Analysis Code for Predicting Caregiver-Reported Antibiotic Administration Among Febrile Under-Five Children in Bangladesh Using Explainable Machine Learning

This repository contains the analysis code used for the study “Predicting Caregiver-Reported Antibiotic Administration Among Febrile Under-Five Children in Bangladesh Using Explainable Machine Learning: A National Analysis of the Bangladesh Demographic and Health Survey 2022.”

The analysis includes cohort construction, data preprocessing and recoding, complex-survey logistic regression, feature selection, predictive modelling, PSU-disjoint development and holdout evaluation, calibration assessment, uncertainty estimation, and explainable machine-learning analyses.

The study used the Bangladesh Demographic and Health Survey (BDHS) 2022 Kids Recode dataset. The original respondent-level BDHS microdata are not redistributed in this repository because access is subject to The DHS Program data-use requirements. Researchers wishing to reproduce the analysis should obtain the BDHS 2022 Kids Recode dataset directly from The DHS Program.

The required source dataset is the BDHS 2022 Kids Recode file,BDKR81FL.SAV).