Objective: In otitis media with effusion (OME), the accuracy of predicting air-conduction hearing-level (HLs) from tympanometry has generally been seen as too poor for use in clinical practice. Previous studies of the relationship have mostly concerned single ears, many using samples with predominantly mild cases of OM and weak statistical approaches. A better understanding of the interrelations between these tests might improve efficiency in testing and decision-making for individuals.
Method: Binaural average HL was adopted as the measure to be predicted most relevant to auditory disability. Multiple regression from modified Jerger tympanogram categories B, C2, C1 and A tympanogram types on 3085 children aged 3(1/4)-6(3/4) years gave formulae which we tested for replication, stability and generalization across distributions differing in severity.
Results: Age-adjusted formulae explained up to 49% of the variance in binaural HL (i.e. a multiple correlation of 0.70), and were robust across phase of disease. Best predictions were seen in a severe sample permitting exploitation of the strong conditioning effect by a B tympanogram in one ear upon the tympanometry/HL relationship in the other. This permits a trichotomous approximation (0, 1, or 2 B-tympanograms) to also perform well.
Conclusions: We name the HL prediction formula "ACET" - Air Conduction Estimated from Tympanometry. We do not recommend replacing audiometry with tympanometry, particularly not at first assessment. However, where the diagnosis is, or likely from history to be, OME (even if fluid is absent on test day), the informativeness of further air-conduction audiometry on the same or later occasion may not always be worth the further effort or cost. It is therefore clinically useful to have a dB measure, from an evidence-based formula justifying a principled estimate. Non-clinical uses include imputation when research data are missing, and non-intensive applications where audiometry is impracticable, e.g. field clinics and large scale or longitudinal research. A companion paper shows how the part of the air-conduction HL variance that is not explicable by ACET, also offers a surrogate, but for bone-conduction HL (BC), where BC testing may be problematic, as in the very young. This surrogate can also define cases needing true BC testing.