Development of the Sasangin Diagnosis Inventory Using the Random Forest Modeling

다국어 초록 (Multilingual Abstract)

Background: Sasang Constitutional Medicine (SCM), originating in Korea, categorizes individuals into four types based on physiological imbalances. Recent studies integrate AI like Random Forest to condense these questionnaires, simplifying diagnosis while maintaining accuracy, enhancing the potential of SCM for personalized medicine and preventive care.
Aim: The purpose of this study is to develop the Sasangin Diagnosis Inventory (SDI), and verify its reliability and validity.
Methods: A questionnaire was developed for constitution diagnosis with reference to the Clinical Practice Guideline, and was subjected to expert consensus and linguistic validation for its final form. In toto, 230 patients were recruited, with 169 assigned to the training group, and 61 to the test group for the questionnaire. The reference standard was agreement in Sasang constitutional diagnosis by at least 2 of 3 experts. A predictive model was built using the Random Forest machine learning algorithm, and based on this, the diagnostic accuracy of the model was assessed in the test group.
Results: The overall Cronbach’s α for the entire set of questionnaire items was 0.72. The diagnostic accuracy of the model for the three constitutions was 68.97 % (kappa coefficient 0.5403). The diagnostic accuracy of the model for the Symptomatology Groups was 47.54 % (kappa coefficient 0.3606).
Discussions and Conclusions: SDI using the model diagnosing the three constitutions has reliability and validity in Sasang constitutional diagnosis.

목차 (Table of Contents)

• List of Tables ............................................................................................................................ iii
• List of Tables of the Appendix .................................................................................................. iii
• List of Figures ........................................................................................................................... iv
• Nomenclature ............................................................................................................................ v
• Abstract ..................................................................................................................................... vi
• Ⅰ. Background ........................................................................................................................... 1
• Ⅱ. Materials and Methods ...................................................................................................... 4
• 2.1. Study Design and Setting
• 2.2. Questionnaire Development Process
• 2.3. Reliability and Validity Test
• 2.3.1. Participants and Sample Size
• 2.3.2. Data Measure
• 2.3.3. Reliability Test
• 2.3.4. Validity Test
• 2.3.4.1. Random Forest Modeling
• 2.3.4.2. Variable Selection
• 2.3.4.2. The Model to Diagnose Sasangin
• 2.3.4.3. The Model to Diagnose Symptomatology Groups
• 2.4. Statistical Analysis
• Ⅲ. Result ................................................................................................................................ 14
• 3.1. Questionnaire
• 3.1.1. Design Concept of the Questionnaire
• 3.1.2. Item Development
• 3.1.3. Questionnaire Construction
• 3.2. General Characteristics of Subjects
• 3.3. Reliability Test
• 3.4. Validity Test
• 3.4.1. Variable Selection
• 3.4.1.1. Response Rate of Items
• ii
• 3.4.1.2. Feature Importance of Explanatory Variables
• 3.4.2. Model to Diagnose Sasangin
• 3.4.3. Model to Diagnose Sasangin except Taeyang Type
• 3.4.4. Model to Diagnose Symptomatology Groups
• Ⅳ. Discussion ......................................................................................................................... 26
• Ⅴ. Conclusion ......................................................................................................................... 31
• Ⅵ. Reference .......................................................................................................................... 32
• Appendix
• A1. Literature Basis of the SDI
• A2. Questionnaire
• A3. Response Rate of All Items
• A4. Feature Importance of Explanatory Variables
• iii
• List of Tables
• Table 1. Sasang Constitutional Diagnosis Factors by Items .................................................... 17
• Table 2. General Characteristic of Subjects ............................................................................. 18
• Table 3. General Characteristic of the Training Group and Test Group .................................. 19
• Table 4. Cronbach’s α of the Questionnaire ............................................................................ 20
• Table 5. Response Rate of Items ............................................................................................. 21
• Table 6. Confusion Matrix of the SDI model to Diagnose Sasangin ....................................... 23
• Table 7. Validation of the SDI model to Diagnose Sasangin ................................................... 23
• Table 8. Confusion Matrix of the SDI model to Diagnose Sasangin except Taeyang Type .... 24
• Table 9. Validation of the SDI model to Diagnose Sasangin except Taeyang Type ................ 24
• Table 10. Confusion Matrix of the SDI model to Diagnose Symptomatology Groups ........... 25
• Table 11. Validation Coefficients of the SDI model to Diagnose Symptomatology Groups ... 25
• List of Tables of the Appendix
• Table A1. Literature Basis of the SDI ............................................................................................... 34−36
• Table A2. Questionnaire ................................................................................................................... 37−39
• Table A3. Response Rate of All Items .............................................................................................. 40−41
• Table A4. Feature Importance of Explanatory Variables ........................................................................ 42
• iv
• List of Figures
• Figure 1. Study design and setting ............................................................................................. 4
• Figure 2. Random forest machine learning model ................................................................... 10
• Figure 3. Random forest and statistical analysis flowchart ..................................................... 12
• Figure 4. Main concept of SDI category ................................................................................. 14
• Figure 5. Feature importance of explanatory variables ........................................................... 22