For participant level analysis, ETDQ-7 and PEq5 scores were compared between symptomatic and control groups using two-sample Wilcoxon test (proc npar1way, SAS v.9.4, Cary, NC). Exact 95% Clopper-Pearson confidence intervals (CI) were obtained for sensitivity and specificity (proc freq, SAS, v. 9.4) and Fisher exact test was used for comparing sensitivity or specificity between different subgroups of patients.
Kendall’s correlation coefficient was used to test association between the PEq5 and ETDQ-7 scores (proc corr, SAS, v.9.4). Association of ETDQ-7 scores with the inadequate PEq5<60% as well as with group assignment were assessed using empirical Area under the ROC curves (AUC).23 (link),24 At ear-level data, statistical analysis (including evaluating Kendall’s coefficient and AUC) was performed using non-parametric bootstrap CI for clustered data with participant as a resampling unit, based on 10,000 bootstrap samples.25