Why patients don’t follow through on treatment plans — and how to fix it

A surgeon tells Steve to wear compression stockings after a procedure. The instruction takes a few seconds. Someone enters it in the chart, and from the clinic’s perspective, the care plan is in place.

For Steve, though, the work has just begun. The initial thought of “how hard can this be?” is quickly replaced by a myriad of questions: Which compression level? Knee-high or thigh-high? Open toe or closed toe? What size, brand, quantity? Does insurance cover them? Which supplier actually stocks the right version? And how long will it take for them to arrive, and will that be too late?

Shortly after the procedure, Steve is still sore and tired, and now encountering a product category he never thought of before. A week later, no one has followed up and the stockings never arrived.

The chart will eventually label Steve’s behavior as nonadherence. I’d argue we should call it a design failure. The recommendation didn’t seem that complicated or hard, for either the clinician or Steve, but clearly something went wrong along the way.

The gap nobody’s dashboard shows

Across my work in health systems and digital health, I see different versions of the same problem. Organizations are treating the patient as if they lived in a world of no constraints, infinite time and unwavering motivation and resilience.

Every clinical recommendation, whether it’s a prescription, a referral, a device or a set of discharge instructions, hands the patient a sequence of tasks that are mainly based on logistics, economics and emotions.

First, the patient has to understand and remember the recommendation well enough to consider it. Then come questions about cost, coverage, scheduling, transportation and what to do if the first attempt fails. Comparing unfamiliar options and determining the best next steps by yourself are already resource-intensive tasks for our chronically exhausted and distracted brains. Now add logistical and administrative friction.

We tend to see the result only after something went wrong. The prescription was never filled. The appointment was missed. The patient stopped using the device. Those data points confirm a breakdown, but they say little about where progress stalled and what could have been done differently to prevent that drop-off.

The gap between a recommendation and follow-through isn’t small. In a study of 195,930 electronic prescriptions, 22% were never filled; among new prescriptions, the number climbed to almost 30%. The study has its limitations, but the point is that writing a prescription and starting treatment are distinctly different events. Adherence researchers distinguish initiation from implementation and persistence for this reason. The same pattern shows up with referrals, home-monitoring devices, behavioral health handoffs, medical equipment and physical or mental health therapy. A treatment plan can’t be followed if it never begins.

Why “just remind the patient” doesn’t work

When follow-through falls short, teams often reach for reminders, education or adherence counseling. Each has a role. I have also seen teams use them for problems they cannot solve.

A patient might feel informed, motivated and ready to act while lacking time, money, transportation, the physical ability, an available appointment or help with an authorization. Consider a patient who understands their hypertension management plan and wants to follow it but cannot find an affordable blood pressure device. A reminder adds one more prompt to a problem the patient already knows exists. Or consider someone who needs physical therapy but cannot get an appointment outside work hours. More information about the benefits of therapy will not create an opening in the schedule.

The COM-B model offers a useful way to think about what enables or prevents behaviors. It describes behavior as a function of capability, opportunity and motivation. Motivation alone is not effective if you are missing the capability or environment that allows that behavior to occur. Research on the intention-behavior gap reaches a similar conclusion. Intending to do something does not ensure the behavior follows, and the missing piece usually isn’t willpower; it’s the type and sheer amount of work that is required.

I think about the type of work as the behavioral workload of care. Some of it is cognitive: interpreting instructions, choosing between products, remembering the sequence of steps. Some is administrative or financial: calls, forms, coverage checks, copays. Some is emotional or physical: fear of a diagnosis or negative health outcomes, perceived stigma, travel while in pain, learning to use equipment with limited dexterity. One treatment or discharge recommendation might involve several kinds of work at once.

Additionally, the same recommendation does not create the same workload for every patient. Pain, fatigue, disability, language, health literacy, income, transportation, digital access, work schedules and social support shape a person’s available capacity. Standard instructions can create an unequal burden because patients begin with different resources and constraints. And the burden doesn’t fall on patients alone; caregivers often absorb a meaningful share of it, coordinating logistics and reassurance that health systems rarely document as part of “the plan.”

Drop-off points and what they mean for building better pathways

Not every step in this workflow carries equal risk. If a team wants to improve a pathway, I would start by looking at what happens during the first few days after a recommendation. How does the patient feel right after a diagnosis or procedure? Where does the patient learn the price? Who helps when a referral reaches an outside office? What happens during the first attempt to use a device at home? These moments combine effort with uncertainty. They also tend to sit between teams, with no clear owner. If you’re mapping a care pathway and trying to decide where to intervene, start there.

I would ask the team to follow a few patients from the recommendation to their first action. Record the steps in their own words. Notice where they pause, call someone, search online or give up for the day. A pathway map made from clinic workflows alone will miss much of this work.

Then address the friction closest to the decision. If a patient needs to monitor blood pressure, give the name of a suitable device, where to get it, what it costs when known and when to take the first reading. If several options are clinically acceptable, say so and explain which one you recommend. If authorization is a common delay, tell the patient who handles it and when they should expect an update.

Measurement needs to begin sooner, too. Start tracking these small in-between steps, like patient understanding and buy-in, readiness and capability to act, potential barriers, or social support availability. Those can be standard measures or sporadic check-ins; either way, they help uncover what is invisible until you see the nonadherence data points. Tracking them also gives care teams a chance to intervene while the problem is still solvable.

The bigger shift

Health care has invested enormous effort in deciding what care to recommend. I believe we should invest comparable attention in what a patient must do after leaving the room.

Acquiring the compression stockings in the opening example did not fail because the patient needed another lecture or reminder on their importance. Steve needed an answer to a practical question, followed by one clear next best action. If we want recommendations to become care, we have to design for those questions and barriers before patients are left to solve them alone.

Anne-Kathrin Eiselt, Ph.D., is a behavioral neuroscientist and founder of ThriveLogic, where she helps health care organizations improve adoption, engagement and patient follow-through. Previously director of applied behavioral science at Teladoc Health and a behavioral scientist at Bayer, she brings 20 years of experience across research, product development and implementation. She also teaches psychology at George Washington University.