Multivariable logistic regression model for the prediction of tinnitus recovery following unilateral cochlear implantation (in the pooled dataset and in the original dataset as sensitivity analysis)
| Predictor | Pooled dataset (15 multiple imputed sets) (n=87) | Original dataset (complete cases) (n=76) | ||||
| OR (adjusted OR) | 95% CI | P values | OR | 95% CI | P values | |
| Preoperative CVC score | 0.978 (0.983) | 0.958 to 0.999 | 0.038 | 0.978 | 0.957 to 0.999 | 0.042 |
| Bilateral tinnitus preoperative | 0.412 (0.501) | 0.151 to 1.124 | 0.083 | 0.490 | 0.171 to 1.402 | 0.184 |
| Difference audiometry at 250 Hz | 1.024 (1.019) | 1.004 to 1.044 | 0.017 | 1.024 | 1.005 to 1.044 | 0.013 |
OR >1: in favour of tinnitus recovery.
Prediction rule of the pooled dataset after internal validation: linear predictor=0247-(0.017*preoperative CVC score)-(0.691*bilateral tinnitus)+(0.019*difference in hearing threshold at 250 Hz).
Adjusted OR: OR corrected for overoptimism after internal validation; CVC, Consonant-Vowel-Consonant test.