CMDAT versus Tzameret: An Exploratory Predictive Study to Testa Categorical and a Quantitative Dietary Analysis Method
By Eyal Shpringer - Online MSc - Advanced Oriental Medicine
Abstract
Background
The Cross Model Dietary-Analysis Tool (CMDAT), classifies patients’ diets according to the attributes of Hot/Cold-Dry/Wet that appear in Traditional Chinese Medicine (TCM) and Ayurveda. However, it is not clear whether the CMDAT output has any correlation with the output of conventional nutrition-analysis tools. Aims and objectives The study aimed to predict the CMDAT results using a Western, conventional dietary-analysis tool. The study’s primary objectives were to select patients’ dietary reports from a dataset, then analyse the musing CMDAT and Tzameret, a conventional nutrition-analysis tool. Thereafter, a logistic regression-based statistical analysis was applied to detect whether the Tzameret outcomes could predict the CMDAT outcomes.
Methodology
The study was an exploratory predictive study using a quantitative analysis of a cross-sectional survey dataset from MABAT, the Israeli National Health and Nutrition survey. The primary challenge of the study was to predict a qualitative and categorical output (CMDAT) from a quantitative, continuous, numerical analytical output (Tzameret).
Findings
120 dietary reports were randomly sampled out of 3,242 MABAT reports. 119 reports (n=119) comprising 50 Hot-Wet and 69 Cold-Wet reports were selected for analysis. Following the CMDAT and Tzameret analysis, a binary logistic regression method was applied to the selected reports. Thereafter, a statistically significant result was identified(p=0.005). The identified model explained 34% of the variance in the CMDAT groups and correctly classified 76.5% of cases. Moreover, four variables were statistically significant: carbohydrates (p=0.042) and copper (p=0.031) were found to have a positive association with the Cold+Wet category, while sodium (p=0.022) and energy (p=0.042) were found to have a negative association with the Cold+Wet category. However, the low odds ratio (OR) values of carbohydrates (1.019), energy (0.996) and sodium (0.999) indicate a very small effect size for these predictors. The OR of copper was much more substantial (7.068), indicating a greater effect size for this predictor; however, its wide confidence interval (1.193-41.917) indicates that generalisation of this predictor from sample to population should be made cautiously.
Conclusion
This exploratory study established a predictive model and demonstrated the association of four predictors with the CMDAT categories. Recommendations for future research include studying the validity of CMDAT, applying similar study designs to other populations, and expanding the nutrients used for establishing predictive models to extend the understanding of the relationship between diets’ content and their attributes.
