Automated ICD Coding of Primary Diagnosis Based on Machine Learning

The accuracy and productivity of ICD coding has always been a concern of clinical practice. Errors of ICD codes may result in claim denials and missed revenue. However, ICD coding process is complex, time-consuming and error-prone. More experienced coders are in need, but there is an increasing lack of supply. Automated ICD coding has potential to facilitate clinical coders for improved efficiency and quality. Model performance of related studies is still far below coders and both the accuracy and interpretability need to be improved in great demand. Besides, studies in Chinese corpus are not sufficient.

In this study, the investigators will implement automated ICD coding study based on inpatient' data collected from electronic medical records from Fuwai Hospital, the world's largest medical center for cardiovascular disease. Feature engineering and machine learning methods will be used to develop classification models with good performance, interpretability and practicability for ICD codes of primary diagnosis.