# Adversarial Learning of Privacy-Preserving Text Representations for De-Identification of Medical Records

## TL;DR

This paper presents a novel adversarial learning approach to generate privacy-preserving text representations for medical records, enabling effective de-identification without sharing sensitive data, thus facilitating research while complying with privacy laws.

## Contribution

It introduces a method to create shareable, privacy-preserving text representations that enable training high-performance de-identification models without exposing PHI.

## Key findings

- Achieved an F1 score of 97.4% with simple models using our representations.
- Our method produces representations that contain no PHI but retain de-identification utility.
- Enables cross-organizational sharing of medical text data for research.

## Abstract

De-identification is the task of detecting protected health information (PHI) in medical text. It is a critical step in sanitizing electronic health records (EHRs) to be shared for research. Automatic de-identification classifierscan significantly speed up the sanitization process. However, obtaining a large and diverse dataset to train such a classifier that works wellacross many types of medical text poses a challenge as privacy laws prohibit the sharing of raw medical records. We introduce a method to create privacy-preserving shareable representations of medical text (i.e. they contain no PHI) that does not require expensive manual pseudonymization. These representations can be shared between organizations to create unified datasets for training de-identification models. Our representation allows training a simple LSTM-CRF de-identification model to an F1 score of 97.4%, which is comparable to a strong baseline that exposes private information in its representation. A robust, widely available de-identification classifier based on our representation could potentially enable studies for which de-identification would otherwise be too costly.