P500 - How Machine-Learning (Ai) can help Identify Individuals at Risk for Discontinuing CGM Therapy

Continuous glucose monitoring (CGM) is an important tool for helping individuals living with diabetes to view their glucose variability. However, when wear-time is limited, CGM is limited in its utility. This poster describes a current study that used a persona-based machine-learning model that identifies patients with specific risk factors for discontinuing CGM. Personalized interventions were developed and evaluated in a sample of individuals living with diabetes. Adults (aged ≥18 y) who ordered CGM supplies from a leading chronic care management company were included in the model. Adults were categorized as adherent, lapsed, or lost to attrition. Data were used to identify specific risk categories for not ordering with scheduled CGM supply reorders. Tailored hyper-personalized interventions were designed to provide specific information relevant to each individual’s needs to prompt re-engagement with their CGM regimens, including communication preferences with respect to channel (text, email, phone). Preliminary testing has been performed to assess the impact of this persona-based approach on CGM reordering and is ongoing. The model included data on 184,036 individuals with type 1 or type 2 diabetes receiving benefits from Medicare, Medicaid, or commercial insurance plans categorized into active (n=46%), lapsing (39%), and lost to attrition (15%) cohorts. Patients fell into 14 distinct Personas based on individual characteristics associated with CGM discontinuation, of which 7 Personas accounted for >90% of patients. Tailored interventions providing information or reminders to patients who are lapsed or lost to attrition were most effective in Medicare and Medicaid participants who seek health information from peers (39-40%), lapsed Medicare patients who trust health advisors (52%), and busy caregivers enrolled in Medicare, Medicaid, or commercial insurance plans (25-48%). Machine-based intervention positively identified individuals at risk for discontinuing prescribed therapy.