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.

The conceptualization of a claim lies at the core of argument mining. The segregation of claims is complex, owing to the divergence in textual syntax and context across different distributions. Another pressing issue is the unavailability of labeled unstructured text for experimentation. In this paper, we propose LESA, a framework which aims at advancing headfirst into expunging the former issue by assembling a source-independent generalized model that captures syntactic features through part-of-speech and dependency embeddings, as well as contextual features through a fine-tuned language model. We resolve the latter issue by annotating a Twitter dataset which aims at providing a testing ground on a large unstructured dataset. Experimental results show that LESA improves upon the state-of-the-art performance across six benchmark claim datasets by an average of 3 claim-F1 points for in-domain experiments and by 2 claim-F1 points for general-domain experiments. On our dataset too, LESA outperforms existing baselines by 1 claim-F1 point on the in-domain experiments and 2 claim-F1 points on the general-domain experiments. We also release comprehensive data annotation guidelines compiled during the annotation phase (which was missing in the current literature).