The Sentence Nobody Wrote: Why Auto-Generated EMR Text Is an Audit Liability

Read Your Note Like a Reviewer Would

Pull up a signed wound care progress note and read it for authorship rather than clinical content. Ask one question of every sentence: who actually wrote this?

In most wound care EMRs, a meaningful number of sentences have the same answer — nobody did. The software did.

"Patient tolerated the procedure well." "Hemostasis achieved with direct pressure." "No known drug allergies." "Advanced therapy is medically necessary and indicated per LCD coverage guidelines for this non-healing wound."

Those sentences look like documentation. Often they're defaults — text a template prints when the underlying field is blank. And under the rules Medicare reviewers actually follow, that distinction decides whether a note defends a claim or undermines it.

What CMS Actually Says About Templates

The relevant text isn't in an LCD. It's in the Medicare Program Integrity Manual, Pub 100-08, Chapter 3 — the chapter that instructs contractors on how to review a medical record. Three passages from §3.3.2.1.1, "Progress Notes and Templates":

"Some templates provide limited options and/or space for the collection of information such as by using 'check boxes,' predefined answers, limited space to enter information, etc. CMS discourages the use of such templates. Claim review experience shows that limited space templates often fail to capture sufficient detailed clinical information to demonstrate that all coverage and coding requirements are met."

"Templates designed to gather selected information focused primarily for reimbursement purposes are often insufficient to demonstrate that all coverage and coding requirements are met."

"...supplier prepared statements and physician attestations by themselves do NOT provide sufficient documentation of medical necessity, even if signed by the ordering physician."

Read that third line twice. The attestation at the bottom of the note — the signed statement that the service was medically necessary — proves nothing on its own. The reviewer is looking past it for clinical facts the clinician recorded.

That reframes what auto-generated text does. It doesn't help. A sentence the software wrote adds no evidentiary weight toward necessity, because there's no clinician behind it. And the same sentence appearing on every note is the specific pattern MACs describe as cloned documentation — treated not as neutral filler but as a misrepresentation of medical necessity, with denial and recoupment as the remedy.

Generated boilerplate lands in the worst quadrant available: no upside on proving necessity, real downside as a cloning pattern, and the clinician's signature underneath it.

The Four Patterns to Look For

Four distinct failure modes show up in wound care documentation. If you audit your own notes, look for these.

1. Defaults dressed as findings

A blank field prints a clinical finding. Empty allergy list → "NKDA (No Known Drug Allergies)." Blank tolerance → "Tolerated well." Blank hemostasis → "Achieved with direct pressure." Blank dressing → "Applied per wound care protocol."

The most consequential version is a defaulted tissue depth on debridement. Depth is what separates CPT 11042 from 11043 and 11044. When a blank field silently prints a depth, a value nobody entered is participating in code selection.

2. Conclusions software isn't entitled to make

"These findings establish medical necessity for ongoing skilled wound management." "E/M service is significant and separately identifiable from the procedure(s) performed on the same date of service."

The second one matters most. Modifier 25 is a standing audit target in wound care, and "significant and separately identifiable" is precisely the phrase a reviewer looks for as evidence the clinician thought about it. Printed automatically on every same-day visit, it manufactures the exact evidence the reviewer is trying to test — identically, every time.

3. Rows answering questions the software never asked

Some notes carry rows like "Pathology sent: No (selective/subcutaneous; not required)." Then look for the screen where a clinician records that answer. If there isn't one, the note is asserting a clinical decision — under a signature — that no clinician was ever given the chance to make.

4. Citations pointing at the wrong policy

LCD numbers printed next to billed services are supposed to signal that the clinician knows the standard. A wrong one signals the opposite. Policy IDs for tracheostomy supplies, seat-lift mechanisms, or DME surgical dressings have no business on a wound care note, and retired or nonexistent IDs are worse than no citation at all.

Every policy number on a note should be current in the Medicare Coverage Database, should actually govern the service on that line, and should carry a verification date somewhere in your system — not be typed into a sentence and left there.

The Rule Worth Adopting

One sentence, and it governs everything else:

The note states only what was documented. Where a field is blank, the note says "Not documented."

This is harder to accept than it sounds, because it makes notes shorter and gaps visible. A page with six "Not documented." lines looks worse than a page of confident prose.

It defends better. A reviewer reading "Not documented." learns something true about the practice: this clinician records what they observe and doesn't pad. A reviewer reading "tolerated well" on all forty of your notes learns something else entirely.

Three Rules More Specific Than Most Teams Realize

Going back to primary sources produces some surprises. Three corrections worth passing along, because all three get misunderstood in the field.

Home visits do not require homebound status

There is no homebound requirement for physician home visits (99341–99350). Noridian states it plainly: "There is no requirement that patient must be homebound." Homebound is a condition of the home health benefit, not of a physician's home visit.

Any template that auto-generates a homebound paragraph — "cannot leave home without considerable and taxing effort" — is adding an unsupported assertion to defend a requirement that doesn't apply to the service being billed.

The "in lieu of an office visit" rule was eliminated in 2019

Plenty of wound care documentation still asks clinicians to justify why a visit happened at home rather than in an office. That requirement is gone. CMS Transmittal R4339CP (CR 11273), implemented 2019, amending Pub 100-04 Chapter 12 §30.6.14.1:

"The requirement that a medical record document is necessary to demonstrate the medical necessity of the home visit made in lieu of an office or outpatient visit is eliminated."

What is required is per-encounter necessity — once per visit

What replaced it is stricter in a different direction. Noridian's home and domiciliary visit policy:

"Visit will be regarded as a social visit unless medical record clearly documents medical necessity for every visit."

"A payable diagnosis alone does not support medical necessity of ANY service."

Every visit — not every wound. That distinction has real design consequences. Necessity for the encounter is one fact about the visit, documented once and printed with the E/M documentation. Asking a clinician the same question three times on a three-wound visit produces three near-identical paragraphs, which is how a well-intentioned prompt turns into a cloning pattern.

What Good Documentation Software Does Instead

Deleting fabricated sentences leaves a gap, and the gap is real: clinicians document at the bedside, often standing, often between stops. A blank narrative box at the end of a long visit gets skipped.

The answer isn't to fill it automatically. It's to ask at the one moment when fixing it is still cheap — after the billing codes are staged, before the note is created.

In Medipyxis, that's a single line: "3 documentation items open — Resolve." Opening it shows one screen listing each open item, what's missing, which wound it concerns, and one button per item.

Items that belong on a form jump straight to that screen — and to the right wound.
Items that need a sentence open a box showing the facts already documented for that visit, with room to answer in the clinician's own words. Dictation is available instead of typing.
Where an example is offered, it's assembled only from facts already in that visit's record and checked before it's displayed — a draft that invents a number, asserts a cause, negates something not recorded as absent, or uses conclusion language is discarded rather than shown. It is never placed into the note automatically. The clinician copies it, edits it, inserts it, and the note records that they did.

Nothing blocks signing; every item is advisory. The one deliberate piece of friction: if an open item affects what can be billed, continuing past it asks for an explicit acknowledgment — a decision that stays with the clinician and is recorded.

The Half Most Teams Skip: Cloning Detection

Removing generated text solves one half of the problem. The other half is text a human wrote once and the system carried forward forever.

Carry-forward is genuinely useful — nobody wants to retype a wound history every visit. But a necessity note word-for-word identical to last visit's is the textbook cloning pattern, and MACs treat it as misrepresentation of medical necessity rather than a formatting quirk.

So the note compares each wound's narrative against the same wound's narrative from the last signed visit and flags an exact match — not to block anything, but to tell the clinician this paragraph looks copied, and a reviewer will read it that way. Updating it or clearing it both resolve it.

The subtle part: a carried-forward value nobody re-examined is an unreviewed copy, not a clinical judgment. So when the system suggests a visit-over-visit assessment — improved, worse, unchanged — it never pre-selects "unchanged" from a value the clinician didn't touch this visit. It suggests "not assessed" and makes the provider choose. A reviewer should be able to tell the difference between a finding and a copy, which means the software has to be able to tell the difference too.

Audit Your Own Notes in Twenty Minutes

You don't need to change software to run this. Pull three signed notes from different patients on different dates, put them side by side, and look for three things.

1. Identical sentences across all three. Especially anything containing "tolerated," "well," "per protocol," "medically necessary," or "separately identifiable." If the same sentence appears on all three, find out which field produced it — and what it prints when that field is blank.

2. Assertions with no matching screen. Pick any definite statement in the note and try to find where in the software a clinician enters it. If you can't find the field, the software wrote it.

3. Policy citations. Take every LCD or policy number printed on the note and look it up in the Medicare Coverage Database. Confirm it's current, and confirm it governs the service on that line.

Then ask your vendor one question, for any sentence in the note: which field produced this, and what does it print when that field is empty? If the answer is that it prints a clinical finding, you've found a liability sitting under your clinicians' signatures.

FAQ

Is auto-generated documentation illegal? It isn't illegal. The issue is evidentiary. CMS Pub 100-08 Chapter 3 §3.3.2.1.1 tells reviewers that reimbursement-focused templates and standalone attestations are insufficient to demonstrate medical necessity. Generated text adds no weight toward proving a service was necessary, and repeating identically across notes creates a cloning pattern MACs associate with denial and recoupment.

Isn't "Not documented." worse than a default sentence? For a reviewer, no. A default sentence is an unsupported assertion carrying a clinician's signature. "Not documented." is an accurate statement about the record. One can be challenged as fabrication; the other shows a gap — which is also the fastest way for a practice to see what it should be capturing.

Do wound care home visits require the patient to be homebound? No. Noridian states there is no homebound requirement for home and domiciliary visit codes. Homebound is a condition of the home health benefit, not of a physician home visit. What is required is documented medical necessity for each encounter.

What should we do about notes already signed with boilerplate in them? Signed notes are permanent records and should never be silently rewritten. The correct remedy is an addendum identifying the specific statement being corrected and why, so the correction reads as a precise amendment rather than a vague edit.

Does removing generated text make notes less complete? Shorter, not less complete. Nothing a clinician documented is removed — only sentences no clinician wrote. Where a note gets shorter, that's a gap that already existed and was previously hidden.

The Takeaway

A progress note is a legal record of what a clinician observed and decided. Every sentence in it should trace back to a person. When software fills silence with plausible clinical language, it isn't helping the clinician — it's adding unsupported assertions to a document they sign, in exactly the pattern Medicare reviewers are trained to look for.

The fix is straightforward, if uncomfortable: print what was documented, say "Not documented." where nothing was, and ask for the two or three sentences that actually carry weight at the moment they're still easy to write.

Want to see documentation guardrails working at the point of care? Book a walkthrough or read more about wound care charting and LCD compliance.