UNDERSTANDING INTERGENERATIONAL MALNUTRITION IN RURAL BANGLADESH: A SYSTEMS APPROACH
Johns Hopkins University
Background: Determinants of malnutrition across the lifecycle are complex. Beyond host characteristics and behavioral choices, the immediate environment and broader societal context remains a challenge to measure. The JiVitA-1 trial in rural northwest Bangladesh reveals intergenerational spatial patterns of malnutrition, based on mid-upper arm circumference (MUAC), that converge with age from early life onward, suggesting possible contextual influences. This dissertation explores these spatial patterns, and investigates interactions among biological, socioeconomic and contextual factors on trajectories of nutritional status from birth to adulthood in this typical rural South Asian population. Methods: Spatial analysis with multiple-linear regression was used to assess independent and synergistic effects of individual, household and community factors, added sequentially, on MUAC among mother-infant dyads. Explanatory improvements and spatial correlation accounted for by characteristics at each level was considered. Results: Multilevel regression models explained 13.2%, 14.5%, and 11.7% of variabilty in MUAC of infants at birth and six months and expectant mothers, respectively. Most variability in MUAC was explained by individual and household-level variables. Community influences accounted for 0.3%, 0.4%, and 0.7% of the variability in MUAC at the two infant ages and among women, respecitvely. Infant growth between birth and six months was guided by initial size (r² = 0.33), and modified further by household and community socioeconomic-status (SES) and maternal nutritional status. The full multilevel model accounted for 40% of the variance in growth rate, 1% of which was attributed to community context. Including individual- and household-level variables provided modest reductions in residual spatial autocorrelation of MUAC, compared to reductions associated with contextual factors. Promising contextual variables included neighborhood economic structure, maternal education, elevation, population density, and travel-time to markets. Many contextual variables were correlated, indicating that those living in wealthier neighborhoods enjoy healthier environments, which may reinforce benefits of greater household and neighborhood SES. Conclusions: A systems science approach revealed multi-level and age-specific influences on infant and maternal nutritional status in this rural Bangladesh setting. Contextual variables explained differences in MUAC and reduced residual spatial autocorrelation more than individual and household variables combined. Correlations between contextual variables also indicated economically-driven population sorting in this typical rural setting.
Undernutrition, Mid-upper arm circumference, Systems approach, Geospatial Information Systems, Bangladesh, Contextual effects