Identification of gestational diabetes mellitus in European electronic healthcare databases: insights from the ConcePTION project

To develop and compare algorithms for identifying gestational diabetes mellitus (GDM) across European electronic healthcare databases and evaluate their impact on the estimated prevalence.

Multi-national cohort study using routinely collected electronic healthcare data SETTING: National and regional databases in five European countries (Norway, Finland, Italy, Spain and France), in primary and/or secondary care.

Pregnancy cohorts resulting in stillbirths or live births between 2009 and 2020, comprising 602 897 pregnancies in Norway, 507 904 in Finland, 374 009 in Italy, 193 495 in Spain and 116 762 in France.

The primary outcome was the prevalence of GDM identified using six algorithms: (1) Only diagnosis; (2) Diagnosis or prescription; (3) Two diagnoses or prescriptions (2DxRx); (4) Diagnosis including unspecified diabetes in pregnancy or prescription (DxRx broad); (5) Diagnosis excluding pre-existing diabetes in pregnancy or prescription; (6) Registration of GDM in a birth registry (BR).

The strictest algorithm (2DxRx) resulted in the lowest GDM prevalence, while the broadest (DxRx broad) resulted in the highest, except in France where it was BR. In the Nordic countries, GDM prevalence varied only slightly by algorithm; greater variations were observed in other countries. The prevalence ranged from 3.5% (95% CI: 3.5% to 3.5%) to 4.6% (95% CI: 4.5% to 4.7%) in Norway; 12.1% (95% CI: 12.0% to 12.2%) to 15.8% (95% CI: 15.7% to 15.9%) in Finland, where prevalence was much higher than elsewhere. The prevalence ranged from 1.3% (95% CI: 1.3% to 1.3%) to 5.4% (95% CI: 5.3% to 5.5%) in Italy; 1.6% (95% CI: 1.5% to 1.7%) to 6.2% (95% CI: 6.1% to 6.3%) in Spain; and 1.7% (95% CI: 1.6% to 1.8%) to 5.8% (95% CI: 5.7% to 5.9%) in France.

In this multinational study, GDM prevalence ranged from 1.3% to 15.8% depending on the algorithm and database. Nordic countries showed smaller differences in prevalence between algorithms, while the other countries showed larger variations, likely due to differences in coding practices, healthcare systems and database coverage.