Study On A Self-rediction Model For Metabolic Syndrome Based On Interpolated Features

Metabolic syndrome(MetS)is a multimetabolic disorder that can cause hypertension,hyperglycemia and hyperlipidemia.MetS is usually diagnosed by indicators from physical examinations,and diagnostic indicators include height,weight,waist circumference,blood pressure,and some blood test data.However,the penetration rate of physical examinations in China is substantially lower than that in developed countries,such a situation demands an approach for predicting MetS that does not depend on hospital health checks.The lack of blood test data in MetS self-care poses challenges to MetS prediction.The research proposed a novel method for MetS prediction based on feature imputation.The dataset for modeling contains 91,420 individuals from the First Affiliated Hospital of Zhejiang University.In this method,the first step predicted blood test data,and the second step used the predicted blood test data as supplemental features for training the MetS predictive model.XGBoost was used to build two additional models to compare the performance of the three methods.The first model used only self-observable features,the second was trained with extra blood test data,and in the third approach,the blood test items were imputed using multivariate imputation by chained equations(MICE)instead of raw data.The results confirmed that our feature imputation model exhibited the highest prediction accuracy vs.the first two models.Moreover,this method is also applicable under different scenarios when earlier blood test data can be obtained.We revealed that with full sets of the three measurements in earlier blood test data,the prediction accuracy of MetS could be further improved.However,the improvement failed to pass the significance level.Our findings demonstrate the feasibility of the feature imputation methods for MetS homecare applications and provide novel ideas for promoting innovative research and health management in MetS.Further validation and implementation of our proposed model might improve quality of life and ultimately benefit the general population.