, and the CST can be presented with the speech and noise separated. By presenting the speech at or near a 00 azimuth and with noise from behind the listener, the dispenser can test the hearing aid for both the omnidirectional and directional settings. An improved percent correct score or signal-to-noise ratio for 50% correct would be expected in the directional setting.
• To obtain information for counseling: Knowledge of a patient's ability to understand speech in noise can be very useful for counseling and in developing realistic expectations. For example, if a patient's aided speech results reveal a HINT RTS of +10 dB, and a QuickSIN SNR-50 of +12 dB, and the patient's favorite listening environment has a signal-to-noise ratio around 0 dB, it is important to tell the patient that even with his new digital directional hearing aids, he will still not understand speech very well in that listening environment.
• To illustrate the benefits of visual cues: As part of routine patient counseling or a more formal rehabilitation program, patients are usually provided information concerning the use of visual cues. It is often helpful to make patients aware of the contributions of visual cues by having them take a speech test with and without visual aids. A speech material that works nicely for this is the CST, as Robyn Cox, PhD, and colleagues at the University of Memphis have conducted research with this test using an audiovisual presentation. The word is that a DVD format of the CST will be available soon. HGM