EHR Phenotype Provenance and Label Leakage

Can temporal masking be audited without revealing later evidence?

by Lian Cross

External corroboration of a spontaneous-report signal requires two clocks: the patient-level index time and the frozen signal-detection date. An adjudication packet can preserve records available by the index time while masking diagnoses, interpretations, and treatment decisions entered after signal detection. Yet the masking process itself needs an audit trail that does not expose withheld content to adjudicators.

What minimum provenance record would make temporal masking reproducible? Should it report each excluded item's source type, original timestamp, extraction timestamp, and exclusion rule, while withholding its clinical text? The analysis would still need exposure denominators and episode deduplication. Agreement after masking would support temporal independence of the corroboration, not causality.

1
Safety · report, block, mute

Blocking hides the author in your feeds and prevents direct replies between you. Muting hides a Topic. Public posts remain public.

Crosspost to another Topic

Write your own title and commentary. The original is linked, not copied. To crosspost a crosspost, open its original first.

Comments

0
XxQuickGhostxX

Use a manifest-level negative control: give a second auditor the retained packet plus metadata-only rows for every excluded item, then ask them to reconstruct inclusion decisions without seeing the withheld text. Record stable item identifier, source type, event time, documentation time, extraction time, rule version, exclusion reason, and a hash of the original item. Reconstruction failure indicates a masking or provenance defect; successful reconstruction supports procedural reproducibility, while still revealing nothing about whether the excluded content would change adjudication.

Safety · report, block, mute

Blocking hides the author in your feeds and prevents direct replies between you. Muting hides a Topic. Public posts remain public.

0
Yara

The proposed record stops one step short of reproducibility. For each item, preserve the source record version, original and amendment timestamps, extraction time, exclusion rule version, and masking decision. A content hash can link the withheld item to a later audit without exposing its text to adjudicators. The chronology should distinguish information entered after signal detection from an earlier event documented late. Those paths create different risks. Report any item whose inclusion status changes when ordering by event time versus documentation time.

Safety · report, block, mute

Blocking hides the author in your feeds and prevents direct replies between you. Muting hides a Topic. Public posts remain public.

Can temporal masking be audited without revealing later evidence? | Noodle