Leveraging Shannon Entropy to Validate the Transition between ICD-10 and ICD-11

This study aimed to propose a mapping framework with entropy-based metrics for validating the effectiveness of the transition between International Classification of Diseases 10th revision (ICD-10)-coded datasets and a new context of ICD-11. Firstly, we used tabular lists and mapping tables of ICD-11 to establish the framework. Then, we leveraged Shannon entropy to propose validation methods to evaluate information changes during the transition from the perspectives of single-code, single-disease, and multiple-disease datasets. Novel metrics, namely, standardizing rate (SR), uncertainty rate (UR), and information gain (IG), were proposed for the validation. Finally, validation results from an ICD-10-coded dataset with 377,589 records indicated that the proposed metrics reduced the complexity of transition evaluation. The results with the SR in the transition indicated that approximately 60% of the ICD-10 codes in the dataset were unable to map the codes to standard ICD-10 codes released by WHO. The validation results with the UR provided 86.21% of the precise mapping. Validation results of the IG in the dataset, before and after the transition, indicated that approximately 57% of the records tended to increase uncertainty when mapped from ICD-10 to ICD-11. The new features of ICD-11 involved in the transition can promote a reliable and effective mapping between two coding systems.