Geospatial Analysis of Malnutrition among Under-Five Children: A Scoping Review

Abstract Nutritional status is one of the important factors that indicate children's proper development and growth. The geospatial analytic approach is useful in describing and analyzing the characteristics, depth, and coverage of the malnutrition burden among under-five children. This current scoping review was performed to systematically map the spatial analytical techniques and approaches applied in nutrition among under-five children. An organized online database search was conducted to identify articles published between 1995 and 2021 on under-five nutrition and spatial statistic in PubMed, Science Direct, Scopus, and Web of Science. A total of 80 distinct articles were identified, of which 34 articles were used for the final review. A spatial statistical correlation was mainly used ( n  = 15), followed by Bayesian spatial modelling ( n  = 7), Global Moran's technique ( n  = 9), and Getis-Ord ( n  = 3). Nine studies in India concerning spatial analysis and undernutrition were conducted based on a national-level demographic health survey. There is a need for future spatial studies related to nutrition and under-five children at the sub-national level in India.


Introduction
The spatial analysis deals with application of various statistical methods for analysis of information related to areaspecific, topographic-specific, and temporal and spatial correlation association between the available facts.The usage of spatial analysis in the field of public health provides enough base to plan for public health intervention based on geographic information and disease.As per United Nations member states, the agenda for Sustainable Devel-opment Goals (SDGs) 2030 is focusing mainly on ending poverty, eliminating malnutrition, and improving health and education. 1Although many studies focus on nationallevel data for interpretation, to avoid diverseness at a subnational level, the SDG targets the progress of targets at the sub-national level.Concentrating effectiveness and monitoring of data at the sub-national level will give more accurate information emphasizing on the usage of spatial tools to provide information at district and sub-district levels. 2 The spatial analysis tool Global Moran's I measures and evaluates the spatial autocorrelation based on both feature attributes, to find out if the patterns are clustered, dispersed, or random. 3A LISA usually has two important features, the first one gives a statistic for each location with an assessment of significance and then initiates a corresponding relationship between the sum of the local and a global statistic. 4lobally, children under five (U5) were wasted to the tune of 45.4 million, and among the severely wasted were 13.6 million.The prevalence of acute malnutrition in South Asia is quite high (14.7%).As per the United Nations International Children's Emergency Fund (UNICEF) report 2021, nearly one-third of the children are still stunted and at least 30% of children were still affected by stunting in 2020.In 30 countries, at least 1 in every 10 children U5 is overweight, with the highest numbers in the Middle East and North Africa. 5The update from the joint team of UNICEF, World Health Organization (WHO), and the World Bank of annual estimates of malnutrition among children U5 is expected to report further increase in all forms of malnutrition, especially in the vulnerable population-limited availability and affordability of nutritious food, disruptions in essential basic supply of resources due to worsening household income, and reduced physical activity. 6In different parts of India, the study shows that underweight among U5 children ranged from 39 to 75%, wasting from 10.6 to 42.3%, and 15.4 to 74% of stunting. 7In the field of health and epidemiological sciences, the spatial statistical method of approach has become a predominant statistical tool to study the geographical distribution with respect to health-related data and outcomes. 8,9n India, stunting is higher among children in rural areas (41% vs. urban, 31%).The prevalence of stunting is highest in Bihar (48%), Uttar Pradesh (46%), Jharkhand (45%), and Meghalaya (44%), and with the lowest in Kerala and Goa (20% each) among children below 5 years.Jharkhand has the highest levels of underweight (48%) and wasting (29%). 10A valuable methodology can be utilized to understand the quality of literature through scoping review. 8The application of spatial statistics in the field of nutrition has grown in recent times.The nationallevel data have been used for spatial analysis and the conclusion is also based on the results of these data, hence the subnational diverseness is not addressed.
A scoping review is a useful approach in evidence synthesis and identifying the knowledge gaps and clarify the concepts of the particular study.This scoping review is aimed to identify and describe the different methods of spatial analysis and its application in the field of nutrition.

Inclusion Criteria
All articles published in English during the period of 1995 to 2021 which used spatial statistic methods such as Moran's indexing technique, LISA analysis, and geographic information system (GIS) were included to analyze nutritional status among children U5.

Exclusion Criteria
The articles that have not used the spatial statistical method and those published outside the 1995 to 2021 period including systematic review and meta-analysis articles were excluded from the review.

Search Method
This review was conducted based on the guidelines of the Preferred Reporting Items for Systematic Review and Metaanalyses (PRISMA) extension for scoping review (PRISMA-ScR). 11The articles published within 1995 to 2021 using the spatial statistical method were organized for literature search, which was done through PubMed/Medline, Scopus, Science Direct electronic database, and Web of Science.The search strategy for the articles was using the following keywords: spatial statistic, spatial modeling, Moron's indexing technique, GIS, LISA analysis, nutritional status, children under five years.Boolean operators "AND/OR" with keyword combinations such as "Spatial Statistics" OR "GIS" OR "Moran Index" AND "Nutrition among under five children" were used for literature search.

Study Selection
All possible studies retrieved were first imported to Zotero and duplicates were removed.Using pre-established inclusion criteria, we screened the titles along with their abstracts.After meeting the inclusion criteria, the full-text article assessment was done, followed by extraction of information fulfilling the objectives of the scoping review from the articles.

Data Extraction
A master template was prepared using Microsoft Excel, which involved information on spatial statistical technique, spatial software used, study focus, and main findings.The categories were made based on spatial application techniques.The summary of study findings was determined using counts and proportions.

Study Features
Overall, 80 distinct articles were identified.Twenty articles were duplicates, hence excluded.Further, based on abstract and titles, six articles were excluded for not fulfilling the eligibility criteria.For the final review, the 34 articles were identified (►Fig. 1).Data extracted from the publications on nutritional status among U5 children have used several spatial methods.Moran's I measures spatial autocorrelation of dataset by correlation coefficient, Getis-Ord statistic gives information on how spatial autocorrelation varies over study location for each area, and local spatial autocorrelation analysis (hotspot analysis) is used to identify the local clusters of high, low, or high-low clusters.These methods are used with the GIS network, which observes the nutritional status U5 by location to estimate values for unobserved locations.Generally, stunting, wasting, underweight, overweight, and obesity are considered nutritional-related problems among children.To detect the status among children, WHO growth standards of measurement were used by calculating the Z-score value of "weight for age" (WAZ), "height for age" (HAZ), and "weight for height" (WHZ). 12In this scoping review, articles fulfilling the eligibility criteria for nutritional status U5 using various methods of spatial analysis at different locations were selected.The stunting value (HAZ) of À2 standard deviation (SD), the wasting value (WHZ) is À2 SD, and the underweight value (WAZ) is À2 SD is considered moderately malnourished.To assess the children's anthropometric failure, the Global Moran's I was used to find out whether it is clustered, dispersed, or distributed randomly in the study.

Focus of the Study
Among 34 articles screened, 12 studies have focused on understanding the spatial patterns of stunting, wasting, and underweight (n = 12).Then, 9 studies (n = 9) focused on the spatial distribution of risk factors and the determinants of malnutrition.Five studies focused on under-nutrition and over-nutrition among children U5 and four studies each focused on the mothers' nutritional dependency on children nutritional status and the intake of diet content and nutritional status (►Table 2).
The articles (n = 34) that were included for full-text assessment after fulfilling the eligibility criteria are from Asia and Africa continents (47.1% and 35.3% respectively), followed by South America (8.8%) and North America (2.9%).Two articles (5.9%) were conducted multi-centric (►Table 3).
Table 2 Main focus of the study.

Number of studies Percentage
Patterns of stunting and wasting 16,19,21,26,28-30,34,44,46-   ►Table 4 depicts the aims of the study and the findings of each article.Most of the studies have focused on spatial distribution of risk factors and their association with malnutrition.Furthermore, most of them focused on spatial pat-terns of stunting, wasting, and underweight.Only few studies explained the dependency of mother's nutritional status on child's nutrition and intake of diet content with its effect on the status of nutrition among children U5.The

Hierarchical Bayesian spatiotemporal modelling
There are trends of variation in all blocks except in three indicators of facility delivery, public facility delivery, and age-appropriate initiation of complementary feeding.

3
A h e t oe ta l 29 Ghana To explore and forecast spatiotemporal patterns in childhood of chronic malnutrition under 5.

Spatiotemporal modelling software
There is substantial spatiotemporal variation in the prevalence of stunting.

Akseer et al 43 Afghanistan
To comprehensively assess geographical disparities and nutritional status among women and children.

Winbugs and Rsoftware
The result found that children in Afghanistan were on average shorter and underweight and slightly more emaciated.There is significant dependence of seasonal and spatial factors on the patterns of growth of children.25  Seboka et al 40 Ethiopia Spatial dependence between anthropometric failure of children under 5 years.

Getis-Ord spatial statistical tool
The spatial analysis revealed that the northern part of the country is at higher risk of anthropometric failure.

India
The spatial analysis to identify key determinants of malnutrition.

Discussion
This review focuses on the application of spatial analysis techniques on nutrition among U5 children across the globe.Spatial autocorrelation and cluster detection using GIS software were the predominant methods applied.Articles in this review used demographic health survey data conducted nationwide, and in India, they used NFHS (National Family Health Survey) data.These data are analyzed using various versions of the Arc/GIS software.Only 2.9% of review articles focus specifically on the intake of nutritional content in their diet, like iron, vitamins, etc.The review found that there are limited articles (5.8%) describing the association of nutritional status among children and maternal health, food security, and household economy, which is significant toward the child's nutritional status.Furthermore, there is a lack of evidence of programs on interventional evaluation on nutritional status among children U5 in different places in the world.Many articles apply spatial analysis of Moran's I correlation coefficient, Getis-Ord statistic, LIZA, and Z-score.All spatial autocorrelations are measured by Moran's index, its values lie within þ1 and À1, wherein the value of À1 is perfect clustering of dissimilar values and the value of þ1 shows perfect clustering of similar values. 13LISA was used to identify local clustering under four categories; high-high and low-low autocorrelation, which comes under positive spatial autocorrelation, then high-low and low-high autocorrelation as negative autocorrelation. 14The study in India focuses on identifying and assessing various key factors and determinants among children below 5 of their anthropometric failure and their spatial dependencies across India.Thereby, to plan a specific program that focuses on the influencing determinant to intervene malnutrition in India, there are limited articles focusing on the evaluation of nutrition programs and redesigning the interventions after finding out the key influencing factors.Overall, articles related to spatial analysis need bio-statistical expertise and innovative research with skills in managing, analyzing, and using GIS to predict spatial patterns from observed patterns to notify implementers and policymakers as much concerned. 15

Limitations
Even after following the guidelines of PRISMA-ScR, we might have overlooked those articles focusing on the application of spatial analysis of nutritional status among children U5 in India as well in other countries.This review had excluded studies that focus on nutritional status other than children, like maternal health, household income, and food security.Most studies have different forms of result presentation and different utilizations of spatial techniques.For example, some have only used Z-score and Moran's I, but others made use of all the techniques, such as Moran's I, LISA, Getis-Ord, and Bayesian model, to come up with an outcome.A few papers were excluded as they were published in a language other than English.

Strength
This review provides the best available knowledge on nutritional status among children U5, with the application of different techniques like spatial analysis to analyze health demographic data in India and other countries.It also focuses on the use of spatial analysis techniques predominated by the spatial autocorrelation method.The review included articles covering almost 26 years (1995-2021) where spatial analytical methods were used to determine the U5 nutrition.

Conclusion
In this descriptive and analytical analysis, there is a need for a standard statistical package for estimating the population's parameters to provide results approximately equal to those obtained from the software.Most of the studies have used Rsoftware, ArcGIS version, and Kulldorff's SaTScan version, etc., in which the detailed information about the characteristics of the software is unknown, leading to a potential pitfall.
Advanced analytical skills are needed to manage and analyze available data to notify policymaker and implementers.Most of the studies have focused on the anthropometric failure of children U5, and only a few studies were meant to assess and identify key indicators of nutritional diet content, maternal health, household income, and food security toward malnutrition among preschool-going children.The spatial analysis is according to a grouping unit that may unintentionally misrepresent or overlooked actual risk variation.To avoid the misrepresentation of the sample and distortion of spatial inference, all studies have utilized demographic health survey data at the national level for the correct interpretation of spatial patterns toward nutritional status among children U5.

Out of 34
articles, 15 articles used spatial statistical clustering/autocorrelation for analysis, 9 articles used Moran's index, LISA, and global Moran's index technique, and 7 implemented a Bayesian spatial statistical model.The details of spatial techniques employed among U5 children are shown in ►Table 1.

Fig. 1
Fig. 1 Flow diagram of the article selection process using PRISMA guidelines.

Table 1
Types of spatial analysis technique adopted for data extraction

Table 3
Continents covered in the studies

Table 4
Study aims and outcomes