External validation of a clinical scoring system for the risk of gestational diabetes mellitus

Abstract

Aim: A prediction rule for gestational diabetes mellitus (GDM) could be helpful in early detection and increased efficiency of screening. A prediction rule by means of a clinical scoring system is available, but has never been validated externally. The aim of this study was to validate the scoring system.

Methods: We used data from a prospective cohort study. Women were assigned a score based on age, BMI and ethnicity. Performance of the scoring system was evaluated in terms of discrimination and calibration (agreement between clinical score and observed probability of GDM). We compared the efficiency of a screening strategy derived from the scoring system with conventional screening.

Results: We studied 1266 women. Forty-seven women had GDM (3.7%). The scoring system discriminated moderately (area under the curve=0.64 (95% CI 0.56-0.72)). Calibration was limited (chi(2)=8.89, p=0.06). The screening strategy derived from the scoring system reduced the number of women needed to be screened with 25% for a comparable detection rate to universal screening.

Conclusion: Despite moderate discriminative capacity and calibration of the scoring system, the screening strategy based on the scoring system appears clinically useful. There is need for better prediction models for GDM.