Skip to main content

Table 2 Mediation of the cohort difference in low frequency hearing thresholds (0.5, 1 and 2 kHz)

From: Explaining better hearing in Norway: a comparison of two cohorts 20 years apart - the HUNT study

  Cohort effecta Natural indirect effect
Estimate (dB) 95% CI Estimate (dB) 95% CI Mediated proportion
All:
 Total −2.79 − 2.64, − 2.94 − 0.76b − 0.85, − 0.68 0.27
 Education    −0.28c − 0.35, − 0.21 0.10
 Recurrent ear infections    − 0.18c −0.21, − 0.15 0.07
 Occupational noise    −0.14e − 0.18, − 0.10 0.05
 Smoking    −0.14e −0.21, − 0.08 0.05
Women:
 Total −2.65 −2.45, − 2.84 −0.70b − 0.81, − 0.59 0.26
 Education    −0.38c −0.48, − 0.27 0.14
 Recurrent ear infections    −0.20c −0.25, − 0.16 0.08
 Occupational noise    −0.04d −0.05, 0.02 0.01
 Smoking    −0.05d −0.09, − 0.02 0.02
Men:
 Total −3.01 −2.78, −3.23 −0.91b −1.06, − 0.77 0.30
 Education    −0.22c −0.31, − 0.13 0.09
 Recurrent ear infections    −0.17c −0.21, − 0.12 0.06
 Occupational noise    −0.27e −0.34, − 0.19 0.09
 Smoking    −0.21e − 0.33, − 0.08 0.07
  1. All models were adjusted for age and sex
  2. a Difference in hearing thresholds between HUNT4 (2017–2018) and HUNT2 (1996–1998)
  3. b Estimated by fitting natural effect models in the R-package medflex using the imputation method including all exposure-mediation interaction terms
  4. c Estimated with gformula in Stata
  5. d Estimated with gformula in Stata with the assumption of no exposure mediation interaction as proposed by Robins and Greenland [32]
  6. e Estimated with gformula in Stata with the assumption of no exposure intermediate interaction together with only linear effects of the intermediate variable as proposed by Petersen et al. [33]
\