Bridging the Version Gap: Multi-version Training Improves ICD Code Prediction, Especially for Rare Codes
Researchers have developed a multi-version training approach to improve the accuracy of automated clinical coding, particularly for rare medical codes. By incorporating data from different versions of the International Classification of Diseases (ICD), such as ICD-9 and ICD-10, the model demonstrated a significant performance boost. This method addresses the challenge of evolving coding systems and the long-tail problem in rare code prediction, leading to better overall metrics with fewer model parameters. AI
IMPACT Enhances accuracy in clinical coding, potentially streamlining administrative tasks and improving rare code identification.