Building and Validating Lifelong Self-Management Capacity with Advanced AI: MyMSMentor - PROJECT SUMMARY In 2018, 51.8% (129 million) of American adults had at least one chronic condition (e.g., diabetes, arthritis). The total cost of chronic conditions is $3.7 trillion; about 84% of healthcare costs are attributed to chronic condition treatment. Self-management is critical to reducing clinical and economic burdens. Multiple sclerosis (MS) is one of the most demanding cases, costing approximately $28 billion for about one million Americans with MS. However, people with chronic conditions, including people with MS (PwMS), often face challenges in applying self-care information and partnering with their clinicians over their illness progression to reflect their ever- changing needs. Unfortunately, the existing self-management interventions and mHealth solutions for PwMS and other patients do not support self-sustained personalization of self-management through lifelong user modeling. To address these challenges, we propose MyMSMentor, a Health Action Process Approach (HAPA)- driven artificial intelligence (AI) agent, to estimate users’ states and provide just-in-time guidance. Overarching goals are: [i] developing personalized, lifelong self-management support that is contextualized in an individual’s daily life and available resources; [ii] making MyMSMentor generalizable for the millions of people who need lifelong self-management of other chronic conditions. The innovation of the study focuses on bridging HAPA, social-cognitive science, and AI algorithms to build a theory-based intelligent healthcare advising agent that can support the adoption and maintenance of health behavior. Specific Aims are: (1a) Build MyMSMentor for proactive patient-centered self-management support; (1b) adopt participatory design to refine and evaluate MyMSMentor iteratively; (2) validate the feasibility and acceptability of MyMSMentor with PwMS. For Aim-1a, we will build [i] lifelong comprehensive user modeling by combining the Information Need Model, Known Information Model, and HAPA status variables and [ii] five function modules (i.e., symptom recording and updating, case summary generation, note-taking, user model inference module, HAPA-based interactive intervention module). For Aim-1b, we will collaborate with patient and clinician advisory boards to co-design and evaluate MyMSMentor iteratively. For Aim-2, 50 PwMS will use MyMSMentor for 30 days. They will use feasibility and acceptance measures to evaluate MyMSMentor and provide user feedback. The study will make a critical and timely contribution to establishing a generalizable novel model to bridge big data, artificial intelligence, and social- cognitive science for precision self-management among people with chronic conditions and support the National Neurological Conditions Surveillance System to advance data-powered research and healthcare services.
