Identification of Traditional Chinese Medicine Constitutions and Physiological Indexes Risk Factors in Metabolic Syndrome: A Data Mining Approach

Abstract

Objective. In order to find the predictive indexes for metabolic syndrome (MS), a data mining method was used to identify significant physiological indexes and traditional Chinese medicine (TCM) constitutions. Methods. The annual health check-up data including physical examination data; biochemical tests and Constitution in Chinese Medicine Questionnaire (CCMQ) measurement data from 2014 to 2016 were screened according to the inclusion and exclusion criteria. A predictive matrix was established by the longitudinal data of three consecutive years. TreeNet machine learning algorithm was applied to build prediction model to uncover the dependence relationship between physiological indexes, TCM constitutions, and MS. Results. By model testing, the overall accuracy rate for prediction model by TreeNet was 73.23%. Top 12.31% individuals in test group (n=325) that have higher probability of having MS covered 23.68% MS patients, showing 0.92 times more risk of having MS than the general population. Importance of ranked top 15 was listed in descending order . The top 5 variables of great importance in MS prediction were TBIL difference between 2014 and 2015 (D_TBIL), TBIL in 2014 (TBIL 2014), LDL-C difference between 2014 and 2015 (D_LDL-C), CCMQ scores for balanced constitution in 2015 (balanced constitution 2015), and TCH in 2015 (TCH 2015). When D_TBIL was between 0 and 2, TBIL 2014 was between 10 and 15, D_LDL-C was above 19, balanced constitution 2015 was below 60, or TCH 2015 was above 5.7, the incidence of MS was higher. Furthermore, there were interactions between balanced constitution 2015 score and TBIL 2014 or D_LDL-C in MS prediction. Conclusion. Balanced constitution, TBIL, LDL-C, and TCH level can act as predictors for MS. The combination of TCM constitution and physiological indexes can give early warning to MS.