Unraveling birth weight determinants: Integrating machine learning, spatial analysis, and district-level mapping

Area of Science:

• Public Health
• Epidemiology
• Machine Learning in Health

Background:

• Low birth weight (LBW) persists as a significant public health issue in Bangladesh, linked to infant mortality and long-term health problems.
• Despite a declining trend, LBW necessitates continued monitoring and analysis of its determinants.

Purpose of the Study:

• To identify factors associated with birth weight in Bangladesh using nationally representative data.
• To map high-risk districts for LBW and inform targeted public health interventions.
• To explore interactions among determinants of birth weight using machine learning.

Main Methods:

• Analysis of nationally representative Multiple Indicator Cluster Surveys (MICS) data from 2012-2013 and 2019.
• Application of regression tree, a machine learning algorithm, to identify key interactions.
• Utilized fixed effect, mixed effect, and spatial dependence models, considering clustering and spatial correlation.

Main Results:

• Average birth weight increased slightly from 2.93 kg to 2.96 kg between survey periods.
• Maternal age, household head's education, and antenatal care were significant factors influencing birth weight.
• District-specific maps identified the southwestern and selected northern regions as persistent high-risk areas for LBW.

Conclusions:

• Machine learning and spatial modeling enhance understanding of LBW determinants and distribution.
• Findings support the need for district-level policy formulation and targeted interventions for vulnerable populations.
• Continued monitoring and data-driven strategies are essential to address LBW in Bangladesh.