Insights / AI in Personal Injury Litigation
Five Ways AI Can Improve Medical Record Review Without Replacing Clinical Judgment
By Breen LNC.
Short answer: AI can help in five practical ways: clearing out repeated text, drafting a dated timeline, linking each fact to its page, searching across the record, and surfacing possible conflicts. In each case the software prepares the work, and a clinician decides what it means.
1. Set aside repetition
Records repeat themselves. At one large health system, half the words across six years of notes were exact duplicates of earlier text about the same patient. Software is good at spotting repeated pages and passages, which saves a reader’s attention for what’s new.
The clinical call: Repetition isn’t the same as irrelevance. “Duplicated” there means exact repeated text, which is not the same as wrong, and the authors call duplication a systemic hazard. A copied line that should have changed can itself be a finding.
2. Draft the timeline
Pulling dates, providers and events into order is tedious work that software handles quickly.
The clinical call: A nurse decides whether the sequence makes clinical sense and whether the timing of an event matters.
3. Tie each fact to its page
The Michigan Bar Journal’s chronology guidance advises recording discrete facts and citing the source document. Software can do this on every line, which makes checking far faster.
The clinical call: A cite proves where the words are. A nurse confirms that the entry fairly represents them.
4. Search across the whole file
A question like “when was this medication held?” can be answered much faster than by paging through the file, once the text has been read and indexed.
The clinical call: A nurse knows the other ways a chart might record the same event, and what its absence might mean.
5. Surface possible conflicts
Software can point to places where two entries seem to disagree, such as a medication list that doesn’t match a note. The legal nurse consultant’s role as AALNC describes it includes identifying omissions and contradictions, and a pointer gets the reviewer there sooner.
The clinical call: Many apparent conflicts have ordinary explanations. Telling them apart takes clinical training.
The common thread
NIST warns about “excessive deference to automated systems.” Each of these five uses avoids that risk by treating the output as a place to look, leaving the conclusion to a person. That is the idea behind Breen: the speed of AI and the judgment of a nurse.
Frequently asked questions
What parts of medical record review can AI handle?
Sorting and deduplicating pages, drafting timelines, attaching page cites, searching text and pointing to possible conflicts. A clinician should still decide what those results mean.
Is duplicated text in a medical record a problem?
It can be. A large study found about half of note text was exact repeats, which measures repetition and not errors, and its authors call duplication a systemic hazard. A reviewer has to consider whether copied information was still accurate.
Sources
- Steinkamp et al., Prevalence and Sources of Duplicate Information in the Electronic Medical Record, JAMA Network Open (2022)
- Greg Krehel, Chronology Best Practices, Michigan Bar Journal
- American Association of Legal Nurse Consultants, FAQs
- NIST AI 600-1, AI Risk Management Framework: Generative AI Profile (July 2024)
General information for legal professionals. This is not legal or medical advice.