Individualized Diagnosis and Prescription in Traditional Medicine: Decision-Making Process Analysis and Machine Learning-Based Analysis Tool Development.

Abstract

While pattern identification (PI) is an essential process in traditional medicine (TM), it is difficult to objectify since it relies heavily on implicit knowledge. Therefore, this study aimed to propose a machine learning (ML)-based analysis tool to evaluate the clinical decision-making process of PI in terms of explicit and implicit knowledge, and to observe the actual process by which this knowledge affects the choice of diagnosis and treatment in individual TM doctors. Clinical data for the development of the analysis tool were collected using a questionnaire administered to allergic rhinitis (AR) patients and the diagnosis and prescription results of TM doctors based on the completed AR questionnaires. Explicit knowledge and implicit knowledge were defined based on the doctors' explicit scoring and feature evaluations of ML models, respectively. There were many differences between the explicit and implicit importance scores in this study. Implicit importance is more closely related to explicit importance in prescription than in diagnosis. The analysis results for eight doctors showed that our tool could successfully identify explicit and implicit knowledge in the PI process. This is the first study to evaluate the actual process by which explicit and implicit knowledge affect the choice of individual TM doctors and to identify assessment tools for the definition of the decision-making process in diagnosing PI and prescribing herbal treatments by TM clinicians. The assessment tool suggested in this study could be broadly used for the standardization of precision medicine, including TM therapeutics.