AI, Aging, and the Medication Adherence Problem
Noah Vandal and Dr. Joseph Yoon discuss medication adherence, aging, polypharmacy, and how AI voice agents could support safer routines.
Aging, Medication Use, and AI
Why medication routines became harder to manage
Modern medicine has given patients better options for chronic disease, post-surgical recovery, cardiovascular care, diabetes, pain, infection, and many other conditions. That progress has also created a practical problem: more people, especially older adults, now manage multiple prescriptions at once. In this episode of the AI and Healthcare Podcast, Dr. Joseph Yoon points out that the number of medications used in everyday care has increased dramatically over recent decades. For a patient with one simple prescription, adherence may be mostly about remembering a daily habit. For an older adult with several conditions, it can become a complicated schedule of morning pills, evening pills, refills, instructions, side effects, and uncertainty about what each medication is supposed to do. That complexity is why medication non-adherence is not a small administrative issue. The episode discusses estimates that poor adherence contributes to approximately 125,000 deaths each year in the United States. Even when that figure varies by source or methodology, the operational point remains clear: missed doses, stopped courses, and misunderstood instructions can change real health outcomes.
Why seniors are especially vulnerable to missed doses
Medication adherence becomes harder when a person is managing polypharmacy, cognitive changes, vision problems, transportation limits, cost concerns, or simple uncertainty about whether a dose was already taken. A pill organizer can help, but it does not answer every question. A phone alarm can remind someone to take something, but it cannot explain why the medication matters or when a symptom should be escalated. Dr. Yoon makes a practical distinction in the conversation: missing one dose is not equally dangerous for every drug. Some medications have more forgiving timing. Others, such as certain blood thinners, can carry higher risk when missed or taken incorrectly. That nuance matters because adherence support should not treat every reminder as the same level of urgency. For healthcare organizations, the challenge is not only "send more reminders." The harder problem is helping patients understand their regimen well enough to follow it and helping clinicians know when confusion, missed refills, or concerning symptoms need attention.
Why education changes adherence
One of the strongest points in the episode is simple: patients are more likely to follow a medication plan when they understand why they are taking each medication. "The doctor told me to" is weaker than "this helps reduce my clotting risk" or "this helps control my cholesterol over time." That does not mean every patient needs a pharmacology lecture. It means healthcare teams should make the purpose of a medication understandable, repeatable, and accessible outside the appointment. A patient may forget the explanation by the time they get home. A family caregiver may not have been present. A front desk or care coordination team may receive follow-up calls that are not urgent but still matter. This is where AI becomes interesting. The value is not just a notification. It is the ability to provide plain-language, patient-specific education within a defined safety boundary.
How AI voice agents could support safer medication routines
AI voice agents can make adherence support more interactive than a one-way reminder. A system could call at scheduled times, confirm whether a patient took a medication, ask if the patient is confused, explain general information from an approved care plan, and route unresolved questions back to a clinician or pharmacist. The safest version of that workflow is bounded. The AI should not improvise medication changes. It should not tell a patient to start, stop, double, or substitute medications. It should be able to say, in effect: "This is what your care plan says. Here is the plain-language reason. If you missed a dose or feel a concerning symptom, I can help connect you with the right person." That distinction is central for SpeechSage. Healthcare voice AI should be designed around approved workflows, auditability, escalation, and human review. For medication adherence, the practical opportunity is to reduce avoidable confusion and surface issues earlier, not to replace clinical judgment.
Where refills, wearables, and dispensers may fit next
The episode also looks ahead to combinations of AI with other tools: wearable devices, smart medication dispensers, refill workflows, and prescription renewal pilots. A voice agent by itself can help with reminders and education. A connected workflow could notice patterns: a refill is late, a patient reports confusion repeatedly, a dispenser shows missed doses, or a wearable signal suggests a concern that deserves follow-up. That future needs careful design. More automation is not automatically safer. Medication data is sensitive, older adults deserve consent and dignity, and clinicians need reliable signals rather than noisy alerts. The right model is a human-in-the-loop system that makes routine support easier while escalating the moments that require professional attention. Utah's AI prescription renewal pilot, discussed in the episode summary, is one example of the broader direction: AI systems may take on more structured intake and renewal support for chronic medications. Whether those workflows succeed will depend on oversight, clear scope, documentation, and the ability to hand off safely when a case is no longer routine.
What healthcare teams should take from this episode
Medication adherence is a workflow problem, an education problem, and a relationship problem. A patient needs to remember the dose, understand the reason, trust the plan, and know what to do when something does not make sense. AI can help with parts of that work. It can repeat instructions patiently, call at the right time, adapt explanations to the patient's level of understanding, and make it easier for staff to see where follow-up is needed. But the goal should never be an autonomous medical authority. The goal should be a safer support layer around patients, caregivers, pharmacists, and clinicians. For healthcare AI builders, medication adherence is a useful test case. If the system can stay grounded, explain clearly, respect scope, and escalate responsibly, it may help solve a problem that has been hiding in plain sight for decades.