A clinical model and nomogram for early prediction of gestational diabetes based on common maternal demographics and routine clinical parameters.

Aim: We aimed to develop a risk prediction model for gestational diabetes mellitus (GDM) based on the common maternal demographics and routine clinical variables in Chinese population.

Methods: Individual information was collected from December 2018 to October 2019 by a pretested questionnaire on demographics, medical and family history, and lifestyle factors. Multivariable logistic regression was performed to establish a predictive model for GDM by variables in pre- and early pregnancy. The consistency and discriminative validity of the model were evaluated by Hosmer-Lemeshow goodness-of-fit testing and ROC curve analysis. Internal validation was appraised by fivefold cross-validation. Clinical utility was assessed by decision curve analysis.

Results: Total 3263 pregnant women were included with 17.2% prevalence of GDM. The model equation was: LogitP = -11.432 + 0.065 × maternal age (years) + 0.061 × pre-pregnancy BMI (kg/m ) + 0.055 × weight gain in early pregnancy (kg) + 0.872 × history of GDM + 0.336 × first-degree family history of diabetes +0.213 × sex hormone usages during pre- or early pregnancy + 1.089 × fasting glucose (mmol/L) + 0.409 × triglycerides (mmol/L) + 0.082 × white blood cell count (109/L) + 0.669 × positive urinary glucose. Homer-Lemeshow goodness-of-fit testing indicated a good consistency between predictive and actual data (p = 0.586). The area under the ROC curve (AUC) was 0.720 (95% CI: 0.697 ~ 0.744). Cross-validation suggested a good internal validity of the model. A nomogram has been made to establish an easy to use scoring system for clinical practice.

Conclusions: The predictive model of GDM exhibited well acceptable predictive ability, discriminative performance, and clinical utilities. The project was registered in clinicaltrial.gov.com with identifier of NCT03922087.