# Identifying Predictors of Utilization of Skilled Birth Attendance in Uganda Through Interpretable Machine Learning

## TL;DR

This study uses machine learning to identify factors influencing skilled birth attendance in Uganda, aiming to improve maternal health outcomes.

## Contribution

The study introduces interpretable machine learning to identify predictors of skilled birth attendance in a low-income setting.

## Key findings

- XGBoost achieved the best performance with an AUC of 0.75 in predicting skilled birth attendance.
- Education level and antenatal care visits were key predictors of skilled birth attendance.
- Region and perceived distance to healthcare facilities significantly influenced SBA use.

## Abstract

Skilled Birth Attendance (SBA) is essential for reducing maternal and neonatal mortality, yet access remains limited in many low- and middle-income countries. This study used machine learning to predict SBA use among Ugandan women and identify key influencing factors. We analyzed data from the 2016 Uganda Demographic and Health Survey, focusing on women aged 15 to 49 who had given birth in the preceding five years. After preparing and selecting relevant features, six tree-based models (decision tree, random forest, gradient boosting, XGBoost, LightGBM, CatBoost) and logistic regression were applied. Class imbalance was addressed using cost-sensitive learning, and hyperparameters were tuned via Bayesian optimization. XGBoost performed best (F1-score: 0.52; recall: 0.73; AUC: 0.75). SHapley Additive Explanations (SHAP) were used to interpret model predictions. Key predictors of SBA use included education level, antenatal care visits, region (especially Northern Uganda), perceived distance to a healthcare facility, and urban or rural residence. The results demonstrate the value of interpretable machine learning for identifying at-risk populations and guiding targeted maternal health interventions in Uganda.