Real-World Drug-Adverse Reaction Prediction of Combined use of TCM Formula Granules and Western Medicine

The standardization of TCM formula granules and the monitoring of adverse reactions can assist clinicians in promoting rational drug use and improving treatment effects. However, the detection and identification of adverse reactions from the combination use of Chinese and Western medicine in the real world remain challenging. This paper proposes a framework that implements an automated DIKW (DataInformation-Knowledge-Wisdom) process by integrating the reaction detection from historical adverse reaction reports from FAERS (FDA Adverse Event Reporting System), as well as the construction of a knowledge graph of adverse reactions to Chinese and Western medicine. Experiments show that the framework achieves the best AUC score 0.82, outperforming a set of baseline models in the temporal adverse event prediction and drug association detection. The performances of feature ablation show that the knowledge represent learning and reaction detection help enhance the accuracy and Robustness of adverse event prediction. The refinement of this framework is expected to effectively assist clinicians in early detection of potential adverse reactions from the use of TCM formula granules and Western medicine, monitor and assess the adverse reactions, and help reduce health risks for patients.