Insights / Medical Evidence, Causation and Case Value
How AI-Assisted Medical Review Can Help Identify Overlooked Evidence
By Breen LNC.
Short answer: AI can help surface buried evidence by searching every page, comparing entries and pointing out repeats and changes. It can also miss things, so treat what it finds as leads. A clinician should check each lead and separately look for what the software did not report.
Why things get overlooked
Volume is a large part of it. At one health system, notes available to an emergency physician grew more than 30-fold over 17 years, which the authors say makes chart review harder. When the same paragraph appears dozens of times, the one line that differs can be easy to miss. Half the words in one system’s notes were exact repeats.
Where software helps
No fatigue. Software doesn’t tire on page 3,800, though it can still skip content, as the next section shows.
Finding the first mention. When did someone first document this complaint? Search answers that across the whole file.
Comparing like with like. Software can line up versions of a repeated section and show where the wording changed, drawing attention to the entry a reader might skim.
Connecting distant pages. A billing line, a medication record and a nursing note may describe the same event in different words, far apart in the file. Pointing out possible matches saves a lot of searching.
Where it falls short
The same tools leave things out. In a 2026 hospital pilot, physicians reported omissions in 25 of 100 rated AI summaries, more than any other kind of problem, though most unedited summaries were judged to carry no potential for harm. A JAMA Internal Medicine study found AI narratives less comprehensive than physician narratives, with similar overall quality scores. Neither was a litigation study. They do counsel against treating an AI output as a complete list of what’s in the file.
Some evidence is by nature invisible to text search. An assessment that was never charted has no words to find. AALNC’s list of what legal nurses do includes identifying omissions, and that takes knowing what a chart should contain.
A two-direction approach
- From the software outward. Take each item the tool surfaces and check it on the page.
- From the case inward. Have a clinician list what should exist given the injury and the care, then look for it. Missing items become questions for the provider, your records requests or discovery.
Running both directions is the safeguard. The first catches what a tired reader would miss, and the second catches what the software never saw.
Keep findings traceable
Whatever you find, keep the page cite and the exact line. A lead you can’t trace back to the record isn’t one you can use. Every flag at Breen comes with the quoted line and the nurse’s reasoning for that reason.
Frequently asked questions
Can AI find everything important in a medical record?
No tool can promise that. Studies of AI clinical summaries report omissions, and text search cannot find something never charted. A clinician should also look for what is missing.
How should a firm use AI to look for evidence?
Treat results as leads and check each on its source page. Separately, have a clinician list what should exist and look for it.
Sources
- Patterson et al., Call me Dr Ishmael: trends in electronic health record notes available at emergency department visits and admissions, JAMIA Open (2024)
- Steinkamp et al., Prevalence and Sources of Duplicate Information in the Electronic Medical Record, JAMA Network Open (2022)
- Grolleau et al., Physician-Reported Safety Outcomes of AI-Generated Hospital Course Summaries, JAMA Network Open (2026)
- Williams et al., Physician- and Large Language Model-Generated Hospital Discharge Summaries, JAMA Internal Medicine (2025)
- American Association of Legal Nurse Consultants, FAQs
General information for legal professionals. This is not legal or medical advice.