Abstract:
Introduction: Cesarean section delivery is one of the most common surgical procedures in the
world. Common problem includes infection of uterine, excessive bleeding, blood clots, and
potential injury to surrounding organs during surgery. Babies born by cesarean section also
experience breathing difficulties, especially if delivered before 39 weeks, and are at a slightly
higher risk of certain health conditions later in life.
Objective: To predict Cesarean section delivery and identify its determinants factor among
reproductive-age women (15–49 years), in sub-Saharan Africa countries, using machine learning
algorithm.
Methods: The data set obtained from a demographic health survey among Sub-Saharan Africa
countries. This study used data on weighted sample of 388,015 women who were in their
reproductive age. Python software version 3.2 utilized for data processing, and machine learning
models such as lightGBM, Random Forest, Decision Tree (DT), XGB (Extreme Gradient
Boosting), K-Nearest Neighbor(KNN), ANN, logistic regression (LR), Naïve Bayes were
implemented. In this work, we evaluated the predictive models' performance using performance
assessment criteria such as accuracy, precision, recall, F1_score and the AUC curve.
Results: In this study, 388,015 reproductive-aged women were included in the final analysis.
Kenya (18.28%), Madagascar (12.11%), Burundi (12.08%), and Tanzania (10.4%) were among
the countries with high rates of cesarean section delivery. LightGBM classifier were the best
performed model with an achievement of accuracy of 85%, precision of 85%, recall of 91%, and
an AUC 89%. The Shapley additive explanation plot were identifying the most significant
determinants of cesarean section delivery. Which is: age, size of child at birth, watching TV, listing
radio, maternal education, health status, marital status, place of residence, pregnancy terminated
were the most predictor variable based on the outperformed model.
Conclusions: Among the machine learning algorithms evaluated, the LightGBM model
demonstrated the highest predictive performance, making it an effective tool for identifying
women at risk of undergoing CS delivery. The most significant predictors for cesarean section
(CS) delivery were identified by shape adaptive explanation model. These significant predictors
provide valuable insights that can inform targeted interventions and policy decisions in SubSaharan
Africa
countries.
V | P a g e
➢ Recommendation: Health providers should use tools like the lihgtBGM model to predict
cesarean section delivery, which is important for timely prediction. Such tools have been
proved to improve clinical decision-making processe
