
Discontinuation almost never arrives suddenly. The signals show up weeks early — scattered across systems that no single function can see across.
The short version: When a patient comes off therapy, the organization experiences it as sudden. It almost never was. The warning signs existed for weeks — a slipped refill, an unanswered call, a coverage change, a side-effect question — but each landed in a different function's system, where, alone, it looked like noise. The pattern that says this patient is leaving only exists across the systems, in the space no one is watching. The signals weren't missing. They were scattered.
The surprise that wasn't
Month seven. A patient who enrolled smoothly, started well, and refilled on time stops therapy. The team finds out after the fact, the way teams usually do — a refill that never came, flagged in a report weeks later. Nobody saw it coming.
Now rewind the tape with everything on one screen.
Week 19: a refill ships three days late. The specialty pharmacy logs it — within tolerance, no action needed. Week 21: the patient doesn't pick up a scheduled support call. The hub notes it — happens all the time. Week 22: a formulary change posts on the patient's plan. Market access catches it — one line in a payer update covering thousands of lives. Week 23: the patient asks the nurse line a question about managing a side effect. Logged, answered, closed.
Four warnings in five weeks. Viewed together, they're not subtle — they're a patient visibly working their way toward the exit. But nobody viewed them together, because there is no together. Each signal lives in the system of the function that caught it, and no two of those systems talk.
Each signal alone is noise
Here's the part that matters for how you fix it: every one of those functions behaved correctly.
A three-day refill slip, alone, is noise — most patients with a late refill persist just fine. A missed call, alone, is noise. So is a coverage line-item, so is a side-effect question. Dismissing each one in isolation isn't negligence; it's statistics. No team ignored a warning, because no team ever saw a warning — they each saw one data point that was rationally dismissible on its own.
The warning only exists as a pattern, and the pattern only exists across functions. Which means the failure isn't anyone's judgment. It's architecture: the organization has instrumented each function's piece of the journey and left the connections between them dark. This is the measurement face of the 90/10 problem — the visible 10% of the journey is tracked by everyone, and the long middle where persistence is actually decided is tracked by no one, because its signals are split four ways.
Why better dashboards don't fix it
The instinct is to sharpen each function's reporting. It won't help, because every function's dashboard is already right. The pharmacy's refill metrics are accurate. The hub's contact rates are accurate. Each dashboard is faithfully describing its slice — and the patient is leaving through the space between the slices. You cannot see a cross-functional pattern by staring harder at one function's data. The discontinuation will keep arriving as a surprise that was never actually a surprise.
The starting point — not the solution
There's a first step that costs nothing and takes about a week. It won't fix anything — its job is to prove, with your own data, that there's something worth fixing.
Take the last ten patients who discontinued. Ask every function to pull what it knew about each of them, and when. Put it on one timeline per patient.
In our experience, the moment a team sees its own gaps assembled in one place for the first time, the room goes quiet — and the signals are almost always there, weeks early, scattered exactly as described. That one exercise changes what the organization believes is possible: discontinuation stops being treated as unpredictable and becomes a pattern that was visible, in pieces, to people who had no way to assemble it. And once leaders believe the signals exist, the real question lands with force: who is responsible for seeing them together — and what would it take?
That's a champion's question. It doesn't accuse anyone, because the exercise just proved no one was at fault. It simply makes the unowned space between functions impossible to unsee.
And that's precisely why this is the starting point and not the solution: it's forensic, manual, and once. It proves the signals existed for ten patients, looking backward, assembled by hand. Seeing them for every patient, forward, in time to act — knowing which signals matter most, where they cluster, and who owns the catch — is the actual work. The exercise doesn't do that work. It makes your organization ready to.
Seeing it together
Answering it is cross-functional work: deciding which signals matter most, where they concentrate, which handoffs they cluster around, and who owns the watch. The 90/10 Framework is the operating model for exactly that — it tells teams where to look, which gaps cost the most persistence, and what to fix first. And it's the work Linked Patient Learning was built for: because we understand the pain points each function faces and the reality of the operating model they work inside, we can assemble the scattered pieces into the one picture of the journey no single team can see.
Patients will keep sending the warnings. They always have. The only question is whether they keep going to four different inboxes — or finally arrive somewhere someone is reading.
Linked Patient Learning helps pharma teams see the early signals of patient discontinuation across functions — assembled, prioritized, and owned, instead of scattered. If your discontinuations keep arriving as surprises, let's talk.
Written by
Linked Patient Learning
Share your details and we'll send you the playbook.