The Use of Artificial Intelligence for Clinical Coding Automation: A Bibliometric Analysis

In hospital settings, all information concerning the patient’s diseases and medical procedures are routinely registered in free-text format to be further abstracted and translated into standard clinical codes. The derived coded data is used for several purposes, from health care management and decision-making to billing and research. However, clinical coding is mostly performed manually, is very time-consuming, is inefficient and prone to error task. We conducted a bibliometric analysis of the scientific production on automated clinical coding using AI methods in the context of the International Classification of Diseases (ICD) classification system. The study aims to provide an overview and evolution of the research on automated clinical coding through Artificial Intelligence (AI) techniques. We did not consider time or language restrictions. Our analyses focused on characteristics of the retrieved literature, scientific collaboration and main research topics. A total of 1611 publications on automated coding during the last 46 years were retrieved, with significant growth occurring after 2009. The top 10 most productive publication sources are related to medical informatics, even though no common publication source or author was identified. The United States had by far the highest number of publications in this field. We found that natural language processing and machine learning were the main AI methodological areas explored for automated coding applications. Automated clinical coding using AI techniques is still rising and will undoubtedly face several challenges in the coming years. The results of the bibliometric analysis can assist the conduction of a more in-depth review to assess the variation of specific techniques and compare the performance of different methodological approaches regarding automated coding applications.