Readers Write: The New Lifestyle Coach: How AI Can Support Adherence for People with Diabetes

The New Lifestyle Coach: How AI Can Support Adherence for People with Diabetes
By Richard Mackey

Richard Mackey, MBA is CTO of CCS of Dallas, TX.

Living with diabetes is more than a full-time job. It’s a 24/7 commitment to monitoring thousands of tiny details related to diet, exercise, sleep, stress, and overall health.

People living with diabetes end up making nearly 200 extra decisions every day to keep on top of their condition, creating a near-constant risk that a split-second of inattention can snowball into a slide away from healthy habits and appropriate adherence to care.

While much of this conversation around diabetes adherence has historically focused on medications, the modern diabetes care plan now often includes devices such as insulin pumps and continuous glucose monitors (CGMs). These devices are supposed to make self-management of care decisions easier, but sometimes they cause unexpected issues when a patient doesn’t understand how to use their device or struggles with staying on top of their routine.

As the diabetes epidemic continues to grow, health plans and their partners cannot expect people living with diabetes to bear these constant burdens all alone. Instead, they need to surround each and every individual with predictive, data-driven support that accurately flags risks of non-adherence — including that involving medical devices — before they become unmanageable and lead to poor outcomes and higher healthcare spending.

Leveraging artificial intelligence (AI) and machine learning to assess members predictively and longitudinally can help health plans identify emerging risks of non-adherence and proactively reach out with support for members to keep them on the right track with their care.

Data-driven risk stratification has become a core component of chronic disease management in recent years, but health plans still face challenges with identifying when and why certain individuals move up and down the risk ladder. Many plans primarily work with claims data, which can be incomplete from an analytics perspective and offers little insight into why members are straying from their care plans. With limited scope and up to several months of lag time, this claims dataset alone is not sufficient to get ahead of the exact moment a person starts to show potential issues that are likely to lead to non-adherence with recommended care best practices.

Instead of relying too heavily on claims data alone, health plans need to integrate datasets that give a more comprehensive and current view of member activities: socioeconomic data to identify non-clinical barriers; pharmacy data to show medication access and adherence patterns; diabetes supply ordering records to indicate therapy adherence; and device data to highlight continual usage of management tools and control of clinical factors, like blood glucose levels.

Together, these and other datasets paint a powerful, holistic, and timely portrait of a member’s ability to participate in their own care from a clinical and nonclinical perspective, enabling health plans and providers to pinpoint potential trouble spots and dynamically predict rising or falling risks of non-adherence.

AI has quickly become an essential tool for making sense of rich and varied healthcare datasets, but it must be deployed intentionally to maximize its impact. That means developing bundled algorithms and services that can identify accessible patient data while also spotlighting what data is actually missing in a patient’s longitudinal record.

For example, the sudden absence of a monthly diabetes supply order or prescription refill or a sporadic tapering off of data reports from a CGM over time are major red flags on the adherence front. AI tools must be sophisticated enough to know when missing data is a sign of an impending problem, which means designing models and corresponding patient outreach and education strategies that support prevention.

After examining these patterns at scale and over time, AI models can accurately assist health plans with identifying the clinical and socioeconomic factors that most directly correlate with these adherence gaps in their unique populations, allowing care management teams to move closer and closer to the non-adherence trigger point for individuals, and ideally, to also be able to predict likely non-adherence events for members before they occur.

For example, some members are providing care for children and aging parents while working full time and may have more limited opportunities to invest in their own care. Offering these members insights and best practices specific to maintaining therapy under a tight schedule can prove helpful. In other cases, financial uncertainty may be impacting a member. Providing these members with education and coaching on tools that allow for flexibility in out-of-pocket costs for medications and/or medical devices so that they can continue therapy without disruption can mean the difference between adherence and non-adherence.

Considering that the costs of caring for people with diabetes consumes more than 20% of the nation’s annual $4.5 trillion healthcare budget, investing in next-generation tools and partnerships to get ahead of non-adherence and negative member outcomes is essential for altering the trajectory of the ongoing diabetes epidemic.

However, identifying impending problems is only half the battle. Plans must be ready and able to conduct meaningful, individualized outreach to members who show signs of non-adherence as soon as possible. Direct engagement and education with members can often uncover the true obstacles, both tangible and emotional, behind non-adherence issues, including underlying issues of trust in the health system that may stem from personal or community experiences. These conversations with extended care teams can shift that narrative for individuals and become an opportunity for plans to provide compassionate, actionable problem solving for members that help build relationships and prevent future issues.

Information that is gathered during these outreach interactions can be structured and folded back into analytics efforts to enrich future insights and enable health plans to become even more predictive, personalized, and prepared to support their members with community-based resources, tailored diabetes education, and specialized training on how to best use their devices and adhere to a recommended care regimen.

Ultimately, AI can help identify non-adherence issues in people living with diabetes before it becomes a full-blown, costly problem. By diving deeper into holistic datasets and member care patterns, AI tools will soon be able to identify the underlying challenges facing members, empowering health plans to address these issues earlier while fostering meaningful outreach activities that surround people living with diabetes with the support and guidance they need to thrive.