| Objective To understand the incidence rate,and its distribution of metabolic syndrome among governmental employees in Hunan Province.To explore the behavioral influences of metabolic syndrome and establish a predictive model for the incidence of metabolic syndrome,and to provide a scientific basis for the prevention,identification and control of metabolic syndrome in this population.Methods This study was based on "Prospective Cohort Study on Common Chronic Non-Communicable Diseases among Governmental Employees in Hunan Province",a sub-project of the National Key Research and Development Program,in which a general hospital from each of 4 cities in Hunan Province was selected as the study site.Organizations of governmental employees undergoing health examinations were selected by cluster sampling.Questionnaires,health checks,and laboratory tests were used for data collection.A total of 13,113 governmental employees were recruited in the baseline survey,of whom 10,996 did not have metabolic syndrome,and 9,706 were followed up after 1 year.Subjects without metabolic syndrome at baseline were divided into 2 groups according to whether they had developed metabolic syndrome by the follow-up survey.Poisson regression models were used to analyze the associations of behavioral factors(i.e.physical activity,smoking,drinking,sleep status,and diet)with the incidence of metabolic syndrome.Doseresponse relationships between them were estimated by restricted cubic spline function.A pathway model was established to explore the pathways of environmental and psychosocial factors influencing the onset of metabolic syndrome through behavioral factors.Machine learning models were constructed to predict the incidence of metabolic syndrome.Results(1)Influence of behavioral factors on the incidence of metabolic syndrome: 1-year incidence rate of metabolic syndrome in Hunan governmental employees were 7.2%.After adjusting for covariates,sleep duration at night 8.1 to 9 hours(RR = 0.61,95% CI: 0.41 to 0.93,P= 0.020)and nap duration 31 to 60 minutes(RR = 0.77,95% CI: 0.62 to0.96,P = 0.022)were protective factors for metabolic syndrome.Current smoking(RR = 1.38,95% CI: 1.12～1.70,P = 0.003),higher amount of daily smoking(RR = 1.01,95% CI: 1.00 to 1.02,P = 0.022),passive smoking(RR = 1.20,95% CI: 1.03 to 1.40,P = 0.017),more years of passive smoking(RR = 1.01,95% CI: 1.00 to 1.02,P = 0.009),more days of passive smoking per week(RR = 1.10,95% CI: 1.05 to 1.17,P < 0.001),higher sleep disturbances component score of PSQI(RR = 1.14,95% CI:1.00 to 1.30,P = 0.042),and not eating cereals(RR = 1.34,95% CI: 1.01 to 1.79,P = 0.046)were risk factors for metabolic syndrome.Doseresponse relationships showed that the metabolic syndrome risk increased with increasing amount of daily smoking,years of passive smoking.It showed U-shaped curve relationships of sleep duration at night and nap duration with metabolic syndrome risk,respectively.(2)Path analysis for the incidence of metabolic syndrome:socioeconomic status of governmental employees influenced the metabolic syndrome risk through direct effect(standardized effect size:-0.023,95%CI:-0.043 to-0.003,P = 0.027)and indirect effect(standardized effect size:-0.008,95% CI:-0.012 to-0.003,P = 0.001),where work intensity,life events,depressive symptoms,and risky behaviors mediated the association.Depressive symptoms influenced metabolic syndrome risk through risky behaviors(standardized effect size: 0.035,95% CI: 0.014 to 0.055,P =0.001).(3)Prediction of incidence of metabolic syndrome: the Bayesian Additive Regression Trees algorithm was used to train the training set.A total of 350 features were incorporated to build the "All variable model",with an AUC,accuracy,sensitivity,and specificity for the test set of 0.899,0.800,0.858 and 0.795,respectively.After selecting the 27 most important features by filtering method,its AUC,accuracy,sensitivity,and specificity were 0.904,0.767,0.895 and 0.756,respectively.The “Questionnaire variable model” incorporated 297 features and had an AUC,accuracy,sensitivity,and specificity of 0.833,0.699,0.821 and 0.688,respectively.After choosing the 13 most important features,its AUC,accuracy,sensitivity,and specificity were 0.843,0.736,0.815,and 0.729,respectively.Conclusions(1)1-year incidence rate of metabolic syndrome in Hunan governmental employees was 7.2%.Sleep duration at night 8.1 to9 hours and nap duration 31 to 60 minutes were protective factors for metabolic syndrome.Current smoking,passive smoking,higher sleep disturbances component score,and not eating cereals were risk factors for metabolic syndrome.The risk of developing metabolic syndrome increased with amount of daily smoking and the years of passive smoking.The relationships of sleep duration at night and nap duration with the risk of metabolic syndrome showed U-shaped curves.(2)Socioeconomic status,risky behaviors,and depressive symptoms could affect the onset of metabolic syndrome.Socioeconomic status had direct and indirect effects on metabolic syndrome,with indirect effect mediated mainly through work intensity,life events,depressive symptoms,and risky behaviors.Depressive symptoms affected metabolic syndrome mainly through indirect effects mediated by risky behaviors.(3)The best algorithm for predicting the individual risk of metabolic syndrome was Bayesian Additive Regression Trees.The AUC,accuracy,sensitivity and specificity were 0.843,0.736,0.815 and 0.729,respectively,for the "questionnaire variable model" using 13 variables.The model is simple to use and has good predictive performance for early identification of people at risk for metabolic syndrome. |
PDF Full Text Request