Start with one patient, because the averages hide this.
She starts a GLP-1 in January, and it works. By spring, her weight is down, her blood pressure is better, and her last A1c is the best she's had in years. In April her copay goes up, and she stops. Nobody records a clinical decision, because nobody made one.
Eleven months later she's in an emergency department with something that doesn't look like a medication problem at all.
Nothing in any system connects those two events.
Two halves, one patient
We've written about each half of this before: persistence in October, and the week after discharge earlier this month. If you've been reading along, it would be natural to treat those as separate problems.
They're not. It's the same patient at two points in time, and this post is about where they meet.
It's also the easiest argument in this series to overclaim, so I'll take it one link at a time and be clear about where the evidence runs out.
Link one: a lot of people stop
Prime Therapeutics found that 60.9 percent of commercially insured adults without diabetes who started Wegovy or Zepbound in the first half of 2024 were still on therapy a year later. That's much better than during the shortage years, and it still means close to four in ten stopped within the year.
We covered why last week. In a 2026 survey of 440 adults who stopped, cost was the top reason at 36.1 percent, and lack of insurance coverage came third at 28.2 percent.
So the first link is solid, and it already tells you something. A lot of these patients didn't stop for clinical reasons.
Link two: stopping gives the gains back
People tend to assume this one without checking. Two trials actually looked.
In the STEP 1 trial extension, people who had lost an average of 17.3 percent of their body weight on semaglutide came off it. A year later they had regained about two thirds of that weight, and most of their cardiometabolic improvements had drifted back toward where they started.
The SURMOUNT-4 analysis in JAMA Internal Medicine last November went further. Of 308 people switched from tirzepatide to placebo, 82 percent regained at least a quarter of the weight they had lost within a year. Waist size, blood pressure, non-HDL cholesterol, A1c and fasting insulin all got worse, and in general, the more weight came back, the worse they got.
That's the detail I'd underline. Stopping doesn't flip a switch. The further a patient drifts, the more of the benefit she gives back.
Link three: worse control means more time in the hospital
This is where I have to be careful, because the best evidence here isn't about GLP-1s.
A study of 9,497 older adults with diabetes in Pennsylvania's drug assistance program for seniors compared people who took their oral diabetes medication consistently with people who didn't, using the standard cutoff of 80 percent of days covered. The ones who didn't had about twice the odds of being hospitalized, 27 percent more hospital visits overall, and stays that ran 24 percent longer.
That's a different class of drug and an older population, so I won't present it as a GLP-1 finding. What it does show is the general pattern: when people with a chronic condition go without their medication, they end up in the hospital more, and they stay longer.
Where the evidence stops
Each of those three links has its own evidence. The chain as a whole is my argument, and no single study proves it.
Nobody has run the trial that follows a GLP-1 patient from a copay change, to stopping, to regaining the weight, to a missed follow-up, to an admission, and then puts a price on it. If someone tells you they have, ask to see it.
I'd rather say that than borrow a statistic that doesn't quite fit. We did the same thing when we wrote that not every readmission is preventable, and it's a big part of why that piece landed with the people we most wanted to reach.
Why nobody connects the two
The pharmacy sees a refill that never happened. The health plan sees claims stop coming in. The clinic sees an appointment nobody rebooked. Eventually, the hospital sees an admission.
Each of them sees its own piece. Nobody sees the line between them, because none of those systems spans both ends.
By the time she's admitted, the missed refill is eleven months old and sitting in another company's database. Nobody is going to connect those by hand, and the reporting on each side only answers questions about its own half.
The cost lands twice
This is the part that should matter to whoever owns the budget.
When therapy stops early, the money already spent bought a few months of benefit instead of a lasting one. That's the first cost, and it shows up in the pharmacy budget.
Then the condition the drug was controlling comes back, along with the visits and admissions that come with it. That's the second cost. It usually lands on a different line of the budget, in a different meeting, with someone who was never part of the formulary decision.
It's one problem. It just shows up in two budgets.
What I'd measure
One comparison would settle most of this, and almost nobody runs it.
Take the patients who stopped their GLP-1 and the ones who stayed on it, and compare their emergency visits and admissions over the same period. Count visits and admissions, not just dollars.
If you're a health plan or an ACO, you already have both groups. Pharmacy data shows who stopped, and claims show who ended up in the hospital. You don't need to wait for a journal to publish the study I just said doesn't exist. You can run your own version on your own patients this quarter.
One caution: people who stick with a therapy tend to be healthier to begin with (researchers call it the healthy adherer effect), so adjust for baseline risk as well as your data allows before you read too much into the gap.
If the two groups look the same, then this argument is wrong for your population, and that's worth knowing too. I expect they won't.
That's why we built Preventra the way we did. Readmission risk and therapy continuity sit on the same patient record, so the people who make the call can see that the woman being discharged today stopped her GLP-1 back in April.
The missed refill and the admission are the same story. They just get told eleven months apart, to two different audiences, by systems that have never met.
Sources
Marshall LZ, Gleason PP, Friedlander N, Farley J, Urick BY. Trends in 1-year persistence and adherence among initiators of high-potency, weight loss-indicated glucagon-like peptide 1 receptor agonists. Journal of Managed Care and Specialty Pharmacy, 2026;32(3):281-291.
Journal of Managed Care and Specialty Pharmacy. Cross sectional survey of reasons for GLP-1 receptor agonist discontinuation, 2026. DOI 10.18553/jmcp.2026.32.8.903.
Wilding JPH, Batterham RL, Davies M, et al. Weight regain and cardiometabolic effects after withdrawal of semaglutide: the STEP 1 trial extension. Diabetes, Obesity and Metabolism, 2022;24(8):1553-1564.
Horn DB, Linetzky B, Davies MJ, et al. Cardiometabolic parameter change by weight regain on tirzepatide withdrawal in adults with obesity: a post hoc analysis of the SURMOUNT-4 trial. JAMA Internal Medicine, published online November 24, 2025.
Pednekar P, Heller DA, Peterson AM. Association of medication adherence with hospital utilization and costs among elderly with diabetes enrolled in a state pharmaceutical assistance program. Journal of Managed Care and Specialty Pharmacy, 2020;26(9):1099-1108.
