Endocrinology, Diabetes & Metabolism (Jan 2023)

Validation of Danish registry‐cases of type 1 diabetes in women giving live birth using a clinical cohort as gold standard

  • Tina Wullum Gundersen,
  • Andreas Ebbehoj,
  • Sine Knorr,
  • Dorte Møller Jensen,
  • Peter Damm,
  • Ellen Christine Leth Løkkegaard,
  • Elisabeth R. Mathiesen,
  • Reimar W. Thomsen,
  • Tine Dalsgaard Clausen

DOI
https://doi.org/10.1002/edm2.374
Journal volume & issue
Vol. 6, no. 1
pp. n/a – n/a

Abstract

Read online

Abstract Introduction The aim of this study was to validate type 1 diabetes in women giving live birth in the Danish national registries against a clinical cohort of confirmed cases (the Danish Diabetes Birth Registry [DDBR] cohort). Methods National registries including diagnosis codes, redeemed prescriptions and background data were combined. Three main algorithms were constructed to define type 1 diabetes in women giving live birth: (1) Any diabetes diagnosis registered before delivery and before age of 30, (2) a specific type 1 diabetes diagnosis registered before delivery regardless of maternal age and (3) a ‘preexisting type 1 diabetes in pregnancy’ diagnosis registered before delivery. In additional sub‐algorithms, we added information on anti‐diabetic medicine and gestational diabetes diagnosis. We calculated positive predictive value (PPV) and completeness using the DDBR cohort as gold standard. Since DDBR included between 75 and 93% of women with confirmed type 1 diabetes giving live birth, we used quantitative bias analysis to assess the potential impact of missing data on PPV and completeness. Results Main algorithm 2 had the highest PPV (77.4%) and shared the highest completeness (92.4%) with main algorithm 1. Information on anti‐diabetic medicine and gestational diabetes increased PPV, on expense of completeness. All algorithms varied with PPV between 65.7 and 87.6% and completeness between 73.6 and 92.4%. The quantitative bias analysis indicated that PPV was underestimated, and completeness overestimated for all algorithms. For algorithm 2, corrected PPV was between 82.1 and 94.6% and corrected completeness between 84.7 and 91.2%. Conclusions The Danish national registries can identify type 1 diabetes in women giving live birth with a reasonably high accuracy. The registries are a valuable source for future comparative outcome studies and may also be suitable for monitoring prevalence and incidence of type 1 diabetes in women giving live birth.

Keywords