Impact and future role of LLMs in clinical coding: A bibliometric review

Abstract

Since their creation, LLMs (Large Language Models) have been thought to be capable of changing the paradigm of clinical coding by partially or even fully automating this extremely time-consuming task. Although the application of these concepts in medicine is relatively recent, it has already gained significant momentum. The objective of this bibliometric analysis is to guide research efforts and identify gaps in knowledge. A search was carried out in the databases of PubMed, Scopus, Web of Science, IEEE Xplore and ACM, followed by a manual selection of the relevant papers from over 1000 search results. No time or language restrictions were considered. After duplicate removal and screening, 181 publications were deemed relevant and selected for analysis. The papers are from 2019 to 2025 with 2024 being the one with the most publications. The United States of America lead in publication volume followed by China and Spain. When considering authorship, the top authors were Dalianis H with 6 articles followed by Boytcheva S, Casillas A and Veredas FJ both with 5 articles. Co-authorship analysis revealed fragmented collaboration patterns, with dense intra-group connections but limited inter-group integration. The application of LLMs in clinical coding is a rapidly expanding but methodologically fragmented field. Although LLMs show promising results in improving the accuracy and efficiency of clinical coding, there is a critical gap in the translation of these technologies into real-world clinical settings. Future research should address interdisciplinary collaboration, real-world implementation, and ethical-technical alignment to advance clinical applicability.