Objective  To construct a metabolic syndrome (MS) factors-based model for predicting the risk of coronary heart disease (CHD) among rural residents of Xinjiang Uygur Autonomous Region (Xinjiang) for CHD prevention and treatment in the population.

  Methods   Complete baseline information were collected through face-to-face questionnaire interview, physical examination and laboratory tests from 16 853 of 18 524 adult rural residents (aged ≥ 18 years) recruited with multistage stratified cluster random sampling at two counties of Xinjiang and a division of Xinjiang Production and Construction Corps during April 2010, December 2012 and November 2016. Subsequent follow-up surveys were carried out among the eligible participants from April 2013 to July 2021 and finally 3 647 participants with a follow-up period of five years and more (averagely 5.28 ± 1.67 years) and without CHD at the baseline survey were included in the analysis, of which, two-thirds (n = 9 155) were randomly assigned into a training set and one-third (n = 4 492) into a verification set. Factor analysis was performed base on the data of 3 206 MS sufferers identified in the participants of training set to explore potential clustering pattern of MS components and probable CHD-related factors. Multivariate Cox proportional risk regression analysis was used to construct prediction model for CHD risk and receiver operating characteristic curve (ROC) analysis was adopted to evaluate the efficiency of the prediction model constructed.

  Results  The cumulative incidence of CHD was 4.94% for all the participants and for those of training and verification set. Training set data-based factor analysis revealed potential CHD-related factors for prediction model construction: obesity, blood pressure, blood lipid/glucose, renal metabolism, blood protein, liver enzyme, myocardial enzyme, and bilirubin and the 8 factors could explain 77.905% of the cumulative variance. Cox proportional hazard regression analysis showed that being female, at older age, obesity, low bilirubin, high blood pressure, and high blood lipid and glucose were risk factors of CHD for the participants of training and verification set. The area under the ROC curve (AUC) was 70.62 (95% confidence interval 95%CI: 0.742 – 0.782) for the prediction derived from training set data and 0.774 (95%CI: 0.742 – 0.805) for the prediction derived from validation set data.

  Conclusion  A MS factor-based prediction model for predicting CHD risk among rural residents in Xinjiang was constructed and the model could be used in CHD risk assessment and management in local population.