ABSTRACT
Objective: In China, doctors at TCM hospitals and clinics often divide patients with a Western medicine (WM) disease into several syndrome classes from the TCM perspective and treat patients in different classes using different principles. A key problem is how to carry out the classification properly. We propose an evidence-based ap-proach for solving the problem where evidence is obtained by analyzing unlabeled symptom data using latent tree models.Method: In previous work, we have shown how latent tree analysis of symptom data can be used to identify TCM syndrome classes among patients with a WM disease. In the paper, we investigate how to establish classification rules for distinguishing between the classes.Results: We have applied the method to a data set about Vascular Mild Cognitive Impairment that involves 93 symptoms and 803 patients. Nine syndrome types are identified, along with the corresponding classification rules. Conclusions: An evidence-based approach to the TCM patient classification prob-lem has been developed. The approach can be used to answer the following questions about a WM disease: What TCM syndrome classes are there? What are the sizes of the classes? What are the statistical characteristics of each class? How can one differentiate between the different classes?