Background: Gestational diabetes (GDM) is becoming more prevalent and with the recent revision of GDM diagnostic criteria, the incidence will further increase. With a significant proportion of GDM patients needing insulin to achieve glycemic control, this will result in higher clinical workload and resource utilization.

Aim: To develop a risk prediction model to identify those who are high risk of needing insulin to help reduce resource utilisation in this population.

Methods: Retrospective data was collected on 684 pregnant women receiving antenatal care in Sunshine Hospital who were diagnosed with GDM in 2013. We collected data from medical records on demographics, BMI, family and past medical history, and their management of GDM.

Women in the dataset were randomly divided into two groups: a derivation group (n = 309) and a validation group (n = 330).  Data from the derivation group were used to develop a simple prediction tool.  Logistic regression was performed to identify the predictors of insulin dependence with corresponding odds ratios (95% confidence intervals) for each risk factor. The prediction model was then assessed in the validation datasets.

Results: Among the variables explored, significant predictors for insulin-requiring GDM which were included in the model include: age >30 (OR 1.34, 95%CI 0.86-2.71), BMI ≥30 (OR 1.14, 95%CI 0.66-1.97) and fasting glucose level ≥ 5.5 mmol/L (OR 4.5, 95%CI 2.3-8.05). The performance of the prediction model on the validation dataset achieved a sensitivity of 43.9% and specificity of 82.3% in identifying women at high risk of needing insulin for their GDM management.

Conclusion: Predicting which patient need insulin for their GDM management early has proven difficult given the poor sensitivity of our prediction model. However, these factors especially fasting glucose ≥5.5 mmol/L at  diagnosis of GDM can be useful to identify those who are high risk of needing insulin and therefore in need of closer monitoring of their GDM.