Objectives: We developed a prediction tool to assist in evaluation of pediatric candidates for cochlear implantation (CI) and to help plan for preoperative and postoperative support.
Methods: Between 1995 and 2005, 277 patients underwent CI at Children's Hospital Boston. Of these 277 patients, 250 had at least 2 years of post-CI follow-up and adequate pre-CI information for rating by our prediction tool. Of the 250, 106 were randomly selected for inclusion. The patients were divided into group A (auditory/oral communicator); group B (auditory/oral communicator with visual assistance), group C (visual/manual communicator with auditory/oral skills assistance), and group D (will not derive communicative benefit from implant). Predictions were performed with clinical assessment and two statistical techniques: regression modeling and classification and regression tree (CART) analysis.
Results: Among patients who became auditory/oral communicators (group A), clinical assessment predicted that outcome accurately 65% of the time, CART analysis had intermediate sensitivity (79%), and regression modeling was the most sensitive (95%). Groups B through D were predicted 45% of the time by regression modeling, 90% of the time by clinical assessment, and 100% of the time by CART analysis.
Conclusions: A combination of speech-language, medical, and educational constructs can provide a reliable prediction of the communication outcome. Our goal for the prognosis tool is to make it part of the overall candidacy process in supporting decision-making about CI and planning for post-CI therapy.