Establishment And Validation Of Risk Prediction Model For Gestational Diabetes Mellitus

Objective: Based on "Zunyi Birth Cohort ",this study aimed to determine the current incidence of gestational diabetes mellitus(GDM)of pregnant women in Zunyi and identify its key influencing factors;establish a visual prediction model of GDM to screen high-risk pregnant women.This model is to provide support for the screening of high-risk pregnant women,prediction of GDM occurrence and provide basis for the formulation of primary prevention strategies of GDM.Methods: Pregnant women who didn’t severe diseases,infectious diseases and birth defects were included.We selected the pregnant women in this cohort from September 2020 to May2021 as the modeling population,from June 2021 to December 2021 as the internal validation population,and the pregnant women from Wuhan Cohort as the external validation population for model validation.Predictive models that were established based on the baseline status of pregnant women,personal history,family history,lifestyle and environmental pollutant exposure,and were divided into demographic-lifestyle factors model(DLFM)and demographic-lifestyle-environmental pollution factor model(DLEFM)based on whether environmental pollutant exposure included.Final model was selected according to better predictive performance of those two models,and visualized based on nomogram.Predictive model was established using the least absolute shringage and selection operator(Lasso)method to screen variables and logistic regression to establish the model.The discrimination and calibration of the model were evaluated respectively by the area under the curve(AUC)and the calibration curve analysis.The net benefit of the model was evaluated by the decision curve.Results: 1.This study included 1756 pregnant women from September 2020 to May 2021,1234 pregnant women from June 2021 to December 2021,and 1200 pregnant women from Wuhan during the same period.The incidence of GDM in each population was 15.5%,12.6%,and 16.3%,respectively.2.Based on prediction results evaluation of two models,this study selected DLEFM as the final model.A total of seven predictors including the history of GDM,family history of diabetes,pre-pregnancy BMI,history of hypertension,sedentary time,MBz P and MEP exposure were included based on Lasso-logistic regression analysis,The results showed that the low pre-pregnancy BMI(BMI≤18.5)was a protective factor of GDM risk,and reduced 10% GDM risk compared with pregnant women with normal BMI,OR was 0.90(95%CI 0.54-1.42).Pregnant women who were overweight or obese before pregnancy would increase 1.71(95%CI 1.27-2.29)times GDM risk compared with normal BMI pregnant women.The OR of pregnant women with a history of GDM is 4.22(95%CI 1.89-9.41).The OR of pregnant women who had family history of diabetes is 2.28(95%CI 1.05-4.71),those who had a history of hypertension is 2.61(95%CI 1.41-4.72).The risk of GDM increased by 16% for each hour of sedentary time.The risk of GDM in pregnant women with MBz P detected was 1.95(95%CI 1.45-2.67)times higher than that in undetected pregnant women,and MEP concentration in Q4 was 1.85(95%CI 1.26,2.73)times compared with that in Q1.The risk of GDM increased by degrees with the increase concentration of MEP.3.The AUC values of the external and internal validation of the DLEFM constructed in this study were 0.827(95%CI: 0.791-0.862)and 0.801(95%CI: 0.765-0.838)respectively,while the AUC values of the DLFM were 0.783(95%CI 0.743-0.824)and 0.760(95%CI:0.720-0.798).The calibration curve results show that the predicted probability of DLEFM is more consistent with the actual probability,and the net benefit of the decision curve of DLEEM is better than that of DLFM.4.This study constructed a visual GDM risk score nomogram based on DLEFM,included seven predictors including pre-pregnancy BMI,history of GDM,history of hypertension,family history of diabetes,sedentary time,MBz P and MEP concentrations.A rapid assessment of GDM risk can be calculated based on these predictors.Conclusion: The study found that the incidence of GDM in pregnant women in Zunyi is relatively high.The model results show that pregnant women with a history of hypertension,family history of diabetes and history of GDM are a high-risk group of GDM,and corresponding measures should be taken to intervene and control;Keeping normal BMI,reducing sedentary time,and minimizing exposure to PAEs during pregnancy are expected to reduce the risk of GDM.The visualized nomogram can provide clinicians with a more intuitive reference and provide preventive strategies for pregnant women.