Predicting Childhood Anaemia in Nigeria: A Machine Learning Approach to Uncover Key Risk Factors.

Area of Science:

• Public Health
• Machine Learning
• Pediatrics

Background:

• Childhood anaemia presents a significant public health issue in Nigeria, disproportionately affecting children under five.
• High prevalence rates necessitate innovative approaches for identification and intervention.

Purpose of the Study:

• To identify key determinants of childhood anaemia in Nigeria.
• To develop and validate a predictive model for childhood anaemia using machine learning techniques.

Main Methods:

• Analysis of 13,136 children (6-59 months) from the 2018 National Demographic and Health Survey (NDHS).
• Evaluation of 16 machine learning algorithms, with the Extra Trees (ET) classifier selected for its superior predictive performance.
• Identification of top 10 anaemia predictors and assessment of model fairness across demographic subgroups.

Main Results:

• The ET classifier achieved an Area Under the Curve (AUC) of 0.8319, accuracy of 0.7565, and recall of 0.7565.
• Top predictors included household size, birth order, child age, media access, maternal health-seeking behavior, child gender, water proximity, financial difficulties, land surface temperature, and population density.
• Model performance varied across regions, wealth quintiles, and ethnic groups, with lower AUCs in the North-East, poorest quintile, and among Hausa/Fulani individuals.

Conclusions:

• Machine learning effectively predicts childhood anaemia in Nigeria and pinpoints critical risk factors.
• Findings support the development of targeted interventions to mitigate childhood anaemia prevalence.
• Future research should explore AI-driven interventions for anaemia reduction.