Documentation That Prevents Lawsuits — and the Kind That Creates Them

By John Mahoney · August 2026 · 11 min read

You write a note for the next clinician. It gets read, years later, by a lawyer, a nurse consultant, an opposing expert, and eventually twelve people who have never set foot in a hospital except as patients. Nothing about the way most physicians are taught to chart accounts for that second audience.

It matters more than most clinicians assume. Candello, the benchmarking arm of the Harvard-affiliated risk-management foundation CRICO, examined more than 65,000 malpractice cases closed between 2014 and 2023 for its 2024 report For the Record. Roughly 20% involved at least one documentation failure — and those failures more than doubled the odds that the case closed with an indemnity payment. The four that moved the odds most: alterations to the record, insufficient documentation of clinical findings, inadequate documentation of clinical rationale, and thin informed-consent documentation. Every one of those is a writing habit, not a clinical skill.

This is not an argument for defensive medicine, and it is not an argument for longer notes. Length is not defensibility. Some of the most indefensible charts in American medicine are enormous — forty lines of auto-populated review of systems wrapped around two sentences of actual thought. The chart that protects you shows a reasonable clinician doing reasonable things for reasons a stranger can follow. The chart that hurts you records conclusions without reasoning, repeats yesterday's findings verbatim, or gets edited after everyone knows how the story ended. Here is the practical version, with the actual language.

Rule 1: Contemporaneous beats complete

A modest note written during or immediately after the encounter is worth more than a thorough one written at 9 p.m. That is not because of any legal rule. It is because every modern electronic record keeps a second, invisible chart — the audit log — and that log knows when you actually typed.

The HIPAA Security Rule requires covered entities to implement mechanisms that record and examine activity in systems containing protected health information (45 C.F.R. § 164.312(b)). In practice, certified systems log user identity, action type, the object acted on, and a server-generated timestamp for essentially every interaction — and the audit-log standard incorporated into federal certification requires recording the time an entry refers to when that differs from when it was typed, along with copy-and-paste as its own action type. The clinical content carries the time you say something happened. The metadata carries the time you wrote it down. The gap between those two numbers is the single most litigated fact in documentation cases — and it is not a fact you can testify your way out of, because it is not your memory against anyone's. It is a database row. Attorneys on both sides now request these logs routinely; our companion piece on how audit trails are read shows exactly what the reviewer sees.

None of that makes late charting wrong. Clinicians chart late on bad nights, and everyone knows it. It makes late charting something to do honestly and visibly — which is Rule 5.

Rule 2: Document the reasoning, not the conclusion

This is the highest-yield change most physicians can make, and it usually costs one extra sentence.

A conclusion tells the reader what you decided. Reasoning tells the reader you were thinking. Standard of care is judged on the decision-making process available to you at the time — not on the outcome, which the reader already knows and cannot un-know. A note that records only the conclusion invites the opposing expert to supply the reasoning for you, and the reasoning they supply will not be flattering.

Creates exposure	Prevents it
"Chest pain, likely musculoskeletal. D/C home."	"Chest pain, reproducible with palpation, no exertional component, no radiation, ECG without ischemic change, HEART score 2. ACS considered and felt unlikely; PE considered, Wells low, no hypoxia or tachycardia. Discussed with patient; return precautions given for any change in character, exertional pain, dyspnea, or syncope."
"Abdominal pain improved. Stable for discharge."	"Pain now 2/10 after fluids, tolerating PO, abdomen soft without rebound or guarding, WBC 8.2, lactate normal. Appendicitis considered — Alvarado 3, no RLQ tenderness on serial exam at 1400 and 1730. Imaging deferred on that basis; explicitly discussed return for worsening or migration of pain."
"Will monitor."	"Will monitor with repeat neuro checks q2h and repeat Hgb at 0600; threshold for CT re-image is any change in GCS, new focal deficit, or persistent vomiting."

Notice what the right column does. It names the dangerous diagnosis, states that it was considered, and gives the basis for excluding it. That single move — documenting the differential you ruled out and why — converts "the physician missed it" into "the physician considered it and reached a defensible judgment on the information then available." Those are entirely different cases.

It also does something subtler. A jury forgives a wrong answer far more readily than an absent question.

Rule 3: The negative findings that actually matter

Not all pertinent negatives are equal. Auto-populated "denies fever, chills, night sweats" across fourteen systems is noise, and worse, it is discreditable noise: if any one of those is contradicted elsewhere in the chart, the whole block loses credibility along with the parts that were true.

The negatives worth typing by hand are the ones that map to the thing that could kill the patient:

The red-flag negative. "No saddle anesthesia, no urinary retention, normal rectal tone" in a back-pain visit. "No neck stiffness, no photophobia, no petechiae" in a febrile child.
The negative that documents a completed action. "Pedal pulses 2+ bilaterally post-reduction, sensation intact in all distributions" — proof the compartment check happened.
Informed refusal. When a patient declines your recommendation, document the recommendation, the specific risks you described, that the patient understood, and the alternative plan. "Patient declined admission for observation. Advised risk of arrhythmia and sudden death; patient verbalized understanding, wishes to go home. Return precautions and follow-up with cardiology in 48h arranged; family present." A declined recommendation you documented is a defense. The same refusal undocumented is a failure to recommend.
Discharge and return precautions, specifically. "Return precautions given" is a template. "Advised to return immediately for fever >38.5, worsening RLQ pain, or persistent vomiting; patient and spouse both verbalized understanding" is evidence. Failure-to-warn theories are common precisely because the generic version proves nothing.
Communication that happened outside the note. The page to the attending, the call to the consultant, the callback to the family: who, when, what was said, what was decided. Missing notification documentation is one of the most reliably case-making gaps in any chart.

Rule 4: Copy-forward is now the default, and that is the problem

Researchers at UC San Francisco analyzed 23,630 signed inpatient progress notes and traced the provenance of every character. Their finding, published as a research letter in JAMA Internal Medicine: about 46% of note text was copied, 36% was imported, and only roughly 18% was entered manually by the clinician (Wang, Khanna & Najafi, 2017). Four-fifths of the average progress note is not new writing.

Copy-forward is not fraud and it is not going away; it is how anyone survives a service with a census. But it produces three specific failure modes that reviewers find quickly:

The exam that did not happen. A detailed neurologic exam that is byte-identical for five consecutive days tells the reader exactly what occurred on days two through five. If the patient was intubated on day three, the copied exam is worse than useless — it is affirmatively false, and it contaminates the credibility of everything you documented that was true.
The stale problem. The assessment carrying forward "resolved cellulitis" while the flowsheet shows rising temperature curves is the contradiction that gets highlighted on a demonstrative exhibit.
Inherited error. A wrong allergy, a wrong laterality, or a misattributed history propagates through fifteen notes and now has fifteen physicians' signatures on it.

The safety literature has settled on a workable standard. ECRI's Partnership for Health IT Patient Safety issued four safe practices for copy and paste: make copied material identifiable, make its provenance available, train staff, and monitor the practice. At the individual level it reduces to one habit: copy the stable scaffolding, retype the assessment and the exam findings that changed. If your daily assessment is genuinely new writing, the rest of the copy-forward in the note is nearly harmless.

Rule 5: Late entries and addenda — done right, they are fine

Physicians hear "never change the record" and conclude they should never add to it. That is not the rule, and the fear causes worse behavior than the correction would.

The Centers for Medicare & Medicaid Services set out the standard for amendments, corrections, and delayed entries in Transmittal 732 to the Medicare Program Integrity Manual (the current manual, Chapter 3 § 3.3.2.5, states the same expectation in condensed form). Three requirements, and they are the whole of it:

Clearly and permanently identify the entry as an amendment, correction, or delayed entry.
Clearly indicate the date and the author of the change.
Clearly identify all original content, without deletion — on paper, a single strike-through that leaves the original readable, initialed and dated.

Applied to a real note, the defensible version looks like this:

"Addendum 03/14/26 0915, Dr. Mahoney: Delayed entry. At approximately 0210 on 03/13 I was called to bedside for hypotension and ordered a 1L bolus and repeat lactate; this was not charted contemporaneously due to a second emergency on the unit. Entered now from memory and from the medication administration record."

That addendum is honest, self-dating, and explains the delay. It is also almost impossible to cross-examine, because it concedes the only fact the metadata would have proved anyway.

The indefensible version is an edit to the body of an existing note days after an adverse outcome, adding reassuring detail, with no addendum label. The audit trail preserves the prior version and the edit timestamp; the reviewer builds a two-column before/after table; and the case stops being about medicine. As our piece on late entries and amended records lays out from the reviewing side, the amendment itself is often more damaging than whatever it was meant to fix. A documentation problem is survivable. A credibility problem generally is not.

One corollary: once you are aware of a claim, stop. Any addition at that point — however truthful — is read as strategic. Route it through your carrier and counsel instead. Our survival guide for named physicians covers that sequence.

Rule 6: The new one — signing an AI draft you did not read

Ambient documentation is now mainstream, and it introduces a risk that did not exist five years ago: the note in your chart may contain sentences no human ever wrote, ever read, and ever verified — carrying your signature.

The failure modes are distinctive: fabricated normal findings for exams that never occurred, omission of the one symptom that mattered, and misattribution of history from a family member to the patient. One validated evaluation of an LLM ambient scribe found hallucinations in 31% of ambient notes versus 20% of physician-written comparison notes (Palm et al., Front Artif Intell 2025) — note that the human rate is not zero either, which is the honest framing.

Your signature attests to accuracy regardless of who drafted it, and CMS has said so directly: guidance effective January 2025 in the Program Integrity Manual's signature section states that practitioner concurrence with the note "is also required when using Artificial Intelligence (AI) technology to capture the transcription of medical record entries." Meanwhile the metadata problem returns with force — time-in-note and edit-count data show whether the eleven-second review happened. We covered the downstream discovery consequences in how AI is quietly changing the medical record; the prevention version is short:

Read the objective section every time. That is where invented findings live.
Delete anything you did not personally perform or verify, even if it reads well.
Write the assessment and plan yourself. The reasoning is the part with legal weight and the part a model is least equipped to invent honestly.

What never belongs in the chart

Blame or criticism of another clinician. "Should have been transferred hours ago" creates a co-defendant, an admission, and a headline. Clinical facts about what was and was not done are appropriate; characterizations are not.
Speculation about causation. Record findings and actions. "Likely secondary to our delay in diagnosis" is a legal conclusion written by someone with no legal training and incomplete facts.
References to incident reports or peer review. Writing "incident report filed" in the chart can pull an otherwise protected document toward discoverability. Report through the safety system; keep the chart clinical.
Arguments with the patient. "Patient is drug-seeking and hostile" reads very differently to a jury than "patient requested additional opioid analgesia; discussed risks and declined to escalate; patient expressed frustration with plan."

The chart is only half of it. The other half is you, under oath.

Survive Your Deposition is a 15-module course for physicians — each module ends with a live AI drill where you answer the questions out loud and get scored on what you actually said.

See the course →

The bottom line

Defensible documentation is not more documentation. It is four habits: chart while it is happening, write the reasoning and not just the result, name the dangerous thing you ruled out and why, and never let the record say something happened that did not. Every one of those makes you a better clinician independent of any lawsuit — which is the only reliable sign that a medicolegal recommendation is worth following.

The chart you write on an ordinary Tuesday is the chart that will be read on the worst day of your professional life. Write it so that the person reading it can see you thinking.