Developing a Risk Stratification Tool to Predict Patients with Gestational Diabetes Mellitus at Risk of Insulin Treatment: A Cohort Study.

Abstract

Objectives: We aimed to develop and validate a simple, easy-to-use risk stratification tool to use in the diagnosis of gestational diabetes mellitus (GDM) to triage those more likely to require insulin treatment. Methods: Using an audit of patients with GDM in 2019, multivariable logistic regression was used to select variables and develop a prediction model for insulin requirement. A stratification tool was developed by dichotomising these selected variables; its performance was assessed with an internal cohort from 2021 and externally from patients managed at a separate hospital. Results: Patients with a higher fasting blood glucose concentration (OR 2.41, 95% CI 1.84-3.15) and higher booking body mass index (OR 1.48, 95% CI 1.07-2.03) were more likely to require insulin therapy whilst a later gestational-weeks-at-diagnosis value gave a lower risk of insulin therapy (OR 0.71, 95% CI 0.62-0.81 per week). The low-risk group for insulin requirement was defined thus: fasting blood glucose < 5.6 mmol/L, booking BMI < 30 kg/m², and gestational weeks at diagnosis ≥ 24 weeks. This classification had a negative predictive value (NPV) of 94% for insulin requirement, with a sensitivity of 84% and specificity of 56% in the development cohort. Similarly, in the internal and external validation cohorts, the NPVs were 93 and 90%, with sensitivity values of 77 and 78%, respectively. Conclusions: This study developed a pragmatic tool with three criteria for stratifying the GDM group not requiring insulin treatment, with successful validation for clinical use.