Beyond De-Identification: A Structured Approach for Defining and Detecting Indirect Identifiers in Medical Texts

TL;DR

The paper tackles privacy in medical text by moving beyond traditional PHI de-identification to address indirect personal identifiers (IPIs) through a nine-category schema. It builds and annotates a 100-document MIMIC-III discharge-summaries corpus at the span level, achieving an overall inter-annotator agreement of $IAA = 0.87$ across 6,199 IPI spans. Baseline detection using BERT yields strong $F_1$ performance (micro ≈ $0.93$) with high recall, while open-source LLMs underperform (best ≈ $F_1 = 0.51$) and struggle with formatting and hallucinations. The work provides a publicly useful dataset, guidelines, and baselines to support privacy-preserving anonymization of unstructured clinical text and motivates future $k$-anonymization frameworks and downstream task evaluations.

Abstract

Sharing sensitive texts for scientific purposes requires appropriate techniques to protect the privacy of patients and healthcare personnel. Anonymizing textual data is particularly challenging due to the presence of diverse unstructured direct and indirect identifiers. To mitigate the risk of re-identification, this work introduces a schema of nine categories of indirect identifiers designed to account for different potential adversaries, including acquaintances, family members and medical staff. Using this schema, we annotate 100 MIMIC-III discharge summaries and propose baseline models for identifying indirect identifiers. We will release the annotation guidelines, annotation spans (6,199 annotations in total) and the corresponding MIMIC-III document IDs to support further research in this area.

Figures (2)

• Figure 1: A snippet of a fictitious discharge summary with annotations according to our IPI schema in red.
• Figure 2: A (generated) discharge summary with annotations based on the proposed schema.