Leveraging Taxonomy Similarity for Next Activity Prediction in Patient Treatment

Researchers developed TS4NAP, an AI approach that uses medical taxonomies and graph matching to predict next treatment steps for patients. The method leverages domain-specific knowledge from ICD-10 medical codes to improve treatment planning recommendations and make predictions more explainable for physicians.

• →TS4NAP combines medical taxonomies (ICD-10-CM and ICD-10-PCS) with graph matching to predict next patient treatment activities.
• →The approach addresses challenges in medical AI including knowledge-intensive data, high variability, and scarcity of medical datasets.
• →Electronic health records are analyzed to recommend suitable next steps in treatment processes.
• →The method was evaluated using event logs derived from the MIMIC-IV medical dataset.
• →Domain-specific taxonomies improve both prediction accuracy and explainability for medical decision-making.