Reconciling a defense record production: the page-level method for medical malpractice teams

An electronic health record has no pages. It is a database, and every copy of a chart is a report someone ran on a given day, with a given template, against a given scope. Your pre-suit set is 1 such report. The defense production is another, run a year later by different people for a different purpose. They will not match page for page.

That explains most of what you will find. The rest is why the work is worth doing: an addendum written after the chart was requested, a flowsheet row that was not there before, a discharge summary your set has and theirs lacks. Cases turn on those pages.

The 8 causes, from most to least common

Legal health record vs designated record set

This distinction explains more mismatches than any other, and most litigators have never heard of it. HIPAA defines the designated record set broadly: it covers "the medical records and billing records about individuals maintained by or for a covered health care provider" and anything "used, in whole or in part, by or for the covered entity to make decisions about individuals" (45 CFR 164.501). The legal health record is a narrower set the facility defines for itself in policy. AHIMA, the health information management profession's body, puts it this way:

"The legal health record is a subset of the entire patient database. The elements that constitute an organization's legal health record vary depending on how the organization defines it."

The same brief says the legal health record "does not affect the discoverability of other information held by the organization", and that "audit trails, metadata, and psychotherapy notes are not included" in either record set. In plain terms: the release desk hands out the legal health record, discovery can reach more, and the audit trail sits outside both unless someone asks for it. AHIMA also lists electronic fetal monitoring strips, diagnostic images and video among the things a facility has to decide whether to include. Many decide not to print them on a routine release.

The 2 sets arrive through different legal doors, and each door shapes what comes through it. Knowing which door your pre-suit set used tells you what it was ever going to contain.

The discovery door is the widest, and it is also the one you control. The pre-suit door is narrower but has 1 property nothing else has: it was opened before anyone at the facility was defending a claim. That is why the pre-suit set is the baseline for the whole reconciliation, even when it is the smaller set.

What the pre-suit set usually lacks

• Full flowsheets. Standard prints show a summary grid or scheduled vitals only.
• The full MAR. A medication summary lists drugs; the administration record shows each dose, time, nurse and PCA entry.
• Orders, including discontinued ones.
• Source data: monitor strips, telemetry, oximetry trends, anesthesia machine data, device logs.
• Anything signed after the print date. A hospital has 30 days after discharge to complete the record under 42 CFR 482.24(c)(4)(viii), so an early copy can lack reports that exist today.

What people get wrong about the 2 sets

Most reconciliation pain is set up by the request for production, weeks before the documents arrive. If the request does not name a form, the rules pick one for you. The federal text:

"(i) A party must produce documents as they are kept in the usual course of business or must organize and label them to correspond to the categories in the request; (ii) If a request does not specify a form for producing electronically stored information, a party must produce it in a form or forms in which it is ordinarily maintained or in a reasonably usable form or forms; and (iii) A party need not produce the same electronically stored information in more than one form."

Why "as kept in the usual course" is slippery for an EHR

For a filing cabinet, "as kept" means in the folders, in order. An EHR keeps data in tables, and the document you receive is a report generated for the production, so "as kept in the usual course of business" tells you little. Our view: always ask in writing which report or print template generated the production, on what date, with what filters. That answer explains more differences than any other.

"Reasonably usable" has teeth

The 2006 committee note to Rule 34 limits what a producing party can do with the reasonably usable option:

"The option to produce in a reasonably usable form does not mean that a responding party is free to convert electronically stored information from the form in which it is ordinarily maintained to a different form that makes it more difficult or burdensome for the requesting party to use the information efficiently in the litigation."

The same note says information kept in a searchable form "should not be produced in a form that removes or significantly degrades this feature." An EHR is searchable, so image-only scans with no text layer and no index are a fair meet and confer topic under that note.

The rules around the production

The 2015 committee note to Rule 34 draws the line on withholding: no detailed log of withheld documents is required, but the producing party must "alert other parties to the fact that documents have been withheld." Privilege claims are logged separately under Rule 26(b)(5)(A).

The biggest decision in a reconciliation is what you compare. Compare pages, and 2 print jobs generate hundreds of false differences: a note that ran 3 pages in 1 template runs 4 in another, and every later page shifts. Compare documents instead: 1 note, 1 report, 1 flowsheet for 1 date range, 1 order, 1 result. Match documents, then compare the pages inside each pair.

Every document in either set ends up with exactly 1 status. We use 5, plus a sixth for things neither side produced:

Match

Same document in both sets, same content, same layout. Log both references and move on.

Reformatted

Same content, different presentation: new header or footer, different print date, different page breaks, different font, a different page count. No clinical word, value, time or signature differs.

Changed

Same document in both sets, but content differs: text added or removed, a value or time different, an addendum present in 1 version only, a different signer or signature time.

Added

In the defense set only. Usually a detail print, an order, a source-data page or a document the pre-suit release did not cover.

Missing

In the pre-suit set only. The defense production lacks a document you know exists because you already hold it.

Absent from both

Neither set has it, but something in the record says it exists: an order for continuous monitoring with no monitor data, "see anesthesia record" with no anesthesia record.

Format differences and content differences

Most errors in reconciliation are classification errors: a reviewer calls a reformatted page "changed" and burns a meet and confer on a font. Hold this line firmly.

1 edge case deserves a rule of its own. When a flowsheet in 1 set shows rows the other lacks, it can be a display filter (reformatted) or a row filed later (changed). You cannot tell from the page. Classify it as changed until someone proves it was a filter, because the opposite mistake can bury the one finding the case needs.

This is the method a paralegal or legal nurse consultant can run with 2 PDFs and a spreadsheet. It is slow on a 1,000-page chart, and it is still the method you want to be able to describe under oath. Before step 1, freeze both sets: keep untouched copies as received, with cover letters, certifications and receipt dates, stamp the pre-suit set with its own prefix (see what is Bates numbering), and record a hash of each file. Never merge or deduplicate the 2 sets before you compare them. A deduplication pass treats 2 versions of the same note as duplicates, and the version it drops may be the evidence.

1. Index each set separately. 1 row per document: date and time, type, author, encounter, note or order ID if printed, page range, print date. Do not look at the other set while indexing. A cited medical chronology of each set works as the index.
2. Match on a stable key. A printed note or order ID is best, then date plus time plus type plus author. Titles are the worst key, because templates rename things. Unpaired rows are findings, not failures.
3. Compare and classify each pair. Check the signature block and times first, then the last lines, where addenda live, then values and times. For a changed pair, quote the words that differ with both page cites.
4. Assign the next step. Reformatted: none. Added: chronology, cited to DEF. Missing: check the request and objections, then the deficiency letter. Changed: version comparison, then chapter 9 if the timing is unexplained. Absent from both: deficiency and preservation letters naming the system.

What to compare first

You will not read 1,000 matched pairs line by line. Our order: every document within 24 hours of the event, every document the defense expert cites, every addendum or corrected result in either set, and the flowsheets and MAR for the event window. Skim the rest for signature block and page count only, and write the scope you chose at the top of the log.

RECONCILIATION LOG
Matter: [CASE NAME / FILE NO.]
Set A (pre-suit): [SOURCE], received [DATE], [PAGES] pages, [PREFIX] [FIRST] to [LAST], print date [DATE], hash [VALUE]
Set B (defense):  [PRODUCING PARTY], received [DATE], [PAGES] pages, [PREFIX] [FIRST] to [LAST], response served [DATE], hash [VALUE]
Scope of line-by-line comparison: [e.g. all documents 03/03 00:00 to 03/05 00:00; all cited by defense expert; all addenda]

COLUMNS
 1  Row #
 2  Document type                [H&P / progress note / flowsheet / MAR / order / result / report]
 3  Date and time of service     [MM/DD/YYYY HH:MM]
 4  Author and role              [NAME, ROLE] (cosigner if any)
 5  Encounter / note / order ID  [ID as printed, or "none shown"]
 6  Set A pages                  [PL 000000 to PL 000000] or "none"
 7  Set A print date / signed    [DATE] / [DATE TIME]
 8  Set B pages                  [DEF 000000 to DEF 000000] or "none"
 9  Set B print date / signed    [DATE] / [DATE TIME]
10  Status                       MATCH / REFORMATTED / CHANGED / ADDED / MISSING / ABSENT BOTH
11  What differs                 "[exact words in A]" vs "[exact words in B]"
12  Evidence it should exist     [for MISSING and ABSENT BOTH: page that implies it]
13  Within request?              [RFP no. X] / [not requested]
14  Objection or log entry       [none / objection text / privilege log line]
15  Next step                    [none / add to chronology / deficiency letter / version comparison / audit trail request]
16  Owner and due date           [NAME, DATE]
17  Closed                       [DATE, how resolved, Bates of any new production]

Steps 1 and 2: index and match

The paralegal indexes 188 pre-suit documents and 241 defense documents. Matching pairs 171, leaving 17 pre-suit and 70 defense rows unpaired. The full log has 258 rows; 12 are shown.

Step 3: the 2 changed documents, side by side

The surgery note is the easier of the 2 to read, because the change announces itself. Here are the 2 versions:

The flowsheet is harder. The pre-suit page PL 000143 shows rows at 20:00 and 00:00, then the event. The defense page DEF 000405 shows the same 2 rows plus 02:00. Neither page shows a filed time. Either the pre-suit view showed scheduled vitals only and the 02:00 row was a PRN check all along, or the row was filed after 05/19/2025. The pages cannot settle it, and the addendum cites that row, so it needs settling.

Step 4: the page ledger

Before any letter goes out, reconcile the page counts so nobody can say a range was skipped. Every page in each set has to land somewhere.

The status tells you what kind of difference you found. It does not tell you what it means. Most differences have an innocent explanation, and the job is to rule those out in order, cheapest first, before anyone uses the word "altered".

The finding table

Applying it to the hypothetical

In the chapter 6 case, the anesthesia graph, orders and full MAR are added pages the standard template left out: chronology, DEF cites. The discharge summary is missing, requested and unobjected: deficiency letter. The monitor data is absent from both: deficiency letter, citing the order at DEF 000298. The addendum passes the label test and fails the timing test, and the row it relies on cannot be dated from the page. Those 2 go to chapter 9.

The deficiency letter is where the crosswalk pays for itself. A letter that says "your production is incomplete, produce the complete chart" invites a reply that says "we produced the chart as kept". A letter that says "the discharge summary dated 03/28/2025, which we hold at PL 000590 to PL 000594, is not in DEF 000001 to DEF 001047 and is responsive to request no. 1" leaves no room for that reply.

The rule behind the letter

In federal court you cannot move to compel without trying to resolve the dispute first:

"The motion must include a certification that the movant has in good faith conferred or attempted to confer with the person or party failing to make disclosure or discovery in an effort to obtain it without court action."

Rule 37(a)(4) adds that "an evasive or incomplete disclosure, answer, or response must be treated as a failure to disclose, answer, or respond," and Rule 37(a)(5)(A) makes expenses, including attorney's fees, the default when a motion to compel is granted or the discovery arrives after it is filed (FRCP 37).

State practice differs. In California, a motion to compel a further response to a demand for production must be accompanied by a meet and confer declaration under Code of Civil Procedure 2016.040, and notice of the motion must be given within 45 days of service of the verified response, or of a supplemental verified response, unless the parties agree in writing to a later date; otherwise the right to compel is waived (CCP 2031.310(b) and (c)). That clock runs from the response, not from the day you finish reconciling, so scope the work to fit it or get a written extension early.

The template below has 1 numbered item per crosswalk row that needs action, then the withholding, privilege and form questions, then dates. Changed documents stay out of it; they belong in the narrower request in chapter 9.

[FIRM LETTERHEAD]

[DATE]

VIA EMAIL
[DEFENSE COUNSEL NAME]
[FIRM, ADDRESS]

Re:  [CASE CAPTION], [CASE NO.]
     Deficiencies in [DEFENDANT]'s production of [DATE], [PREFIX] [FIRST] to [LAST]

Dear [NAME]:

We have compared [DEFENDANT]'s production against the records of
[PATIENT] that we obtained before suit ([PREFIX] [FIRST] to [LAST],
printed by [FACILITY] on [PRINT DATE]). We write to meet and confer
about the items below, which are responsive to Plaintiff's [FIRST]
Request for Production served [DATE] and are not in the production.

A. Documents we hold that were not produced
 1. [Document type], dated [DATE TIME], author [NAME/ROLE].
    Responsive to Request No. [X]. We hold it at [PL RANGE].
    No objection in your response of [DATE] covers it.
 2. [...]

B. Records the chart shows exist, produced by neither side
 3. [Data/system, e.g. continuous pulse oximetry data, DATE RANGE].
    [DEF PAGE] shows [the order / the reference] for it.
    Your response objects that it is "[QUOTE]". Please identify the
    system that captured it, its retention setting, and whether it
    still exists, and produce it in native or CSV form with any
    legend needed to read it.

C. Withholding and privilege
 4. For each objection in your response, please state whether any
    responsive material is being withheld on the basis of that
    objection (Fed. R. Civ. P. 34(b)(2)(C) [or STATE RULE]).
 5. If anything was withheld as privileged, please provide the
    description required by Rule 26(b)(5)(A) [or STATE RULE].

D. Form of production
 6. Please confirm whether documents were produced as kept in the
    usual course of business or organized and labeled to the
    requests (Rule 34(b)(2)(E)(i)), and identify the report or print
    template, the print date and any filters used.
 7. Please produce [FLOWSHEET NAME] for [DATE TIME] to [DATE TIME]
    showing, for each row, the value time, the filed time and the
    user, and a production index listing Bates range, document type,
    date, author and encounter for each document.

Please respond in writing by [DATE]. We are available to confer by
phone on [DATE/TIME], [DATE/TIME] or [DATE/TIME]. This letter does
not waive any other deficiency or any right to seek relief.

Sincerely,
[ATTORNEY NAME]
[BAR NO., CONTACT DETAILS]

Before the letter goes out

• Both sets frozen, stamped and hashedYou may have to prove which copy you compared.
• Page ledger totals match both Bates rangesNo unaccounted page in either set.
• Each missing item checked against the requestDo not demand what you never asked for.
• Each missing item checked against the objectionsObjected items get the withholding question.
• Evidence cited for each absent-from-both itemAn order or reference showing the data should exist.
• Changed documents held for chapter 9Not in a general deficiency list.
• Motion deadline calendaredSome state clocks run from the response date.
• Preservation letter sent for any named systemDevice data and logs rotate.
• Log rows updated with the letter dateA row closes when the item arrives or the court rules.

0 of 9 checked

A PDF shows that 2 versions of a note differ. It cannot show when the new text was typed, by whom, or whether a flowsheet row was filed at 02:00 or 2 months later. That lives in the audit trail and revision history, which AHIMA places outside both record sets. You get it only by asking for it, by name, in discovery.

Our view is that the request should be narrow and early. Ask for the audit trail and every saved revision of the specific note IDs and flowsheet rows your crosswalk flagged, for a date range running from admission to the latest production, in native or CSV form with the vendor's data dictionary. Narrow requests are harder to resist on proportionality under Rule 26(b)(1) and faster to read. How to read what comes back is covered in the EHR audit trail guide; how to tell a proper addendum from an alteration is in the altered medical records guide.

REQUEST FOR PRODUCTION NO. [X]

For each of the following items in the electronic health record of
[PATIENT], encounter [NO.]:

  (a) [Note type], dated [DATE TIME], author [ROLE], produced at
      [DEF RANGE] and previously released at [PL RANGE];
  (b) [Flowsheet name], rows timed [DATE TIME] to [DATE TIME],
      produced at [DEF RANGE] and previously released at [PL RANGE];

produce, for the period [ADMISSION DATE] through the date of your
response:

  1. The audit trail of every create, view, modify, sign, cosign,
     addend, file, retract, print and release event for the item,
     showing for each event the user, the user's role, the date and
     time with time zone, the action, and the workstation or device.
  2. The full text of every saved version or revision of the item,
     including superseded and retracted versions, with the date and
     time each version was saved.
  3. For flowsheet rows, the value time and the filed time of each
     row, and the user who filed it.
  4. The release-of-information and print log for the chart,
     including the release printed on or about [PRE-SUIT PRINT DATE].
  5. A data dictionary or legend sufficient to read the event codes
     and field names in items 1 to 4.

Produce items 1, 3 and 4 in native or CSV form, not as images.

Every citation now has to pick a set. An expert who cites PL 000161 while defense counsel holds DEF 000432 will spend 10 minutes at deposition establishing they are looking at the same note, and may find out mid-answer that they are not.

Our rule: the defense production is canonical. Cite the pre-suit set only where a document exists nowhere else, and cite both, with the difference stated, wherever a document changed. Keep 1 chronology, not 2 that drift apart, and attach the crosswalk to the expert's materials reviewed.

When the next production arrives

Rule 26(e)(1)(A) requires timely supplementation, and a deficiency letter usually produces a second set. Treat it as a third set: freeze, index, match against the log, add rows. Never renumber earlier sets. Record the new Bates range in the closed column of the row it answers, tell the expert which rows are still open, and if the defense expert cites a document your log shows as changed, say so with both cites in the expert report rebuttal.

Reconciliation has a mechanical half and a judgment half. The mechanical half is reading 1,600 pages, writing an index row for every document, and noticing that the note on page 432 has 4 more lines than the note on page 161. The judgment half is deciding whether a difference is format or content, and what it means. Legal AI tools are useful on a good part of the first half. They should not be trusted with the second.

What a large language model does well here

• Indexing. Clinical natural language processing can pull date, type, author and encounter from every document. An AI medical chronology of each set, cited to the page, is most of step 1.
• Finding labels. "Addendum", "late entry", "corrected" and "entered in error" are easy for software to find and easy for a tired reader to miss.
• Surfacing near-duplicate notes and highlighting the lines that differ.
• Spotting what should exist, such as a monitoring order with no monitoring data.

Where it fails

• OCR on faxes and scans. Optical character recognition can read "02:00" as "08:00", and timing is often the whole finding.
• Handwritten notes on downtime forms, anesthesia records and flowsheets.
• Print templates. A text comparison flags every reformatted page unless someone defines what counts as format.
• Copy-forward and ambient AI scribes. Notes built from the same template look like versions of each other, so a tool can pair the 03/04 note with the 03/05 note and report a "change".
• Hallucination. Ask a general chatbot what differs between 2 productions and generative AI may invent differences or skip whole ranges without saying so.
• Metadata. No tool reading PDFs can tell you when text was typed. That is in the audit trail.

Near-duplicate detection in AI ediscovery and technology assisted review platforms was built for email and office files. It can work on EHR prints with a good text layer and fails quietly on bad OCR. Check a sample of its "no difference" calls by hand.

Why page-level citations decide whether the output is usable

A reconciliation finding you cannot trace to a page is worthless. A deficiency letter lands only if every gap it names points to a page in the production. Grounded, page-level citations give you that pointer, and human-in-the-loop review confirms the gap is real before the letter goes out. Courts have made the cost of skipping the check concrete. In Mata v. Avianca, Inc. (S.D.N.Y. 2023), lawyers were sanctioned under Rule 11 for filing a brief built on case citations a chatbot had fabricated. The same duty applies to a deficiency letter that cites a page, or an expert report that quotes a note. Open every cited page before it goes out.

Choosing legal AI tools for this work

For AI document review in reconciliation, including agentic AI that promises to run the whole workflow, rank the requirements in this order:

For how AI review compares to a person on accuracy, see is AI accurate enough for court and HIPAA-compliant AI medical record review.

Medrecords AI is medical record review software: you upload records, and it drafts a cited chronology, summaries and flags, every line linked to its source page. On this topic the limits come first.

What it does not do

• It does not reconcile 2 productions document by document. Supplemental record review removes pages you already have from a new batch and flags where the new pages agree or conflict with the file. It will not tell you which documents are missing from 1 side or map one Bates range onto the other.
• It does not build a Bates crosswalk. The crosswalk in chapter 6 is built by a person, using chapter 5.
• It does not parse EHR audit trail exports, so it cannot say when text was typed or a row was filed.
• It does not decide whether a difference is innocent, a record was altered or a production is deficient. It gives no legal advice and does not retrieve records.

Where it helps the person doing the reconciliation

• A cited chronology of each set, separately. Upload each set as its own file and get a cited chronology of each, every entry linked to its PL or DEF page. Those are the 2 indexes of step 1, ready for a reviewer to match.
• Missing expected records within a file. Missing records identification flags visits, providers and date ranges the file implies but does not contain, cited to the page that implies them.
• Date gaps and undated pages. The chronology shows a jump from 00:00 to 04:40, and undated document flagging keeps undated pages in their own bucket.
• Near-duplicate notes within a file. Record alteration detection puts near-duplicate pages that differ side by side.
• Stable page numbers and low-confidence pages. Bates stamping gives your pre-suit set its own prefix, and OCR marks pages read with low confidence, including handwriting.

Flags are signals, not verdicts. A near-duplicate that differs may be a pended note, a template difference or a real change, and the software cannot tell which. A human decides. The work runs under SOC 2 and HIPAA with a signed BAA (see security and HIPAA). The chain of custody log records who accessed your uploaded file; it is not a hospital EHR audit trail. Self-Service bills 10 cents a deduplicated page, down to 5 cents at volume, duplicates free; Enterprise On-Prem is an annual license.

Why don't the defense records match the records we got before suit?

The 2 sets are different reports run on the same database, with different release scope, print dates and EHR print templates. The release desk prints the facility's legal health record; discovery reaches more. Later addenda and source data add more differences. Most are innocent, so compare documents, not pages, and classify each difference before drawing conclusions.

What is the difference between the legal health record and the designated record set?

The designated record set is HIPAA's term, defined at 45 CFR 164.501: medical and billing records and anything used to make decisions about the patient. The legal health record is a narrower set each facility defines for release. AHIMA places audit trails and metadata outside both.

What does FRCP 34(b)(2)(E) require for the form of a production?

Documents come as kept in the usual course of business or labeled to the request categories. Electronically stored information, if the request names no form, comes in the form ordinarily maintained or a reasonably usable form, and only 1 form is required. State rules differ.

What should a deficiency letter for a records production include?

1 numbered item per missing document, with its date, type, author, the request it answers and proof it exists, such as the page in your pre-suit set. Add the withholding question for each objection, the privilege log question, the print template question, and dates for an answer and a call.

When should we ask for the EHR audit trail?

When a changed document has no label, when an addendum about the disputed hours was written after the records request or notice of claim, when flowsheet rows appear in 1 version only, or when a signature time moved. Ask for those note IDs only, in native or CSV form.

Which set should our expert cite?

Usually the defense production, because every party shares its Bates numbers. Cite the pre-suit set where a document exists nowhere else, cite both versions where a document changed, and attach the crosswalk to the expert's materials reviewed.

Can AI compare 2 medical record productions?

AI tools can help with parts: indexing each set, finding addenda and surfacing near-identical notes. They struggle with OCR errors in times, handwriting, template noise and copy-forward text, and a chatbot may invent differences or skip pages. A person confirms every difference at the page.

Does Medrecords AI build the Bates crosswalk between 2 productions?

No. It does not build a crosswalk or map Bates ranges between sets. When a new batch arrives, supplemental record review removes pages already in the file and flags where the new pages agree or conflict with it. It also flags missing expected records, date gaps and near-duplicate notes, each cited to its page. A person matches the sets and decides.

Is it HIPAA compliant to upload medical records to AI software?

It can be, if the vendor signs a business associate agreement, can show verifiable controls such as a SOC 2 report, and does not train on your records. Get the BAA before the first upload, and keep consumer chatbots without one away from client records.

Can ChatGPT tell us whether the defense altered the records?

No. A chatbot reading 2 PDFs cannot see audit trails, filed times or retracted entries, and it may answer confidently without support. A tool can point to 2 versions of a note that differ. Whether that is a print setting, an addendum or an alteration takes metadata, an expert and counsel.

Rules, regulations and statutes were checked against primary sources in September 2026, and quoted text is verbatim. The AHIMA brief is quoted in 3 short passages and otherwise paraphrased. The case in chapters 6 and 7, including every date, page number, page count and clinician, is hypothetical. Product facts come from this site's product pages. Nothing here is legal or medical advice, and state rules differ from the federal rules described here.

• Federal Rule of Civil Procedure 34: (b)(1)(C), (b)(2)(A) to (E), and the 2006 and 2015 committee notes.
• Federal Rule of Civil Procedure 26: (b)(1), (b)(5)(A) and (e)(1)(A).
• Federal Rule of Civil Procedure 37: (a)(1), (a)(4) and (a)(5)(A).
• 45 CFR 164.501, definition of designated record set.
• 45 CFR 164.524, right of access: exceptions in (a)(1), timing in (b)(2), form and format in (c)(2).
• 42 CFR 482.24, hospital medical record services: (c)(1) and (c)(4)(viii).
• California Code of Civil Procedure 2031.280 and 2031.310.
• AHIMA, "Fundamentals of the Legal Health Record and Designated Record Set", Journal of AHIMA 82, no. 2 (February 2011).
• Mata v. Avianca, Inc., S.D.N.Y., 22 June 2023: Rule 11 sanctions for filing fabricated AI-generated case citations.

Related guides: what is Bates numbering, altered medical records, EHR audit trails in medical malpractice, how to get medical records for a lawsuit, and medical malpractice record review.