Analysis Of The Relationship Between Impaired Fasting Glucose And Blood Indicators Based On Random Forest And Logistic Regression

ObjectiveThis study aims to screen the predictors of impaired fasting glucose from the blood test data commonly used in physical examination by using the data of health examination.To explore the relationship between blood test indexes and their changes and impaired fasting blood glucose,identify the risk factors of impaired fasting blood glucose,and conduct risk prediction research,so as to provide theoretical basis for health education and health promotion of impaired fasting blood glucose and diabetes.Through the development of appropriate health guidance strategies,reduce and slow down the incidence and development of impaired fasting blood glucose and diabetes.MethodsThis study is a cross-sectional study,using convenience sampling.Taking health management center in jinan health check-up,2 medical check-up,1 year time intervals of six months or more for health check-up when there is no history of diabetes and normal fasting glucose from January 2018 to December 2019 as the research object.Collect basic information(gender,age),general examination information from 2018 to 2019(height,weight,body mass index,systolic blood pressure,diastolic blood pressure,disease history,surgical history,etc.)and blood test results.Calculating the gender and age distribution of the study subjects,and calculating the number of people who progressed to impaired fasting blood glucose and diabetes within 1 year of each age.The χ2 test was used to compare the changes of fasting blood glucose in each age group within 1 year.Take the second year fasting blood glucose level as the dependent variable,and use the first year general inspection and blood test indicators,and the second year general inspection and blood test indicators as the independent variable random forest model.The research objects were randomly divided into training set and test set according to the ratio of 7:3,and the sensitivity and specificity were calculated based on the OOB misclassification rate.The top 15 important factors selected by random forest are used as independent variables,First,a single-factor Logistic regression analysis is performed,and the statistically significant factors of the single-factor analysis are incorporated into the multi-factor Logistic regression model.Using maximum partial likelihood estimation for variable screening and parameter estimation,the factors related to abnormal glucose metabolism and their effect values[OR(95%CI)]are finally obtained.Results1.A total of 4416 subjects were included in this study,including 2560 males(57.97%),1856 females(42.03%),1792 persons under the age of 40(40.58%),1839 persons aged 40-59(41.64%),and 785 persons aged 60 and above(17.78%).At 1 year follow-up,3.45%of those with normal baseline blood glucose developed prediabetic blood glucose and 0.42%developed diabetic blood glucose.2.The sensitivity and specificity of predicting the change of fasting blood glucose after 1 year using blood test indexes was 67.1%and 100.0%.Uric acid,glutamyl transpeptidase,creatinine,AST/ALT,and estimated glomerular filtration rate were associated with changes in fasting glucose at 1 year.Each lumol/L increase in uric acid was associated with an increased risk of 0.3%(OR=1.003,95%CI:1.002-1.005,P<0.001).Each 1U/L increase in glutamyltranspeptidase was associated with a 0.6%increased risk(OR=1.006,95%CI:1.003-1.010,P<0.001).At the same time,each lumol/L decrease of creatinine was associated with a 3.0%increased risk(OR=0.971,95%CI:0.957-0.985,P=0.017).The risk of developing disease increased by 62.9%(OR=0.614,95%CI:0.411-0.917,P<0.001)for each 1 unit decrease of AST/ALT.Each 1%decrease in the estimated glomerular filtration rate was associated with an increased risk of 2.9%(OR=0.972,95%CI:0.958-0.988,P=0.025).Age,systolic blood pressure and body mass index were also associated with changes in fasting glucose at 1 year.The risk of impaired fasting glucose at 1 year increased by 2.4%(OR=1.024,95%CI:1.014-1.035,P<0.001)for each additional year of age.Each 1mmHg increase in systolic blood pressure was associated with a 1.8%increased risk(OR=1.018,95%CI:1.011-1.025,P<0.001).Each lkg/m2 increase in BMI was associated with a 7.9%increased risk(OR=1.079,95%CI:1.036-1.125,P<0.001).3.The sensitivity and specificity of predicting the change of fasting blood glucose after 1 year by using the changes of blood test indexes was 65.2%and 99.7%.The changes of triglyceride,alkaline phosphatase,albumin and glutamic oxalacetic transaminase were related to the changes of fasting blood glucose.The risk of impaired fasting blood glucose increased by 15.2%(OR=1.152,95%CI:1.057-1.257,P<0.001)for each 1mmol/L increase of triglyceride.The risk was increased by 2.1%(OR=1.021,95%CI:1.012-1.030,P<0.001)with the increase of alkaline phosphatase level by 1U/L.The risk of disease increased by 8.7%(OR=1.087,95%CI:1.029-1.148,P=0.003)for each g/L increase of albumin level.Each 1U/L increase in the level of asparaginase increased the risk of disease by 0.6%(OR=1.006,95%CI:1.003-1.009,P<0.001).The changes of body weight was also related to the changes of fasting blood glucose.Each 1kg increase in weight gain was associated with a 4.9%increased risk(OR=1.049,95%CI:1.019-1.081,P=0.001).Conclusion1.General results of systolic blood pressure,body mass index and blood test results such as uric acid,glutamyl transpeptidase,creatinine,AST/ALT,and estimated glomerular filtration rate have predictive significance for the risk of impaired fasting glucose one year later.For people with normal fasting blood glucose at present,if the above general examination and blood test indicators are abnormal,health managers should popularize diabetics’ knowledge,improve their vigilance against impaired fasting blood glucose and the onset of diabetes,timely adjust bad lifestyle and conduct,and regularly monitor blood glucose.In case of hyperglycemia and other phenomena,timely medical treatment can reduce the risk of diabetes and realize early screening and early diagnosis of diabetes.2.Regular monitoring of blood test indicators such as triglycerides,alkaline phosphatase,albumin,aspartate aminotransferase and changes in body weight are also important for the prediction of impaired fasting blood glucose.If the fasting blood glucose is impaired and the population at high risk of diabetes has the increase of the above indexes,the risk of the disease will increase accordingly.At this time,health management should be strengthened,health education should be carried out,and bad life behavior should be adjusted in time.Reduce the risk of impaired fasting blood glucose and diabetes by adjusting diet and increasing exercise.