Predictors of Contemporary under-5 Child Mortality in Low- and Middle-Income Countries: A Machine Learning Approach

Area of Science:

• Public Health
• Data Science
• Pediatrics

Background:

• Child mortality (CM) remains a significant global health challenge, particularly in low- and middle-income countries (LMIC).
• Existing data holds potential for identifying underlying causes and guiding effective interventions.

Purpose of the Study:

• To identify and rank distal causes of child mortality (CM) using a machine-learning approach.
• To pinpoint the top 10 most potent distal causes of CM.
• To identify household conditions associated with reduced CM and inform policy.

Main Methods:

• Utilized data from the Multiple Indicators Cluster Survey (MICS) encompassing 27 LMIC (N = 275,160).
• Applied a machine-learning model to rank 37 distal causes of CM based on predictive power.
• Retrospectively validated findings by comparing CM variations with improved household conditions across MICS survey periods.

Main Results:

• Identified and ranked 37 distal causes of child mortality (CM).
• Determined the top 10 most predictive distal causes of CM.
• Linked improved household conditions to reduced CM, validating the model's insights.
• Highlighted lesser-known, preventable distal causes contributing to CM.

Conclusions:

• Machine learning offers a powerful tool for extracting operational insights from large datasets to combat child mortality (CM).
• The identified distal causes are amenable to social, educational, and physical interventions.
• Findings can guide policymakers and intervention programs to effectively reduce child mortality in LMIC.