Research using UK Biobank data has shown ethnic inequalities in hearing health; however, the hearing test used may exhibit a disadvantage for non-native language speakers.
To validate the results of the UK Biobank hearing test (Digit Triplet Test, DTT) against self-reported measures of hearing in the dataset and create classifications of hearing health. To observe if language proficiency and migration age have the same effect upon hearing health classification as upon the DTT in isolation. Our hypothesis is that language proficiency acts differently upon the DTT, demonstrating that the DTT is biased for non-native speakers of English.
Latent classes representing profiles of hearing health were identified from the available hearing measures. Factors associated with class membership were tested using multinomial logistic regression models. Ethnicity was defined as i) White, native English-speaking, ii) ethnic minority, arrived in UK aged <12 or iii) ethnic minority, arrived aged =>12.
UK Biobank participants with valid hearing test results and associated covariates (N=151,268)
DTT score, self-reported hearing difficulty, self-reported hearing difficulty in noise and hearing aid use
Three classes of hearing health were found, ‘Normal’, ‘Generally Poor’ and ‘Only Subjectively Poor’. In a model adjusting for known confounders of hearing loss, a poor or insufficient hearing test result was less likely for those with better language (OR 0.69, 0.65-0.74 95% CI) or numerical ability (OR 0.71, 0.67-0.75 95% CI) but more likely for those having migrated aged >= 12 (OR 3.85, 3.64-4.07 95% CI).
The DTT showed evidence of bias, having greater dependence upon language ability and migration age than other hearing indicators. Designers of future surveys and hearing screening applications may wish to consider the limitations of speech-in-noise tests in evaluating hearing acuity for populations that include non-native speakers.