2023, Vol. 27 ›› Issue (5): 657-662

Risk factor analysis of metabolic syndrome and construction of a nomogram prediction model in middle-aged and elderly people

Shen Lianwei, Zhu Hongliu, Wang Wei   

Abstract: BACKGROUND: With population aging and life-style variation in China, the prevalence of metabolic syndrome in the middle-aged and elderly is increasing. Therefore, it is necessary to screen the risk factors of the disease and construct a prediction model of the disease.
OBJECTIVE: To investigate the risk factors of metabolic syndrome in middle-aged and elderly people in China and to construct the nomogram prediction model of metabolic syndrome, so as to provide reference for the prevention and control of metabolic syndrome in such populations.
METHODS: The data were from the 2015 follow-up data of the China Health and Retirement Longitudinal Study (CHARLS), with metabolic syndrome as a dependent variable. Nine variables including left hand muscle strength, right hand muscle strength, standing time, walking speed, time to get up from the chair, sex, age, smoking and drinking were involved to explore the related factors of metabolic syndrome in middle-aged and elderly people. Propensity score matching and descriptive analysis were performed on SPSS. The original data set was divided into training set and verification set with Rstudio. The training set was subject to univariate and multivariate regression analysis, nomogram model construction and internal verification, whereas the verification set was subject to external verification.
RESULTS AND CONCLUSION: (1) A total of 7 384 middle-aged and elderly people were screened, including 518 (7%) with metabolic syndrome and 6 865 (93%) without metabolic syndrome. (2) 6 865 subjects without metabolic syndrome was paired by 1:1 with propensity score matching to screen out 518 subjects as the control group and those with metabolic syndrome acted as the case group. The control group and the case group constituted the original data set. The original data set was randomly divided into the training set (728 subjects) and the validation set (308 subjects) at 7:3. (3) Construction of the model: according to the binary logistic regression analysis results of the training set (modeling cohort), five variables such as right hand muscle strength, walking speed, sex, drinking and standing time were selected to build the model. Evaluation of the model: the area under the receiver operator characteristic curve of the model was 0.877 (95% confidence interval: 0.851-0.902), indicating that the model has high discrimination and good fitting of the calibration curve, indicating that the model has a high calibration degree. (4) Internal verification of the model: the bootstrap method was used to generate new samples after 1000 repeated sampling with return, and the generated calibration curve fitted well. (5) External verification of the model: the model was tested by the verification set (verification queue), the area under the receiver operator characteristic curve was 0.89 (95% confidence interval: 0.854-0.926), indicating that the model has high discrimination and good fitting of the calibration curve, that is, the model has high calibration degree and good efficiency in the verification queue. (6) The prediction model of metabolic syndrome in the middle-aged and elderly constructed in this study has good reliability. The nomogram obtained from this can predict the probability of metabolic syndrome in the middle-aged and elderly according to the right hand muscle strength, walking speed, sex, drinking, age and standing time, which is conducive to the early detection, diagnosis and treatment of metabolic syndrome in the middle-aged and elderly and can be popularized in clinical practice.
Key words: middle-aged and elderly people, metabolic syndrome, nomogram prediction model, risk factor, regression analysis