Nutrition gap in children, urban-rural: the key education and food. Colombia, 2015

ABSTRACT OBJECTIVE: To analyze the nutritional situation of children under five years old from both urban and rural areas of Colombia. METHOD: Analytical study, based on cross-sectional data, collected from ENSIN-2015. The sample consisted of 12,256 children aged between 0 and 4 years old. We calculated the prevalence ratios (PR) with their respective 95% confidence interval (95%CI). PR were assessed by binomial regression models with malnutrition or overweight as the dependent variable and geographic area as the explanatory variable. We used context variables to adjust the estimated PR and control the confounder within. RESULTS: Acute malnutrition (weight-for-height) had a prevalence of 1.6%, while overweight had a 5.6% rate. No differences per geographic zone in the weight-for-height indicator were found. Stunted growth – chronic malnutrition – was higher in the rural area (PR = 1.2; 95%CI 1–1.53; p = 0.050). Prevalences adjusted by variables related to structural, social and economic developement showed that both the household chief's educational level and the food insecurity of the area account for malnutrition. CONCLUSION: The height-for-age indicator works better to establish development level. Measures against coverage, relevance and quality of education and access to food can harm the nutritional status of the children.


INTRODUCTION
According to the Food and Agriculture Organization of the United Nations (FAO), the prevalence of undernutrition increased from 10.6% in 2015 to 10.8% in 2018 worldwide. In South America, it increased from 4.9% in 2015 to 5.5% in 2018; and stunted growth slowly decreases while overweight, obesity and malnutrition quickly increase 1 . In upper-middle-income countries, such as Colombia, stunted growth in under-five was at 6.3%, wasting at 0.6% and overweight at 7.4% in 2018 1 . Colombia faces changes in its economic, demographic and social structures since the 1960s 2 . Consequently, the transition is visible and dynamic regarding variables related to the health 3 , nutritional 4 , and food 5 status. Despite income increments and poverty reduction in children under five years old (64% in 1999 3  The development of Colombia has been mediated by armed conflict and forced displacement for the past 60 years [8][9][10] . Recently, both advance or retard in the structural and economic development are credited to market economy and trade agreements 11 . According to the report "The Changing Wealth of Nations 2018," issued by the World Bank 12 , Colombia is the fifth country in the region with the highest wealth per capita and holds one of the most unequal wealth distribution rates in the region: the GINI index was at 0.53 in 2018 3 . The poverty decreased from 37.2% in 2010 to 27.8% in 2015 13 . The nutritional situation assessed by the Encuestas Nacionales de Situación Nutricional de Colombia (ENSIN -Colombia National Survey of the Nutritional Situation) is result of an environment that comprehends the poverty-obesity paradox 14 , the return of communicable diseases and the rise of non-communicable chronic diseases 15 , the increase of income and social tension, the persistence of armed conflict, the enhancement of inequality and differences between the rural and urban development 3,6,16 . Thirty two per cent of the Colombian population lives in rural areas where rural social dynamics prevail 9 .
In low-and middle-income countries, despite the gaps between urban and rural, the overweight disparities are reduced. The speed with which overweight occurs is faster in rural areas, while malnutrition in these areas is more relevant than underweight 1 . According to the United Nations (UN), the road map to end the obesity epidemic is multilevel, multidimensional and multisectoral, and it considers the transformation of the food systems throughout education and availability of nutritious and healthy food, both accessible and affordable 17 . Grounded on it, the UN proclamed the Decade of Action on Nutrition in 2016 17 .
This study analyzes the nutritional situation of children under five years old from both urban and rural areas of Colombia.

METHODS
Analytical study, carried out in Colombia, based on cross-sectional data, collected in the 2015 -2016 period by ENSIN-2015.
ENSIN were designed to represent 99% of the population, through stratified multistage sampling. The ENSIN-2015 comprehended 44,202 households, representing 4,739 groups from 177 strata. Data from ENSIN-2015 are anonymized, public and can be accessed through a well-founded request to the Colombian Ministry of Health.
Trained personnel interviewed the household chief regarding sociodemographic, food security and household economic status. The anthropometric measurements were taken by nutritionists and qualified interviewers, using standard techniques and calibrated equipment. Children under 24 months old had their height taken in supine position 18,19 . A portable stadiometer was used (Shorr Board Producctions LCC, Olney, MD, USA), and value was rounded to the closest milimeter. Weight was measured by a SECA scale (model no. 874) and value was rounded to 100g divisions. Age was based on date of birth. Age, gender, height and weight data were analyzed in Stata Release 14.1 20 , according to the child growth standards, interpreted by a Z-score classification system, issued by the World Health Organization (WHO) 19 . All children scoring Z<-2 in the indicator weight-for-height fell under the wasting (acute malnutrition) classification. In the same indicator, those scoring Z > 2 were classified as overweight. Children scoring Z<-2 in the height-weight indicator were classified as stunted growth (chronic malnutrition).
The geographic zones were classified according to the population density. The ENSIN-2015 divided the populated centers into four categories, based on their density: a) A Area, 0 -100,000 inhabitants, b) B Area, 100,000 -1 million inhabitants, c) C Area, over 1 million inhabitants. These three areas were classified as urban area, and d) "others", such as the rural area. The "others" category encompassed outskirts close to small cities, rural areas far from small cities and population either spread out or too far from rural areas.
The variables considered were the nutritional situation based on the Z-score, according to the weight-for-height, height-for-age and geographic zone indicators. The covariables: gender, age, household chief's educational level, health insurance, household food security, wealth index, ethnicity and geographic region where individuals live were also considered. The household chief's educational level was based on the formal schooling years attended. The type of health insurance fell under three categories: no health insurance, subsidized coverage, and contributory regime 21 . The food security status was assessed according to the Escala Latinoamericana y Caribeña de Seguridad Alimentaria (ELCSA -Latin American and Caribbean Food Security Scale) 22 . The wealth was established according to index designed for international survey on demographics and health 23 . The geographic region is a set of several geodemographic units and, overall, the most relevant aspects of the food culture are represented within them 24 . The occurrence of acute diarrhea and acute respiratory infection two weeks before the survey was self-reported by the household chiefs or their spouses.
Data were analyzed in Stata Release 14.1 20 . The sampling weights of ENSIN-2015 design were incorporated in Stata's complex analysis routines 20 . The analysis was first conducted to identify the demographic structure, through gender and socioeconomic status, within each geographic zone. This established the level of structural and economic development, which determines to great extent the nutritional and health situation of under-fives. The second phase established the raw prevalence, for both wasting and overweight, considering the weight-for-height and height-for-age indicators and each one of the covariables. The third phase of the analysis established Prevalence Ratios (PR), both raw and adjusted, by using the prevalences divided by zone and category of each covariable. For such, the RP and their respective intervals were estimated at 95%CI. The RP were estimated using a binomial regression model having either wasting or overweight (yes/no) as the dependent variable and the geographic zone as the explanatory variable.
Informed consent form was granted during field work. The analyses were compliant with the Declaration of Helsinki 25 . This is a "non-risk" research according to the Resolution no. 8430 from 1993, issued by the Health Ministry of Colombia 26 . Authorization from the Research Ethics Committee of the Universidad Industrial de Santander was not required, since this is a secundary research of population studies, employing anonymized data.
In the households where the chief had less than high school education, the children living in the rural area presented a smaller rate of poor weight-for-height growth than those living in the urban area, p = 0.013. The same was true for children with no health insurance or under the contributive regime and for those living in places with INSA mild and moderate ( Table 2). The children living in households having moderate INSA presented more overweight than those living in the urban area (PR = 3.8; 95%CI 1.7-8.8; p = 0.002). The same was true for those children falling in the lowest tercile of the economic status and those living in the Atlantic region ( Table 3).
The stunted growth was greater in the rural area (PR = 1.24; 95%CI 1-1.53; p = 0.050); and the same was found by comparing the prevalence of poor weight-for-height growth in the rural area against some categories of covariables; while those under one year old, overweight children whose parents had a low educational level, those living in households with moderate INSA and in the central region presented greater stunted growth rates if living in the rural area (Table 4).

DISCUSSION
Based on ENSIN-2015, the prevalence of wasting and overweight was estimated in under-fives with the weight-for-height indicator and stunted growth with the height-for-age indicator. The prevalence of stunted growth increased from 13.2% in 2010 7 to 10.8% in 2015. However, both wasting and overweight increased, with acute malnutrition going from 0.9% in 2010 7 to 1.6% in 2015 and overweight from 5.2% in 2010 7 to 5.6% in 2015. In addition, geographic zone had not presented statistical significance to the prevalences of wasting and overweight in the weight-for-height indicador, but rather had presented it to the height-for-age indicator.
The weight-for-height indicator in Colombia is smaller than the average reported by FAO for 2018 in Asian countries, which is 22.7%, or in African countries, which is 30%, but higher than average in high-income countries, which is 3% higher, and in Latin American, which is of 9%. The overweight (weight-for-heigh) rate in Colombia is higher than the Asian countries average of 2.2%, or the 4.9% in African coutries, and smaller than the Latin America average of 7.5%, and than high-income countries, such as the United States, which is 7.2%.
According to the World Bank and the United Nations Children's Fund (UNICEF), the prevalence of stunted growth is decreasing in all upper-middle-income countries of the region, such as Colombia 27  Additionally, the overweight in the same indicator is slightly gerater than reported by this study than rates reported in high-income countries: 9.9% in Argentina (2005) The classification of urban and rural areas in the ENSIN-2015 is based on the concentration of population unaware of the territorial concept and of the seven dimensions constitutive of it 9,28 . Yet, the rural concept, as tradicionally known, encompasses the relationships of the individuals living there, which are predominantly towards farming 2,28 . Poverty and other variables -here the nutritional situation -reaching similar figures in rural and urban areas are theoretically differents, given they resulted from structurally dissimilar processes in the geographic zone. For example, a) goods and services are more expensive in urban than rural area, b) urban survival requires goods and services not needed in the rural environment, c) money is more important in the urban than rural area, d) urban households are more vunerable than those of the rural areas; e) the position of the household chief in urban area is determined by the labor market, whereas the same does not occurr in rural area, f) development impacts urban areas, whereas has little or no impacto in rural areas and g) access to food in urban areas relies almost exclusively on money, whereas the same does not occurr in the rural areas. Besides, we established that the educational level of the household chief and of household's INSA, in addition to basic infrastructure variables, such as water supply and sewage networks availability, are different in each geographic zone (Table 1).
Regarding to the wasting in the weight-for-heigh indicator, the educational level of the household chief -less than high school -, type of health insurance -no health insurance or contributive regime -, and INSA -mild or moderate, were associated as expected. The PR in these variables show that acute malnutrition is smaller in the rural area. When prevalences are adjusted, despite reduced individual capacities, such as education 29 or institutionality, such as health insurance, a smaller wasting rate in the weight-for-heigh indicator is associated with mild and moderate INSA in rural areas, rather than the urban ones (Table 2). This finding diverges from what we already know about the Colombian set and invites us to rethink the inequalities by geographic zone. The current dynamics in study variables become potencial confounders of the wasting prevalence.
Those classified as poorer by the economic index, the households with mild INSA and the Atlântico region, of which poor departments such as La Guajira are part, presented overweight rates higher in rural than in urban areas ( Table 3). The overweight associated with greater poverty was observed when the nutrition transition was declared in Latin America 14 , with the ENSIN-2010 4 data in Colombia and in Colombian children 30 .
The poor growth in the height-for-weight indicator corroborates that minor determinants of social development (the educational level of the household chief), which lead to reduced food access (INSA), combined to an enviroment of institutional inequalities (reduced health coverage), are shown at an early age (in children under one year old) in the form of stunted growth. Biologically, the increased obesity vicious cycle in rural children, along with stunted growth, illustrate what was already described in the 1960s 31 .
The Colombian rural area presents chronics inequalities, which strongly limitate the human, social, structural and economic development; their effects being visibles on both individual and collective levels. This chronic state of urban underdevelopment is also evidenced by the nutritional situation, and even though poverty was reduced, the same is not true for inequality. The develpment and underdevelopment ways change and yet, the nutritional situation remains sensitive to these. The greater contribution of this study is not whether there are inequalities between rural and urban areas, as for over 70 years these are indeniable in Colombia. The evidences are shown in all countries in the region, including Brazil 32 and other low-income countries 33 . Its greatest contribution is to establish a set of variables expressing the edges of development in Colombia, factors that currently affect the nutrition situation. To increase individual capacities -through education as a mid and long-term strategy -and to guarantee food access -using short, mid and long-term measures -are the key variables to positively impact the nutrition situation of children under five years and general population. This corroborates both the intent of the Colombian state 2,3,28 and the UN 17 .
The analyzed data were extracted from a national survey and through multivariate analysis, and the potential confounder effect of variables traditionally associated with the nutrition situation of under-five was cleared. The methods used to determine both the nutrition status and geographic zones classifications are universally accepted and remain sensitive to inequality under the current conditions and level of development reached in Colombia. The main limitation found is the cross-sectional data source. However, the coherence with the knowledge accumulated over the past 100 years assures external validity to the study.
The prevalence of nutrition situation in the indicators follows the regional and global trends. Geographic zones cause no differences in the prevalences of wasting and overweight in the weight-for-height indicator, only in the height-for-age indicator. Heightfor-age is the best indicator to establish the level of geographic development and the gaps between rural and urban areas. Some expressions of the socioeconomic development were related to the nutrition situtation, being the household chief educational level and household food insecurity the most relevant. The nutrition transition in Colombia is dynamic. These findings will enable the design and adjustment of related public policies to achieve cost-effectiveness.