Intelligent De-Identification of Medical Discharge Summaries Using Hybrid NLP Techniques

Abstract

Medical discharge summaries are vital documents in healthcare, often containing Personally Identifiable Information (PII), raising concerns regarding privacy and regulatory compliance. This article proposes a cutting-edge approach that utilizes intelligent data de-identification to address this challenge. This article employs Natural Language Processing (NLP) techniques such as Named Entity Recognition (NER), a hybrid approach that integrates Machine Learning (ML) models, Regular Expressions (REGEX)-based recognizers, and extensive lists of names and addresses. The proposed method focuses on achieving a delicate balance between extracting valuable insights from data and safeguarding sensitive information. The evaluation against benchmarks demonstrates significant improvements in de-identification performance, particularly in discharge summaries. We present findings from our system’s evaluation of synthesized discharge summaries, the OntoNotes dataset, and the CoNLL-2003 dataset, demonstrating its effectiveness in anonymizing diverse medical text sources.