Improved first trimester selective screening for gestational diabetes mellitus in nulliparous women

To develop a clinical tool for first trimester risk-prediction of gestational diabetes mellitus (GDM) based on clinical risk factors among nulliparous women. The study population was drawn from singleton live births in California from 2007-2012 in the birth cohort file maintained by the California Office of Statewide Health Planning and Development, which includes linked birth certificate and mother and infant hospital discharge records. The sample was restricted to 1,156,708 nulliparous women without preexisting type 1 or type 2 diabetes mellitus. Patients with GDM were identified by ICD-9 codes from the maternal hospital discharge record and risk factors for GDM were identified from birth certificate and hospital discharge records. A parsimonious risk-prediction model was built using logistic regression in a random 2/3 of the cohort and tested in the remaining 1/3. In the model testing subset, model accuracy was assessed using discrimination (area under ROC curve (AUROC)) and calibration (calibration plots and Hosmer-Lemeshow goodness-of-fit test). To address the critical clinical issue of risk stratification for selective glucose testing, women were categorized as being low (<5% predicted GDM risk), average (5-10%), or high risk (>10%). The final prediction model included five risk factors (age at delivery, pre-pregnancy BMI, race/ethnicity, preexisting hypertension, and family history of diabetes) that were found to be most influential to GDM prediction through classification tree analysis (table). In the model testing subset, the risk-prediction model yielded an AUROC of 0.73 (95% CI 0.73-0.73) and showed good fit (figure). Using a risk stratification framework, 13.6% of nulliparous women (high risk) would undergo first trimester glucose testing, of whom 15.9% would develop GDM. Although evidence suggests that early diagnosis and treatment of GDM can greatly improve pregnancy outcomes, tools for early identification of risk for GDM aren’t commonly used in clinical practice. We demonstrate that incorporating routinely collected clinical risk factors in a tool aimed at predicting GDM risk early in pregnancy for nulliparous women, with unique risk profiles, can have important clinical value.View Large Image Figure ViewerDownload Hi-res image Download (PPT)