Does This Patient Have Hearing Impairment?

A 77-year-old previously healthy woman in your primary care practice comes to your office for a routine medical examination. She feels well, has no complaints, and remains active. During a routine functional inquiry, you inquire whether she feels she has any difficulty hearing. She tells you yes. She says that her husband occasionally complains that she is not hearing him very well. With further questioning, she also reports that she sometimes has difficulty understanding people at work. In social situations, she often has to ask others to repeat themselves. What do you do next?

You have been treating a 69-year-old man in your primary care practice for hypertension and diabetes for the last 10 years. He comes to see you in your office for a routine annual checkup. You decide to screen him for hearing impairment and ask him whether he feels that he has difficulty hearing. He replies no. What do you do next?

Hearing impairment, a common chronic condition in the older American population, affects between 25% and 40% of the population aged 65 years or older.^{1 - 5} The prevalence increases dramatically with age, affecting 40% to 66% of patients older than 75 years and more than 80% of patients older than 85 years.^{6 - 8} Several studies highlight the negative social and emotional effects of hearing impairment. Hearing-impaired elderly individuals who do not use hearing aids are more likely to report sadness and depression, worry and anxiety, paranoia, decreased social activity, emotional turmoil, and insecurity compared with their hearing-impaired peers who use hearing aids.^{9 - 14}

Formal audiological tests are required to diagnose hearing impairment. These involve testing the patient's ability to hear sound tones in a soundproof room with standardized equipment. While these tests are not invasive and are easy to conduct, they require expensive equipment that is not widely available for mass screening, require dedicated staff, and are time consuming. Therefore, tests that are readily performed by clinicians as part of a general physical examination are useful to minimize the number of patients who require these formal tests. The objective of this article is to inform clinicians which bedside tests are useful for this purpose and to provide estimates of their accuracy and precision.

The ear is divided into 3 structural parts: external, middle, and internal (Figure 1). The external ear consists of the pinna, or auricle, located on the head surface, and extends through the external auditory canal to the tympanic membrane (eardrum). It functions to capture sound waves for conduction to deeper ear components. In addition, the external ear protects the air-filled space of the middle ear, which contains the medial side of the tympanic membrane, the proximal end of the eustachian tube, and 3 bony ossicles (malleus, incus, and stapes). The ossicles convert sound vibrations from the air into mechanical waves for the inner ear.

The inner ear, or labyrinth, begins where the footplate of the stapes fits into the oval window and includes the cochlea, the semicircular canals, and the distal end of the auditory nerve (cranial nerve VIII). As part of the cochlea, the organ of Corti is composed of hair cells (sensory transducers) and a complex assortment of supporting cells and is the end-organ of hearing. Perilymph fluid surrounds the membranous labyrinth of the semicircular canals and the cochlea. Located within the cochlea are the cochlear ducts, which contain endolymphatic fluid. Movement of the footplate of the stapes causes vibrations of the perilymph and endolymph, thus transmitting mechanical energy from the footplate at the oval window to the organ of Corti. Movement of the hair cells in the organ of Corti allows for the conversion of mechanical waves into electrical potential. This electrical potential is transmitted via the auditory nerve to the brain for interpretation.

Disruption of the auditory pathway anywhere from the pinna to the brain may result in hearing loss. Hearing loss is typically classified as conductive, sensorineural, or mixed. Conductive hearing loss results from pathologic changes of either the external or the middle ear structures, preventing the sound waves from reaching the fluids of the inner ear. Common causes include cerumen or foreign-body impaction, perforated tympanic membrane, otitis media, otosclerosis, cholesteatoma, tumor, or disarticulation of the ossicular chain due to trauma.

Sensorineural hearing loss results from pathologic changes of inner ear structures such as the cochlea or the auditory nerve and prevents neural impulses from being transmitted to the auditory cortex of the brain. Sensorineural hearing loss can be genetic, resulting from a mutation in a single gene or from a combination of mutations in several genes, or acquired, due to prolonged exposure to loud noises, exposure to ototoxic substances such as aminoglycosides, inner ear infections, Meniere disease, and other systemic disease such as diabetes mellitus. Patients with sensorineural hearing loss typically have difficulty filtering background noise, which makes listening especially challenging in common social settings. Degeneration of the hair cells in the organ of Corti causes sensorineural hearing loss related to aging (presbycusis). Presbycusis is the most common cause of hearing loss in the United States and is typically gradual, bilateral, and characterized by high-frequency hearing loss. Mixed hearing loss comprises elements of both conductive and sensorineural hearing loss.

Sound is described in terms of frequency (or pitch, measured in Hertz) and intensity (or loudness, measured in decibels). Conversational speech usually occurs between 500 and 3000 Hz and between 45 and 60 dB. A person with normal hearing perceives sounds that have frequencies between 20 and 20 000 Hz.

Impairments in hearing can involve impairments of sensitivity to frequency, intensity, or both. Since hearing impairment is not a “yes or no” phenomenon but rather a matter of type and degree, there is no universally accepted case definition for hearing impairment.^{15 - 18} However, all case definitions require formal audiometric testing that includes pure-tone assessment, speech audiometry, and impedance studies. In pure-tone audiometry, individual tones of different frequencies (ranging from 250 to 8000 Hz) are presented in a soundproof room at various intensities (ranging from 5 to 120 dB) to each ear. Air conduction thresholds, which are a measure of both conductive and sensorineural hearing, are determined by presenting pure tones (using headphones) that must travel via the outer and middle ear before getting to the cochlea and auditory nerve. Bone conduction thresholds, which test only sensorineural hearing, are determined by placing a bone oscillator on the mastoid bone; this stimulates the skull, which in turn stimulates the cochlea directly, bypassing the outer and middle ear. The threshold for each tone frequency is determined by finding the intensity level (in decibels) at which the individual can detect the tone 50% of the time. An audiogram graphically displays the threshold for each frequency. In presbycusis, the pure-tone audiogram characteristically shows hearing impairment for higher-frequency sounds (1000 to 8000 Hz).

The traditional definition used to classify sensorineural hearing impairment is a pure-tone average, called the speech frequency pure-tone average, greater than 25 dB in the better ear at 500 Hz, 1000 Hz, and 2000 Hz.^{17 - 18} In other words, an individual has sensorineural hearing impairment if the bone threshold average at 500 Hz, 1000 Hz, and 2000 Hz is greater than 25 dB. Numerous variations to this standard criterion have been proposed. For example, a definition proposed by Ventry and Weinstein¹⁶ and widely accepted by physicians and audiologists defines hearing impairment as the inability to hear a 40-dB tone at 1000 or 2000 Hz in both ears, or a 40-dB tone at 1000 and 2000 Hz in one ear. Clinically, a 25-dB threshold is usually considered mild sensorineural hearing impairment; a 40-dB threshold, moderate; and a greater than 60-dB threshold, severe.

To diagnose conductive hearing impairment, an air-bone gap is calculated by subtracting the bone conduction threshold (in decibels) from the air conduction threshold at each tested frequency. When the air conduction threshold is greater than the bone conduction threshold, conductive hearing impairment exists. The greater the air-bone gap, the greater the magnitude of conductive hearing impairment. Clinically, an air-bone gap of 25 dB is considered mild conductive hearing impairment; of 40 dB, moderate; and of greater than 60 dB, severe.

While audiometric tests provide a quantitative measure of hearing loss, they do not reflect the impact of such a loss on an individual's life. Hearing handicap is used to denote a change in hearing that interferes with performing activities of daily living.^{15 - 16} Some individuals with mild hearing loss experience a substantial disability and handicap, whereas others with moderate hearing loss may not exhibit any form of disability or hearing handicap. The most commonly used test to quantify hearing handicap is the screening version of the Hearing Handicap Inventory for the Elderly (HHIE-S).¹⁹ Although initially designed to identify hearing handicap, this test has also been used to screen individuals for hearing impairment.

Self-reported data to assess presence of diseases and disorders have been used frequently in large-scale epidemiologic survey studies, such as the Health Interview Survey²⁰ or the National Health and Nutrition Examination Survey.²¹ Similarly, screening patients for hearing impairment using self-reported screening questions involves asking the patient whether they feel they have hearing impairment. Numerous variations in the wording of the sentence exist, including “Do you feel you have a hearing loss?” or “Do you have a hearing problem now?” or “Would you say you have any difficulty hearing?” A yes or equivocal response to this question is considered a positive screen for hearing impairment.

The HHIE-S ( Article )¹⁹ is a 10-item, self-administered questionnaire developed to measure social and emotional handicap secondary to hearing impairment. The HHIE-S can be administered easily in a primary care office setting.

Item No./Question

Individual questions are scored as yes (4 points), sometimes (2 points), or no (0 points). Scores on the HHIE-S range from 0 (no handicap) to 40 (maximum handicap). Different scores have been proposed as a cutoff above which individuals are identified as hearing handicapped.

Clinical textbooks describe several tuning fork tests, including the Weber, Rinne, Bing, and Schwaback. Over the years, the Bing and Schwaback tests have gone out of favor due to inadequate evidence of their performance. The Weber and Rinne tests, on the other hand, are still widely taught. To perform both tests, the tuning fork (256 or 512 Hz) is struck gently on a hard rubber pad, the elbow, or the knee about two thirds of the way along the tine.²²

Weber Test. To conduct the Weber test, the base of the vibrating fork is placed on the vertex (top or crown of the head). Alternative locations are the bridge of the nose, upper incisors, or forehead.²³ The patient is asked if the sound is heard and whether it is heard in the middle of the head (or in both ears equally), toward the left, or toward the right. In a patient with normal hearing, the tone is heard centrally. In asymmetric/unilateral hearing impairment, the tone lateralizes to one side. Lateralization indicates an element of conductive impairment in the ear in which the sound localizes, a sensorineural impairment in the contralateral ear, or both.

The Weber test detects differences in air and bone conduction thresholds in one ear compared with those in the other. Therefore, individuals with bilateral conductive or bilateral sensorineural hearing impairment should, like individuals with normal hearing, have a Weber test result that does not demonstrate lateralization, as these individuals have no difference in air and bone conduction thresholds between the 2 ears. Thus, the Weber test is not useful to identify individuals with bilateral conductive hearing impairment or bilateral sensorineural hearing impairment.

Rinne Test. The Rinne test can be performed in 2 ways to detect conductive hearing impairment.

Loudness Comparison Technique. The base of the vibrating fork is placed on the mastoid bone, behind the ear and level with the canal, to assess bone conduction. To assess air conduction, the fork is then quickly placed close to the ear canal with the “U” of the fork facing forward to maximize the sound for the patient. The patient is asked if the sound is louder by bone conduction or by air conduction. Normally, the air conduction is louder than bone conduction (often abbreviated in written records as AC>BC).

Threshold Technique. The base of the vibrating fork is placed on the mastoid bone, behind the ear and level with the canal, to assess bone conduction. To assess air conduction, the patient is asked to tell the examiner as soon as he or she can no longer hear the sound, then the fork is quickly placed close to the ear canal with the “U” of the fork facing forward. The patient is asked if he or she can hear the sound again. Normally, the sound is heard longer through air conduction than through bone conduction.

In conductive hearing loss, sound is heard through bone as long as or longer than (by threshold technique) or as loud as or louder than (by loudness comparison technique) it is through air (written in records as BC>AC). In sensorineural hearing loss, the Rinne test result is the same as that achieved in normal hearing, with sound heard longer (by threshold technique) or louder (by loudness comparison technique) through air.

The examiner stands 2 ft (0.6 m; arm's length) behind the patient's field of vision (to prevent lipreading); the examiner then whispers while gently using the end of his or her finger to occlude and rub the external auditory canal of the patient's nontested ear. It is imperative to rub, as occlusion alone does not provide sufficient masking. The examiner should take in a full breath, exhale, and then whisper a set of 3 random numbers and letters (eg, “5, B, 6”). To confirm that the patient understands the instructions, a trial run using a loud voice and a simple number such as 99 is often worthwhile.

Patients with normal hearing will repeat back all 3 numbers/letters correctly. If they respond incorrectly or not at all, the test should be repeated once more using a different combination of 3 numbers/letters. It is important to use a different combination each time to exclude the effect of learning. Overall, the patient is considered to have passed the screening test if they repeat at least 3 out of a possible total of 6 letters/numbers correctly. The other ear should be then assessed in a similar manner, again using a different combination of numbers/letters.²⁴

An audioscope is a rechargeable battery–powered, lightweight, handheld instrument that combines a pure-tone screening audiometer and otoscope into a single unit. It contains built-in integrated circuitry for producing pure tones at 500, 1000, 2000, and 4000 Hz, at loudness levels of 20, 25, and 40 dB, as well as a halogen fiber optic otoscope for otoscopic examination. The cost of this apparatus is approximately US $550.

The patient sits with an elbow propped on the armrest of a chair and with the hand in the form of a gentle fist.²⁵ The patient is then instructed that he or she will hear faint tones of different pitches and should raise a finger as soon as a tone is heard and then lower the finger as soon as the tone is no longer heard. The patient should repeat back the instructions to ensure that they have been completely understood. Patients who are not able to respond with arm or finger movements due to physical disability should be instructed to answer yes when they hear the test tone.

The patient's ear canal and tympanic membrane should first be visualized with an appropriately fitting speculum attached to the audioscope. The speculum should be inserted into the ear canal to get a tight seal between the speculum and the ear canal. The audioscope is then used to present pure tones of random loudness (in decibels) to prevent patients from anticipating the loudness of the next presented tone (each tone has an on-time of 1.5 seconds and an off-time of 1.5 seconds). The results for that ear should be recorded and the contralateral ear should be tested in the same fashion.²⁶ The audioscope screening can be performed in less than 90 seconds in office settings.

A structured MEDLINE and EMBASE database search including the years 1966 through April 2005 was conducted to identify English-language articles examining the accuracy or precision of bedside screening tests for hearing impairment. Medical Subject Headings or keywords used in the search included hearing loss, hearing handicap, hearing tests, tuning fork, deafness, physical examination, sensitivity, specificity, audiometry, tuning fork tests, Rinne, Weber, audioscope, Hearing Handicap Inventory for the Elderly–Screening version, whispered voice test, sensorineural, and conductive. This computerized search was supplemented with a manual review of the bibliographies of all identified articles, additional “core” articles (identified a priori as articles used to develop recent guidelines for the screening of elderly patients for hearing impairment), a commonly used clinical skills textbook,²⁷ and contact with experts in the field.

One of the authors (A.B.) screened all potential articles and then reviewed and abstracted data from all articles that were identified as relevant. A second author (P.T.) independently reviewed and abstracted data from the same articles. Both authors together reviewed these extracted data for inclusion; differences were resolved by consensus. Articles were included if they were original studies on the accuracy and/or precision of bedside screening questions or physical examination maneuvers for hearing impairment. Studies on screening of both symptomatic and asymptomatic individuals were included. Articles were excluded if they evaluated individuals younger than 16 years, used another screening test evaluated in this article (eg, audioscope) as a reference standard, or contained insufficient or incomplete data to allow calculation of likelihood ratios (LRs).

Study quality was assigned based on a grading scheme previously used for this series.²⁸ Level 1 studies were blind independent comparisons of a test with a valid reference standard in a large number (≥200) of consecutive patients. Level 2 studies were similar to level 1 studies, but with fewer than 200 patients. Level 3 studies were also blind independent comparisons of the test with a reference standard, but the patients were enrolled in a nonconsecutive fashion using a subset or smaller group who may have had the condition, and the studies generated results on both the test and the reference standard. Level 4 studies were nonindependent comparisons of a test with a valid reference standard among a “grab” sample of patients believed to have the condition in question. Level 5 studies were nonindependent comparisons of a test with a reference standard of uncertain validity.

We included studies of the self-reported screening question and HHIE-S that were level 1 or level 2 in quality. For other screening tests, we included the best available studies, which consisted of several studies having quality levels of 3 or 4.

All analyses were performed using Comprehensive Meta-analysis version 2.023 (Biostat Inc, Englewood, NJ). We used the raw data in the articles to calculate the LRs and diagnostic odds ratios (DORs) associated with test results. Likelihood ratios are a method of converting pretest probability into posttest probability.²⁹ A positive LR is the ratio of the chance of an abnormal test result in people who have the disease condition relative to people who do not have the disease. Similarly, a negative LR is the ratio of the chance of a normal test result in people who have the disease condition relative to people who do not have the disease. The pretest probability is the prevalence of hearing impairment in the general adult population. For a hearing loss prevalence or pretest probability of 25%, a positive LR of 10 raises the posttest probability to 77%, a positive LR of 5 raises it to 63%, and a positive LR of 3 raises it to 50%. Similarly, negative LRs of 0.15 and 0.10 lower the posttest probabilities to 5% and 3%, respectively.

The DORs are a global measure of test performance that tell examiners how likely they are to correctly identify hearing impairment or normal hearing. Several studies determined test characteristics (sensitivity, specificity, positive and negative LRs) for each test against varying hearing impairment definitions of the reference standard. For hearing impairment definition at a particular intensity level (in decibels), the testing frequency that provided the best DOR was included in the accuracy tables presented herein and used to calculate the summary LRs.

When there were at least 2 studies with similar definitions of hearing impairment and similar screening tests, we used random-effects measures to calculate the summary LRs. For most studies, despite statistical heterogeneity, the confidence interval around the point estimate for the LR was sufficiently narrow that we could make appropriate inferences about the usefulness of the test.

Most studies reported precision (intraobserver and interobserver) as either a κ statistic or Pearson r coefficient. There were insufficient data for each test to combine results and provide summary precision scores.

A total of 924 studies were identified by the search strategy, of which 24 met the inclusion and exclusion criteria (Table 1). The included studies accounted for a total of 12 645 patients. The studies were published between 1966 and 2004 and involved patients both with and without ear symptoms who were seen at outpatient otolaryngology or primary care clinics as well as geriatric medicine inpatient settings.

Most studies used pure-tone thresholds from an audiometer alone or in combination with speech reception and speech recognition thresholds as the reference standard; however, the case definitions differed. The differences were in both the tone frequencies used for testing and the decibel threshold used to classify someone as hearing impaired. For most tests, the testing frequencies used had small nonsignificant effects on the calculated LRs. Therefore, in studies that evaluated the screening test against varying hearing test frequencies, only the frequency definition that provided the best DOR is presented in the corresponding screening-test table and used to calculate the likelihood summary estimate.

No studies reported the precision of the self-reported single-question screening test, though it seems likely that patients would provide consistent results given the test's simplicity and ease of administration. The HHIE-S is a standardized test with preset questions that is assumed to have high reproducibility and precision. The HHIE-S performed at 6-week intervals displays high test-retest reliability (Pearson r = 0.84 when completed on paper and r = 0.96 when conducted face-to-face⁵⁰ ). Similarly, the HHIE-S performed at a physician's office and then repeated at a hearing center shows high test-retest reliability (r = 0.84, P<.001).⁴⁰

No studies evaluating the precision of the Weber test were identified. The reproducibility of the Weber test may be limited by the lack of a standardized force used to strike the tuning fork, the tuning fork frequency used, and the precise location at which the base of the fork is placed.

Only 1 study evaluated the reliability of the Rinne test. Burkey et al⁴⁸ showed that the sensitivity improved considerably when the test was performed by an otolaryngologist as compared to an otolaryngology postgraduate trainee. The variability in the test accuracy when performed by 2 different examiners on the same patient population suggests that the test may not be reliably reproduced. Furthermore, the variability in other test factors similar to the Weber test limits the expected reproducibility.

Several studies examined the precision of the whispered-voice test. Macphee et al³⁵ compared the results of a geriatrician and an otolaryngologist and found an agreement of 88%, while Uhlmann et al⁵¹ compared the results of an audiologist with those of an otolaryngologist and found a correlation of 0.67. However, in the study by Eekhof et al,³⁶ in which the results of 6 examiners were compared with those of the first examiner, the interobserver reliability (measured by the Cohen κ) ranged from 0.16 (poor) to 1.0 (perfect). The authors attributed the large variation between examiners' outcomes to the differences in loudness of the whispering. A systematic review of the whispered-voice test⁵² reinforced the fact that standardizing the loudness of the whisper is the greatest challenge. Few studies state that the whispered sequence occurred after a full expiration. In addition, they suggest that the most appropriate letters, numbers, or words for testing also need further investigation. Furthermore, presbycusis in older patients results in difficulty hearing high-frequency sounds such as consonants; hence, using different consonants and vowels could also alter the reproducibility of the test.⁵³

Two studies examined the precision of the audioscope. In the study by Lichtenstein et al,⁴⁰ in which the results of the audiologist in the hearing center were compared with those of the internist in the office, the interobserver reliability measured by the κ score ranged from 0.41 (poor) at 500 Hz to 0.74 (very good) at 2000 Hz. Bienvenue et al²⁵ showed that the correlations between repeated audiometric screenings taken about one-half hour to 1 hour apart were 0.996 at 500 Hz, 0.988 at 1000 Hz, 0.998 at 2000 Hz, and 0.989 at 4000 Hz. Since these correlations were quite high, the authors concluded that the results of the audioscope test are reliable.

Self-reported Screening Question. Six studies assessed the accuracy of a self-reported screening question against a reference standard of pure-tone average threshold across a broad range of patients (Table 2). The studies were similar in design but differed slightly in the wording of the self-reported question. The screening question is only moderately useful to detect subtle hearing impairment (≥25 dB). Patients who report difficulty hearing have a summary LR of 2.2 (95% confidence interval [CI], 1.8-2.8), while those who claim no problem have a summary LR of 0.45 (95% CI, 0.36-0.56). Patients who report normal hearing are much less likely to have moderate to severe hearing impairment (≥40 dB), with a summary LR of 0.13 (95% CI, 0.09-0.19). A family member's assertion that the patient has difficulty hearing may be as important as, or more important than, the patient's own recognition of hearing loss. Although we found no evidence that addresses this issue in the studies we reviewed, we suggest that clinicians should decide individually on how to incorporate a family member's perception of their relative's decreased hearing when deciding on formal audiology testing.

The HHIE-S. Seven studies met our study criteria, and all used pure-tone average as the reference standard (Table 3). The presence of hearing handicap (ie, an HHIE-S score >8) increases the probability of a hearing impairment of at least 40 dB (summary LR, 3.8; 95% CI, 3.0-4.8). However, a more severe hearing handicap (ie, an HHIE-S score >24) does not further improve the probability of detecting hearing impairment (summary LR, 4.0; 95% CI, 2.6-6.2). The absence of a hearing handicap, defined as an HHIE-S score of 8 or less, marginally lowers the probability of hearing impairment (summary LR, 0.38; 95% CI, 0.29-0.51).

Weber Tuning Fork Test. In the only study of the Weber test that met the study criteria,⁴⁴ the reference standard test and definition of hearing impairment were not explicitly described in terms of air-bone thresholds. Neither the tests performed using the 256-Hz tuning fork nor those performed using the 512-Hz tuning fork are useful to either increase or decrease the probability of identifying individuals with unilateral sensorineural or unilateral conductive hearing loss. An abnormal Weber test result has an LR of only 1.6 (95% CI, 1.0-2.3) to 1.7 (95% CI, 1.0-2.9), while a normal result lowers the probability, with an LR of 0.70 (95% CI, 0.48-1.0) to 0.76 (95% CI, 0.57-1.0), making it an inaccurate test for screening purposes.

Rinne Tuning Fork Test. Although numerous studies have evaluated the Rinne tuning fork test for assessing conductive hearing loss, only 5 studies met the study criteria (Table 4). All 5 studies were of lower quality (level 3 or 4). Although Chole et al⁴⁶ found better characteristics with the 512-Hz tuning fork, most studies found that a 256-Hz tuning fork is more accurate either alone or in combination with other tuning forks.^{43 - 44 ,49 ,54 - 55} The reference standard was air-bone threshold gaps determined using pure-tone audiometry. Variability in the degree of air-bone threshold gaps used to define conductive hearing impairment prohibits the combination of data across studies to calculate summary estimates. An abnormal Rinne test result increases the probability of conductive hearing loss but has a wide range of LRs (2.7 [95% CI, 2.0-3.5] to 62 [95% CI, 3.9-970]), with a majority of the studies showing LRs greater than 15. On the other hand, a normal Rinne test result may be less useful in dismissing hearing impairment, as indicated by the LR range from 0.01 (95% CI, 0-0.15) to 0.85 (95% CI, 0.76-0.95), with the majority of the LRs greater than 0.30. Several studies have attempted to report the minimal air-bone threshold gap required for the Rinne test result to change from normal to abnormal (results range from from 17 dB to 40 dB^{23 ,49 ,54 ,56 - 59} ). Again, the large variance in the results precludes using the Rinne test as an accurate screening tool.

Whispered-Voice Test. Four studies met our study criteria (Table 5). Pure-tone threshold audiometry was the reference standard in all 4 studies. The prevalence of hearing impairment ranged from 26% to 61%. Inability to repeat back a letter/number combination whispered at a distance of 2 ft (0.6 m) increases the probability of a 30-dB or greater hearing impairment (summary LR, 6.1; 95% CI, 4.5-8.4). However, normal perception of the whispered voice dramatically lowers the likelihood of hearing impairment (summary LR, 0.03; 95% CI, 0-0.24). Inability to perceive the whispered voice when the test is performed at 6 in (15 cm) from the patient, instead of the more commonly used distance of 2 ft, sharply increases the likelihood of hearing impairment (LR, 67; 95% CI, 4.3-1062); however, accurate perception of the whispered voice at 6 in only marginally reduces the likelihood of hearing impairment (LR, 0.27; 95% CI, 0.19-0.39).³⁵ Thus, screening for hearing loss with the whispered voice test at 6 in will more likely fail to detect affected patients than will performing the test at 2 ft.

Six studies evaluating the accuracy of the audioscope test were identified (Table 6). Three of the studies were quality level 1 or 2, while the remaining 3 studies were level 3 or 4. The prevalence of hearing impairment ranged from 26% to 69%. Four studies^{8 ,36 ,40 ,42} used a 40-dB pure-tone average threshold (reference standard), while the remaining 2 studies used a 30-dB²⁵ and a 45-dB⁴⁵ threshold, respectively.

Despite slight differences in the definition of hearing impairment, the sensitivity to detect hearing impairment was consistently high, ranging from 87% to 100%. On the other hand, the specificity was variable, ranging from 42% to 90%. Thus, normal hearing perception on audioscope screening makes hearing impairment very unlikely (summary LR, 0.07; 95% CI, 0.03-0.17). Patients with abnormal hearing perception on audiocope screening have a moderately increased probability of hearing impairment (summary LR, 2.4; 95% CI, 1.4-4.1).

Given this women's age, her pretest probability of hearing impairment is between 40% and 66%. Her yes reply to the question as to whether she feels she has hearing impairment indicates that she is symptomatic and hence increases the probability of her having significant hearing impairment (summary LR, 2.5; 95% CI, 1.7-3.6) to more than 80%. This patient should be referred both for formal audiological testing to determine the degree and type of hearing impairment and for planning for hearing amplification. No further bedside testing is required in this patient, since even a pass on the whispered-voice test (summary LR, 0.03; 95% CI, 0-0.24) only lowers the posttest probability to approximately 10%.

If we assume this patient's pretest probability for hearing impairment is approximately 50%, his no response to the single screening question (LR, 0.45) only reduces his chance of greater than 25-dB impairment to about 30%. Since the single question has not confidently ruled out hearing impairment, a screening whispered-voice test will be very useful in this patient. A passing result on this test makes hearing impairment much less likely (summary LR, 0.03; 95% CI, 0-0.24), and no further testing will be required. However, not passing the test suggests hearing impairment (summary LR, 6.1; 95% CI, 4.5-8.4), and the patient should then be referred for formal audiometric testing. Although the audioscope and whispered-voice tests both have similar accuracy test characteristics, the whispered-voice test has a better DOR compared with the audioscope test. Clinicians who are concerned about the reliability of using their own whispered voice might opt to screen with an audioscope.

To screen elderly individuals for hearing impairment in general practice, they should first be asked whether they feel they have hearing impairment (Figure 2). Patients who provide a yes or equivocal response should be referred directly for formal audiometric testing. If the reply is no, then they should be further screened with a whispered-voice test. Accurate perception of a whispered letter/number combination significantly lowers the likelihood of hearing impairment, and these individuals do not require further evaluation. Patients unable to perceive the whispered combination should be referred for formal testing. The audioscope has diagnostic accuracy characteristics similar to those of the whispered-voice test and may be preferable to clinicians who are unsure of their ability to perform the whispered-voice test reliably. Further research is required to improve standardization of the technique of conducting the whispered-voice test, including ensuring that it is performed after full expiration and determining a set of letters and numbers that will most reliably screen for hearing impairment. Both the Rinne and Weber tuning fork tests, which are limited by inaccuracy and lack of precision data, are not recommended for routine screening and should no longer be part of the medical curriculum.