ABSTRACT Older adults with Alzheimer’s disease and related dementias (ADRD) require care from numerous specialists and clinical teams to manage ADRD-related symptoms and other comorbidities. The majority of patients with ADRD have their healthcare managed by non-specialists who often lack the time, confidence, and expertise to manage ongoing ADRD needs, which leads to significant referral-based care that often suffers from a lack of coordination. Supporting primary care providers in their ongoing management of care for patients with ADRD by promoting deliberate organization of care activities and information sharing among clinical teams is a critical opportunity to limit unintended gaps and ensure that patients with ADRD receive the high-quality multidisciplinary care necessary for long-term wellbeing. Few solutions exist to measure and identify gaps in care coordination. Current approaches primarily rely on single payor claims data to evaluate patient sharing relationships between providers, which neglects to provide granular insight necessary to improve local healthcare delivery. Applying advanced statistical modeling to EHR usage and communication data will provide critical insight into healthcare delivery patterns necessary to accurately model and optimize referral-based care coordination. In the proposed project, I will apply innovative knowledge representation and machine learning to improve referral-based care coordination by developing intelligent approaches that monitor coordination activities and recommend actionable opportunities for improvement. Under the guidance of a multidisciplinary team of mentors, I will receive training to expand my knowledge in healthcare delivery to promote healthy aging, further my knowledge of state-of-the-art machine learning techniques, and will develop a deeper understanding of quantitative approaches to investigate complex sociotechnical systems. I will apply this training to address knowledge gaps related to the formation of referral-based clinical teams in the first two aims: (1) model and identify patterns of collaboration among healthcare providers teams treating patients with ADRD that contribute to improved healthcare delivery; and (2) apply natural language processing to messages sent via patient portal understand how patient and caregiver interactions influence care patterns. In aim 3, I will combine insights and collaboration networks from the first two aims to develop explainable machine learning models to identify optimal patterns of care coordination. I will use these optimal care coordination patterns to highlight features that cause deviation in a patient’s treatment pathway and identify actionable steps for improvement. This career development award will provide the rigorous training and mentorship necessary to become a fully independent principal investigator. The research will benefit from a PI who has a strong background in information science, knowledge representation, and collaborati...