OP13.09: Validating the machine learning model for first trimester prediction of pre‐eclampsia using a cohort from Spain

Le résumé fourni par la source

The objective of this study was to validate an artificial intelligence (AI) model that predicts the risk of pre-eclampsia (PE) based on maternal demographic, medical and pregnancy history, and the use of first trimester biomarkers. We used a previously trained machine learning model that predicted early (< 34 weeks of gestation) and preterm (< 37 weeks of gestation) PE versus no PE using a data set from two UK hospitals. The model was trained on demographic characteristics, medical and pregnancy history, and first trimester biomarkers. We applied the model to a Spanish data set for external validation, scaling the biomarker values for consistency. We measured the accuracy of the model using the area under the receiver-operating-characteristics curve (AUC) and the detection rate at various false-positive rates (FPRs) with and without race data. We found that the detection rates for early and preterm PE were 0.844 (0.672 to 0.947) and 0.778 (0.664 to 0.867), respectively, when using the triple test biomarker (uterine artery pulsatility index, mean arterial blood pressure, and maternal blood placental growth factor) and that removing race data lowered the accuracy. The corresponding AUCs were 0.92 and 0.913. We observed that the accuracy of the model was similar in Spain and the UK, and that PAPP-A did not increase the detection rate. The AI model accurately predicts the risk of PE based on maternal characteristics and first trimester biomarkers, and that it can be applied to diverse populations with some scaling required for biomarker values. It has potential for use in other geographical regions, but the model must be widely trained if new biomarker analysers are added. Consistent data collection is important for intercohort validation. The use of the AI model in clinical settings has the potential to improve the early detection of PE, allowing for timely interventions to improve maternal and fetal outcomes.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.