Generation and De-Identification of Indian Clinical Discharge Summaries using LLMs
Sanjeet Singh, Shreya Gupta, Niralee Gupta, Naimish Sharma, Lokesh Srivastava, Vibhu Agarwal, Ashutosh Modi
TL;DR
The paper addresses the critical problem of protecting patient privacy in India's expanding electronic health records by evaluating cross-institution de-identification performance when models are trained on non-Indian data. It shows poor cross-institution generalization and limited effectiveness of off-the-shelf commercial systems on Indian PHI, and proposes generating synthetic Indian discharge summaries with LLMs to augment training. The study demonstrates that fine-tuning on synthetic data (ICDS_G^l) yields near-perfect ICDS_R de-identification performance, and combining synthetic with real data improves cross-institution transfer to other datasets (e.g., n2c2-2014) with high micro-F1 scores. It also highlights challenges in on-demand LLM de-identification and underscores the value of on-premise, customizable LLMs for privacy-preserving Indian healthcare data, while outlining future active-learning approaches to scale evaluation and generalization. Overall, the work provides a practical pathway to robust PHI de-identification in India through synthetic data augmentation and local model fine-tuning, with significant implications for secure digital health adoption.
Abstract
The consequences of a healthcare data breach can be devastating for the patients, providers, and payers. The average financial impact of a data breach in recent months has been estimated to be close to USD 10 million. This is especially significant for healthcare organizations in India that are managing rapid digitization while still establishing data governance procedures that align with the letter and spirit of the law. Computer-based systems for de-identification of personal information are vulnerable to data drift, often rendering them ineffective in cross-institution settings. Therefore, a rigorous assessment of existing de-identification against local health datasets is imperative to support the safe adoption of digital health initiatives in India. Using a small set of de-identified patient discharge summaries provided by an Indian healthcare institution, in this paper, we report the nominal performance of de-identification algorithms (based on language models) trained on publicly available non-Indian datasets, pointing towards a lack of cross-institutional generalization. Similarly, experimentation with off-the-shelf de-identification systems reveals potential risks associated with the approach. To overcome data scarcity, we explore generating synthetic clinical reports (using publicly available and Indian summaries) by performing in-context learning over Large Language Models (LLMs). Our experiments demonstrate the use of generated reports as an effective strategy for creating high-performing de-identification systems with good generalization capabilities.