Omission auditor
Finds what the note left out.
Checked against clinicians
TPR/FPR with CI on clinician-labelled omissions; compare to generic judge baseline.
Where a person decides
Flags go to clinician; never auto-edits the record.
Works on
Scribes, discharge and prior-auth summaries.
Delivered inside a Quality Audit today. Reviewed .
The problem it solves
The most common scribe error is what's missing. Missing content is hard to see because the note still reads well.
How it works
List the facts
The agent reads the visit transcript and lists every clinically relevant fact.
Check each fact
It looks for each one in the note and marks it present, partial or missing.
Flag what matters
Missing items are ranked by clinical importance for a clinician to confirm.
An example, illustrative
Transcript: 'allergic to penicillin'. Note has no allergy line.
Flagged: allergy omitted. Severity: high.
Transcripts and notes, from real or simulated visits.
It can misjudge what counts as relevant, so clinicians review its flags and we report how often it's wrong.
What it helps measure
Quick answers
What does the omission auditor do?
Enumerate source facts, check each against output, flag missing material.
How is the omission auditor checked against clinicians?
TPR/FPR with CI on clinician-labelled omissions; compare to generic judge baseline.
How is this different from reading the note?
It works from the source, listing what should be there first. Reading only the note rarely reveals what's absent.
Does it work without transcripts?
It needs some source, such as a transcript, chart or clinician-written reference.
What can the omission auditor not do?
Flags go to clinician; never auto-edits the record. No agent makes a clinical decision.