- Templates are not the problem. Reviewers expect structured notes and repeated section headings.
- The problem is identical clinical content — the same observations, the same response, the same plan, session after session.
- Cloning is usually a workload signal, not a dishonesty signal. Clinicians who are behind write from memory, and memory repeats.
- Catching it internally costs a supervision conversation. Catching it externally applies to every note in the range.
A clinician carrying twenty-eight sessions a week does not sit down on Friday intending to write the same note nine times. They sit down intending to catch up, and by the fourth note the sentences that worked are the sentences that appear.
This is how most cloned documentation happens. Not fraud — fatigue, plus a template that made repetition easy and a system that never flagged it.
What a reviewer is actually looking for
Medical necessity is established by showing that a specific patient, at a specific point in their treatment, needed the specific service that was billed. Every element of that sentence has to appear somewhere in the note.
When two notes nine days apart are word-for-word identical in their clinical content, the record fails that test twice over. It cannot show what changed, and it cannot show that the second session addressed anything the first did not. A reviewer does not need to allege that the session did not happen. They only need to conclude that the documentation does not support billing it separately.
Structured notes repeat by design. What has to differ is what the patient did, said, and responded to.
Where the line sits
| Expected to repeat | Expected to differ |
|---|---|
| Section headings and field structure | Presenting concern that session |
| Modality and duration where genuinely unchanged | Patient response to the intervention |
| Diagnosis and treatment plan reference | Observed affect, engagement, and content |
| Standard risk screening language | Progress against a named objective |
| Credential and signature block | Plan for the next session |
A note where everything in the left column repeats and everything in the right column varies is a well-templated note. A note where the right column repeats too is a clone, regardless of how good the care was.
Group notes are the highest-risk surface
A twelve-person group produces twelve notes describing the same session. The shared clinical content — the topic, the curriculum, the interventions delivered — genuinely is the same for everyone, and writing it twelve times is waste.
But if the individual response section is also shared, you have produced twelve identical records of twelve different people's treatment. That is the fastest way to lose a group claim, and it is why the shared content and the individual response should be structurally separate fields rather than one free-text box that gets duplicated.
Treat it as a supervision metric
The useful response to cloning is not a policy memo. It is a number that a clinical director can see weekly.
Score each signed note against the same clinician's recent notes for the same patient. Set a similarity threshold — 85% is a reasonable starting point — and route anything above it to a review queue with the two notes shown side by side. Most flags will resolve one of three ways:
- An amendment. The clinician adds what actually differed. Two minutes, and the record is defensible.
- A caseload conversation. The pattern is concentrated in one clinician who is carrying too much. That is a staffing decision, and it is cheaper than a recoupment.
- A template problem. The form has too much prefilled text and not enough required individualisation. Fix the form and the pattern disappears across everyone.
Similarity scoring lands badly if it is introduced as a fraud-detection measure aimed at clinicians. It works when it is introduced as what it is: an early-warning system that protects the people delivering care from a finding that will otherwise be discovered by someone with recoupment authority.
The asymmetry that makes this worth doing
An internal flag costs a review and, sometimes, an amendment. An external finding applies retroactively across the episode, can be extrapolated across a claim population, and arrives with a demand attached.
The two are not comparable, and the difference between them is a comparison that takes a computer a fraction of a second.