Automated Audiometry: Then And Now



Author: Robert H. Margolis

“The number of audiometric examinations made today has grown to such a magnitude that it is only natural that some of the techniques of measurement should become automated.”
— Rudmose (1963)

Then
The first description of pure-tone threshold audiometry published in the scientific literature appeared in The Laryngoscope during the Coolidge administration (Jones & Knudsen, 1924). Electronic audiometers had recently been introduced and were finding their way into otology practices. The description of pure-tone audiometry in that germinal report is remarkably similar to procedures used today. Audiometers have evolved through the technological epochs defined by vacuum tubes, then transistors, then integrated circuits, and now microprocessors, but their functionality is essentially the same.

Automated measurement of pure-tone thresholds probably started with Georg von Bekesy’s “tracking” device which allowed the listener to track their thresholds by depressing a button when the tone was audible and releasing it when it was not (Bekesy 1947). The resulting pattern, written on a chart recorder, was used to infer the threshold of audibility. But it was the observation that patients with different auditory disorders produced different tracking patterns that clinicians found useful. At one time, Bekesy audiometry was one of the “special auditory tests” that were used to distinguish between cochlear and retrocochlear hearing loss. Clinicians routinely obtained audiograms manually with the “Hughson-Westlake” method, and then, when indicated, would perform the automated Bekesy test for differential diagnosis. Manual pure-tone audiometry has outlived many technical developments and diagnostic tests that made cameo appearances in the evolution of our field. 1

Pure-tone threshold measurement in research laboratories has a different history. One of the first applications of laboratory computers by hearing scientists was to automate pure-tone threshold testing. Bekesy’s Nobel-prize winning discovery of cochlear travelling waves that mapped frequency onto the cochlear partition led to the use of pure-tone threshold testing to determine the function of the cochlea in various regions. The ability to make quantitative measurements of the function of microscopic structures in the most inaccessible location in the body by audiometry has been one of the most important scientific tools in neuroscience. Hearing scientists wanted a procedurally robust, efficient method for measuring audiometric thresholds, thereby revealing all sorts of cochlear processes.

When I arrived at the University of Iowa in 1970 to begin my doctoral work, my first research assignment was in Dr. Arnold Small’s psychoacoustics laboratory working with post-doctoral research fellow, Dr. Harry Rainbolt. Using a pre-computer system of vacuum tube switches, timers, and amplifiers that we affectionately called the “monster”, we automatically measured thresholds at frequencies that were sometimes a few Hz apart. To obtain the data in Figure 1, three threshold determinations were obtained at 55 test frequencies for three subjects, a total of 990 threshold determinations. The masking patterns represent the excitation of specific locations of the cochlea for two different types of masking noise. The automated method provided efficiency and procedural consistency that was difficult to achieve with manual testing.

These “adaptive psychophysical procedures” were used in hundreds of research studies but would not find their way into audiology clinics.

Figure 1. Automated measurement of pure tone thesholds. From Margolis & Small (1974). Reprinted with permission from J. Acoust. Soc. Amer. 56, 692-694. Copyright, 1974, Acoustical Society of America.”


Automated audiometry has been employed for decades in industrial hearing conservation programs, first with Bekesy audiometry and later with “automated Hughson-Westlake” methods. Like the automated laboratory methods that began to emerge in the 1960’s, these automated methods have not been employed in audiology clinics.

An important reason that the automated methods employed in laboratories and hearing conservation programs have not been used in audiology clinics is that they have not included bone-conduction testing and contralateral masking. The diagnostic audiogram (air conduction and bone conduction with masking) is significantly more difficult to automate than air-conduction only tests. The rules for unmasked air-conduction threshold testing require that the signal-level be varied around the threshold until it can be determined with some level of certainty that threshold has been found. Bone-conduction testing requires an additional transducer, usually placed in two different locations on the head. Masking requires the determination of appropriate masker levels and additional testing to ensure that the signal is not audible in the non-test ear. The vast majority of errors made in pure-tone audiometry are associated with bone conduction and masking.

Now
The first automated system for diagnostic pure-tone audiometry was introduced by Tympany Inc. in 2003. The “Otogram” employs an automated method very similar to the “Hughson-Westlake” method. Masking during air-conduction and bone-conduction is implemented when the system determines it is appropriate and bone-conduction testing is performed with forehead placement of the vibrator. A validation study has been published that indicates good agreement between thresholds obtained with the automated Otogram system and manual audiometry (Ho, Hildreth & Lindsey, 2009). The “Otogram” is currently marketed by Ototronix Diagnostics.

In 2000, Audiology Incorporated (AI) was founded to develop and commercialize automated hearing tests. Funded by the National Institutes of Health Small Business Technology Transfer program, AI developed AMTAS®, a system for obtaining an automated diagnostic pure-tone audiogram. (See audiologyincorporated.com for a video demonstration). The system has been validated in a series of studies comparing automated audiometry with manual testing by audiologists (Margolis, Saly, Le, & Laurence, 2007; Margolis, Glasberg, Creeke, & Moore, 2010; Margolis, Frisina, & Walton, 2011; Margolis & Moore, 2011). At that time I was director of the University of Minnesota Hospital Audiology Clinic and it became clear that staff time was much too valuable to spend on tasks that were perfectly amenable to automation. We had three Ph.D. and five M.A. audiologists at the time. All of the M.A. audiologists have since earned Au.D. degrees. Two comments made to me by people outside our profession crystallized the issue for me. An adult patient who I had just tested asked why I had to sit and push those buttons. And a hospital administrator who insisted that the staff was not sufficiently productive suggested that we should automate some of our testing. Those comments got me thinking about advantages that could result from automated pure-tone and speech audiometry. An analysis of our activities revealed that we spent more time doing those tests than any other activity – by a wide margin. Let’s look at the potential advantages of automating pure-tone and speech audiometry.

Use of Professional Time. The transformation of audiology to a doctoral profession was based on the finding that the skills and knowledge necessary to practice audiology were equivalent to those of other doctoral health professions. Our persistent use of 90-year old methodology that can be automated is completely inconsistent with that finding. If our current practice patterns were meeting the needs of our patients there would be no need to change how we do things. But they are not. A report published by ASHA this year indicates that there are 29 million Americans with untreated hearing loss (http://www.asha.org/Aud/Articles/Untreated-Hearing-Loss-in-Adults). If our profession was to take care of these potential patients each audiologist would have to see an additional 2500 patients per year. If we really want to be the profession that takes care of the hearing-impaired population as we like to say we are, we have to radically change the way we deliver our services. A good start would be to maximize the effectiveness of our professional time by automating tasks that can be automated.

Standardization. We like to say that we have a standard pure-tone audiometry method – the Hughson-Westlake method. But the original article, published five months before D-Day (Hughson & Westlake, 1944) leaves many procedural variables to the discretion of the tester. Later publications, including the classic article by Carhart and Jerger (1959) and the American National Standard Methods for Manual Pure-Tone Audiometry (ANSI S 3.21-2004) provide additional guidance but it is not accurate to say that we have a standard method for pure-tone audiometry. The varying skill levels of testers and their idiosyncratic choices of methods result in widely varying accuracy. As a result it is very common for a patient to bring a recent audiogram to a clinic only to have it repeated, because of lack of confidence in the original test. Automation provides an opportunity to standardize the hearing test for the first time.

Accuracy. Experienced audiologists learn certain tricks that insure accuracy. They are sensitive to cues that suggest that the patient may not be responding consistently and accurately. The behaviors that provide these cues can be tracked by a computer and procedural safeguards can be incorporated into the automated test. For example, an outlier (a threshold that differs from those surrounding it) can be automatically retested. The validity of a test can be quantitatively estimated by tracking variables that are known to correlate with accuracy and using the measured behaviors in a regression formula that estimates the differences between measured and expected thresholds (see Margolis et al, 2007).

Analysis. Computerized automation can provide an analysis of results that can help us communicate consistently with our patients and other professionals. It is common in our clinical reports to describe audiometric results in terms like “symmetrical, mild-moderate, sloping, sensorineural hearing loss”. The problem is that none of those terms have standard definitions so the same audiogram can be described differently by two audiologists. In our studies comparing the judgments of audiologists, we found agreement in describing configuration, severity, site of lesion, and symmetry of 68%, 83%, 74%, and 77%, respectively (Margolis & Saly, 2007, 2008a). Computers can make these judgments with complex rules in milliseconds, providing consistency and validity in how we describe hearing loss. The results for each category of hearing loss can then be described in patient-friendly language in a report that the patient can take home and discuss with family members.

Access. Perhaps the most compelling reason that we should automate our procedures is to make hearing testing accessible. Figure 2 shows the gap between the capacity and need for hearing tests in the U.S. In 2000, all of the audiologists and hearing-aid dispensers working full time could provide just a little more than half of the needed hearing tests. As the population grows and ages, the gap will grow rapidly. And that is in the U.S. where hearing tests are more available than anywhere else in the world. Most countries don’t have any audiologists. In the U.S. there is one audiologist for 26 thousand population. When you compare that figure to lawyers (1 per 258) and dentists (1 per 2060) it is not surprising that most people don’t know who we are. Due to our small numbers we don’t reach most of the people who could benefit from our services. Because it is not realistic to expect a significant increase in the number of audiologists, our only hope to provide services to the 29 million with untreated hearing loss is to provide more efficient pathways to service. Automation is an essential tool in increasing access.

The AMTAS report shown in Figure 3 illustrates how automated features can add value to the audiogram. The Quality Assessment analysis to the right of the audiogram shows measured values for quality indicators that are correlated with accuracy. The percentile (%ile) values indicate the subjects’ performance in relation to a large database that was acquired during the validation phase of AMTAS development. The quality indicators that are highlighted in red are those that exceed the 80th percentile and are “red flags” with regard to the accuracy of the audiogram. In spite of two “red” quality indicators, the predicted accuracy is “Good”, based on a predictive formula that estimates the overall accuracy of the audiogram. Although the predictions from this method aren’t perfect, they are highly predictive of the accuracy of the results.

Figure 2. Capacity and need for audiograms in the U.S.
The need is based on the prevalence of hearing loss (11.5% in 2010), an assumption that people with hearing loss should be tested once every two years, and the finding that 25% of patients seen at the University of Minnesota Hospital Audiology Clinic had normal hearing in both ears (Margolis & Saly, 2008b). The growth in need over time is based on projected population growth and the increase in hearing loss prevalence due to the aging of the population. The capacity for audiologists is based on the workforce data from the Bureau of Labor Statistics and the assumption that every employed audiologist performs three audiogram per work day. The increased capacity when hearing aid dispensers are grouped with audiologists is based on an estimate of the number of licensed non-audiologist hearing aid dispensers (Karl Strom, personal communication). The workforce capacity is assumed to increase at a rate of 1% per year. This is a revision of the analysis in Margolis and Morgan (2008).


Below the audiogram is the table of masker levels that were present when the signal level was at the indicated threshold. A post-hoc analysis of masker levels is performed and potential undermasking and overmasking situations (masker alerts) are determined. The number of masker alerts is one of the quality indicators. <br /><br />At the bottom of the report is a classification of the audiogram based on severity, configuration, site of lesion, and symmetry. The method (AMCLASS) has been validated against the classifications provided by a panel of experts (Margolis & Saly, 2007, 2008a). The hearing loss in the right ear is classified as a moderate, flat, sensorineural hearing loss. The left ear has a moderate-severe, sloping, sensorineural hearing loss. The hearing loss is bilaterally asymmetric. The classification system employs a complex set of 161 rules that maximize the agreement between the AMCLASS outcomes and the categories provided by the expert judges. Although the agreement between AMCLASS and experts is not perfect – some may call the hearing loss configuration in the right ear “peaked” rather than “flat” – the method provides a tool for standardizing our descriptions of audiometric patterns, thereby improving communication with patients and other professionals and reducing confusion that occurs when the same hearing loss is described differently by two audiologists.

Figure 3. AMTAS report.


Obstacles
So why is automated audiometry not widely used in the hearing professions? There are a number of obstacles that have hindered the development and deployment of commercially-available, automated systems for diagnostic pure-tone audiometry.

1. Culture. Manual pure-tone audiometry is well accepted in the hearing professions and in other professions that rely on hearing test results. Many health disciplines refer their patients to audiologists for hearing tests and use their results in their management plans. Obtaining a pure-tone audiogram is the standard of care for hearing aid selection, determining candidacy for certain surgeries, evaluating treatment outcomes, educational placement, assessing the effects of occupational noise, and evaluating age-related hearing changes, among others. It is an essential research tool for a wide variety of studies of auditory function, pharmacological effects, disease progression, and genetic syndromes. As a colleague astutely pointed out to me, the pure-tone audiogram is one of the most powerful tests in medicine (David Zapala, personal communication). When a procedure is powerful and effective there is resistance to changing it.

In addition, the prospect of turning over to automation an activity in which we spend so much of our professional time is threatening. Figure 2 and the ASHA report of 29 million Americans with untreated hearing loss should assure audiologists that we must increase access to hearing testing and using automation to do so will strengthen the profession.

2. Availability. Currently the only commercially-available system for automated diagnostic audiometry is the Otogram Hearing Diagnostic System from Ototronix. A number of factors have discouraged manufacturers from making automated audiometry more available, including the push-back from the professions that occurred when the Otogram was introduced in 2003 and the reimbursement problem discussed below.

3. Reimbursement. When the Otogram was introduced in 2003, there was a vigorous response from our professional organizations, based mostly on anecdotal evidence of inaccurate results. This negative response was, at least in part, related to the perception that the device was being marketed as a means to eliminate audiologists from the hearing care process. The Centers for Medicare and Medicaid Services (CMS) responded in 2008 by issuing a communication stating the policy that Medicare and Medicaid programs prohibited reimbursement for automated hearing tests. In an unusual, if not unprecedented, communication the Otogram specifically was named as a non-reimbursable procedure. The policy was clearly aimed at a specific product in addition to the general class of automated procedures. Ototronix sought revisions to the 2008 transmittals that singled out the Otogram as a non-reimbursable procedure and petitioned for new CPT codes for computer-assisted and automated audiometric testing. In June 2009, the American Medical Association CPT Advisory Panel approved the CPT-III codes for automated audiometry including
  1. 0208T Pure tone audiometry (threshold), automated; air only
  2. 0209T air and bone
  3. 0210T Speech audiometry threshold, automated;
  4. 0211T with speech recognition
  5. 0212T Comprehensive audiometry threshold evaluation and speech recognition
  6. (0209T, 0211T combined), automated
Category III codes are temporary codes created to provide a means for evaluating emerging technologies. These codes became effective on January 1, 2010. In May, 2010, CMS issued new transmittals, revising those issued in 2008, allowing automated audiometric procedures to be eligible for coverage and reimbursement under the new Category III codes. Reimbursement for these codes is at the discretion of regional Medicare contractors.

At a time when the challenges faced by clinical practices require solutions to maximize efficiency and direct valued resources to high value activities, coding of “automated”, “computer-assisted” and manual audiometry continues to be confusing. With the advent of more and more computer-integrated audiometers, audiologists must be careful to use the Category III codes when the computer is performing some or all of the testing.

A Look Ahead
In 1994 I did an analysis of the number of each diagnostic procedure that was performed in the University of Minnesota Hospital Audiology Clinic. By a wide margin, pure-tone audiometry eclipsed all other diagnostic procedures. Many of the procedures we perform, like auditory brainstem response, otoacoustic emissions, and tests for pseudohypacusis are done because we can’t get an accurate audiogram. But with one audiologist per 26,000 population in the U.S. and even smaller ratios in other countries, there is not sufficient access to basic hearing testing. It should not be acceptable for the hearing professions to claim that we take care of people with hearing disorders when we can’t reach most of the people who can benefit from our services.

The National Institutes of Health has taken notice of the shortage of accessible and affordable hearing services. A workshop help in 2010 led to a new grant program aimed at developing new technologies and service delivery models that would increase access and affordability. The workshop findings can be viewed at www.nidcd.nih.gov/funding/programs/09HHC/Pages/summary.aspx.

Audiology as a profession needs a keener sense of awareness to the shortage of hearing services and to our inefficient delivery models. Recent attempts from outside our profession, such as the hiHealthInnovations program (https://www.hihealthinnovations.com/) should not be dismissed as an attack on our profession. Twenty-nine million Americans with untreated hearing loss need creative approaches to hearing care.
Robert H. Margolis earned bachelor’s and master’s degrees from Kent State University (1968, 1969) and a Ph.D. degree from the University of Iowa (1974). After a post-doctoral research fellowship at the University of Wisconsin, he joined the faculty of the UCLA Medical School in 1975. In 1980, he was appointed associate professor of communication sciences and disorders and director of the Gebbie Hearing Clinic at Syracuse University. In 1988 he became professor and director of Audiology at the University of Minnesota Medical School. In 2000 he established AUDIOLOGY INCORPORATED. Margolis has over 130 publications in scientific and clinical journals and textbooks. His research has focused on development of methods for evaluating disorders of hearing. He has been awarded research grants from the Deafness Research Foundation, NATO Division of Scientific Affairs, and the National Institutes of Health. He has received honors and awards from several organizations. Dr. Margolis’s current research focuses on the development and clinical validation of automated tests of hearing.

References
ANSI S3.21 (2004).American National Standard Methods for Manual Pure-Tone Audiometry. American National Standards Institute.

Carhart R & Jerger J (1959). Preferred method for clinical determination of pure-tone thresholds. J Speech Hearing Dis 24,330–345.

Ho ATP, Hildreth AJ, & Lindsey L (2009). Computer-assisted audiometry versus manual audiometry. Otol &Neurotol 30, 876-883.

Hughson W, Westlake H (1944). Manual for program outline for rehabilitation of aural casualties both military and civilian. Trans. Amer. Acad. Ophthal. Oto-laryng. Suppl. 48, 1-15.

Jones IH, Knudsen VO (1924). Functional tests of hearing. Laryngocope 34, 673-686.

Margolis RH, Frisina R, Walton JP (2011). Automated method for testing auditory sensitivity: II. Air Conduction Audiograms in Children and Adults. Int J Audiology 50, 434-439.

Margolis RH, Glasberg BR, Creeke S, Moore BCJ (2010) AMTAS® - Automated Method for Testing Auditory Sensitivity: Validation Studies. Int J Audiology 49, 185-194.

Margolis RH, Moore BCJ (2011). Automated method for testing auditory sensitivity: III. Sensorineural hearing loss and air-bone gaps. Int J Audiology50, 440-447.

Margolis RH, Morgan DE (2008). Automated Pure-Tone Audiometry: An Analysis of Capacity, Need, and Benefit. Amer J Audiology 17, 109-113.

Margolis RH, Saly G, Le C, Laurence J (2007). Qualind™: A Method for Assessing the Accuracy of Automated Tests. J Amer Acad Audiol 18, 78-89.

Margolis RH, Saly GS (2007). Toward a standard description of hearing loss. Int. J Audiology 46, 746-758.

Margolis RH, Saly GL (2008a). Asymmetrical Hearing Loss: Definition, Validation, Prevalence. Otol &Neurotol 29, 422-431.

Margolis RH, Saly GL (2008b). Distribution of Hearing Loss Characteristics in a Clinical Population. Ear & Hear, 29, 524-532.

Margolis RH, Small AM (1974). Masking with narrow band FM noise. J AcoustSoc Amer56, 692-694.

Rudmose W (1962). Automated Audiometry. In J Jerger (ed) Modern Developments in Audiology, New York (Academic Press)

Von Bekesy, G (1947). A new audiometer.ActaOto-Laryngol. 35, 3411-3428.

Disclosure
The author is president of Audiology Incorporated, a company that develops automated hearing tests. Some of the technologies described in this article may become commercial products.

Acknowledgements
The author is grateful to Samantha Stiepan and George Saly for their helpful reviews of the manuscript and to Michael Spearman for information related to reimbursement for automated tests.    


1 The only use of Bekesy audiometry for routine clinical hearing testing known to this author was by Dr. Alan Feldman, a pioneer of clinical audiology. In his private practice in Syracuse, New York, he routinely obtained air-conduction audiograms by Bekesy audiometry to determine the need for further testing.