Considering the age of 60 years as the dividing line between elderly and non-elderly, we observed based on IBGE data (Brazilian Institute of Geography and Statistics) that there has been percentage increase in number of elderly people (60 years and over) in Brazil, from 5.07% in 1970 to 8.56% in 20001. The number of elderly patients went up from 3 million in 1960 to 7 million in 1975, to 14 million in 2002 (a 500% increase within 40 years). It is estimated that it will reach 32 million by 20202. If there are further advances in mortality drop at advanced ages, this process will be even further enhanced1.

Among pathologies whose frequency increases with aging, we can include those related with inner ear. Hearing loss is the third most prevalent chronic condition among American elderly patients, third to hypertension and arthritis3. In Brazil, studies have shown a prevalence of hearing loss in the elderly that ranges from 20 to 85%4–7. Hearing loss has been associated with negative psychosocial impact8, with inability to perform heavy home chores9 and increase in occupational accidents10. Auditory sensorial loss not corrected by hearing aids is associated with loss of self-sufficiency in daily living activities and social relationship impairment in the elderly, plus increased mortality in males11. The wish to wear a hearing aid is not associated with severity of loss, but rather with functional status of the subject, being higher among more independent subjects8.

Despite the high prevalence among elderly patients, hearing loss is one of the problems not investigated during routine medical examination in this age range12. Screening can be useful in identifying primary health care, given that the onset is insidious and that patients are frequently unaware of it. Audiometry is the gold standard test, but its conduction is sometimes hindered in some regions owing to problems of access, reference and reimbursement. Thus, many clinicians rely on self-administered questionnaires13. As to research studies, large-scale trials of auditory status may provide clues on temporal trends of prevalence of hearing loss and contribute to the identification in geographical areas and subgroups of risk populations (gender, race and ethnicity). These investigations may be a quick and cheap way of providing estimates to large populations, in which expenses (audiometric equipment and trained personnel) and time restrictions are a concern14. Generic isolated issues on hearing have also been used in epidemiological studies. Self-report can also be an indicator of hearing loss and it is quick and cheap to administer15.

Considering that audiometry is the gold standard for hearing loss detection and requires trained team, soundproof booth and equipment, hindering its execution in large scale, our final goal was to determine whether one single global question would be valid to be used in epidemiological research studies. This paper aimed at reviewing studies whose purposes were to compare results obtained with use of isolated questions and results obtained through pure tone audiometry in assessment of hearing loss in the elderly.

The present study aimed at comparing the prevalence estimates of hearing loss obtained using two methods: self-report and audiometry. According to the analysis of the studied literature, the single global question seems to be an acceptable indicator of hearing loss which is sensitive and reasonably specific, especially when the loss is identified as being pure tone mean up to frequencies 2 or 4 kHz, at 40 dBHL on the best ear.

The comparison of hearing loss prevalence among the studies is hindered by the differences in investigated populations and in audiometric criteria used for its definition14,20. Prevalence of hearing loss is smaller in women and younger people20. Moreover, elderly people selected from a nursing home23 or hospital setting8 represented a fragile selected group, presenting more comorbidities than healthy elderly people from the general population, leading to increase in prevalence of hearing loss in comparison to population-based studies8,20. The discrepant results found by Hashimoto et al. (2004)24 and the other studies may be related to the population studied by the former, which comprised industry workers. Therefore, part of the studied people could have preferred to say they did not have a hearing loss in fear of losing their jobs, malingering the audiometric performance. In addition, there is also the possibility that the results had been affected by the healthy-worker effect, even though the prevalence had been comparable to that of other Japanese studies24.

Definitions of hearing loss assessed by pure tone mean in the literature range according to the ear used to classify the subject (i.e., best, worst, both, right or left ear) and to frequencies included in the pure tone mean to determine the best and the worst ear22. The use of worst ear to define hearing loss results increased prevalence14. Inclusion of frequency of 4 kHz has also increased prevalence of hearing loss14,23 given that frequencies 0.5 to 2 kHz are considered important in the identification of the disability related with hearing, frequencies 3 and 4 kHz are among the first ones to show hearing decrease associated with age and they are important for speech understanding, especially in noisy environment20. Even though many different definitions of hearing loss had been proposed, there is no universal acceptance, and stricter definition (low intensity as cut-off point, such as 25 dBHL) leads to higher prevalence of hearing loss19.

As to self-report of hearing loss comparing to pure tone audiometry classification criteria, the use of the best ear is justified by the fact that the worst ear tends to be compensated by the function of the side that has the best subjective perception24. Self-perception of hearing loss seems to be in agreement with pure tone mean of medium frequencies (1, 2, 3 and 4 kHz)16 and it is more frequent in subjects with moderately-severe hearing loss than among those with mild loss8,23.

As to questions used, multiple-choice questions14,24 presented lower sensitivity when compared to two-option questions. The type of question used by Hashimoto et al. (2004)24 may have caused confusion given that the answer “same as before” may have been used to mean “no hearing problem since then”, even though when they classified “same as before” as having no hearing loss, there were no statistically significant differences in sensitivity and specificity of the question. Moreover, even though two-option questions had been similar one to the other, such as those used by Gates et al. (199019 and 200313), they may have given the impression to respondents that they were more serious difficulties, which could have reduced the estimated prevalence of hearing loss14.

Voeks et al. (1993)23 considered mistaken the response “I don't know” as a positive indication of hearing loss because they consider that this strategy conveys more sensitivity to the question, whereas other studies considered them as missing values14,15. However, in the group studied by Voeks et al. (1993)23 there were some subjects with cognitive deficit, which could have hindered the identification through a questionnaire and reduced sensitivity, given that among the three above-referred studies, the lowest sensitivity was reached by Voeks et al. (1993)23. In addition, the study by Voeks et al. (1993)23 detected more wrong responses in the group without hearing loss in the audiometry, increasing the number of false positives and justifying the low specificity found by them. They also referred that the response “yes” conveys certainty about the hearing loss, whereas “I don't know” response conveys a little more than 50% certainty23.

According to Hashimoto, Nomura and Yano (2004)24, simple questions such as “Do you feel you have a hearing loss?” are based on subjective assessment of individual health status, and they may be distorted by other psychosomatic symptoms and mental health status. They observed higher proportion of false positives in those with more than two complaints of somatic symptoms, and wondered whether these patients would not be the most sensitive ones to all somatic sensations - they could have exaggerated about their hearing acuity. Thus, it might be reasonable to say that subjective hearing difficulties without objectively diagnosed impairment may reflect psychosocial problems experienced in daily communication situations at the workplace.

Screening approaches try to increase the likelihood a person with a specific dysfunction has to be identified (sensitivity), excluding those without the dysfunction (specificity). In practice, not all cases are identified (false negative), and some people without dysfunction will be misdiagnosed (false positive)13. Therefore, the more sensitive the test is for hearing loss, the higher the likelihood of false positives, and among those, many might have had some degree of dysfunction, even if they had not reached the specific criteria set by audiometry, benefiting from the referral to complete audiometric assessment13. Hashimoto et al. (2004)24 referred that discrepancy of results in relation to other studies should be related with high number of false negative results. Given that subjects in the study were relatively younger than the other studied subjects and that normal speech is produced at frequencies of about 1 kHz, even if the studied subjects had early signs of sensorineural hearing loss, they would not have been recognized in daily verbal communication. They explained the number of false negative results in the study by the fact that the proportion of subjects that complained of auditory difficulties was smaller in the group of people that had hearing loss only in the frequency of 4 kHz, compared to the group whose affection comprised 1 and 4 kHz. They reported that since audiometry was conducted yearly, they might have experienced some learning bias, and in addition, people might have stated that they had auditory difficulties because they remembered the results of hearing loss detected by previous year audiometry. It may artificially reduce false negative results and actual sensitivity could be even lower24.

As to results of the studies based on gender and age, Gates et al. (1990)19 warned to the possible role of noise exposure as basis for different etiology between men and women, whereas Uchida et al. (2003)22 observed that men tended to underestimate their hearing difficulty more than women. Sindhusake et al. (2001)15 did not find statistically significant differences concerning gender and age, and Nondahl et al (1998)14 did not describe whether the differences found by them had been statistically significant. The greater sensitivity to the question found in the younger group in these two studies14,15, might have been explained by low self-perception or denial of problems faced by the elderly8, considering that there is a popular belief that hearing loss is a normal part of aging and not a health problem that deserves special attention14, making them underestimate their auditory difficulty22.

As to both studies that did not present results of sensitivity, specificity and predictive values, Reuben et al. (1998)21 conducted a study in the 70's and the traditional trend poses a risk to generalization of data on prevalence and validity of the study, because they used a relatively young sample, maximum age of 74 years, which should have reduced the number of subjects with positive screening. Uchida et al. (2003)22 obtained reasonable performance in the question used by them to stratify subjects by hearing loss level.

Gates et al. (1990)19 stated that hearing is poorly described through one single parameter, be it self-report or pure tone thresholds, and it is highly prevalent among elderly people and increases in older groups. Few studies with representative population samples used audiometry as a hearing assessment method. A population-based study allows generalization of results, provided that geographical areas, ethnics and noise exposure levels are the same. Studies based on convenience or clinical samples, with or without estimates derived from other geographical regions, may lead to false results concerning hearing loss. If the purpose of hearing loss measurement is to associate it with other factors, then knowledge of characteristic of mistakes comparing questionnaire and audiometry through validation study is quite convenient.

We should also add that we did not find any Brazilian studies that compared one single question questionnaire to pure tone audiometry, and we do not have estimates that can be generalized to the Brazilian population.

One single global question has good performance to identify elderly people with hearing loss and it may be recommended for epidemiological studies that cannot apply audiometric measurements. Considering that we do not have estimates that can be generalized to the Brazilian population, it would be recommendable to conduct a validation study of one single global question compared to audiometry, allowing the use of this useful tool in Brazilian epidemiological studies, with the advantage of being able to study a large number of subjects.