# Modeling diarrhea in children under five in Somaliland: A machine learning analysis using SLDHS 2020 data

## TL;DR

This study uses machine learning to analyze factors contributing to childhood diarrhea in Somaliland, highlighting the need for region-specific public health strategies.

## Contribution

The study applies multiple machine learning models to identify key determinants of childhood diarrhea in Somaliland using SLDHS 2020 data.

## Key findings

- The overall prevalence of diarrhea in children under five was 7.2% with significant regional variation.
- Nomadic households had a higher incidence of diarrhea compared to rural and urban households.
- Machine learning models showed high accuracy, but sensitivity for predicting diarrhea cases was low.

## Abstract

Diarrhea remains a leading cause of morbidity and mortality among children under five years of age, particularly in low- and middle-income countries. This study investigated the prevalence and determinants of diarrhea in Somaliland using nationally representative data from the 2020 Somaliland Health and Demographic Survey (SLDHS) 2020.

We employed a cross-sectional study design and analyzed data from 1,112 women (aged 15–49) and their children under the age of five from six geographic regions in Somaliland. Variables were selected based on data availability in SLDHS 2020, including socioeconomic, demographic, and environmental factors. We employed descriptive statistics and binary logistic regression to identify significant associations of the variables with childhood diarrhea. Additionally, supervised machine-learning models (Logistic Regression, Probit Regression, Random Forest, Decision Tree, and SVM) were used to identify key determinants of diarrhea.

The overall prevalence of diarrhea was 7.2%, with significant regional variation (Togdheer: 12.5%; Awdal: 4.24%). Nomadic households had a significantly higher incidence (8.62%) than rural (2.41%) and urban (5.16%) households. Logistic regression analysis highlighted region, household wealth index, and sanitation access as significant predictors. Interestingly, maternal educational level was not significantly associated with the prevalence of diarrhea. The Decision Tree model achieved the highest accuracy (92.3%) and sensitivity (33.3%), while Logistic Regression had specificity >97%.

This study underscores the importance of region-specific public health strategies focused on improving access to water and sanitation, especially in nomadic and low-income populations. Despite the high overall accuracy, the machine-learning models indicated that the predictive accuracy for positive diarrhea cases could be further refined. Efforts to alleviate diarrhea among young children in Somaliland should prioritize the enhancement of infrastructure and sanitation resources in underserved communities.