Background: Preeclampsia is one of the most dangerous complications in pregnancy. It is characterised by hypertension, lower limb oedema, proteinuria, and thrombocytopenia. Preeclampsia can cause various complications that impact many body systems.
Aim and Objectives: The main aim of this study is to predict preeclampsia in mothers and find a possible relationship with neonatal outcomes.
Materials and Methods: This hospital-based retrospective case-control study involves 1231 pregnant women. They were randomly allocated into the testing (N=246) and training (N=985) groups, which were set at a ratio of 1:4. All medical records were analysed. Various machine learning models such as logistic regression (LR), K nearest neighbours (KNM), support vector machine (SVM), discriminant analysis (DA), extreme gradient boosting (EGB), and random forest (RF) algorithm were used to predict preeclampsia. The performance of these models was evaluated using standards such as accuracy, precision, recall, correct classification, misclassification, and F score using XLSTAT 2024.
Results: Out of the 1231 women, 368 were diagnosed with Preeclampsia. Factors such as kidney disease, blood pressure (BP) ≥ 120/80 mmHg in early pregnancy, chronic hypertension, family history of hypertension, and diabetes were found to be significant contributors to preeclampsia. The random forest model showed the highest performance.
Conclusions: Renal disease, chronic hypertension, family history of hypertension and diabetes and BP ≥120/80 mmHg early in pregnancy were the most important factors for predicting preeclampsia. The random forest model is the best feasible approach for screening and predicting preeclampsia because it helps obstetricians identify high-risk pregnancies and thus prevent adverse outcomes. Preeclampsia in mothers strongly affects neonatal outcomes, and this can result in a higher percentage of low birth weight, prematurity and intrauterine growth retardation.
