Questionnaire and LGBM Model for Assessing Health Literacy levels of Mongolians in China

Abstract

Background

It is difficult to accurately assess the health literacy(HL) level of Mongolians by using Chinese conventional HL questionnaire, due to their particularity in language, culture and living environment. Therefore, it is very important to design an exclusive HL questionnaire for them. In addition, the existing statistical models cannot meet the requirement of HL assessment with high precision, so it is necessary to study a new HL assessment model.

Methods

A HL questionnaire with 68 questions is designed by combing the HLS-EU-Q47and the characteristics of Mongolians in China. 742 Mongolians aged 18 to 87 in Inner Mongolia of China answered the questionnaire. A data set with 742 samples is constructed, where each sample has 68 features and 1 target. Based on it, the XGB and LGBM regression models are respectively constructed to assess the HL levels of respondents, and their evaluation effects are compared. The impact of each question on the HL level is quantitatively analyzed by using the feature-importance function in LGBM model to verify the effectiveness of the questionnaire and to find the key factors for affecting HL.

Results

The HL questionnaire has the high reliability, which is reflected by the high internal consistency (Cronbach’s coefficient=0.807) and test-retest reliability (Mutual Information Score= 0.803). The validity of the HL questionnaire is obtained by solving KMO and Bartlett Spherical Test Chi-square Value, which are 0.765 and 2486 ( \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p<0.001$$\end{document} ), respectively. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$R^2$$\end{document} index and the absolute error obtained by using the HL assessment model based on LGBM are 0.98347 and 11, which are better than ones by applying the model based-XGB, respectively. The quantitative analysis results show that all 68 questions have influence on HL level, but their degree are different. The first three factors are age, salary level, the judgment ability for the HL information in media, respectively. The HL level distribution of the respondents was 66.71 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\%$$\end{document} excellent, 25.74 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\%$$\end{document} good and 7.54 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\%$$\end{document} poor, respectively.

Conclusions

The presented HL questionnaire with 68 questions and LGBM regression model can obtain the HL level assessment results with high precision for Mongolians in China. The impact of each question in the questionnaire on the final assessment results can be quantified by using the feature-importance function in LGBM model, which is better than the existing qualitative analysis methods.