Screening for handicapping hearing loss in the elderly
TABLE 3
Sensitivity and specificity for the HHIE-S and the global question, “Do you have a hearing problem now?” in identifying people with hearing loss
| Referred, % | Sensitivity, % | Specificity, % | LR+ | LR– | PPV, % | NPV, % | |
|---|---|---|---|---|---|---|---|
| HHIE-S* | 15.2 | 36 | 92 | 4.7 | 0.70 | 63 | 80 |
| Global Question | 39.5 | 71 | 72 | 2.5 | 0.40 | 48 | 87 |
| Both positive† | 14.2 | 34 | 93 | 5.0 | 0.71 | 65 | 79 |
| Both negative† | 40.4 | 72 | 71 | 2.5 | 0.39 | 48 | 87 |
| *Cutoff score of 0–8 vs 10. | |||||||
| †See text for a detailed description of “both positive” and “both negative.” | |||||||
| HHIE-S, Hearing Handicap Inventory for the Elderly–Screening; LR+, positive likelihood ratio; LR–, negative likelihood ratio; NPV, negative predictive value (percentage with a negative screening test who did not have hearing loss); PPV, positive predictive value (percentage with a positive screening test who had hearing loss). | |||||||
Discussion
Screening for any disorder attempts to increase the likelihood that people with the disorder will be identified (sensitivity) and exclude those without the disorder (specificity). In practice, not all cases will be identified by screening (false negatives), and some people without the disorder will be incorrectly labeled (false positives). The more sensitive the screening method to the presence of the disorder, the greater the probability of false-positive results. Thus, there is an inherent and unavoidable tradeoff between sensitivity and specificity.
The goal of the screening program dictates the approach to managing this tradeoff. From our perspective, the goal of hearing screening in the elderly is to identify people likely to benefit materially from amplification. The current data suggested a clear choice. The global measure was considerably more sensitive (71%) than the HHIE-S (36%) for detecting the criterion handicapping hearing loss, but would have over-referred more false-positive cases (28%) than the HHIE-S (8%).
The global question method would nearly double the capture rate of the screening process at the cost of a 20% difference in over-referral. Given that many of the over-referral cases will have some degree of hearing loss, albeit less than the criterion, that some will have central auditory dysfunction (where speech understanding is poorer that would be predicted by the hearing threshold criterion), and that all would likely benefit from evaluation and counseling, this apparent over-referral rate does not seem objectionable.
Combining both screening measures, although intuitively attractive, proved to be counterproductive and arguably not worth the extra effort to administer and score the instrument. The anomaly whereby combining the strengths of both approaches was not fruitful can be attributed to the nonlinear association of HHIE-S scores and hearing level: many people with high HHIE-S scores had good hearing and vice-versa. This suggests over-concern, on the one hand, and denial, on the other. For the group of people who deny their hearing loss on the single question or the HHIE-S, referral cases can be based on the clinical examination or the families’ or caregivers’ comments and concerns.13
This report specifically excluded people with hearing aids because the purpose of the instrument is to identify people with unrecognized hearing loss.
Conclusions
Based on this report, we recommend using the question, “Do you have a hearing problem now?” as a global measure on the intake or annual history form for geriatric practices. Others have found high sensitivity for the single history question.7,14 A positive response to this question in this population identified all the people with the criterion hearing loss who responded to the highest probability HHIE-S category (from 26 to 40)5 and 95% of the people in the middle category (from 12 to 24). Moreover, 40% of respondents in the lowest probability HHIE-S category (from 0 to 8) who responded yes to the global question had a criterion hearing loss that would not have been identified by the HHIE-S.
Acknowledgments
Aimee Verrall assisted with data management and manuscript preparation.
Corresponding address
George A. Gates, MD, Virginia Merrill Bloedel Hearing Research Center, University of Washington 357923, Seattle, WA 98195-7923.
ggates@u.washington.edu.
