Predicting Birth Mode Using Ensemble Machine Learning: Evidence from Bangladesh Demographic and Health Survey

Background and Aim: Maternal mortality and childbirth complications remain significant challenges in many developing countries, particularly in rural areas. Early identification of childbirth risks is crucial for reducing mortality rates. This study employs a range of machine learning techniques to identify key features related to Birth Mode and determine the most effective predictive model. The analysis is primarily based on data from the BDHS 2017–2018, aiming to improve risk prediction and ultimately enhance maternal health outcomes. Methods To identify key factors influencing birth mode, the Boruta algorithm was applied. The optimal predictive framework was then determined by evaluating several machine learning algorithms, including Random Forest (RF), Bagging, GradientBoost (GB), XGBoost (eXtreme Gradient Boosting), and AdaBoost (Adaptive Boosting). The best model was selected based on performance metrics such as precision, sensitivity, specificity, F-score, accuracy, and AUC value. Results The chi-square test revealed a significant association between birth mode and factors such as division, place of residence, age at first birth, number of living children, husband's education, respondent's employment status, wealth index, iron intake during pregnancy, education level, and BMI, all at a 1% significance level. The Boruta algorithm identified 13 key features, including those from the chi-square test, plus two additional ones. Ensemble models (RF, Bagging, GB, XGBoost, AdaBoost) were trained using these features. Bagging exhibited the highest accuracy (0.7611), precision (0.7746), F-score (0.8278), and sensitivity (0.8887). XGBoost achieved the highest specificity (0.8962), while AdaBoost had the best AUC (0.7833). Overall, Bagging outperformed the other models across four out of six metrics. Conclusions The study demonstrates the effectiveness of the Bagging algorithm for predicting birth mode, providing important insights into the factors associated with caesarean deliveries. The findings indicate a higher likelihood of caesarean sections among women from wealthier socioeconomic groups compared to those from lower-income backgrounds. This observation underscores the importance of increasing awareness regarding the potential risks and overuse of caesarean procedures, particularly among higher-income populations. The results suggest that incorporating these identified risk factors into clinical decision-making and public health strategies may support healthcare providers and policymakers in Bangladesh in reducing the prevalence of unnecessary caesarean deliveries.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-19
DOI
https://doi.org/10.21203/rs.3.rs-9726982/v1
Primary Topic
Artificial Intelligence in Healthcare
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Predicting Birth Mode Using Ensemble Machine Learning: Evidence from Bangladesh Demographic and Health Survey

Md. Musharraf Hossain, Farhana Afrin Duty, Nahid Salma, Tasneem Binte Reza et al.
Research Square
Artificial Intelligence in Healthcare
preprint

Predicting Birth Mode Using Ensemble Machine Learning: Evidence from Bangladesh Demographic and Health Survey

Md. Musharraf Hossain, Farhana Afrin Duty, Nahid Salma, Tasneem Binte Reza, M Mahbub
preprint en

Abstract

Background and Aim: Maternal mortality and childbirth complications remain significant challenges in many developing countries, particularly in rural areas. Early identification of childbirth risks is crucial for reducing mortality rates. This study employs a range of machine learning techniques to identify key features related to Birth Mode and determine the most effective predictive model. The analysis is primarily based on data from the BDHS 2017–2018, aiming to improve risk prediction and ultimately enhance maternal health outcomes. Methods To identify key factors influencing birth mode, the Boruta algorithm was applied. The optimal predictive framework was then determined by evaluating several machine learning algorithms, including Random Forest (RF), Bagging, GradientBoost (GB), XGBoost (eXtreme Gradient Boosting), and AdaBoost (Adaptive Boosting). The best model was selected based on performance metrics such as precision, sensitivity, specificity, F-score, accuracy, and AUC value. Results The chi-square test revealed a significant association between birth mode and factors such as division, place of residence, age at first birth, number of living children, husband's education, respondent's employment status, wealth index, iron intake during pregnancy, education level, and BMI, all at a 1% significance level. The Boruta algorithm identified 13 key features, including those from the chi-square test, plus two additional ones. Ensemble models (RF, Bagging, GB, XGBoost, AdaBoost) were trained using these features. Bagging exhibited the highest accuracy (0.7611), precision (0.7746), F-score (0.8278), and sensitivity (0.8887). XGBoost achieved the highest specificity (0.8962), while AdaBoost had the best AUC (0.7833). Overall, Bagging outperformed the other models across four out of six metrics. Conclusions The study demonstrates the effectiveness of the Bagging algorithm for predicting birth mode, providing important insights into the factors associated with caesarean deliveries. The findings indicate a higher likelihood of caesarean sections among women from wealthier socioeconomic groups compared to those from lower-income backgrounds. This observation underscores the importance of increasing awareness regarding the potential risks and overuse of caesarean procedures, particularly among higher-income populations. The results suggest that incorporating these identified risk factors into clinical decision-making and public health strategies may support healthcare providers and policymakers in Bangladesh in reducing the prevalence of unnecessary caesarean deliveries.

Research Square
Jahangirnagar University (BD)
Artificial Intelligence in Healthcare
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon