Assessing hearing loss in older adults with a single question and person characteristics; Comparison with pure tone audiometry in the Rotterdam Study

Hearing loss (HL) is a frequent problem among the elderly and has been studied in many cohort studies. However, pure tone audiometry-the gold standard-is rather time-consuming and costly for large population-based studies. We have investigated if self-reported hearing loss, using a multiple choice question, can be used to assess HL in absence of pure tone audiometry. This study was performed within 4,906 participants of the Rotterdam Study. The question (in Dutch) that was investigated was: 'Do you have any difficulty with your hearing (without hearing aids)?'. The answer options were: 'never', 'sometimes', 'often' and 'daily'. Mild hearing loss or worse was defined as PTA0.5-4(Pure Tone Average 0.5, 1, 2 & 4 kHz) ≥20dBHL and moderate HL or worse as ≥35dBHL. A univariable linear regression model was fitted with the PTA0.5-4 and the answer to the question. Subsequently, sex, age and education were added in a multivariable linear regression model. The ability of the question to classify HL, accounting for sex, age and education, was explored through logistic regression models creating prediction estimates, which were plotted in ROC curves. The variance explained (R2) by the univariable regression was 0.37, which increased substantially after adding age (R2 = 0.60). The addition of sex and educational level, however, did not alter the R2 (0.61). The ability of the question to classify hearing loss, reflected in the area under the curve (AUC), was 0.70 (95% CI 0.68, 0.71) for mild hearing loss or worse and 0.86 (95% CI 0.85, 0.87) for moderate hearing loss or worse. The AUC increased substantially when sex, education and age were taken into account (AUC mild HL: 0.73 (95%CI 0.71, 0.75); moderate HL 0.90 (95%CI 0.89, 0.91)). Self-reported hearing loss using a single question has a good ability to detect hearing loss in older adults, especially when age is accounted for. A single question cannot substitute audiometry, but it can assess hearing loss on a population level with reasonable accuracy.