Exploring machine learning algorithms to predict acute respiratory tract infection and identify its determinants among children under five in Sub-Saharan Africa

• Published in: Frontiers in pediatrics Vol. 12; p. 1388820
• Main Authors: Yehuala, Tirualem Zeleke, Fente, Bezawit Melak, Wubante, Sisay Maru, Derseh, Nebiyu Mekonnen
• Format: Journal Article
• Language: English
• Published: Switzerland Frontiers Media S.A 20.11.2024
• Subjects: acute respiratory infection; hyper parameter tuning; machine learning; Pediatrics; prediction; SHAP (SHapley additive exPlanations); Sub-Saharan Africa; prediction; SHAP (SHapley additive exPlanations); acute respiratory infection; machine learning; Sub-Saharan Africa; hyper parameter tuning
• ISSN: 2296-2360; 2296-2360
• DOI: 10.3389/fped.2024.1388820

The primary cause of death for children under the age of five is acute respiratory infections (ARI). Early predicting acute respiratory tract infections (ARI) and identifying their predictors using supervised machine learning algorithms is the most effective way to save the lives of millions of children. Hence, this study aimed to predict acute respiratory tract infections (ARI) and identify their determinants using the current state-of-the-art machine learning models. We used the most recent demographic and health survey (DHS) dataset from 36 Sub-Saharan African countries collected between 2005 and 2022. Python software was used for data processing and machine learning model building. We employed five machine learning algorithms, such as Random Forest, Decision Tree (DT), XGBoost, Logistic Regression (LR), and Naive Bayes, to analyze risk factors associated with ARI and predict ARI in children. We evaluated the predictive models' performance using performance assessment criteria such as accuracy, precision, recall, and the AUC curve. In this study, 75,827 children under five were used in the final analysis. Among the proposed machine learning models, random forest performed best overall in the proposed classifier, with an accuracy of 96.40%, precision of 87.9%, F-measure of 82.8%, ROC curve of 94%, and recall of 78%. Naïve Bayes accuracy has also achieved the least classification with accuracy (87.53%), precision (67%), F-score (48%), ROC curve (82%), and recall (53%). The most significant determinants of preventing acute respiratory tract infection among under five children were having been breastfed, having ever been vaccinated, having media exposure, having no diarrhea in the last two weeks, and giving birth in a health facility. These were associated positively with the outcome variable. According to this study, children who didn't take vaccinations had weakened immune systems and were highly affected by ARIs in Sub-Saharan Africa. The random forest machine learning model provides greater predictive power for estimating acute respiratory infections and identifying risk factors. This leads to a recommendation for policy direction to reduce infant mortality in Sub-Saharan Africa.