A clinical de-identification model lost accuracy on intensive-care notes
Source check: Source checked. Not independent scientific confirmation.
Finding: EdgeDeID, trained on synthetic notes, scored 0.478 F1 on 2,434 MIMIC intensive-care nursing notes, compared with 0.994 on 100 synthetic notes. A separately trained clinical text model scored 0.675 on the intensive-care set. The test corpora differ in writing style and annotation conventions.
Technology: EdgeDeID; Qwen3-0.6B; clinical entity recognition
Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- Fine-tuned Qwen3-0.6B detects surrogate identifiers in clinical notes.
- Test and result
- External notes use replaced identifiers and human labels. The paper also reports an analysis limited to the 823 Safe Harbor spans.
- Limitation / still unknown
- The evaluation does not establish safe redaction or portability. Style and label differences both affect the synthetic-to-clinical comparison.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://arxiv.org/abs/2610.33839v1 · https://arxiv.org/html/2610.33839v1
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: linked separately.
Review date: 21 Nov 2026. Review scope: v1 §4.1 and Table 4 external clinical-text evaluation.
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.