Dynamic Prediction of Perinatal Outcome in Pregnancies After Stillbirth or Perinatal Death: A Pilot Study to Establish Proof-of-Concept and Explore Method Feasibility

Objective To establish proof-of-concept for the dynamic prediction of adverse outcome in women with a history of stillbirth, repeatedly throughout the pregnancy. Methods A retrospective cohort study of women in a subsequent pregnancy following perinatal loss at a tertiary hospital between 2014 and 2019 was used as the basis for exploratory prognostic model development. Models were developed to repeatedly predict adverse outcome using the findings of sequential ultrasound scans for fetal biometry, umbilical and uterine artery Doppler. Results 506 participants were eligible, of whom 504 were included in analysis. 110 (22%) experienced an adverse pregnancy outcome. Ability to predict outcome using repeated head circumference and fetal weight measurements improved as the pregnancy progressed (e.g. Area Under Receiver Operating Characteristic Curve improving from 0.59 at 24 weeks to 0.74 at 36 weeks’ gestation), supporting proof-of-concept. Predictors to include in dynamic prediction models were identified. Conclusions The present study supports proof-of-concept for dynamic prediction in pregnancy following stillbirth, which could be used to identify risks earlier in the pregnancy, while highlighting methodological challenges and requirements for subsequent large-scale model development studies.