Method · Medical-legal review

How to build a defensible medical-record chronology (and where AI helps vs. where it can't)

A chronology is the spine of a medical-legal opinion and the thing you will be cross-examined on. Here is a working method for building one that survives Daubert — page-cited, conflict-flagged, gap-aware — and an honest line on what AI should and should not touch.

Ask any seasoned expert where the hours go on a personal-injury or med-mal case and the answer is the same: reading the bundle. Hundreds — sometimes thousands — of pages of records, imaging reports, depositions, and prior disclosures, condensed into a timeline you can actually reason from. It is the least glamorous part of the engagement and the most consequential, because every opinion you offer ultimately traces back to it.

A defensible chronology is not just a tidy list of dates. It is one where each entry can be traced to its source page, where contradictions are surfaced rather than smoothed, and where the line between record fact and expert inference never blurs. Below is a method, then the part most articles skip: where AI genuinely helps the review, and where it must not go.

What a defensible chronology actually captures

For every material event, capture four things — and keep them in four distinct columns so a reader (and a cross-examiner) can see exactly where each came from:

• Date — normalized and sortable, with date ranges where care spans a period.
• Event or finding — what the record says, in neutral language. Quote the operative phrase when it matters ("denies loss of consciousness").
• Source citation — the exact file and page, ideally with the Bates number, so the entry can be verified in seconds.
• Flag — whether this entry is clean, or needs your judgment because it is ambiguous, contradicted, or raises a causation question.

The citation column is what separates a defensible chronology from a summary. A page-cited line looks like this:

When every line carries its source, you verify in seconds instead of re-reading the bundle, and your eventual report's "basis and reasons" are auditable end to end.

Flag conflicts and gaps — don't resolve them silently

The most dangerous chronology is the one that reads cleanly because someone quietly reconciled the contradictions. Three categories deserve an explicit "needs expert review" flag rather than a confident entry:

• Internal contradictions. The ED note says no loss of consciousness; the deposition says "might have blacked out for a second." That conflict is a fact about the record — flag it, cite both, and let the expert reconcile it. Don't pick a side in the timeline.
• Causation and apportionment questions. An MRI showing degenerative change of "indeterminate age" is not evidence of traumatic causation. Whether a finding is acute or pre-existing is an expert determination, never a record fact to be asserted in a chronology.
• Treatment gaps. A three-month gap between visits is routinely raised on causation and mitigation. Surface it so you can address it affirmatively in your report rather than be surprised by it on cross.

Where AI helps — and where it can't

AI is genuinely useful for the mechanical bulk of record review. It is genuinely dangerous when it is allowed to cross from organizing facts into forming opinions. The line is not subtle, and keeping it bright is what makes AI-assisted review defensible rather than a liability.

The right mental model is a force multiplier on your review, not a replacement for it. AI compresses 40+ hours of chronologizing into a verifiable draft; you click each line to confirm it against its source page, you make every causation call, and you author every opinion. Pasting privileged records into a public chatbot, by contrast, keeps no disclosable audit trail and may compromise privilege — which is the opposite of defensible.

A short build checklist

• Every entry has a date, a neutral event description, a source-page citation, and a flag.
• No contradiction is silently resolved — both sources are cited and the conflict is flagged.
• Causation and apportionment are marked as expert questions, not asserted as record facts.
• Treatment gaps are surfaced, not buried.
• You have clicked through and verified each flagged line against its source.
• There is an exportable record of how the chronology was built.