A Japanese five-constitution model: validation and characterization using the constitution in Chinese medicine...

The 60-item Constitution in Chinese Medicine Questionnaire (CCMQ), based on a traditional nine-constitution framework, has been widely used in China for health assessments. However, its length poses a practical barrier to its widespread clinical use. A previous study in Japan suggested that a five-constitution model was more optimal for the Japanese population and subsequently developed a simplified 18-item questionnaire. This study aimed to validate the structure of this five-constitution model and establish the standalone diagnostic accuracy of the 18-item questionnaire.

This was a secondary analysis of a cross-sectional dataset comprising 851 healthy Japanese participants who completed the 60-item CCMQ. We first used hierarchical clustering (the NbClust package) on the 18-item data to determine the optimal number of clusters. We then used principal component analysis (PCA) to analyze the structure of the clusters. Finally, a decision-tree model was developed to classify participants based on the 18 items, and its performance was evaluated using 10-fold cross-validation.

Clustering analysis of the 18-item data confirmed that five was the optimal number of clusters. PCA revealed that the five groups were clearly separated along a primary "health axis". The decision-tree model, using only 18 items, successfully classified participants into five constitution types (a structure validated on the 18-item data itself), with an overall accuracy of 81.6% and a balanced accuracy of 81.3%.

The 18-item simplified questionnaire is a valid and reliable instrument for constitutional assessment in Japan. This concise tool effectively retains the structural and predictive information of the original 60-item version, offering a practical solution for large-scale research and personalized health management in clinical settings.