Prescription is not persistence

Health plans and health systems are investing heavily in GLP-1 programs right now. Formulary strategy, prior authorization criteria, utilization management, clinical pathways, contracting, and cost models that run years into the future.

Nearly all of that work rests on one assumption that deserves more attention than it gets: The patient is still taking the medication.

Real world evidence suggests that assumption is fragile. A 2024 study of 4,066 commercially insured adults with obesity and without diabetes who initiated GLP-1 therapy found persistence of 46.3 percent at six months and 32.3 percent at one year, which means roughly two thirds were no longer persistent after twelve months. The picture in type 2 diabetes is better and still significant: a separate claims analysis of 4,791 patients starting injectable GLP-1 therapy found 47.7 percent had discontinued by twelve months.

We spend enormous energy deciding who should start therapy and considerably less understanding who is beginning to fall out of it. That asymmetry is where both the clinical and the operational opportunity now sit.

The number almost nobody talks about

Discontinuation is the easy part to understand. The patient was receiving therapy, then the patient stopped, and eventually the claims data tells you.

The more interesting figure in that 2024 study is the one that gets far less attention: mean proportion of days covered was 51.0 percent. "Proportion of days covered" is the standard adherence measure, stewarded by the Pharmacy Quality Alliance and used widely across quality programs. It asks a straightforward question. During the period when a patient was expected to have medication available, what share of those days were actually covered?

In that cohort, the average was barely half.

That describes a more complicated problem than clean discontinuation. A patient does not usually move in one step from taking the medication to not taking it. The deterioration is gradual: a refill arrives late, then the next one arrives later, a follow up appointment never gets scheduled, an authorization drifts toward expiration, a visit gets missed, and eventually a claim simply stops arriving. From a distance that patient shows up in a report as a discontinuation. Up close, they were signaling trouble for weeks.

Patients rarely disappear from therapy overnight

The industry tends to discuss persistence as a question of patient motivation. Commitment, engagement, follow through. Sometimes that is exactly what it is, but as a general explanation it is far too convenient, because it puts the problem somewhere a program cannot reach.

Look at what happens between one successful prescription and the next one that never comes. A refill runs nine days late, then fourteen. An authorization expires and needs documentation from a physician's office already working through hundreds of administrative tasks. The follow up visit that would have surfaced a side effect, a cost problem, or a dosing question was never booked, because whoever handled the initial visit assumed somebody else was handling it.

None of those events are hidden. The prescriber assumes the pharmacy has visibility, the pharmacy assumes the patient will call, the health plan sees claims, the care team sees appointments, and the authorization team sees the coverage problem.

Everyone has part of the story. Nobody has the patient.

That is the operational gap. Three or four systems can each hold a perfectly accurate piece of information while the organization as a whole never registers that one person is drifting out of treatment. By the time discontinuation is obvious, the useful intervention window has usually closed.

Prior authorization is one piece of a larger problem

Administrative friction deserves its own attention here. In the American Medical Association's 2025 survey of 1,000 practicing physicians, 95 percent said prior authorization delays access to necessary care and 79 percent reported that patients abandon treatment because of authorization challenges.

That survey covered prior authorization broadly and was not specific to GLP-1 therapy, so it should not be read as a GLP-1 finding. The operational lesson still applies directly, and it is the one worth holding onto: treatment can fail without anyone ever making a clinical decision to stop it. A patient can fall out through coverage friction, a missed follow up, a refill gap, an unresolved authorization, a scheduling breakdown, a cost change, or several of those at once. Those events are recorded somewhere. The trouble is that somewhere is not the same as together.

Persistence is something a program manages

Seen this way, persistence stops being only a characteristic of the patient and becomes a measure of the program, which is a harder thing to say because it makes the problem yours.

You cannot manage someone's motivation with a dashboard. You can manage whether a fourteen-day refill gap gets noticed, whether an expiring authorization has an owner, whether a patient who started therapy months ago and has never come back for follow up gets surfaced to somebody, and how quickly a detected problem turns into an actual phone call.

That changes what a serious program should be measuring. Starts matter, but starts alone say almost nothing about whether treatment is working operationally. The four I would want on the report are persistence at three, six and twelve months, proportion of days covered across those same windows, time from a detected interruption to real outreach, and the share of discontinuations with a documented reason. The space between prescribed and persistent is where quiet failure lives, and none of those four measures require data a program does not already have.

The missing layer is therapy continuity intelligence

Refill activity exists. Appointment history exists. Authorization status exists. Claims history exists. What is missing is the layer that assembles those signals around one patient, recognizes when the pattern is changing, explains why that person needs attention, and surfaces it while somebody can still act.

That is what we are building at Preventra, and it is deliberately not another dashboard showing who received a prescription, or another risk score telling a care team that somebody might discontinue. The objective is therapy continuity intelligence: identify the patients whose treatment is starting to drift, show the specific signals behind the concern, and give teams a chance to intervene before interruption becomes discontinuation.

The distinction is not academic. "High discontinuation risk" is a score. "Refill is fourteen days overdue, no follow up since therapy started, authorization expires this month" is something a care coordinator can act on before lunch. That is the difference between prediction and execution.

The prescription was never the outcome

GLP-1 programs have become sophisticated at deciding who should receive therapy. The next advantage belongs to the organizations that know what happens afterward: who stays persistent, who is intermittently covered, who is starting to drift, why, and how early anyone can reach them.

A prescription tells you that treatment started. It does not tell you that treatment continued.

Starting therapy was never the result. Persistence is where the result has a chance to happen.