Precision Medicine in Cochlear Implantation: Leveraging Artificial Intelligence and Novel Variables for Performance. - PROJECT SUMMARY / ABSTRACT The goal of the proposed research is to develop and validate a reliable clinical decision support tool (e.g. prediction model) for early risk stratification for cochlear implant (CI) speech perception performance. Hearing loss (HL) is the fourth most common disability worldwide, it considerably affects a patient’s quality of life (QOL), directly linked to cognitive decline, and it poses an enormous financial burden to society (~$980 billion/year).1 CI has become the standard of care for patients that no longer benefit from hearing aids. However, despite the enormous success of CIs, there is a wide inter-individual variability with respect to clinical and biological features and their impact on post-implantation speech perception performance. The heterogeneity in CI performance outcomes is a pervasive challenge for clinicians and families, as the management of poor CI performers is largely reactive, without reliable clinical decision support tools for early risk stratification and/or prognostic assessment at time of CI. If available, such predictive models would provide critical opportunity for risk stratification to improve patient counseling, recommendations for auditory rehabilitation, considerations of device-related underperformance (e.g., mapping, troubleshooting, early device failure), and ultimately future CI clinical trials. There is a critical need for better clinical decision support tools for early risk stratification and prognostication of CI performance. Traditional approaches have shown limited success because we often focus on one or two clinical or biological markers to stratify CI performance trajectory, hindering progress in the development of precision medicine delivery. Second, traditional approaches have shown limited success because current preoperative factors can only account for 10-20% of the observed variance.3,4 In response to these challenges, we propose to develop models to support clinical decision making by (1) integrating novel biomarkers that better account for the variability in CI performance including electrocochleography (ECochG), patient comorbidities, and social determinants of health; and (2) utilizing artificial intelligence (AI)-based approach to analyze how multiple clinical and biological markers interact, are associated with sub-phenotypes of CI performance, and can be used to develop superior decision support tools / prediction models. My long-term goal is to be an independent physician-scientist and deploy innovative and autonomous AI-based decision support tools that can seamlessly integrate multiple variables to improve delivery of precision medicine in CI patient care. This K23 will provide me with the protected time required to develop unique skillsets in AI-based techniques in healthcare, quantitatively compare to modern regression techniques to understand translational clinical value gained, and mixed methods research to improve implementation of AI-based precision medicine models. The proposed research program addresses the mission of the NIDCD Auditory Strategy Themes 3 through 6: to promote precision medicine approaches, to translate and implement scientific advances, to facilitate best practices in biomedical data science, and to harness advanced technology.5
