|Year : 2022 | Volume
| Issue : 3 | Page : 264-268
Factors Affecting Stunting in Children under 5 Years of Age in Indonesia using Spatial Model
Zurnila Marli Kesuma1, Latifah Rahayu Siregar2, Edy Fradinata3, Aliya Fathinah4
1 Associate Professor, Department of Statistics, Faculty of Mathematics and Natural Sciences, Indonesia
2 Senior Lecturer, Department of Statistics, Faculty of Mathematics and Natural Sciences, Indonesia
3 Senior Lecturer, Department of Industrial Engineering, All Universitas Syiah Kuala Banda Aceh, Indonesia
4 Researcher, Department of Statistics, Faculty of Mathematics and Natural Sciences, Indonesia
|Date of Submission||26-Oct-2021|
|Date of Decision||20-Dec-2021|
|Date of Acceptance||29-Dec-2021|
|Date of Web Publication||22-Sep-2022|
Zurnila Marli Kesuma
Department of Statistics, Faculty of Mathematics and Natural Science, Universitas Syiah Kuala, Banda Aceh 23112
Source of Support: None, Conflict of Interest: None
| Abstract|| |
Background: Stunting in children under 5 years of age is a condition where they have a length or height that is less than −2 standard deviations of the growth standard of Indonesian children. Stunting is caused by chronic malnutrition in the first 1000 days of life. The spatial panel data method was developed to solve problems related to spatial objects that are measured periodically by involving elements of area and time. Objectives: The purpose of this study was to determine the best model and factors that influence stunting in children under 5 years of age in Indonesia using spatial panel data. Methods: The data used were from the website of the Central Statistics Agency and the publications of the Ministry of Health of the Republic of Indonesia in 2015–2019. Determination of the selected model is done by comparing the random effect spatial autoregressive model and spatial error model (SEM) random effect based on the value and Akaike information criterion (AIC). SEM random effect produces the largest value and the smallest AIC. Results: The selected spatial panel data model in determining the factors that influence stunting in children under 5 years of age in Indonesia is the SEM random effect based on the largest and AIC compared to other models. Conclusion: Based on the selected model, children under five with malnutrition and poor nutrition, receiving Vitamin A, and the average monthly per capita expenditure on food have a significant effect on the percentage of stunting in children under five in Indonesia.
Keywords: Spatial error model random effect, spatial panel data, stunting
|How to cite this article:|
Kesuma ZM, Siregar LR, Fradinata E, Fathinah A. Factors Affecting Stunting in Children under 5 Years of Age in Indonesia using Spatial Model. Indian J Public Health 2022;66:264-8
|How to cite this URL:|
Kesuma ZM, Siregar LR, Fradinata E, Fathinah A. Factors Affecting Stunting in Children under 5 Years of Age in Indonesia using Spatial Model. Indian J Public Health [serial online] 2022 [cited 2023 Feb 3];66:264-8. Available from: https://www.ijph.in/text.asp?2022/66/3/264/356596
| Introduction|| |
Indonesia is in the second position with the highest prevalence of stunting in the Southeast Asia region after Timor-Leste. The results of the 2018 Basic Health Research (Riskesdas) showed that the prevalence of stunting in Indonesia has decreased to 30.8% from the previous 37.2%. However, this result is still above the reasonable threshold set by the WHO, which is 20% for each country.
There are several factors that cause stunting, namely poor parenting behavior, including aspects of physical growth and development of independence that began in the womb. Lack of knowledge of mothers about health and nutrition before and during pregnancy as well as after the mother gives birth. Limited health services include antenatal care, delivery care, and postnatal care and quality early learning. In addition, there are still a lack of access for households/families to nutritious food and a lack of access to clean water and sanitation.
The problem of stunting not only has a major impact on health but also on the economy of a country. According to the National Development Planning Agency (Bappenas), untreated stunting causes losses of 2%–3% of gross domestic product (GDP). With a GDP in 2017 of Rupiah (IDR) 13,000 trillion, it is estimated that losses due to stunting will reach around IDR 300 trillion. This amount includes the cost of overcoming stunting and the loss of potential income due to the low productivity of children who grow up with stunting conditions.
Observation of stunting conditions in Indonesia is not enough if it is not only observed on units of observation at one time but also needs to be observed over several time periods. In statistics, the analysis that can be used is panel data analysis., Panel data are data from observations that are observed in several consecutive time periods. The spatial panel data model is applied to analyze cross-individual data that are observed at each observation location from time to time periodically. By applying spatial data panel analysis, a better model can be obtained because it includes the influence of data over time and the spatial relationship between observation locations.
Indonesia has a varied geographical location, both between provinces and between districts/cities, causing different levels of difficulty in accessing health facilities. People who live in underdeveloped areas of the border and outermost islands generally have difficulty accessing quality primary and secondary health services. This is due to the geographical conditions, transportation, and communication access experienced., The existing literature proves that there is a need to examine the spatial distribution of stunting in children under 5 years of age in Indonesia, especially at the provincial level. Mapping the diversity of stunting under-five cases can help in increasing the allocation of limited resources to existing provinces, as well as the need for high distribution of health care. Therefore, this study tries to examine the spatial grouping of under-five stunting and the factors that influence contextually including the condition of the poor, the amount of family income per month used for food, access of pregnant women to health facilities, and family sanitation facilities.
When the panel data regression model is applied to several countries, districts, or villages, it is feared that it will show spatial autocorrelation between regions. Spatial autocorrelation is a condition that shows the similarity of an object based on the region. Spatial autocorrelation also indicates that the value of a variable in a region is influenced by the variable in other regions that are close together. If the adjacent regions are similar, then the region can be said to have a positive autocorrelation. To find out whether there is a spatial autocorrelation (spatial effect) can use the Moran index test statistic.
Research using panel data regression has been conducted to analyze the factors that influence the percentage of stunted children under 5 years of age in Indonesia in 2015–2018 using the fixed effect model (FEM); it was found that per capita expenditure on food and under-five malnutrition had a significant effect on the percentage of stunted children under 5 years of age.
Another study in modeling spatial panel data was also conducted in Papua Province, Indonesia, to determine the factors that influence the health of children under 5 years of age. The results showed that fixed effect spatial autoregressive (SAR) panel model proved to be better than fixed effects spatial error panel model, (SEM), and General Spatial Model (GSM) judging from the coefficient of determination.
Based on this background, it is important to conduct this research to explore spatial heterogeneity and identify the factors that influence stunting in children under 5 years of age in Indonesia. With good results, there will be the best available evidence for decision-makers to mitigate this problem. Mapping the diversity of stunting cases in children under 5 years of age can also assist in the allocation of appropriate resources in each target area.
| Materials and Methods|| |
We use secondary data from the website of the Central Statistics Agency and publications from the Ministry of Health of the Republic of Indonesia (Indonesian Health Profile), which consists of 34 provinces as cross-section units and time series units from 2015 to 2019.,,,
The Indonesian Health Profile is a publication of health data and information compiled based on routine data and surveys from technical units within the Ministry of Health and other related institutions such as the Central Statistics Agency, Social Security Administration Agency, Ministry of Home Affairs, and the Population Agency and National Family Planning.
At the initial stage, descriptive data exploration was carried out to see an overview of stunting conditions in children under 5 years of age in Indonesia during the period 2015–2019.
Panel data analysis is done by estimating the parameters for the common effect model and FEM. Chow test is applied to select the selected model between these models.
The next step is to estimate the random effect model and perform tests to choose between a FEM and a random effect model using the Hausman test.
The procedure is continued by analyzing the spatial panel data by determining the spatial-weighting matrix and normalizing the matrix. The spatial-weighting matrix used is inverse distance weighted (IDW) by performing line normalization.
Data involving regions or areas need to be seen whether they have spatial autocorrelation or not. Spatial autocorrelation is a condition that shows the similarity of an object based on the region. Spatial autocorrelation also indicates that the value of a variable in a region is influenced by the variable in other regions that are close together. If the adjacent regions are similar, then the region can be said to have a positive autocorrelation.
The spatial autocorrelation test procedure was carried out using the Moran index to see whether or not there was spatial dependence in the panel data regression model. The next step of analysis is to estimate the parameters for the regression model equation for the spatial panel data.
Data management and analysis
The variables of this study include several important variables covering the demographic area and characteristics of children under 5 years of age related to their experiences, namely: getting early initiation of breastfeeding, poor nutritional conditions, getting vitamin A, mothers getting blood-boosting tablets, maternal visits to health care facilities during pregnancy, the average protein consumption per capita per day and average calories per capita per day.
Data were analyzed using ArcMap for mapping the area and R (R Foundation for Statistical Computing, Vienna, Austria version 4.1.1) for data analysis. Checking the variables to be free from multicollinearity is done by using the variance inflation factor value. If the value is more than 10, it indicates that there is multicollinearity in the independent variable, and we have to exclude the variable in the analysis.
The formation of three panel data models is carried out for the common effect model, the FEM, and the random effect model. Chow test and Hausman test were conducted to select the model to be used.
Map of the distribution of stunting in children under 5 years of age in Indonesia was carried out using ArcMap to present thematic map. The selected spatial panel data model is SEM random effect because it produces the largest R2 value and the smallest Akaike information criterion (AIC) when compared to the SAR random effect model.
After adjusting for socioeconomic and characteristics of children under 5 years of age, the final model was chosen based on keeping only variables, the model with or without which would not be significantly different by spatial regression model.
| Results|| |
[Figure 1] shows the decline in the number of provinces with high stunting status in children under five years of age. This can be seen from the color changes in each province. The colors red, orange, and yellow each represent the percentages in the high, medium, and low categories with different percentage ranges.
|Figure 1: Map of the distribution of stunting in children under 5 years in Indonesia from 2015–2019|
Click here to view
It can be investigated that provinces with stunting for children under 5 years of age who are categorized as high are decreasing. However, in 2018, there was an increase in the number of provinces with stunting under-five children in the high category. In Aceh Province, 2 years earlier, the condition of stunting had been categorized as moderate but returned to the high category in 2018. Central Aceh is an area in Aceh Province that has a high stunting rate and is one of the 100 priority districts for stunting intervention.
In addition, the thematic map also shows that the percentage of stunting under-five children in geographically adjacent provinces has similar values. Adjacent provinces will have a color that tends to be the same. This indicates that the value of observations in a province has a spatial dependence. In order to make sure whether a region is independent or not with other regions, a spatial autocorrelation test is carried out.
[Table 1] shows that the average percentage of stunting in children under 5 years of age in Indonesia during 2015–2019 was 27.15%, with the lowest percentage being 13.60% (Bali) in 2017. The highest percentage was 43.82% (Nusa Tenggara Timur) in 2019.
|Table 1: Percentage of children under 5 years of age who suffer from stunting|
Click here to view
Furthermore, it will be seen whether there is an auto spatial correlation in the annual data used. It would be done by using the concept of IDW, to calculate the value of the spatial autocorrelation Moran index test in each year. The test results show that there is a spatial autocorrelation in the annual data used, thus stunting conditions in Indonesia from 2015 to 2019 are interrelated between provinces.
The procedure for determining the best model is carried out through three approaches, namely the common effect, fixed effect, and random effect models. Furthermore, the selection of variables on the spatial model of the panel data is carried out.
[Table 2] shows the significant independent variables of the SAR random effect model. There are three influential variables, namely malnutrition and poor nutrition in children under 5 years of age, the average consumption of calories per capita per day, and the average monthly per capita expenditure on food.
|Table 2: Estimation of random effect spatial autoregressive model parameters for all significant independent variables|
Click here to view
[Table 3] shows the significant independent variables of the SEM random effect model. There are three influential variables, namely malnutrition and poor nutrition in children under 5 years of age, children who received Vitamin A, and the average monthly per capita expenditure on food.
|Table 3: Estimated random effect spatial error model parameters for all significant independent variables|
Click here to view
The selected spatial panel data model is SEM random effect because it has the largest R2 value and the smallest AIC.
The selected spatial panel model is SEM random effect. The model equation is formed as follows.
yit = 0.7975582 (children under 5 years of age with malnutrition and poor nutrition)it – 0.0408743 (receiving Vitamin A) it – 0.0000129 (average monthly per capita expenditure on food) it + μi+uit
From the model, we could say that the spatial error coefficient (λ) of 0.8447775 indicates that the percentage of stunting under-five children in each province has an effect of 0.8447775 times the average percentage of stunting in the surrounding provinces. If there is an increase in undernutrition and malnutrition by 1%, it will increase the percentage of stunting under-five children by 0.7975582% with the assumption that other explanatory variables remain. Then, if children receive Vitamin A by 1%, it will reduce the percentage of stunting toddlers by 0.0408743% with the assumption that other explanatory variables remain. Meanwhile, if the average monthly per capita expenditure on food increases by Rp. 1, it will reduce the percentage of stunting under-five children by 0.0000129%, assuming that other explanatory variables remain.
Children under 5 years of age with malnutrition and the average monthly per capita expenditure on food and children under 5 years of age who received Vitamin A had a significant effect on stunting.
| Discussion|| |
The condition of stunting in Indonesia still needs attention, although in general, the percentage of stunting sufferers is decreasing.
From Asmat, Papua, there were 73 fatalities consisting of stunting under-five children who died from September 2017 to January 2018 due to malnutrition and measles. As a result, this case was immediately categorized by the government as an extraordinary event. Limited access to health, geographical conditions, and the difficulty of traveling to Asmat are the causes of malnutrition and measles outbreaks in Asmat. This condition is suspected to be one of the causes of stunting in the high category in Papua in 2018.
According to these findings, the condition of under-fives with severe malnutrition and undernourished children, the average monthly per capita expenditure on food, and infants receiving Vitamin A contributed to the increased risk of stunting in children under 5 years of age. This result is in accordance with previous research conducted by Sulistianingsih and Yanti which stated that there was a significant relationship between Vitamin A and iron intake and the incidence of stunting. Vitamin A functions in the maturation of new cells. Vitamin A deficiency causes impaired growth function. Moreover, in line with previous research conducted by Pangaribuan et al. which also stated that the percentage of malnourished and undernourished children under 5 years of age and the average monthly per capita expenditure on food has a significant effect on the percentage of stunting under-fives.
By knowing the prevalence of stunting in children under five years of age and the factors that influence this condition, Appropriate interventions could be applied to families. Educating parents on good nutrition is critical to get a positive change in behavior related to the preparation of family menus. In the end, it could bring significant changes in improving the nutritional status of children under five. Furthermore, policymaking related to stunting can carry out the formulation and implementation of policies in the field of public health for all elements of family health, nutrition, promotion, and community empowerment. In addition, it can be considered in improving the nutritional status of under-five children and for improving regulations on related variables that are effective in reducing stunting rates in Indonesia. These results will aid policymakers by highlighting which areas are most vulnerable as well as which factors contribute the most to creating resilient food systems.
Overall, our findings have significance for interested policymakers, foundations, and multinationals in targets such as the sustainable development goals 2 (SDGs 2).
The results of the 2018 Basic Health Research (Riskesdas) showed a decrease in the prevalence of stunting under-five children at the national level by 6.4% over a 5-year period, from 37.2% (2013) to 30.8% (2018). The 2016 Global Nutrition Report noted that the prevalence of stunting in Indonesia was ranked 108 out of 132 countries.
The target for reducing stunting prevalence in Indonesia is aligned with global targets, namely the World Health Assembly target to reduce stunting prevalence by 40% in 2025 from conditions in 2013. In addition, the target for the TPB/SDGs is to eliminate all forms of malnutrition by 2030. For this reason, efforts are needed to accelerate the reduction of stunting from the current conditions so that the prevalence of stunting in children under 5 years of age falls to 19.4% in 2024.
| Conclusion|| |
The prevalence of stunting in the very high category with a different pattern of determinants in each province is spread out evenly in the western and eastern parts of Indonesia. Even the prevalence of stunting in Indonesia has decreased, the category is still above the reasonable threshold set by the WHO, which is 20% for each country. The government needs to be considered improving complete basic immunization coverage and increasing the number of Puskesmas in a subdistrict as an effort stunting prevention.
The limitations of this study are that the selection of independent variables is still limited, both in terms of knowledge and available data. Further, researchers should be able to develop this problem by adding other variables that are estimated to affect the proportion of stunting under five in Indonesia, both from longer time intervals, and by looking at the picture of annual movement patterns using panel data.
Despite the aforementioned limitation, our findings suggest that malnutrition, average monthly per capita expenditure on food, and Vitamin A had a significant effect on stunting in under-five children of age.
The authors would like to thank Universitas Syiah Kuala for funding this research and Health Centers in Kota Banda Aceh for their support and coordination. Special thanks to the team in Biostatistics Laboratory for providing assistance in data analysis.
Financial support and sponsorship
We would like to thank the Ministry of Education, Culture, Research, and Technology, Universitas Syiah Kuala, for fully funding this research.
Conflicts of interest
There are no conflicts of interest.
| References|| |
Kemenkes RI. Bulletin of Health Data and Information: The Situation of Stunting in Under Five Children of Age. Jakarta Selatan: Center for Data and Information Ministry of Health RI; 2018.
Bappenas. Guidelines for Planning the National Movement for the Acceleration of Nutrition Improvement for the First Thousand Days of Life. Jakarta: Bappenas; 2013.
Bappenas. Strengthening the Health Service System. Jakarta Pusat: Directorate of Public Health and Nutrition, Deputy for Human Development, Society and Culture, Ministry of National Development Planning/Bappenas; 2019.
Baltagi BH. Econometric Analysis of Panel Data. 3rd
ed. Chicester, England: John Wiley & Sons Ltd; 2005.
Caraka RE, Yasin H. Spatial Data Panel. Jawa Timur: WADE Group; 2017.
Hikmah IR, Hikmah Y. Panel data spatial modeling to determine factors affecting health in Papua Province. STATMAT (Jurnal Statistika dan Matematika) 2020;2:73-85.
Kemenkes RI. Indonesia Health Profile 2015. Jakarta: Center for Data and Information Ministry of Health RI; 2016.
Kemenkes RI. Indonesia Health Profile 2016. Jakarta: Center for Data and Information Ministry of Health RI; 2017.
Kemenkes RI. Profil Kesehatan Indonesia 2017. Jakarta: Center for Data and Information Ministry of Health RI; 2018.
Kemenkes RI. Profil Kesehatan Indonesia 2018. Jakarta: Center for Data and Information Ministry of Health RI; 2019.
Kemenkes RI. Profil Kesehatan Indonesia 2019. Jakarta: Center for Data and Information Ministry of Health RI; 2020.
Anselin L. Spatial Econometrics: Methods and Models. Dordrecht: Kluwer Academic Publishers; 1988.
Indonesia Ministry of National Development Planning and the United Nations Children's Fund. Achieving the SDGs for Children in Indonesia: Emerging Findings on Trajectories for Reaching the Targets. Jakarta: BAPPENAS and UNICEF; 2019.
Pangaribuan AC, Dhanutama KC, Wijaya MO, Navasha PT, Nooraeni R. Analysis of determinants of stunting in under five age children in Indonesia 2015-2018 with panel data regression. J Econ Educ 2020;8:56-61.
Sulistianingsih A, Yanti DA. Lack of food as a cause of stunting in under five children of age. J Dunia Kesehatan 2016;5:71-5.
[Table 1], [Table 2], [Table 3]