Performance of machine‐learning approach for prediction of pre‐eclampsia in a middle‐income country

Abstract

ObjectivePreeclampsia (PE) is a serious complication of pregnancy associated with maternal and fetal morbidity and mortality. Early‐onset and preterm preeclampsia have long‐term health implications for both mothers and infants. Current prediction models have limitations and may not be applicable in resource‐limited settings. Machine learning (ML) algorithms offer a potential solution for developing accurate and efficient prediction models.MethodsWe conducted a prospective cohort study in Mexico City to develop a first‐trimester prediction model for preterm preeclampsia (pPE) using ML. Maternal characteristics and locally developed multiples of the median (MoMs) for mean arterial pressure (MAP), uterine artery pulsatility index (UtA‐PI), and serum placental growth factor (PlGF) were used for variable selection. The dataset was split into training, validation, and test sets. An elastic net method was employed for predictor selection, and model performance was evaluated using area under the curve (AUC) and detection rates (DR) at 10% false positive rates (FPR).ResultsThe final analysis included 3,050 pregnant women, of whom 124 (4.07%) developed PE. The ML model showed good performance, with AUCs of 0.897, 0.963, and 0.778 for pPE, early‐onset preeclampsia (ePE), and any type of preeclampsia (all‐PE), respectively. The DRs at 10% FPR were 76.5%, 88.2%, and 50.1% for pPE, ePE, and all‐PE, respectively. The ML model outperformed previous prediction models and showed better performance than external validations of existing algorithms.ConclusionsOur ML model demonstrated high accuracy in predicting pPE and ePE using first‐trimester maternal characteristics and locally developed MoMs. The model could provide an efficient and accessible tool for early prediction of preeclampsia, facilitating timely interventions and improved maternal and fetal outcomes.This article is protected by copyright. All rights reserved.

Prediction model of preeclampsia using machine learning based methods: a population based cohort study in China

“…PLGF and UtA-PI could only increase the AUC to 0.83 for PE prediction and to 0.88 for preterm PE prediction. J. Torres-Torres’s study ( 22 ) also found that in PE prediction, the AUC was 0.786 when the predictive factors included maternal characteristics and MAP. However, when PLGF and UtA-PI were added, the AUC decreased to 0.778.…”

“…Melinte-Popescu AS’s study ( 21 ) included four machine learning-based models: decision tree (DT), naïve Bayes (NB), support vector machine (SVM), and random forest (RF) for PE screening in the first trimester, the study indicates that machine learning-based models could be useful tools for PE prediction in the first trimester of pregnancy. Torres-Torres J ( 22 ) study also finds that elastic net regression offers a potential solution for developing accurate and efficient prediction models for PE and offers significant clinical benefits. The predictive model performance of PE may vary among different ethnic groups.…”

Understanding the Pathophysiology of Preeclampsia: Exploring the Role of Antiphospholipid Antibodies and Future Directions

“…Various omics, including miRNAs, which are small, single-stranded RNAs approximately 18-24 nucleotides long, are implicated in placental formation. miRNAs play a crucial role in regulating gene transcription, with deregulation of certain miRNAs implicated in PE, particularly in angiogenesis mediated by VEGF [68].…”

A review on Machine Learning Deployment Patterns and Key Features in the Prediction of Preeclampsia

“…across all included subgroups in the studies. Considering AUC and recall values, which emerged across 10 out of 14 studies leading to the most used performance metrics, Torres-Torres et al achieved the highest AUC of 0.96 as well as a DR of 88% at a FPR of 10% in predicting early-onset PE (<34 weeks of gestation), utilizing Elastic Net Regression [21]. Torres-Torres et al did not report a recall value, hence the highest recall value for early-onset PE was achieved by Gil et al at 84%.…”