Accuracy and variability in clinical predictions of speech recognition outcomes for cochlear implant users

An improvement in speech perception is a major well-documented benefit of cochlear implantation (CI), which is commonly discussed with CI candidates to set expectations. However, a large variability exists in speech perception outcomes. We evaluated the accuracy of clinical predictions of post-CI speech perception scores.

In an online survey, clinicians involved in CI care were asked to predict 6-month performance on AzBio sentences and CNC words for 15 anonymised CI patients based on demographic, medical, and audiological data, and preoperative speech recognition scores.

Sixty-five participants (72% audiologists, 14% neurotologists, 14% speech-language pathologists, and CI researchers) provided predictions for two or more cases, and 43 provided predictions for all 15 CI cases.

Clinical predictions had low accuracy and high variability (i.e. median interclass correlation coefficients [ICC] 0.11-0.13). No relationship was observed between clinicians' confidence and prediction accuracy.

The present findings are consistent with previous research and indicate a concerning discrepancy between clinical predictions and actual speech perception outcomes for adult post-lingual CI users. Future research may explore incorporating predictions based on statistical models into speech outcome prognostication to reduce variance and improve counselling and expectations setting for individual CI candidates.