Implementation and Evaluation of ICD-10 Artificial Intelligence Coding Assistance System

The International Statistical Classification of Diseases, 10th Revision (ICD-10), developed by the World Health Organization, is a coding system designed to classify diseases and ensure international comparability, thereby facilitating the statistical analysis of morbidity and mortality. ICD-10 is widely recognized as the standard for diagnostic classification. Taiwan adopted ICD-10 as its primary coding standard in 2016. With approximately 150,000 diagnostic and procedure codes-significantly more than the 17,000 codes in the previous ICD-9-this increased complexity has raised the workload for physicians and medical coders. Current observations indicate that inpatient cases often have fewer recorded diagnoses, resulting in the underestimation of critical risk factors, comorbidities, and complications. To address these challenges, this study proposes the development of an automated disease coding recommendation system using deep learning and natural language processing techniques. This system aims to assist physicians in coding decisions, optimize HIS interface design, improve the quality and completeness of medical record documentation, reduce the time physicians spend on coding tasks, and alleviate the overall workload associated with medical claim submissions.