From the Desk of Gus Mueller
This is our second month at 20Q celebrating the 10-year anniversary of the introduction of modern-day frequency lowering in hearing aids. I use the word “celebrating” loosely—more on that in a moment. Last month you heard from Josh Alexander on these pages, where he provided a detailed review of what is new, and what has changed with frequency lowering technology over the years, which is now available from all major manufacturers. What Josh purposely didn’t talk about was the research surrounding the benefit we might expect when our patients are using this technology, as that is the topic of this second edition.
When new hearing aid technology is introduced, it is typically studied in a systematic manner in three general stages, starting with efficacy, then moving to effectiveness, and finally efficiency is examined. As reviewed last month, I think we pretty much know that under ideal situations (efficacy), frequency lowering performs as advertised. That is, higher frequencies are lowered to the desired destination without excessive distortion. But what about the next level of research, effectiveness? How does this feature work with our patients in the real world on an everyday basis? You’d think that after ten years we would know the answer. But do we?
To help us understand what we do know, we’ve called on Ryan McCreery, PhD, who has been involved with clinical research with this technology since its introduction. Dr. McCreery is the Director of the Center for Audiology at Boys Town National Research Hospital in Omaha, Nebraska. He also has a research lab at Boys Town where he studies how hearing aids affect speech recognition in children and adults. Some of his research has been working with his mentor, Pat Stelmachowicz, studying how to optimize frequency lowering for children and adults who wear hearing aids. Dr. McCreery serves on numerous committees for both the AAA and ASHA (most everything that has to do with pediatrics), has received the Career Contribution to Research award from the ASHA, and is in the process of completing a book on pediatric amplification for Plural Publishing.
While Ryan spends much of his time as a researcher, you probably know him best for his informative presentations and easy-to-read clinically insightful articles that frequently appear in our audiology journals—the one that you see here is no exception. He tells us that besides frequency lowering research, he does have three other avocations: fishing, beer and searching for warm-weather vacation destinations, none of which are too surprising for someone from Wyoming/Nebraska.
Gus Mueller, PhD
Browse the complete collection of 20Q with Gus Mueller CEU articles at www.audiologyonline.com/20Q
20Q: Frequency Lowering Ten Years Later - Evidence for Benefit
- Readers will be able to explain challenges of applying research findings regarding frequency lowering to individual patients.
- Readers will be able to list factors that may affect the benefit of frequency lowering.
- Readers will be able to discuss results of studies on sound quality and frequency lowering.
- Readers will be able to provide some general fitting tips for frequency lowering that are supported by research.
- Readers will be able to list clinical questions about frequency lowering that have not yet been answered by the research to date.
1. How would you describe the current state of clinical research about frequency lowering?
Highly variable and changing rapidly. As highlighted in Josh Alexander’s 20Q of last month, all of the major hearing aid companies now offer some kind of frequency lowering algorithm in their products or are developing new iterations of their previous algorithms. My colleagues and I published an evidence-based systematic review of frequency lowering as a signal processing strategy for children back in 2012 and identified nine studies that met the inclusion criteria up through 2011 (McCreery, Venediktov, Coleman, & Leech, 2012). Just since then, the number of studies that would be included under the same search criteria is around 20. This is great scientific progress in a very short period of time, but my concern is that clinicians have more questions than ever about which patients in their clinic might benefit from frequency lowering and how to prescribe the best settings for different patients. In addition to lots of studies examining the effects of frequency lowering on speech recognition, some of the more recent studies have focused on how frequency lowering algorithms affect outcomes like ratings of sound quality and objective measures of listening effort.
2. All of that sounds incredibly exciting. With all of those studies, there must be some rock-solid evidence that clinicians can use to decide whether to use frequency lowering with their patients?
I think we obviously know a lot more about how frequency lowering affects speech recognition than we did a few years ago, but there are a few key challenges that have made firm conclusions difficult. Many of the studies related to speech recognition have shown mixed results. Some listeners benefit from frequency lowering, some people perform the same with frequency lowering, and a few might even perform worse with the feature activated versus off. If you do what we researchers tend to do, and just average all those people together, the effects of frequency lowering are not impressive. About 90% of the recent research on frequency lowering has focused on Phonak’s SoundRecover algorithm, which means the independent data on other manufacturers' processors is limited.
3. If on average, frequency lowering doesn’t make a difference, why should I even think about using it?
It can be difficult to predict benefit for an individual patient from average group performance. Just because something doesn’t work on average, doesn’t mean it won’t benefit the specific individual hearing aid patient sitting in your office. One of the challenges that anyone who has conducted hearing aid research can tell you is that no two people with hearing loss are exactly alike. Because frequency lowering is designed to improve access to high frequency sounds, we probably should expect the amount of benefit that a listener receives to depend on how much additional speech energy we can give them access to with frequency lowering. Some of the variation in outcomes within studies of frequency lowering probably has to do with individual differences in high frequency audibility.
4. That seems like a fairly simple thing that some researcher could have tested?
Yes, but unfortunately, just like many things in science, the results are not as straightforward as we would like. In one of our studies with Phonak SoundRecover (McCreery et al., 2014), we estimated the amount of speech information that was being lowered using a modification of the Speech Intelligibility Index. We looked at which frequency bands were inaudible for the listeners with conventional hearing aid processing, and then used one of Josh Alexander’s handy fitting assistants to estimate what the audibility gains were when SoundRecover was activated. We did a monosyllabic word recognition task with children and adults who had hearing loss and found an overall benefit for SoundRecover compared to conventional processing. The effect was small, but so was the amount of high frequency information that was made audible. However, some listeners did not show an improvement with SoundRecover, and two listeners had poorer scores with SoundRecover, even though we had measured the fact that we had given them more audibility.
5. Why do you think that some of the listeners had no benefit, and some did worse, with frequency lowering in that study?
It’s important to remember that any kind of frequency lowering, by definition, will add distortion to the speech signal. In some cases, that distortion provides a helpful acoustic cue. But in other cases, improvements in audibility can be offset by that distortion or the distortion can make speech less intelligible than having nothing at all. Pam Souza has done some interesting work examining trade-offs between increasing audibility and increasing distortion with frequency lowering (Souza, Arehart, Kates, Croghan, & Gehani, 2013). If improving audibility was the only consideration, we could just give everyone the strongest setting of frequency lowering to move as much information as possible into the audible range. Instead, it’s more like Goldilocks where you don’t want too much lowering or too little lowering; we need it to be just right. By doing so, we are improving audibility with the least amount of distortion possible.
6. That makes sense, but how do you figure out the Goldilocks setting for frequency lowering?
Susan Scollie and an army of collaborators from Western University in Ontario published a paper on verification for frequency lowering earlier this year (Scollie et al., 2016). They did not use the Goldilocks analogy, but presented the same idea as the weakest audible setting. The weakest audible setting is the weakest frequency lowering setting that will make high-frequency sounds like /s/ and /sh/ audible. They have calibrated signals to estimate this during the verification process. Unfortunately, many of the recent studies of frequency lowering have not documented whether or not the lowering improved audibility or how much audibility was improved.
7. But, didn’t you just say that some of the listeners in your study didn’t show improvements in speech recognition with lowering, even with the improvements in audibility?
Yes, that’s right. Improving audibility with lowering is an important first step to improving speech recognition. If frequency lowering does not improve audibility, but adds some distortion to the speech signal, we should expect equivocal or poorer performance compared to conventional processing. However, even when audibility is improved by frequency lowering, there may be other factors that limit benefit. Audibility is necessary, but insufficient by itself to improve speech recognition.
8. What other factors may limit benefit?
One major limiting factor may be the ability for the listener to use the lowered speech information to support speech recognition. We know that sensorineural hearing loss results in poorer spectral resolution. Even with conventional amplification processing that is not lowered, there is a point (usually when thresholds start to get around 70 dB HL) where increasing audibility of the signal doesn’t have the same positive effects as it might for listeners with less hearing loss. We think that loss of the ability to differentiate between differences in the frequency of two sounds or changes in frequency over time can negatively affect a listener’s ability to use those lowered cues. Also, many types of lowering use compression to lower the input to the hearing aid. Compressing speech in the frequency domain can make detecting those differences in frequency that we use to understand speech even more challenging.
9. Do people need aided listening experience to learn to use lowered speech information for speech recognition?
Intuitively, you might expect that it would take some time to learn to understand speech when the frequency information has been altered. There is certainly a precedent for this idea in the cochlear implant literature. However, the few studies that have looked at this issue have not fully supported that prediction. Jace Wolfe, for example, has done a lot of work on the benefits of frequency lowering for children. In one study from Jace and his colleagues, school-age children showed increases in the benefit over time with frequency lowering on speech recognition in noise, but not in quiet (Wolfe et al., 2011). Danielle Glista and her colleagues at Western University in Ontario examined changes in benefit from frequency lowering over time using a wide range of speech recognition tasks - from high-frequency detection tasks, to plural and word recognition (Glista, Scollie, & Sulkers, 2012). Consistent improvements in performance were not routinely observed after about two months of experience with frequency lowering for any of the tasks. Individual listeners did vary in their performance for some specific tasks. In our study that I mentioned earlier (McCreery et al., 2014), we saw improvements in word recognition immediately for children and adults who had never used frequency lowering before. Based on these studies, the issue about acclimatization is still very much an open question and might depend on the type of task being used in the study. For clinicians, this means that initial results with frequency lowering may not be indicative of the eventual benefit that a patient might receive.
10. You mentioned that the benefit of frequency lowering might depend on the type of stimuli that are used to assess it in the clinic?
There are a number of things we have learned from research studies on frequency lowering that can help to inform decisions about how to assess frequency lowering benefit in the clinic. I think it’s important to point out that frequency lowering benefits have been documented with a wide range of different stimuli including phonemes (Glista et al., 2009), words (McCreery et al., 2014), and sentences (Wolfe et al., 2011). However, there are a few additional considerations. First of all, if the patient has performance near 100% on the task with conventional processing, it will be difficult to measure any improvements that may happen with lowering. This may require adding background noise to decrease the speech recognition performance. Ideally, performance around 50%-70% in the conventional condition offers the ability to document a clinically significant improvement in speech recognition with frequency lowering. Stimuli that have a larger amount of high-frequency phonemes, such as the California Consonant Test (Owens & Schubert, 1977), may provide a more sensitive measure of high-frequency audibility than a stimulus set with more phonemic balance.
Now, these are clinical measures, and of course what we really want to know is what is the benefit of frequency lowering in the real world? We know that clinical speech tests often are poor predictors of real world benefit and satisfaction. A few studies have examined real world benefit with frequency lowering by having people wear the devices outside of the laboratory. For example, Bohnert and colleagues (2010) asked adults to complete questionnaires about listening satisfaction and found significant improvements with frequency lowering compared to conventional processing. However, the participants in that study were not blinded to the processing manipulation in the study. As you can imagine, these types of real world outcome studies are difficult to do in children. Ruth Bentler and I, along with several of our colleagues (2014), found that children with frequency lowering activated in their hearing aids had comparable speech and language outcomes to children with conventional processing. For children, speech and language outcomes are probably the most important real world outcome measure we can have.
11. Is there any way to predict which individual patients will show the most benefit from the processing?
I think this is a very important question that could be extended to a number of different hearing aid features. The expectation that benefit should be uniform for all people with sensorineural hearing loss is kind of absurd when you think about it. But, then the question that clinicians always ask is, “Who are my patients who could benefit?” As with many of the studies we have talked about with lowering, this is a difficult question to answer based on the current research. Some studies suggest that listeners with the greatest degrees of hearing loss in the high-frequencies are most likely to benefit from frequency lowering (e.g., Glista et al., 2009). However, some of our research (McCreery et al., 2014) and that of others (Wolfe et al., 2011) suggests that listeners with mild or moderate high-frequency hearing loss can and do experience limitations in the audibility of high-frequency sounds and that a small amount of lowering can provide a small benefit in those cases. Based on these data, we believe that listeners who have limited audibility of high-frequency sounds may benefit from frequency lowering, but as we talked about earlier, it’s not a guarantee. Another factor that has received less attention in the research would be to use measures of spectral resolution along with audibility to identify which listeners might not be able to use the lowered speech cues even if we make them audible. We have a paper in preparation that shows that listeners with better spectral resolution are more likely to benefit from frequency lowering than listeners with poorer spectral resolution. Some of the work from Pam Souza that I mentioned earlier also suggests that individuals with stronger working memory capacity may be less susceptible to distortion than adults with poorer working memory capacity. Aside from the findings related to degree of hearing loss, I’m not sure that any of this is ready for clinical use, but they are factors to think about.
12. What about sound quality with frequency lowering? Doesn’t messing with the output of the hearing aid make it sound bad?
Good question. After all, we could improve audibility but make the hearing aid sound so bad that the patient would never want to wear it. Fortunately, it seems as though frequency lowering can improve audibility without major compromises in sound quality. My colleague Marc Brennan from Boys Town published a recent paper (Brennan et al., 2014) that used paired-comparisons of speech and music samples. The samples were processed with either conventional amplification with a restricted bandwidth of about 5 kHz, conventional amplification with an extended bandwidth up to about 10 kHz, and frequency lowering designed to provide a similar bandwidth as the conventional extended condition. We included a wide range of different types of music and asked the children and adults in the study to pick the sample from a paired comparison that had the most favorable sound quality. Ratings for speech and music were similar for conventional amplification with extended bandwidth and the frequency lowering condition. Both of those conditions were rated more favorably than the conventional amplification condition with restricted bandwidth for speech and music. Other studies have shown that increasing the amount of lowering to a point where distortion is increasing without adding audibility can result in much poorer ratings of sound quality than conventional amplification (Parsa, Scollie, Glista, & Seelisch, 2014). Like the results from the speech recognition studies, there are a lot of individual differences in listener’s preferences for frequency lowering. I’d suggest that clinicians incorporate some questions about sound quality into the fitting process if they are using frequency lowering.
13. What should a clinician do if a patient with frequency lowering complains about the sound quality? Just turn it off?
That’s certainly one approach. A less drastic approach might be to try to select a weaker setting that might provide less audibility, but also less distortion. It’s good to counsel patients about the impact that these programming modifications might have for their ability to understand speech. In some cases, patients have been willing to try new processing in cases where their initial perception of the sound quality might have been negative if we explain to them that reducing the strength of the frequency lowering might compromise audibility. It’s an important balance to strike, since as I mentioned earlier, reductions in sound quality might reduce the likelihood that someone will use their hearing aid. That defeats the entire purpose of fitting a patient with a hearing aid, obviously.
14. Are the results you’re describing for speech recognition and sound quality the same across all of the different manufacturers’ frequency lowering algorithms?
As I mentioned before, most of the research in frequency lowering has focused on the first iteration of Phonak SoundRecover with many fewer studies for the other manufacturers’ products. There are even fewer studies that directly compare one manufacturer’s frequency lowering algorithm to another. Josh Alexander and some of my Boys Town colleagues (Alexander et al., 2013) compared Phonak SoundRecover to Widex Audibility Extender. Benefits for frequency lowering were observed for SoundRecover, but decrements in performance were noted for Audibility Extender. Another more recent study by Miller, Bates and Brennan (2016) found no benefit for SoundRecover and reduced performance for two other different kinds of frequency transposition. My take-home point here would be that it’s a bad idea to generalize research from one type of frequency lowering algorithm to another. These algorithms function differently and those differences are likely to affect whether or not speech recognition and sound quality can be enhanced. But, that also means that we have a lot of research on the first version of Phonak’s SoundRecover, but not as much on other types of processors, including the new version of SoundRecover (SoundRecover2) that engages adaptively based in the spectrum of the input to the hearing aid.
15. Is there any indication that children may benefit more from frequency lowering than adults?
There was quite a bit of work by my mentor, Pat Stelmachowicz, showing that children may need more bandwidth than adults to achieve similar levels of speech recognition (Stelmachowicz et al., 2004). If you extrapolate the predictions about high-frequency audibility to frequency lowering, you might predict that children might benefit more from frequency lowering than adults. However, some of the work that we have done suggests that if adults are not already at ceiling levels of performance with conventional processing, then the benefits from frequency lowering can be similar for children and adults (McCreery et al., 2014). Comparisons between children and adults in the literature are also confounded by the fact that adult prescriptive formulae typically prescribe less audibility for adults than for children. Furthermore, differences in the etiology and duration of hearing loss between children and adults make such comparisons challenging. In our clinic, we activate frequency lowering for patients of any age who have restricted bandwidth with conventional processing.
16. You said that you activate frequency lowering for some patients - don’t most manufacturers default the processor to be on automatically?
Phonak has had SoundRecover on by default in their fitting software, whereas other manufacturers may require the clinician to activate frequency lowering. Regardless of what the default setting is in the software, as audiologists, we need to make individualized decisions about whether or not to keep frequency lowering activated or to activate it. I’ve heard audiologists say that they think that frequency lowering should be “off” by default, but if we are evaluating it on a patient-by-patient basis based on how much audibility the listener has in the high-frequencies, I’m not sure the default setting is that important.
Just because frequency lowering is activated does not mean it has been optimized to provide audibility. We could do an entire 20Q on frequency lowering verification, but for the purposes of this discussion, there are a few steps I’ll mention. First, the listener’s maximum audible output frequency (MAOF) should be measured by doing a real ear or simulated real ear measurement and visually determining the highest frequencies where speech is audible with the hearing aid. The MAOF is the frequency range between where the mean of the average speech spectrum and the peaks intersect the patient’s audiometric thresholds. Susan Scollie recently published a paper describing the fine tuning process for selecting the best parameters to enhance audibility and minimize the likelihood of distortion in the Journal of the American Academy of Audiology (Scollie et al., 2016).
17. Listening effort is a hot topic in our field right now. Are there studies on whether or not frequency lowering affects listening effort?
In looking at the research that’s been published, I’m not aware of any data about how frequency lowering could affect listening effort. Based on some posters I’ve seen at scientific meetings, I know there are labs that are interested in those questions. Predictions could go either way. I could see a hypothesis that by increasing audibility, frequency lowering might make listening less cognitively taxing. Alternatively, the fact that frequency lowering adds distortion might require the listener to exert greater listening effort than conventional amplification. I will be interested to see how those results turn out, since that might provide additional information that could assist candidacy decisions for clinicians.
18. I have heard conflicting recommendations about whether or not frequency lowering should be prescribed for each ear individually, or whether both ears should have a single frequency lowering setting based on the better ear?
I know that Andrea Hillock-Dunn is working on gathering data to answer that question. For a long time, there was a recommendation that having stronger frequency lowering settings in one ear might lead to complaints about sound quality. Our rationale for fitting frequency lowering is to maximize audibility for each ear individually, which might necessitate different settings for an asymmetrical hearing loss. Consequently, we have a number of patients with frequency lowering active in one ear, and not in the other, or who have stronger frequency lowering settings in one ear than in the other ear. Despite these differences between ears for some of our patients, we have had few complaints about asymmetric fitting. I think Andrea’s data might help us to understand whether that our practice is supported by data.
19. Is there any way to objectively assess frequency lowering benefit in children who are too young to participate in speech recognition assessment?
It probably is not quite ready for clinical use, but Benjamin Kirby who worked in our lab for the last few years did his dissertation at the University of Iowa studying the use of cortical evoked potentials as a way to objectively document benefit from frequency lowering. He studied the sensitivity of the acoustic change complex to perceptible differences in frequency lowering (Kirby & Brown, 2016). The acoustic change complex was sensitive to changes in the compression ratio for frequency lowering, suggesting that further research could optimize this objective approach for validating benefit from frequency lowering.
20. What do you think are the most pressing research needs for future frequency lowering studies?
As I have already mentioned, more research on differences between frequency lowering algorithms in terms of benefits and candidacy are badly needed. Comparative research of hearing aid algorithms across manufacturers is not a high priority for funding agencies or manufacturers, but I think it’s important to support clinicians with that type of information. I would also like to see more research on the factors that help to predict individual differences in frequency lowering benefit. That type of research not only helps us to understand what factors affect speech recognition with frequency lowering, but can be used by clinicians to make informed decisions about candidacy. Finally, I think it would be great to see more studies used standardized verification methods for frequency lowering (such as those described by Scollie et al., 2016). This might lead to more consistent findings across studies if the amount of audibility being provided by frequency lowering can be more carefully controlled and described.
Alexander, J. M., Kopun, J. G., & Stelmachowicz, P. G. (2013). Effects of frequency compression and frequency transposition on fricative and affricate perception in listeners with normal hearing and mild to moderate hearing loss. Ear and Hearing, 35(5), 519-532.
Bentler, R., Walker, E., McCreery, R., Arenas, R.M., & Roush, P. (2014). Nonlinear frequency compression in hearing aids: Impact on speech and language development. Ear and Hearing, 35(4), e143.
Brennan, M.A., McCreery, R., Kopun, J., Hoover, B., Alexander, J., Lewis, D., & Stelmachowicz, P.G. (2014). Paired comparisons of nonlinear frequency compression, extended bandwidth, and restricted bandwidth hearing-aid processing for children and adults with hearing loss. J Am Acad Audiol., 25(10), 983-998. doi: 10.3766/jaaa.25.10.7
Bohnert, A., Nyffeler, M., & Keilmann, A. (2010). Advantages of a non-linear frequency compression algorithm in noise. European Archives of Oto-rhino-laryngology, 267(7),1045-1053.
Glista, D., Scollie, S., Bagatto, M., Seewald, R., Parsa, V., & Johnson, A. (2009). Evaluation of nonlinear frequency compression: Clinical outcomes. International Journal of Audiology, 48(9), 632-644.
Glista, D., Scollie, S., & Sulkers, J. (2012). Perceptual acclimatization post nonlinear frequency compression hearing aid fitting in older children. Journal of Speech, Language, and Hearing Research, advanced online publication.
Kirby, B.J., & Brown, C.J. (2015). Effects of nonlinear frequency compression on ACC amplitude and listener performance. Ear and Hearing, 36(5), e261-e270.
McCreery, R.W., Alexander, J., Brennan, M.A., Hoover, B., Kopun, J., & Stelmachowicz, P.G. (2014). The influence of audibility on speech recognition with nonlinear frequency compression for children and adults with hearing loss. Ear and Hearing, 35(4), 440-447.
McCreery, R.W., Venediktov, R.A., Coleman, J.J., & Leech, H.M. (2012). An evidence-based systematic review of frequency lowering in hearing aids for school-age children with hearing loss. Am J Audiol., 21(2), 313-328. doi: 10.1044/1059-0889(2012/12-0015)
Miller, C. W., Bates, E., & Brennan, M. (2016). The effects of frequency lowering on speech perception in noise with adult hearing-aid users. International Journal of Audiology, 55(5), 305-312.
Owens, E., & Schubert, E.D. (1977). Development of the California Consonant Test. Journal of Speech and Hearing Research, 20(3), 463-474.
Parsa, V., Scollie, S., Glista, D., & Seelisch, A. (2013). Nonlinear frequency compression: Effects on sound quality ratings of speech and music. Trends in Amplification, 17(1), 54-68.
Scollie, S., Glista, D., Seto, J., Dunn, A., Schuett, B., Hawkins, M., . . . Parsa, V. (2016). Fitting frequency-lowering signal processing applying the AAA pediatric amplification guideline: Updates and protocols. Journal of the American Academy of Audiology, 27(3), 219-236.
Souza, P.E., Arehart, K.H., Kates, J.M., Croghan, N.B., & Gehani, N. (2013). Exploring the limits of frequency lowering. Journal of Speech, Language, and Hearing Research, 56(5),1349-1363.
Stelmachowicz, P.G., Pittman, A.L., Hoover, B.M., Lewis, D.E., & Moeller, M.P. (2004). The importance of high-frequency audibility in the speech and language development of children with hearing loss. Archives of Otolaryngology–Head & Neck Surgery, 130(5), 556-562.
Wolfe, J., John, A., Schafer, E., Nyffeler, M., Boretzki, M., Caraway, T., & Hudson, M. (2011). Long-term effects of non-linear frequency compression for children with moderate hearing loss. International Journal of Audiology, 50(6), 396-404.
McCreery, R.W. (2016, October). 20Q: Frequency lowering ten years later - evidence for benefit. AudiologyOnline, Article 18370. Retrieved from www.audiologyonline.com