Charting by exception: how to read WDL entries and blank flowsheet cells in a nursing negligence case

Charting by exception (CBE) is a bargain. The facility writes down in advance what "normal" means for each part of a nursing assessment: lungs, heart, neuro, skin, pain, sedation, the wound. If the patient meets that standard, the nurse marks the row with a checkmark or "WDL" and moves on. If not, the nurse charts the finding in detail. That entry is the exception, and it is the only narrative the system asks for.

An early description, "Charting by exception: meeting the challenge of cost containment" (Murphy, Beglinger and Johnson), ran in Nursing Management in 1988. Electronic flowsheets made the idea standard, and most inpatient nursing documentation you read in discovery is now a flowsheet grid, whatever the policy is called. So the useful question is rarely "did this hospital use CBE?" It is "what did each mark and each blank mean under the policy that governed this unit on this date?"

The vocabulary you need before the first deposition

WDL

Within defined limits. The finding met the facility's written definition for that row. WDL points at a document.

WNL

Within normal limits. Older and looser, often with no written definition behind it. When a chart says WNL, ask whose normal.

Defined normal (parameters)

The written standard for a row, for example "respirations 12 to 20 a minute, regular, unlabored". It may live in a policy binder, in EHR hover text that never prints, or both.

Flowsheet row

1 line of the grid: a body system, a scale score, a vital sign. Each row has an EHR build definition with its allowed values and any help text.

Exception note

The narrative a nurse writes when a finding departs from the defined normal, often linked from an asterisk, "see note" or comment icon.

Assessment frequency

How often a row must be reassessed, set by policy, protocol or order. The single most important number in a CBE case.

Focused reassessment

A narrower recheck of 1 system between full assessments, often after a medication or procedure.

Copy forward

An EHR feature that carries the last value into the next column. A copied WDL looks exactly like a fresh one on paper.

Device-validated value

A number that flowed from a monitor and was accepted by a nurse. It proves the monitor read something, not that anyone looked at the patient.

What CBE is not

CBE is not permission to skip assessments. It covers documentation of normal findings only. The assessment still has to happen at the required frequency, and a changed condition still has to be charted and, where policy or the standard of care calls for it, reported to a provider.

No federal regulation permits or forbids charting by exception. Experts measure a CBE chart against general record and nursing rules, then the facility's own policy, then the standard of care.

"All patient medical record entries must be legible, complete, dated, timed, and authenticated in written or electronic form by the person responsible for providing or evaluating the service provided, consistent with hospital policies and procedures."

Plaintiffs quote "complete". Defendants quote "consistent with hospital policies and procedures". The regulation defers to the policy, which is why the policy is the case. The nursing Condition of Participation makes that policy binding on the nurses:

"All licensed nurses who provide services in the hospital must adhere to the policies and procedures of the hospital."

If the policy says sedation is reassessed every 2 hours on a patient-controlled analgesia pump, a 4-hour gap departs from a rule the hospital wrote for itself, before any expert speaks to the standard of care.

Joint Commission standards and professional guidance also get cited, but their wording changes by edition, so get the version in effect from your expert. A nursing home chart also has a federal assessment calendar that a hospital chart lacks, which gives you fixed dates to check.

"Not documented, not done" is a teaching maxim, not a statute or rule of evidence. It captures a jury intuition: a careful professional would have written it down. CBE is built on the opposite premise. A nurse who found everything normal was told not to write it down.

Rule 803(6) admits regularly kept business records. Rule 803(7) lets a party use the absence of an entry as proof:

"Evidence that a matter is not included in a record described in paragraph (6) if: (A) the evidence is admitted to prove that the matter did not occur or exist; (B) a record was regularly kept for a matter of that kind; and (C) the opponent does not show that the possible source of the information or other circumstances indicate a lack of trustworthiness."

Condition (B) is where CBE changes the analysis. Under CBE, the regular record of a normal finding is the checkmark, and of an abnormal finding the exception note. That produces 2 arguments from the same rule:

• A blank where the policy required a mark (a row due at 03:00 with nothing in it) fits condition (B) well: the facility regularly recorded that assessment, and it is not there.
• The absence of an exception note fits condition (B) for the defense: the facility regularly recorded abnormal findings, none is recorded, so the argument goes that nothing abnormal was found.

Rule 406 then lets either side offer routine practice:

"Evidence of a person's habit or an organization's routine practice may be admitted to prove that on a particular occasion the person or organization acted in accordance with the habit or routine practice."

This is the defense's bridge across a blank cell: "I always assess every 2 hours on a PCA patient." Whether that is admissible habit is a fight for the evidence motions. Many states have their own versions of 406 and 803(7) with different wording.

What a blank means under each system

Every article on CBE and the law cites Lama v. Borras, 16 F.3d 473 (1st Cir. 1994). A 1994 nursing-law newsletter ran it as "Court holds 'charting by exception' policy negligent," and that framing stuck. The opinion is narrower and more useful.

What the court actually decided

The jury returned a $600,000 compensatory verdict against the surgeon and the hospital, and the First Circuit affirmed the denial of judgment as a matter of law. On the hospital, 4 points carry the holding:

1. Duty and breach were conceded. The hospital did not contest that a 1986 Puerto Rico Department of Health regulation required qualitative nurses' notes each shift, or that its nurses followed the CBE policy instead. The regulation was not even in the appellate record (footnote 10).
2. Causation was the only live question. The judge instructed that violating a regulation "is not in itself enough to constitute negligence absent proof of the existence of the proximate cause" (footnote 11).
3. The evidence of what went unrecorded was specific. A former nurse testified that under the policy she would not chart pain if she gave no medication or only an aspirin-type drug.
4. The delay was quantified. The jury could infer a delay of at least 24 hours and perhaps 72, and an expert testified that 24 hours can make a difference in controlling an infection.

"... the jury could have reasonably inferred that intermittent charting failed to provide the sort of continuous danger signals that would be the most likely spur to early intervention by a physician."

How to read it

For plaintiffs, Lama is the cleanest statement of the CBE causation theory: charting only exceptions can hide a trend, and a hidden trend can delay the physician. It is also a model of proof: a regulation, a witness who described what the policy let nurses leave out, and a timeline that turned the gap into hours of delay.

For defendants, Lama is weak authority against CBE as such. It is 1 federal appellate case applying Puerto Rico law, on the deferential standard for overturning a verdict, and the breach rested on a regulation the hospital did not dispute. The court agreed with the trial judge that the case was "by no means the strongest proposition for medical malpractice against ... a hospital." A modern flowsheet with written definitions, frequent reassessment and escalation rules is a different system from the 1986 practice.

We did not find a later published appellate opinion on charting by exception that we could verify. If opposing counsel cites one, read the opinion, not the summary.

Printed flowsheets were designed for a screen, then flattened onto paper by a print routine nobody in the courtroom has seen. The mock page below shows the 7 things to check on every page before you draw an inference from a blank.

A reading procedure that holds up in a deposition

1. Identify the unit, the date range and the policy family. Med-surg, ICU and nursing home sheets run on different rules.
2. Find the governing frequency for each row. Policy, protocol, then physician orders; the tightest applies. Write it on your working copy: "sedation q2h on PCA per order 11/14 19:10".
3. Lay out the expected columns. List every time an assessment was due. This skeleton is how you tell a gap from a quiet stretch.
4. Fill the skeleton from the flowsheet. For each due time: charted, late, device only, or blank. Cite the page for every cell.
5. Pull every exception note into the same timeline. Narrative and event notes, the MAR, pump and rapid response records, call and paging logs.
6. Mark trends inside the limits. Values that stay WDL while moving in 1 direction (chapter 6).
7. List the questions only the policy or metadata can answer. That list becomes your discovery request.

CBE grades each assessment on its own. A respiratory rate of 16, then 14, then 12 is WDL 3 times if the definition says 12 to 20. Oxygen saturation of 97%, 95%, 94%, 93% is WDL 4 times if the floor is 92%. No single cell triggers a note. The trend lives only in the numbers.

That is the Lama theory in a modern chart: the numbers are there, but nobody wrote down the direction. Plaintiffs should chart the trend themselves, cell by cell, with page cites. Defendants should be ready to explain what the policy told the nurse to do about direction as opposed to thresholds. Many policies say nothing.

Frequency itself is evidence

Nurses change their behavior before they change their notes. A study at a large urban academic medical center tested that.

Collins and colleagues (American Journal of Critical Care, 2013) found that optional comments and vital signs beyond what was required were associated with death and cardiac arrest, reflecting nurse concern. Extra vital signs, all WDL, with no exception note, suggest someone was worried and did not write down why. Neither pattern proves the standard of care by itself.

Copy forward: when WDL is a keystroke

The best-known measurement of copying comes from UCSF physician progress notes:

In a flowsheet the equivalent is the "previous value" function, and a WDL entry has no wording to compare. If the same 9 rows read WDL every 4 hours for 3 days, the print cannot tell you whether a nurse assessed 18 times or pressed a key 18 times. Many EHRs store how a value was entered. Ask for it by name (chapter 9).

Scale rows hide trends too

The Braden scale for pressure injury risk has 6 subscales, with cumulative scores from 6 (highest risk) to 23 (lowest risk), per Alderden and colleagues (2017). A score drifting down 2 points a day may cross no threshold for days. The case often turns on the first date it crossed the facility's threshold, which the facility sets itself, and whether the ordered interventions followed.

The 1 narrative note

Page 219, a nursing note labeled "late entry", entered 06:25 for 05:40: "Pt difficult to arouse on rounds, RR 8, SpO2 86% on 2L NC. PCA paused, HOB raised, sternal rub, RRT called 05:44. Naloxone given per RRT 05:46. Pt restless and snoring since approx 03:30, repositioned x2."

The last sentence is the most important line in the chart for both sides. It puts the nurse at the bedside during the gap, and it records restlessness and snoring that a nurse expert may say called for a sedation assessment. None of it reached the flowsheet.

The reconstruction

How each side reads it

Both readings need what neither has yet: NUR-DOC-114 as of 11/14/2025, the sedation scale's action threshold, the monitor's alarm limits and history, how the 04:12 value was validated, and whether any 01:00 values were copied forward. Chapter 9 turns that list into a request.

Run every blank cell through the same questions in the same order, and log the answer. Each step can end the inquiry.

What a finished log looks like

For the chapter 7 hypothetical, the sedation and respiratory log has 6 due times:

FLOWSHEET GAP LOG
Matter: [CASE NAME]        Patient: [INITIALS]        Unit: [UNIT]
Date range: [MM/DD/YYYY HH:MM] to [MM/DD/YYYY HH:MM]
Reviewer: [NAME]           Prepared: [DATE]

ROW REVIEWED: [e.g. Sedation score]
Frequency: [q2h]   Source of frequency: [policy no./rev. date, protocol, or order + page]
Defined normal for this row: [text, or "POLICY PENDING"]
Policy meaning of a blank cell: [text, or "NOT STATED" / "PENDING"]

Due time | Charted? | Filed time | Filed by | Value | Page | Other source in window (type, time, page) | Status
[03:00]  | [No]     | [n/a]      | [n/a]    | [n/a] | [214]| [Note 05:40 describes 03:30 bedside, p. 219] | [Presence, no assessment]

STATUS CODES
  Charted on time / Charted late (>[X] min) / Device value only /
  Gap, covered by other source / Presence, no assessment /
  Gap, unexplained / Not due / Off unit / Policy pending

TRENDS INSIDE LIMITS (value, time, page for each point)
  [Row]: [16 @19:00 p.212] -> [14 @21:00 p.212] -> [12 @23:00 p.213]

EXCEPTION NOTES
  Flag on flowsheet (time, page) | Linked note found? (page) | Note time vs filed time | Provider notified? (page)

QUESTIONS FOR THE POLICY / METADATA / DEPOSITION
  1. [e.g. How was the 04:12 SpO2 validated, and by whom?]

A HIPAA access request or a records subpoena gets you the chart, not the policy. The right of access in 45 CFR 164.524 reaches the patient's designated record set, and a nursing policy is not patient information. The policy, defined normals and flowsheet build come through party discovery, a subpoena, or a state pre-suit statute. Our guide on how to get medical records for a lawsuit covers the chart side.

Ask in the first request for production, not after the nurse's deposition. In federal court the other side has 30 days to respond under FRCP 34(b)(2)(A). A request for "all policies and procedures" produces the current binder, which may be years newer than the care. Name the date.

Why the version in effect is the whole request

When a facility revises its defined normals, the same checkmark changes meaning. Here is a hypothetical revision of the chapter 7 respiratory row.

Under the 2021 text, the 04:12 saturation of 91% is an exception that called for a note, and with no 2-hour rule the 03:00 blank is not a gap. Under the 2024 text the 91% is WDL and the blank is a gap. Ask for every version from a year before the care to the present, with effective dates.

DEFINITIONS
"Date of Care" means [MM/DD/YYYY] through [MM/DD/YYYY].
"Unit" means [UNIT NAME / NUMBER] and any unit where [PATIENT] received care
during the Date of Care.
"Flowsheet" means any structured nursing documentation template, grid or
row-based form, paper or electronic, used on the Unit.

REQUEST NO. [__]
Every version of each policy, procedure or protocol governing nursing
documentation on the Unit in effect from [DATE 12 MONTHS BEFORE] to the
present, including any policy on charting by exception, flowsheet
documentation or "within defined limits" (WDL) charting, with effective,
revision and retirement dates.

REQUEST NO. [__]
The written definitions of "within defined limits," "within normal limits"
or any equivalent entry for each assessment row in [PATIENT]'s Flowsheets,
as in effect during the Date of Care, including EHR help text, hover text
or row descriptions.

REQUEST NO. [__]
Documents sufficient to show the build of each Flowsheet row in [PATIENT]'s
record during the Date of Care: allowed values, embedded definitions,
defaults, copy-forward settings and device-integration settings.

REQUEST NO. [__]
Every policy, protocol or standing order in effect during the Date of Care
setting the frequency of assessment for: vital signs; pain; sedation during
opioid therapy [or PCA]; neuro checks; skin; fall risk; [OTHER ROWS].

REQUEST NO. [__]
Any policy in effect during the Date of Care on the meaning of a blank
Flowsheet cell, and on late entries, addenda and corrections.

REQUEST NO. [__]
The legend for every symbol or marker on the printed Flowsheets produced.

REQUEST NO. [__]
Training and competency records for [NURSE NAMES / ROLES] on the policies
above, dated within [24] months before the Date of Care.

REQUEST NO. [__]
Monitor data, alarm settings and alarm history, and infusion or PCA pump
history for [PATIENT] during the Date of Care, in native or exported form.

REQUEST NO. [__]
For each Flowsheet value from [TIME] to [TIME] on [DATE], data sufficient
to show when it was entered, by which user, and whether it was typed,
selected, copied forward or accepted from a device.

The last request overlaps with EHR metadata discovery; see the EHR audit trail in medical malpractice guide, and altered medical records if an entry looks changed after the fact.

Depose the facility on the policy

Only the organization can tell you what its policy meant. Under FRCP 30(b)(6) you "describe with reasonable particularity the matters for examination", and the designee must testify about information "known or reasonably available to the organization." Lock in definitions before the individual depositions.

MATTERS FOR EXAMINATION, [FACILITY NAME]

1. The nursing documentation policies for [UNIT] during [DATE OF CARE],
   their approval, effective dates and revisions from [DATE] to present.
2. The meaning under those policies of "WDL", a checkmark, "see note",
   a blank cell, and each symbol in the Flowsheet legend.
3. The written normal-finding definition for [LIST ROWS], and where nurses
   could see it (policy manual, EHR help text, other).
4. Required assessment frequencies for [LIST ROWS] for a patient in
   [PATIENT]'s condition, and the source of each.
5. How the Flowsheet system records when a value was entered, by whom, and
   whether it was typed, selected, copied forward or accepted from a device.
6. Criteria for charting an exception and notifying a provider.
7. Training and competency validation for nurses assigned to [UNIT].
8. Audits of Flowsheet documentation on [UNIT] in the 12 months before
   [DATE OF CARE], to the extent not privileged under [STATE STATUTE].
9. The search for and production of documents responsive to
   [REQUEST NOS.].

Before you serve it

• Date of Care defined to the hour where it countsShift changes and transfers split coverage.
• Every unit the patient passed through namedED, PACU, ICU and med-surg run different flowsheets.
• Version window runs from before the care to the presentYou want the change history too.
• Embedded EHR definitions asked for by nameHover text is not in the binder.
• Meaning of a blank cell requested separatelySilence in the policy is also useful.
• Frequency sources listed row by rowPolicy, protocol, order set, order.
• Legend for print symbols requestedSymbols differ across systems.
• Device and pump data requested in native formThey often sit outside the chart and rotate data.
• Entry-method metadata requested for the key window onlyNarrow is harder to resist on proportionality.
• Peer review and quality privilege checked for your stateThe policy itself is usually not protected.

0 of 10 checked

The defense usually runs 3 arguments together: each WDL is an affirmative record of a normal assessment under a written definition; the nurses followed a routine practice of assessing on schedule (Rule 406 or a state equivalent); and CBE is an accepted method. The paper behind the grid decides which of them holds.

Which facts make a CBE defense hold

This order is our view of how juries weigh the evidence, and it tracks Lama. Defense counsel should audit the chart against it before deciding whether to defend the documentation or concede it and fight causation.

Where plaintiffs attack

• The definition. Vague, missing, not in force that day, or never shown to the nurse.
• The frequency. Due times with nothing charted, or batch-charted at shift end.
• The trend. WDL values moving in 1 direction with no note.
• The exception. Late, unescalated, or contradicted by other sources.
• The keystroke. Identical WDL entries across shifts, filed seconds apart.

A. BACKGROUND AND TRAINING
 1. When were you trained on [FACILITY]'s flowsheet documentation? By whom?
 2. Were you shown the written definitions of WDL for each row? Where were
    they kept? (policy manual, EHR help text, unit binder)
 3. When did you last read them before [DATE OF CARE]?

B. THE DEFINITIONS (show Exhibit [__], policy rev. [DATE])
 4. What did WDL mean for the [ROW] row on [DATE]? Read it into the record.
 5. What does a blank cell mean under this policy?
 6. What does [SYMBOL, e.g. (V)] mean on the printed flowsheet?

C. FREQUENCY
 7. How often were you required to assess [ROW] for this patient? Source?
 8. Show Exhibit [__] (gap log). At [03:00], what did you assess?
    Where is it recorded?
 9. Is there any record of an assessment at [03:00] other than your memory?

D. THE ENTRIES
10. For the [01:00] column: did you assess each row, or carry values forward?
11. How does the "previous value" function work on your unit?
12. The [04:12] SpO2 value: how was it validated? Did you see the patient?
13. When did you chart the [05:40] entry? Why was it a late entry?

E. THE NOTE (Exhibit [__], p. [219])
14. You wrote "restless and snoring since approx 03:30." What did you see?
15. Did you assess sedation or respirations at that time? Where is it?
16. Did you notify a provider before [05:40]? Where is that recorded?

F. MEMORY VS HABIT
17. Do you remember this patient, or are you describing your usual practice?
18. How many patients were you assigned that night? (staffing record)

A 5-day admission can produce hundreds of pages of grids, with the deciding facts scattered across them: a blank due time on page 214, the pump history on page 238, 1 sentence in a late entry on page 219. Legal AI tools are good at the finding half of that work. Do not trust them with the judging half.

What a large language model does well here

• Putting every entry in time order. An AI medical chronology that merges flowsheet values, notes, the MAR and device records, each line cited to its source page, builds the chapter 7 timeline in minutes instead of days.
• Finding the exception notes. Clinical natural language processing picks out "late entry", "RRT called" and "provider notified" across thousands of pages.
• Surfacing date and time gaps. Stretches with no entries, late-filed entries, notes that contradict the grid.
• Answering narrow questions with citations. "List every sedation score from 19:00 to 07:00 with page numbers" is a good prompt. "Was the nurse negligent?" is not.
• Expanding abbreviations. HOB, NC, RRT, POSS and unit shorthand; see medical abbreviations for legal professionals.

Where it fails

• OCR on grids. Optical character recognition can shift a value into the wrong column on a dense or faxed flowsheet, and the wrong column is the wrong time.
• Blank versus unreadable. A faint cell read as empty creates a gap; an empty cell filled from its neighbor erases one. Low-confidence pages must go to a person.
• Handwritten notes and paper downtime forms. Handwriting recognition still needs someone who knows the unit's shorthand.
• What the print does not carry. Hover-text definitions, copy-forward metadata, device validation. No tool reading a PDF can see them.
• Hallucination. Ask a general chatbot what WDL means in this chart and generative AI will offer a textbook definition as if it were the facility's.
• Judgment. Breach, credibility of habit testimony, whether a trend called for escalation: expert and counsel questions.

Why page-level citations decide whether the output is usable

In a charting-by-exception file, the question is always what the flowsheet shows on a given shift. Grounded, page-level citations send the nurse expert straight to that flowsheet, and a human-in-the-loop check of every row decides what the blank boxes mean. In Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023), lawyers were sanctioned under Rule 11 for filing case citations a chatbot had invented. The same duty to verify applies to a chronology an expert relies on or a gap log attached to a motion.

Ambient AI scribes now draft some clinical notes from recorded conversation, and an accepted draft can read as complete while describing an assessment in general terms. Ask in discovery whether any AI drafting tool was used on the unit, and check a scribe-drafted note's times against the grid.

Choosing legal AI tools for flowsheet work

The market for AI document review, technology assisted review and agentic AI tools for lawyers is crowded. For CBE work, rank requirements in this order:

See also is AI accurate enough for court, HIPAA-compliant AI medical record review and the AI-native legal nurse consultant manual.

Medrecords AI is medical record review software. You upload the records; it drafts a cited chronology, summaries and flags, every line linked to its source page. On a flowsheet chart:

• A cited medical chronology of flowsheet values, notes, medication and rapid response records in time order, with a citation on every line.
• Date and time gaps flagged: stretches with no entries, and records that should exist and were not produced, through missing records identification, each flag cited to its evidence.
• Inconsistencies flagged: a note that contradicts the grid, a late-filed entry, an entry dated outside the admission.
• Low-confidence pages marked for a person, including dense grids, faxes and handwriting, through OCR and handwritten record extraction.
• Questions answered with citations through Q&A: "every sedation score on 11/14 with pages".
• New productions compared to the file you have through supplemental record review.

Flags are signals, not verdicts. A time gap may be a missed assessment, a patient off the unit or a page missing from the production. A human decides which.

What it does not do: it does not know a unit's required assessment frequency or read the policy against the chart; the due-time skeleton is your step. It does not read EHR exports or entry-method metadata, so it cannot tell a copied WDL from a typed one. It does not decide breach, score case merit, give legal advice or retrieve records from providers. It works on the files you upload, under SOC 2 and HIPAA with a signed BAA (see security and HIPAA). Self-Service bills 10 cents a deduplicated page, down to 5 cents at volume, duplicates free; Enterprise On-Prem is an annual license.

See how this applies to hospital nursing negligence, nursing home neglect and legal nurse consultant work.

What is charting by exception in nursing?

A method in which the facility writes down what a normal finding is for each assessment area, and nurses chart in detail only findings that depart from it. The assessments still happen at the required frequency; only the documentation of normal findings is shortened.

What does WDL mean in a nursing chart?

Within defined limits: the finding met the facility's written definition for that row. You need the definition in force on the date of care, which may sit in EHR help text that never prints.

Does "not documented, not done" apply to charting by exception?

It is a nursing maxim, not a rule of law. A blank cell at a time an assessment was due is still a gap under CBE. Federal Rule 803(7) lets a party use an absent entry if that kind of record was regularly kept. State rules vary.

Is charting by exception negligent?

No court we found has held the method negligent in itself. Lama v. Borras (1st Cir. 1994) affirmed a verdict where the hospital conceded its CBE practice broke a Puerto Rico regulation, and the jury linked the missing notes to a delayed diagnosis.

What should I request to understand a CBE flowsheet?

Every version of the documentation policy from before the care to the present, the defined normals including EHR help text, frequency policies and orders, the policy on blanks and late entries, the symbol legend, training records, and monitor and pump data.

Can a nurse testify to her routine practice to fill a blank?

Often, yes, under Federal Rule of Evidence 406 or a state equivalent. The court decides whether the testimony qualifies as habit, and a jury may give it little weight against a missing entry.

Can AI read a nursing flowsheet?

AI tools can extract values, merge them with notes and medication records into a cited timeline, and flag time gaps. They struggle with dense grids after OCR, faxed cells and handwriting, and cannot see definitions or metadata that never printed. A person checks every cited value.

Can software tell whether a blank cell means the assessment was not done?

No. It can show that a cell is empty and nothing else in the file covers the window. Whether the assessment was due and happened depends on policy, other records and testimony.

Is it HIPAA compliant to upload medical records to AI?

It can be, with a vendor that signs a business associate agreement, holds a SOC 2 report, and commits in writing not to train on your data. A consumer chatbot without a BAA is a different matter.

Can ChatGPT summarize a flowsheet for a lawsuit?

It can produce a fluent summary, but without page-level citations you cannot check it, and it may fill gaps with plausible values. Lawyers have been sanctioned under Rule 11 for filing invented AI citations (Mata v. Avianca, S.D.N.Y. 2023). Verify every line before you rely on it.

Rules, regulations and the Lama opinion were read in full text from primary sources in September 2026; quoted text is verbatim. Study figures come from the published papers. The flowsheet, policy, patient, times and pages in chapters 5 and 7 to 10 are hypothetical. Product facts come from this site's product pages. Nothing here is legal or medical advice.

• Lama v. Borras, 16 F.3d 473 (1st Cir. 1994), decided February 25, 1994: facts, verdict, footnotes 3, 10 and 11, causation holding.
• Tammelleo, "Court holds 'charting by exception' policy negligent", Regan Report on Nursing Law, 1994.
• Federal Rule of Evidence 803(6) and (7) and Rule 406.
• Federal Rule of Civil Procedure 34(b)(2)(A) and Rule 30(b)(6).
• 42 CFR 482.24(c)(1) and 42 CFR 482.23(b)(4) and (b)(6), hospital Conditions of Participation.
• 42 CFR 483.20(b)(2), (c), (d), (g) and (j), nursing home resident assessment.
• 45 CFR 164.524, HIPAA right of access.
• Murphy, Beglinger and Johnson, "Charting by exception: meeting the challenge of cost containment", Nursing Management 19(2), 1988.
• Collins et al., "Relationship between nursing documentation and patients' mortality", American Journal of Critical Care 22(4), 2013: 15,000 acute care patients, 145 cardiac arrest patients.
• Wang, Khanna and Najafi, "Characterizing the Source of Text in Electronic Health Record Progress Notes", JAMA Internal Medicine 177(8), 2017: 23,630 notes, 18% typed, 46% copied, 36% imported.
• Alderden et al., "Midrange Braden Subscale Scores Are Associated With Increased Risk for Pressure Injury Development Among Critical Care Patients", J Wound Ostomy Continence Nurs, 2017: Braden range 6 to 23.
• Nurses Service Organization, "Charting by exception: the legal risks": insurer guidance to nurses on CBE documentation.
• Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023): Rule 11 sanctions for filing fabricated AI-generated case citations.

Related: legal nurse consultant reports and medical malpractice record review.