Labor and delivery complications are the difficulty or abnormality that occurs during the process of labor or delivery. They include complications such as postpartum hemorrhage (PPH), preeclampsia/eclampsia, perinatal asphyxia, fetal distress, placenta previa, placenta abruption, cephalopelvic disproportion, uterine rupture, failure to progress, rapid labor, and preterm labor. These complications could cause severe health problems or death to the mother or to the baby(1).
According to a World Health Organization (WHO) report, worldwide maternal mortality is very high, especially in developing countries. Globally, in 2015 alone approximately 303,000 women died due to pregnancy complications, 99% (302,000) of whom were in developing regions and sub-Saharan Africa alone accounting for approximately 66% (201,000) (2). Hemorrhage, hypertensive disorders, sepsis, abortion, and embolism are the highly-ranked causes of maternal death (5).
In Ethiopia, the main direct causes of maternal death are obstetric complications such as hemorrhage, obstructed labor/ruptured uterus, pregnancy-induced hypertension, puerperal sepsis, and unsafe abortion. Hemorrhage has been the leading cause of maternal death followed by HDP and sepsis while obstructed labor and abortion are reduced. The most indirect causes were anemia and malaria (3) (4).
Maternal complications have psychological, social, and economic consequences. The psychological consequences (depression, suicidal), social consequences (low self-care, inability to resume housework, afraid sexual and social relations), and economic consequences (catastrophic expenditure, family’s job affected due to delivery) are common for mothers with complicated deliveries compared to mothers with uncomplicated deliveries . Women with severe maternal complications experienced an increased risk of mortality and morbidity, a higher risk of depression, physical symptoms such as difficulty in competing daily household work, and financial problems (5) (7 PH). Furthermore, obstetric interventions and complications such as emergency cesarean section, postpartum hemorrhage, labor induction, instrumental vaginal delivery, and obstetric anal sphincter injury affect childbirth satisfaction (6). Different studies on predictive analysis were conducted using machine learning to predict maternal health issues. Predictive analysis involves the use of data, statistical algorithms, and machine learning techniques to forecast the likelihood of future outcomes based on historical data (7).
Multu et al. (8) used machine learning algorithms particularly KNN, decision tree, light gradient boosting Machine (Light GBM), CatBoost, random forest, KNN to determine maternal health risk on a dataset collected from IoT-based risk monitoring systems. Age, systolic blood pressure, diastolic blood pressure, breathing speed, and heart rate features were considered for model development. Performance metrics such as accuracy, specificity, sensitivity, false positive rate, false discovery rate, false negative rate, and F1 score test were used to measure the model performance. An accuracy of 89.16% was achieved with a model developed using the DT classifier.
Escobar et al. (9) developed a predictive model for obstetric and fetal complications using gradient boost and logistic regression algorithms with electronic health data. Python 3.6.3 and sickit-learn 0.91 were utilized to develop the models. The gradient boost-based model performed 79.0% accuracy which was better than that of the logistic regression based model.
Rodríguez et al. (10) proposed the use of machine learning techniques to build a classifier for severe maternal mortality to support medical workers in decision-making at an early stage. Logistic regression with a 5-fold stratified cross validation technique was used, while analysis of variance was conducted to filter features for modeling. Preeclampsia, voluntary interruption of pregnancy maternal cause, scholarship, diabetes, origin, socioeconomic status, perinatal mortality, and htacronica had significant effects on the occurrence of severe maternal mortality.
Raza et al. (11) developed artificial neural network-based system for predicting maternal health risks using DT-BiLTCN that uses decision trees, a bidirectional long short-term memory network, and a temporal convolutional network that provided a set of features by support vector machine with 98% accuracy. Diastolic and systolic blood pressure, heart rate, and age of pregnant women were health conditions that were the strongest indications of health risk during pregnancy.
Labor and delivery complications cause mortality, morbidity, and socioeconomic problems. To reduce these labor and delivery complication-induced problems, it is very important to have an appropriate method for the early detection and prediction of labor and delivery complications during the prepregnancy and antenatal periods. Hence, using machine learning-based predictive models, this study intended to predict and identify determinants of labor and delivery complications from antenatal care (ANC) follow-up data.