Running a telehealth program at scale, Part 4: the refill lapse problem
Churn is the metric everyone watches. Lapse is the one that causes it. Part 4 of our series for programs at scale is about the days between a patient’s last dose and their next delivery — how to see them, and how to close them.
Lithos StaffA refill lapse is the gap between when a patient’s supply runs out and when the next shipment arrives. Lapse rate — the share of active patients who lapsed in a period — predicts churn, drives support volume, and in GLP-1 programs can mean restarting titration. It goes unmeasured because it spans systems: the pharmacy knows ship dates, the clinician knows decisions, billing knows renewals, and nobody joins them. Compute it per patient from last ship date plus days’ supply versus next delivery date. Causes cluster into five: the check-in never happened, the clinician queue was slow, payment failed, the pharmacy or carrier was late, or the address or state changed. The fix is a cadence engine: trigger the check-in at supply-minus-ten days, authorize within hours, ship by supply-minus-three, and watch the shipment to the door, with every step an event.
What a lapse is
A patient on a 28-day supply who received a shipment on the 1st runs out on the 29th. If the next shipment arrives on the 27th, there is no lapse. If it arrives on the 3rd of the following month, that is a four-day lapse — four days without medication, and in a GLP-1 program, potentially a restart of titration if the gap stretches. Multiply across a thousand patients and you have the single largest hidden cost in the program.
Lapse rate is the share of active patients who lapsed in a period. It is not churn — the patient is still subscribed — but it is churn’s leading indicator. Patients who run out lose momentum, feel symptoms return, and cancel. By the time the cancellation appears in your dashboard, the lapse that caused it is weeks old.
Why nobody measures it
Because it spans systems. The pharmacy knows when it shipped. The carrier knows when it delivered. The clinician platform knows when the renewal was authorized. Billing knows when the subscription renewed. Your app knows when the patient completed a check-in. No single system knows the patient ran out on the 29th and received medication on the 3rd, so no single system reports it. Programs that do measure lapse are the ones whose infrastructure emits every one of those moments as an event they can join.
How to compute it
For each active patient, maintain a supply model:
| Field | Source | Meaning |
|---|---|---|
| Last shipment date | Pharmacy / carrier event | When the current supply started |
| Days’ supply | Prescription | How long it lasts at the current dose |
| Run-out date | Computed | Last shipment + days’ supply |
| Next delivery date | Carrier event (or forecast) | When the next supply arrives |
| Lapse days | Computed | max(0, next delivery − run-out) |
Lapse rate for a month is the share of patients whose run-out date fell in the month and whose lapse days exceeded zero. Report the distribution too: many one-day lapses is a shipping-buffer problem; a few ten-day lapses is a queue or payment problem.
The five causes
- The check-in never happened. The protocol requires a follow-up before a renewal — weight, side effects, adherence — and the patient did not complete it. Either the prompt came too late, was too long, or came through a channel they ignore.
- The clinician queue was slow. The check-in was submitted on time and sat. Turnaround is a lapse cause, which is why Part 3’s capacity planning is retention work.
- Payment failed. The card expired or declined and the renewal stalled behind billing. Dunning that starts after run-out is too late.
- The pharmacy or carrier was late. Compounding backlogs, a cold-chain hold over a weekend, a failed delivery attempt.
- The address or state changed. The patient moved; the new state needs a different licensed clinician or a different pharmacy, and nobody knew until the shipment bounced.
Each has a different owner — product, clinical operations, billing, pharmacy, support — which is another reason lapse goes unowned. Give it one owner and one number.
The cadence engine
The fix is to run the renewal on the patient’s supply clock rather than the calendar. A cadence engine holds each patient’s run-out date and triggers work backward from it.
| When | What | Who |
|---|---|---|
| Run-out − 14 days | Payment method verified; state and address confirmed | Billing, automated |
| Run-out − 10 days | Check-in prompted, in the channel the patient actually uses; agent-run where possible | Product / agent |
| Run-out − 7 days | Check-in overdue → second prompt; still missing at −5 → human outreach | Support |
| On check-in | Renewal encounter submitted with drafted summary | Agent / product |
| Within hours | Clinician authorizes, adjusts, or declines | Clinical |
| Run-out − 3 days | Shipped, cold chain if needed, carrier events flowing | Pharmacy |
| Delivery | Confirmed; supply model updated; next cycle scheduled | Automated |
Two design rules. Every step is an event, so a delay anywhere is visible immediately rather than at run-out. And the buffers are real: shipping by run-out minus three absorbs a weekend and one failed delivery attempt. Programs that run this loop typically see lapse rate fall within the first full cycle, because most lapses were never clinical — they were scheduling.
Where agents fit
The check-in is the step most often missed, and it is exactly the step an agent does well: recognize the supply clock, reach the patient where they are, collect the protocol’s follow-up conversationally, and submit the renewal with a summary the clinician can confirm in a minute. The clinician still decides; the agent makes sure the decision is asked for on time. Our guide to refilling medication with AI is the detailed version, and the titration entry covers why the check-in matters clinically for GLP-1 programs.
Metrics
- Lapse rate, monthly, with the distribution of lapse days.
- Check-in completion by run-out − 5, the leading indicator for next month’s lapses.
- Renewal turnaround, submission to authorization, p90.
- Ship-by adherence, the share of renewals shipped by run-out minus three.
- Lapse-attributed churn, cancellations within thirty days of a lapse. This is the number that makes the case for the work.
Lithos emits the moments the supply model needs — encounter decisions and order status (created, processing, completed) — as webhooks, and every prescription carries days_supply and a state that flips to depleted, so run-out is computable per patient; and the AI Care Coordinator add-on runs the step most programs miss: adherence check-ins, side-effect questions, titration guidance, and refill coordination for chronic-care programs, with physician escalation when clinically needed. Programs on Lithos measure lapse because they can. Next in the series: migrating a live patient base without a refill gap.
Frequently asked questions
What is refill lapse rate?
The share of active patients in a period whose medication supply ran out before their next shipment arrived. It is computed per patient from the last shipment date plus days’ supply against the next delivery date, then aggregated.
Why does refill lapse matter more than churn?
Because it comes first. Most churn in subscription telehealth follows a lapse: the patient ran out, lost momentum or symptoms returned, and cancelled. Lapse is measurable weeks before the cancellation and is fixable; churn is a lagging indicator you can only count.
What causes refill lapses in telehealth programs?
Five things account for most: the follow-up check-in was never completed, the clinician review queue was slow, the payment failed, the pharmacy or carrier was late, or the patient’s address or state changed. Each has a different owner and a different fix.
How do you prevent refill lapses?
With a cadence engine keyed to each patient’s supply: trigger the check-in before supply runs out, authorize renewals within hours, ship with a buffer before run-out, and treat every delay as an event someone acts on. Programs that do this see lapse rate fall within the first cycle.
How is lapse rate different from adherence?
Adherence is whether the patient takes the medication they have. Lapse is whether they have it. A program can have perfectly adherent patients who lapse because the next shipment was late — that is the program’s failure, not the patient’s.
Get the Journal by email
Our best guides on building compliant telehealth programs — a couple a week, unsubscribe any time.
Put a medical practice behind your product
Book a call and we'll scope your program together, or get free sandbox access and start building.