Redesigning Pharma Adherence Around Patient Behavioral Signals and AI-Native MCP Rails

Suzy Jackson, Commercial Strategy Lead at Translate Health

AI’s biggest contribution to adherence can be a new model: reading the signals patients leave before they stop, and responding with support built for that moment, but only if pharma decides who it wants to be to that patient.

In 2003, the World Health Organization reported that adherence to long-term therapy in developed countries averages about 50%. Since then, the industry has taken on that number in waves: first as a compliance problem, then as a cost problem, then as an infrastructure problem, with hubs, copay programs and nurse support built around individual brands. Today, manufacturers are in the beginning phases of building direct relationships with patients, with Eli Lilly reporting more than a million patients engaging with its LillyDirect platform in 2025.

Alongside experimentation with different tactics, the rise of AI has spawned a new breed of possibilities for driving easier patient access, with a large amount of money and attention spent on automation of reimbursement support, prior authorization and benefits verification across digital health providers and pharma in-house.

However, getting a patient to a first fill is only one piece of the puzzle, and keeping that patient on therapy at month six, month twelve and month thirty-six is another. Conversations with commercial and patient services leaders at large manufacturers over the past 18 months suggest most AI investment has been aimed at the first job, which is why the next breakthrough sits in the second.

The varied challenges of keeping patients adherent

In a JAMA randomized trial of 9,501 patients at three US health systems, text message reminders or behavioral nudges sent to people who had fallen behind on cardiovascular prescriptions did not improve refill adherence at 12 months, suggesting generic outreach does not move the needle. In severe asthma, a 2026 analysis of more than 16,000 US patients on biologics found 22.3% discontinued treatment, with adherence associated with the socioeconomic conditions of the areas where patients live (sadly, a surprise to nobody in the pharma industry). In oncology, a large claims analysis linked higher out-of-pocket costs to nonadherence in most of the cancer types studied (only 51.9% were adherent overall). For brevity, these are just three example findings, but they typify two realities the industry must grapple with simultaneously: blanket outreach does not meaningfully impact adherence, and adherence is multi-faceted, varying by social determinants of health, dosing burdens, side effects, and cost, among other things.

Patients leave clues before they stop

Nonetheless, there are reasons for optimism about AI-powered breakthroughs in the adherence space post-medication access. Historically, companies have relied on claims data as a widely used indicator for adherence, which is a thin and delayed source of information. However, given AI’s ability to consume and make sense of large swaths of data, there lies untapped potential in new data usage and pattern recognition. While recognizing that compliance considerations and interpretations around the appropriate and non-problematic uses of PII remain increasingly fraught with concern and confusion, and vary across states, organizations should reconsider their data architecture to pull stop signals out of refill and consented first-party behavioral data in order to offer hyper-personalized support at exactly the right moment. To reduce exposure, organizations should lean into a robust consent architecture, and keep patient identifiers with the covered entity, so end users like manufacturers see only aggregate results. De-identification lets analysts study patterns without naming patients, and tokenization is one way to link these datasets into more meaningful cohorts.

When data sources are applied better, a small amount of evidence is emerging that new algorithms can move the needle. In a 2025 JAMA Network Open trial of 5,421 patients, an algorithm identified people with low adherence, their primary care physicians received tailored clinical decision support, and pharmacists reached out. Adherence to blood pressure medications improved. However, the benefits didn’t extend to statins or non-insulin diabetes medications, which implies that program design will need to be as specific as the medication and condition it supports.

A question worth answering first

While it may seem obvious that post-first-fill adherence is a clear place to invest for service providers and patient teams within pharma organizations, there’s a broader question that the pharma industry is still grappling with that underpins the direction those dollars will need to flow, and to whom: what does pharma want to be to its patients? In general terms, pharma needs to choose among three options, at either the enterprise level or the therapeutic area (TA) level: 

Step out of the identity of being a manufacturer and into the model of being a direct, patient facing unit. Invest in a tapestry of partnerships and technologies to compliantly serve the patient more directly through first fill and well beyond, even in high-science therapeutics Take ownership of the patient relationship by being the conduit behind consumer-facing brands that people are already going to, like the 230 million people going to OpenAI every week for health answers. Build the right MCPs, connectors, and AI-ready information to take as much advantage as possible of those pre-existing distribution models Continue in the model of manufacturer, handing the patient relationship entirely to digital health D2C innovators and continuing to focus on medication access and cost control

Patients will not wait for the answer, and illness and medication needs certainly won’t. Each portfolio or company can credibly shore up strategies behind any of these options, but not choosing is the one option that won’t hold up. The industry has spent the last few years considering new channels, investing in access, and grappling with the realities of AI emerging in its midst. The next few years can, and should, focus on helping people stay on the right medication, because the data, the technology, and even a little evidence are starting to line up in the cause’s favor.

---

About Suzy Jackson

Suzy Jackson is an independent commercial strategy and digital health advisor and co-leads Translate Health, an innovation and build studio focused on health and life sciences. She is the former Chief Commercial Officer of RVO Health, a UnitedHealth Group JV. Before that, she spent over a decade at Accenture, where she was, most recently, a Managing Director leading North America Life Sciences Commercial Consulting. She is an ICF-certified coach, a PM360 Elite 100 Transformational Leader (2026), and serves on the board of Safe Alliance and the advisory board of Equality Now.

---

Sources

In 2003, the World Health Organization reported that adherence to long-term therapy in developed countries averages about 50% – WHO, Adherence to Long-Term Therapies: Evidence for Action (Geneva, 2003 In a JAMA randomized trial of 9,501 patients at three US health systems, text message reminders sent to people who had fallen behind on cardiovascular prescriptions did not improve refill adherence at 12 months, suggesting generic outreach does not move the needle – Ho PM et al. JAMA. 2025 Jan 7;333(1):49-59. doi: 10.1001/jama.2024.21739 In severe asthma, a 2026 analysis of more than 16,000 US patients on biologics found 22.3% discontinued treatment, with adherence associated to the socioeconomic conditions of the areas where patients live – wiatek J, Deb A, et al. “Associations Between Social Determinants of Health and Adherence to Biologic Therapies in Patients with Severe Asthma.” Journal of Asthma and Allergy, published 25 June 2026 In oncology, a large claims analysis linked higher out-of-pocket costs to nonadherence in most of the cancer types studied (only 51.9% were adherent overall) – Vyas A et al. “Predictors of adherence to oral anticancer medications: An analysis of 2010-2018 US nationwide claims”, J Manag Care Spec Pharm, 2023 In a 2025 JAMA Network Open trial of 5,421 patients, an algorithm identified people with low adherence, their primary care physicians received tailored clinical decision support, and pharmacists reached out. Adherence to blood pressure medications improved. However, the benefits didn’t extend to statins or non-insulin diabetes medications – Raebel M et al. “Clinical Decision Support and Cardiometabolic Medication Adherence: A Randomized Clinical Trial.” JAMA Netw Open. 2025 Jan 2;8(1):e2453745 230 million people going to OpenAI every week for health answers – OpenAI, “Introducing ChatGPT Health,” January 7, 2026