Spatiotemporal heterogeneity of malnutrition indicators in children under 5 years of age in Bangladesh, 1999–2011

Abstract

We identified divisions where malnutrition indicators (stunting, underweight and wasting) remain highly clustered and other divisions where they are more widely spread in Bangladesh. This has important implications on how interventions for malnutrition need to be delivered (geographically targeted interventions v. random interventions) within each division of the country.

A Review of Spatial Analysis Application in Childhood Malnutrition Studies

“…Most of the selected articles studied more than one type of childhood malnutrition. For subject of the study, 23 articles used children aged under 5 years old as their subject ( n = 23, 85.2%) ( 15 – 18 , 20 , 21 , 23 – 25 , 27 – 39 , 41 ), one study used children aged under 2 years old ( n = 1, 3.7%) ( 19 ), one study used children aged under 3 years old (0–35 months) ( n = 1, 3.7%) ( 26 ), one study used children aged 3 years old–5 years old ( n = 1, 3.7%) ( 40 ) and one study did not mention the age of children used in the study ( n = 1, 3.7%) ( 22 ).…”

“…For information regarding the demographic, socioeconomic status, health and nutrition of the subjects or respondents, most of the articles used data from Demographic and Health Surveys (DHS) ( n = 15, 55.6%) ( 16 , 19 , 21 , 23 , 24 , 26 , 30 , 31 , 33 , 35 , 36 , 38 , 39 , 40 , 41 ), four studies used data from National Family Health Survey (NFHS) ( n = 4, 14.8%) ( 15 , 17 , 22 , 37 ), two studies obtained data from Food Security and Nutrition Unit (FSNAU) ( n = 2, 7.4%) ( 29 , 32 ), two studied from National Nutritional Survey (NNS) ( n = 2, 7.4%) ( 20 , 34 ), two studies from Nutritional Status Information System (NSIS) ( n = 2, 7.4%) ( 25 , 28 ), one study from Feed the Future Survey (FTF) ( n = 1, 3.7%) ( 18 ), one study from Multiple Indicator Cluster Surveys (MICS) ( n = 1, 3.7%) ( 24 ), one study from Core Welfare Indicators Questionnaire (CWIQ) ( n = 1, 3.7%) ( 24 ) and one study conducted their own survey to obtain the data ( n = 1, 3.7%) ( 27 ).…”

“…slope and elevation) was used by two studies ( n = 2, 7.4%) ( 19 , 35 ). For software packages for spatial statistics, R packages were commonly used ( n = 9, 33.3%) ( 15 , 16 , 20 , 21 , 23 , 27 , 29 , 32 , 36 ), with four articles using Integrated Laplace Approximation ( R -INLA) packages ( 16 , 29 , 32 , 36 ). The second most commonly used computer software package was ArcGIS ( n = 8, 29.6%) ( 17 , 19 , 22 , 28 , 30 , 31 , 39 , 40 ), followed by GeoDa ( n = 4, 14.8%), WinBUGS ( n = 3, 11.1%), SaTScan ( n = 2, 7.4%), BayesX ( n = 2, 7.4%) and QGIS ( n = 1, 3.7%).…”

“…Local indicator of spatial association (LISA) were applied by five studies ( n = 5, 18.5%) ( 15 , 17 , 22 , 25 , 38 ). Local Getis-Ord G index (LGi) was also used in three studies ( n = 3, 11.1%) ( 31 , 39 , 40 ), followed by Kulldorf’s spatial scan analysis ( n = 3, 11.1%) ( 26 , 30 , 35 ) and semivariogram was used in two studies ( n = 2, 7.4%) ( 21 , 27 ).…”

“…Applying this analysis to childhood malnutrition studies can facilitate improved regional targeting by stratifying areas of higher risks and identifying physical locations that are important for the design of nutrition intervention programmes in target regions ( 39 ). Targeted regional needs for interventions can have the utmost benefit in meeting the needs of the population relative to random intervention ( 21 ). The use of this method also has limitations, particularly in the selection of significant clusters.…”