Editorial guide
OCR for Medical Records in Litigation: What to Test Before Trusting the Text
A document-class approach to OCR validation for scans, faxes, rotations, stamps, tables, and handwriting in medical-record review.
By Jake Bauman. Updated August 12, 2026. Under Seal engineering review; not a performance claim for client records. No named legal review is claimed unless identified here.
Short answer
OCR for litigation should be evaluated page by page and document class by document class. The workflow must preserve the page image, expose confidence and coverage, hold uncertain text for human review, and prevent held pages from silently supporting exported facts.
Key takeaways
- An empty OCR string is an unknown page, not proof that the page is blank.
- Tables and handwriting need separate acceptance rules from clean typed scans.
- Character and word error rates do not replace visual source review.
Editorial methods are identified as such. Numbered links beside a paragraph or section point to the authority or framework relevant to that claim; a source does not certify Under Seal or replace current jurisdiction-specific review.
Why an average score is unsafe
Basis: editorial method
Medical files combine cover sheets, forms, billing tables, handwritten notes, stamps, fax generations, and native reports. An aggregate score can be excellent while a small but important class fails. Report coverage, hold accuracy, and error rates for each class.
The minimum evaluation set
Basis: editorial method
Hash the fixtures, retain engine and configuration versions, and keep the expected labels stable. A failing class is useful release evidence. Changing the expected state solely to make a run green destroys the value of the test.
- Born-digital text layer
- Clean typed scan
- Low-contrast fax or repeated-copy artifact
- Rotated or skewed page
- Stamped or annotated page
- Table-heavy bill or medication list
- Handwriting that remains a human-review hold
Downstream controls
Basis: editorial method
The OCR result should carry its page number, retained image, path, confidence, and coverage state into retrieval and chronology review. If the page needs a human read, the downstream system should display it as a gap and block it from silently becoming accepted source support.
Sources for this section: [1] Under Seal OCR evaluation method (opens in a new tab)
Authorities and frameworks
Each source is classified below. It may govern a specific legal point, offer professional or government guidance, or provide a voluntary framework. The collection does not validate every editorial method in this guide and does not resolve a firm-specific question.
- [1]Under Seal OCR evaluation method Opens in a new tab. — Under Seal · Editorial method
Connect the general guidance to the product record
Continue the review
Under Seal is pre-release software, not a law firm. Do not submit matter details or client documents through this website.
Request a no-client-data readiness review