Systematic Review and External Validation of Clinical Prediction Models for Adverse Pregnancy Outcomes Using Routinely Collected Pre-Conception and Early Pregnancy Data

The study systematically reviewed and externally validated existing clinical prediction models for gestational diabetes, pre-eclampsia, stillbirth, and small-for-gestational-age outcomes. The research analyzed 48 models from 30 studies using data from 1.58 million pregnancies in women aged 14-49 from UK primary care between 2000 and 2020. Most models demonstrated poor discrimination ability and significant calibration errors when applied to the UK population. For example, the C-statistic for gestational diabetes ranged from 0.29 to 0.77, for pre-eclampsia from 0.42 to 0.71, while models for stillbirth and small-for-gestational-age showed weaker performance. The authors concluded that most existing models are not suitable for direct implementation in UK clinical practice without further calibration or re-development using large representative datasets.