Impact of Breastfeeding on Child Nutrition in India: A DHS Regression Analysis

Nagavarshini R N, Roshni Vasan, Sneha Ashok Rao

ABSTRACT

Child health is a crucial indicator of a country’s economic development. According to the

World health organisation (WHO), the number of children dying before the age of 5 was halved
from 2000 to 2017, and more mothers and children are surviving today than ever before.
However, there is a lot of improvement to be done by developing countries and India is one
among them. Several factors play important role in determining child nutrition such as living
standard, sanitization facilities, breast feeding, access to healthcare facilities and household
wealth. WHO defines malnutrition as deficiencies or excesses in nutrient intake, imbalance of
essential nutrients or impaired nutrient utilization. This paper utilizes stunting, wasting and
underweight as the indicators of malnutrition and analyses the impact of breast-feeding
concerning child nutrition. Breastfeeding has been linked to improved performance on
intelligence tests among children, reduced likelihood of being overweight or obese, and lower
risk of diabetes in later life. Also, women who breastfeed experience a decreased risk of breast
and ovarian cancers. The aim of the paper is to examine the impact of breast feeding on children
aged 0-5 years in India using the data from the Demographic and Health Surveys 2019-21.

INTRODUCTION

Breastfeeding has an important role in the prevention of different forms of childhood

malnutrition, including wasting, stunting, over- and underweight and micronutrient
deficiencies. Malnutrition is responsible for approximately 60% of the 10.9 million deaths
reported annually among children under 5 years of age. The World Health Organization (WHO)
and the United Nations Children’s Fund (UNICEF) strongly recommend Exclusive
Breastfeeding during the first 6 months of life. Well-established research shows that
recommended breastfeeding practices have a substantial impact on neonatal, infant, and child
mortality. Factors contributing to this decline include various nutrition interventions, with the
Integrated Child Development Services (ICDS) playing a significant role. Maternal nutrition
positively influences the reduction in stunting and underweight, while utilization of ICDS has
a negative association with undernutrition. Despite selectivity bias, ICDS contributes to a
considerable reduction in undernutrition among children. Strengthening ICDS services is
crucial to break the detrimental cycle of undernutrition. (Katoch OR, 2021).
It has long been recognized that the well-being of a population is not solely captured by
measures of consumption and income. Social indicators such as life expectancy, health, and
education serve an important complementary function, although these dimensions of wellbeing
are often difficult to measure. Malnutrition in children leads to permanent effects and to their
having diminished health capital as adults (Strauss and Thomas, 1998; Alderman, Hoddinott,
and Kinsey, 2006). Malnutrition in early childhood is associated with functional impairment in
adult life as malnourished children are physically and intellectually less productive when they
become adults (Smith and Haddad, 1999).

Many empirical studies have investigated issues of child malnutrition in India. However, little
effort had been devoted to examining the India specific socioeconomic determinants of
malnutrition among under-five children like mother’s education, family composition etc
(Bharati et. al; 2010, Borooah; 2005, Griffiths; 2002, Pal; 1999, etc.). Anthropometric
measures, weight-for-age (underweight), height-for-age (stunted) and weight-for-height
(wasted) are used for the assessment of nutritional deprivation among the children. Stunted is
an indicator of chronic deficiency, wasted is an indicator of acute undernutrition and
underweight is a composite measure of both chronic and acute undernutrition [Gillespie and
McNeill 1992, Arnold et al 2003].

As per reports from the World Health Organization (WHO), in 2020 globally, 149 million
children less than 5 years of age were estimated to be stunted (too short for age), 45 million
were estimated to be wasted (too thin for height), and 38.9 million were overweight. The report
further revealed that around 45% of deaths among children less than 5 years of age is linked to
undernutrition, which occur in low- and middle-income countries. Estimates made by United
Nations Children's Fund (UNICEF), WHO, and the World Bank indicate that in 1990, the
prevalence of stunting and underweight were 39.6% and 25%, respectively, which dropped to
32.7% and 20.1% in 2000. In 2015, it was estimated that 23.2% of children under 5 years were
stunted, 7.4% wasted, and 13.9% underweight, thus posing a serious threat to human
development. Analysis from the third and fourth rounds of the National Family Health Survey
data (NFHS-3, NFHS-4) show that the prevalence of underweight (low weight for age) among
children <5 y decreased from 42.5% in NFHS-3 (2009–2010) to 35.1% in NFHS-4 (2015–
2016). Stunting (low height for age) also registered a decline from 48% in NFHS-3 to 38.4%
in NFHS-4. Wasting (low weight for height) increased from 19.8% to 21% for the same period.
LITERATURE REVIEW

The study on malnutrition in children has garnered considerable attention in the academic
community, resulting in a plethora of scholarly works. Bal Govind Chauhan and Ambarish
Kumar Rai (2015) studied malnutrition among children under the age of five in two Indian
cities, Meerut and Chennai. In Meerut, a significant portion of children (28.2%) is born to
mothers aged 30 or older, while only 2.2% are born to mothers under 20 years old. Additionally,
55% belong to the Hindu religion, with 18% being from SC/ST castes. A sizable portion (7.4%)
comes from poor economic backgrounds, and 34% live in slum areas. Meerut exhibits higher
rates of underweight and stunting compared to Chennai, though Chennai has a higher
prevalence of wasting. The educational and economic statuses of mothers are significant factors
affecting child nutrition status, with higher education and economic status leading to better
nutrition outcomes. In Chennai, poorer children tend to suffer from higher rates of underweight
and stunting, while wealthier children are more affected by wasting.

Salah E.O. Mahgoub, Maria Nnyepi and Theodore Bandeke(2006) focussed on the nutritional
status of children under three years old in Botswana, involved 400 households representing 23
Health Regions. Key findings include a 5.5% rate of wasting, 38.7% stunting, and 15.6%
underweight. Boys were significantly more affected than girls. Factors like maternal education
and income, employment in agriculture, and breastfeeding practices were found to influence
nutritional status. Economic engagement without suitable childcare may worsen malnutrition,
while maternal education can be a proxy for women's social status, impacting children's health.
The study suggests focusing on factors like maternal income, education, and creating job
opportunities to address child malnutrition, particularly in single-parent households. M. R.
Prashanth, Savitha M. R. and Prashantha B (2017) compared sociodemographic and nutritional
risk factors for Severe Acute Malnutrition (SAM) between cases and controls. Regarding
sociodemographic factors, SAM was more prevalent in children below 2 years (82.5%) and
among mothers below 20 years (33%). Children from families with 5-8 members, low monthly
income, multiple siblings, shorter birth spacing (< 2 years), lacking toilet facilities, and living
in kucha houses were more prone to SAM. Regarding nutritional factors, SAM cases exhibited
poorer appetite, higher rates of prelacteal feed, shorter duration of exclusive breastfeeding,
early introduction of complementary feeding (< 6 months), increased bottle feeding, calorie
deficit, protein deficit, and lower intake of fruits and vegetables. The study highlights the
importance of addressing both sociodemographic and nutritional risk factors to reduce
childhood malnutrition.

Kakoli Rani Bhowmik and Sumankanti Das (2017) examined child malnutrition in Bangladesh
using various statistical models, including linear regression, logistic regression, and multiple
classification analysis (MCA) on data from the Bangladesh Demographic and Health Survey
(BDHS) 2011. The key findings highlight the significance of age, birth weight, birth interval,
mother's education and BMI, household economic status, family size, residential area, and
regional disparities in child malnutrition. The study concludes that MCA provides additional
insights and helps in ranking predictors, facilitating the design of effective interventions to
reduce child malnutrition in Bangladesh.

Om Raj Katoch (2021) in a recent study found that maternal education and household income
as crucial determinants of child malnutrition. Educated mothers tended to have higher incomes
and better knowledge of child nutrition, impacting their children's nutritional status positively.
Low-income households faced challenges in providing adequate food, health services, and a
healthy environment, affecting child nutrition negatively. As the child's age increased, the risk
of malnutrition also increased. Child malnutrition was associated with factors like
breastfeeding practices, cooking areas, fuel used, and socioeconomic status. Policies targeting
maternal education, household income, and maternal health could effectively improve child
nutritional status. This paper tries to utilize the data from Demographic and Health Surveys
2019-21 to study the factors determining malnutrition in Children in India.

Study on Mamma Bambino cohort (2021) in University of Catania underscores the critical
importance of breastfeeding for both infants and mothers, highlighting its unique nutritional
and health benefits. Studies often associate higher educational levels and employment status
with increased rates of breastfeeding adherence, suggesting the influence of socioeconomic
factors on breastfeeding practices. However, challenges such as returning to work and maternal
health conditions like obesity and smoking can impact breastfeeding duration. Targeted
interventions, particularly for less educated mothers, are recommended to promote exclusive
breastfeeding and address cultural beliefs that influence infant feeding choices. Limitations in
research, such as sample size constraints and self-reported data, warrant further investigation
into factors affecting breastfeeding initiation and continuation.

Nicholas Metheny and Rob Stephenson (2019) highlight the significant impact of intimate
partner violence (IPV) severity on breastfeeding practices. Their analysis indicates that severe
physical violence is associated with lower odds of exclusive breastfeeding among mothers of
infants 0 to 6 months. However, the association diminishes for older infants, suggesting a
nuanced relationship that may require targeted interventions for mothers with younger infants
to mitigate the negative effects of early weaning. Additionally, factors such as wealth quintile
and husband's education also influence breastfeeding behaviour, highlighting the multifaceted
nature of this issue.

DATA & METHODOLOGY

For the present study, the Survey data from the Demographic and Health Surveys (DHS 7)
Program have been used that was conducted during 2019-21. The DHS Program collects data
that are comparable across countries using model questionnaires. These raw data are then
processed into standardized data formats known as recode files. The required variables are
extracted from children’s recode file (Children’s data). This dataset has one record for every
child of interviewed women, born in the five years preceding the survey. It contains the
information related to the child's pregnancy and postnatal care and immunization and health.
The data for the mother of each of these children is included. This file is used to look at child
health indicators such as immunization coverage, and recent occurrences of diarrhoea, fever,
and cough for young children and treatment of childhood diseases. The unit of analysis (case)
in this file is the children of women born in the last 5 years (0-59 months). We use this dataset
to study the impact on child anthropometric measures due to factors like place of residence,
mother’s education and so on.

Data description

Dependent variable

The three standard indices of physical growth given by World Health Organization (WHO) are
used as dependent variables to describe nutritional status of children. Weight for age
(underweight) is an indicator of either current or past nutrition, whereas Height for age
(stunting) is an indicator of past nutrition. Weight for height (wasting) is a sensitive indicator
of current nutrition status and the degree of wasting.

Most commonly the z scores the above-mentioned indicators are used in order to compare the
results on the same scale. The z score of each of the indicator ranges between -3(poorly
nourished) and 3(well nourished). For example, a stunting z score of -2 indicated sign of
prominent stunting in a child and a stunting z score of 2 indicated the child is moderately
nourished.

Independent variable

Birth order:

Birth order shows the order in which the child is born within a particular family. Children are
numbered (like 1,2,3 and so on) from the oldest to youngest. Higher birth order children (those
born later in the family) tend to be more prone to malnutrition compared to those born earlier.

wealth index:

The wealth index is a composite indicator of a household's overall living standard. The wealth
index is computed using simple data on a household's possession of specific items, such as
televisions and bicycles; the materials used in housing construction; and the sorts of water and
sanitation services. DHS separates all interviewed households into five wealth quintiles namely
poorest, poorer, middle, richer, richest to compare the influence of wealth on various
population, health and nutrition indicators.

Age mother:

This variable indicates the current age of the mother at the time of survey. Several research
papers indicate that young maternal age is a risk factor for child undernutrition.

Rural urban:

This shows the child’s type of place of residence (rural, urban). The place of residence plays a
vital role in determining the availability of necessities and resources for child’s development
and growth.

Month breastfed:

Months breastfed represents the number of months the mother breastfed her children.
Breastfeeding is a important factor for preventing child malnutrition as it set the foundation for
optimal health and acts as a first vaccine to the children.

Parents education

Parents education represents the number of years of education the parents have attended. The
years of education of the parents plays a vital role in determining the nourishment the child
will receive. The more educated the parents are the more aware they will be of the nutritional
necessities of the child.

DESCRIPTIVE STATISTICS:

Tabulation of Z-score of Weight -Height Ratio: Wasting

Z-score Freq. Percent Cum.

-3 1834 1.25 1.25

-1.99 21586 14.75 16.00

-1 56543 38.63 54.63

0 48384 33.06 87.69

1 18025 12.31 100.00

Total 146372 100.00

The above table displays the z score of weight-height standard deviation of children between
0 to 5 years of ages. This enables us to understand the proportion of children who are
moderately and severely wasted. DHS defines any children whose Weight-Height Z -score is
between -2 and -3 as severely wasted. In children under 5 years of age it appears that 1.25%
had values of standard deviation in the weight height ratio below -2 and therefore had a state
of malnutrition present in a very high degree. However, 12.3% had values for Standard
deviation weight-height ratio above 1 indicating over nutrition.

Tabulation of Z-score of Weight-Age Ratio: Underweight

Z-score Freq. Percent Cum.

-3 1529 1.10 1.10

 -1.99 18743 13.42 14.52

-1 57341 41.07 55.59

0 45386 32.51 88.09

1 16623 11.91 100.00

Total 139622 100.00

The above table displays the z score of weight age standard deviation of children between 0 to
5 years of age. This enables us to understand the proportion of children who are moderately
and severely underweight. DHS defines any children who weight-age ratio z-score lying
between -2 to -3 as severely underweight. In children under 5 years of age it appears 1.10 %
are severely underweight whereas 13.42 % of children whose z score lies below 2 are said to
be moderately underweight. The rest seem to be the correct weight for their age and hence is
not discussed below.

Tabulation of Z-score of Height-Age Ratio: Stuntedness

Z-score Freq. Percent Cum.

-3 3225 2.31 2.31

-1.99 17130 12.27 14.57

-1 54268 38.86 53.43

0 50029 35.82 89.26

1 15005 10.74 100.0

0

Total 139657 100.00

The above table displays the z-score of height age standard deviation between 0 to 5 years of
age. This enables us to identify the degree of stunt in their growth. DHS defines any children
whose height-age ratio z-score is between -2 to -3 is severely stunted and any child whose ratio
is below -2 is said to have moderately stunted growth. In children under 5 years of age it appears
2.31% are severely stunted and 12.27 % are moderately stunted. The rest appear to have no to
less stuntedness in their growth and hence it is not discussed below.

Composition of sex of the child by area of residence in the dataset

type of place of residence

sex of child urban rural Total

male 17816 65725 83541

female 15763 56551 72314

Total 33579 122276 155855

The dataset consists of 53% male children out of which 21% live urban societies whereas 32%
live in rural households. It can be clearly seen that most of male children live in rural areas
when compared to urban areas. The dataset consists of 47% female children out of which 21%
live in urban areas while the rest reside in rural areas. It can be seen for female children as well
that their predominant residence is rural areas. To sum up the dataset is dominated by children
whose prominent area of residence is rural households.
The above bar graph shows the mean malnutrition score in children below age 5. It can be
inferred that children of one year of age are more prone to malnutrition than children of other
age groups. It can also be observed that an average one year can be moderately to severely
malnourished i.e. they can have stuntedness in growth, be underweight or be wasteful or
overweight.

The graph above shows the mean malnourishment score of children who have been breastfed.
Mothers who did not breastfed their children have low malnourishment score when compared
to children who have been breast for 20 and 40 months respectively. The graph shows that
malnutrition amongst children who have been breastfed for 20 to 40 months is high however
lower when not breastfed or breastfed for more than 40 months. The data speaks contrary to
popular belief and hence must be tested to identify the significance of breastfeeding for
nutrition in children.

UNDERWEIGHT

According to World Health Organization (WHO), child underweight belongs to a set of

indicators whose purpose is to measure nutritional imbalance and malnutrition resulting in
undernutrition (assessed by underweight, stunting and wasting) and overweight.

Underweight: Weight-for-age z-score is below -2.0 standard deviations from the median on the
WHO Child Growth Standards.

The model for underweight is given as follows.

Underweight = α +β1 breastfeeding + β2 mother’s education + β3 mother’s age+ δ1 rural + δ2

poorer + δ3 middle + δ4 richer + δ5 richest + δ6 (“2-3”) +δ7 (“4_5”) + δ8 (“6+”)
The intercept in the multiple linear regression model indicates an estimated baseline z-score of
approximately -0.2824158 for the weight-to-age ratio, reflecting the expected value when all
independent variables are either zero or not applicable. The analysis reveals a statistically
significant positive association between breastfeeding duration and the weight-to-age ratio z-
score, with each additional month of breastfeeding correlating with a roughly -0.117764 unit
increase in the z-score, holding other variables constant. This implies that longer breastfeeding
durations are linked to healthier weight-to-age ratios in children, reducing the likelihood of
them being underweight.
Moreover, there is a negative relationship between maternal education level and the risk of
child underweight. A one-unit increase in the mother's education level corresponds to an
approximately 0.0174408 unit increase in the z-score, after controlling for other factors. This
suggests that higher maternal education, likely leading to better understanding of nutritional
needs, contributes to preventing children from being excessively thin relative to their age.

Regarding maternal age, each unit increase is associated with a 0.0133603 unit change in the
child's underweight possibility, all else being equal. This underscores the nuanced impact of
maternal characteristics on child health outcomes within the context of the weight-to-age ratio
z-score

Relationship between place of residence and child underweight

The coefficient of the dummy variable 'rural' (with the base group being 'urban') indicates a
significant difference in z-scores between these two settings, showing a decrease of
approximately -0.966186 units for children living in rural areas compared to urban areas. This
finding suggests that residing in rural areas is associated with a notable decrease in the z-score
of weight-to-age ratio, indicating a higher likelihood of children being underweight compared
to their counterparts in urban settings.
Relationship between child underweight and Wealth index
The wealth index categories show distinct associations with the z-score of weight to age ratio,
reflecting varying levels of likelihood for children to have a healthy ratio and avoid
underweight.
Poorer: Children in the "Poorer" wealth index category, compared to the base category
("Poorest"), exhibit a statistically significant increase in the z-score of weight to age ratio, with
an approximate increase of 0.0929728 units. This suggests that children in the "Poorer"
category are less likely to be underweight than those in the "Poorest" category.
Middle: Moving up to the "Middle" category shows a higher z-score by approximately
0.1559766 units compared to the base group. This indicates a further improvement in the
likelihood of a healthy weight-to-age ratio.
Richer: Children in the "Richer" category demonstrate a higher z-score by approximately
0.1966299 units compared to the reference category. This signifies an even greater likelihood
of maintaining a healthy ratio.
Richest: The "Richest" category exhibits the highest increase in the z-score, approximately
0.3325118 units higher than the base group. This substantial increase underscores a
significantly higher probability of children in this category having a healthy weight-to-age
ratio.
Overall, the trend indicates a progressive improvement in the likelihood of a healthy weight-
to-age ratio as one moves from poorer to richer wealth index categories. This implies that socio-
economic status plays a crucial role in determining the nutritional status of children, with higher
wealth levels associated with a greater likelihood of maintaining a healthy ratio and avoiding
underweight.
Relationship between child underweight and Birth order
The birth order categories reveal notable differences in the z-score of weight to age ratio
compared to the base group, i.e., firstborn children.
2-3 Birth Order: Children in the 2-3 birth order category show a decrease in the z-score by
approximately -0.1022399 units compared to the base group. This indicates a slight decrease
in the likelihood of maintaining a healthy weight-to-age ratio compared to firstborns.
4-5 Birth Order: Moving to the 4-5 birth order category, there is a statistically significant
decrease in the z-score compared to the reference category. Children in this group exhibit a
decrease in the z-score by approximately -0.1518624 units, highlighting a higher likelihood of
being underweight compared to firstborns.
6+ Birth Order: Similarly, for children in the 6+ birth order category, there is a significant
decrease in the z-score compared to the reference category. The z-score decreases by
approximately -0.226351 units, indicating a notable increase in the likelihood of being
underweight among children born as the 4th child or higher.
These findings suggest that as the birth order increases, there is a progressive decrease in the
likelihood of maintaining a healthy weight-to-age ratio. Children born as 4th child or higher
are particularly vulnerable to being underweight compared to firstborns, highlighting the
importance of considering birth order as a factor in assessing nutritional status.
Interaction terms
We conducted a test to evaluate the impact of the interaction between months of breastfeeding
and the place of residence of the child on the z-score for underweight. The negative coefficient
(-0.0657702) from the regression model indicates that the influence of breastfeeding duration
on underweight is less pronounced in rural areas compared to urban areas.
The interaction term between wealth_index and rural_urban regression
Months Breastfed: Each additional month of breastfeeding is associated with a decrease in the
z-score for underweight by approximately -0.0117838 units, holding other variables constant.
This suggests that longer durations of breastfeeding are linked to healthier weight-to-age ratios
in children, reducing the likelihood of them being underweight.
Maternal Education: An increase in the mother's education level by one unit corresponds to an
increase in the z-score by approximately 0.0174354 units. This indicates that higher maternal
education is associated with a reduced risk of children being underweight, likely due to better
nutritional knowledge and practices.
Maternal Age: Each unit increase in maternal age correlates with a z-score increase of
approximately 0.0135058 units. This implies that older mothers may have better strategies or
resources to prevent their children from being underweight.
Place of Residence (Rural vs. Urban): Living in a rural area is associated with a decrease in the
z-score for underweight by approximately -0.1444841 units compared to urban areas. This
suggests that children in rural areas are more likely to be underweight compared to their urban
counterparts.
The interaction term between breastfeeding and place of residence
Months Breastfed: Each additional month of breastfeeding is associated with a decrease in the
z-score for underweight by approximately -0.0086307 units, holding other variables constant.
This suggests that longer durations of breastfeeding are linked to healthier weight-to-age ratios
in children, reducing the likelihood of them being underweight.
Maternal Education: An increase in the mother's education level by one unit corresponds to an
increase in the z-score for underweight by approximately 0.0174393 units. This indicates that
higher maternal education is associated with a reduced risk of children being underweight.
Maternal Age: Each unit increase in maternal age is associated with an increase in the z-score
for underweight by approximately 0.013413 units. This suggests that older mothers may have
better strategies or resources to prevent their children from being underweight.
Place of Residence (Rural vs. Urban): Living in a rural area is associated with a decrease in the
z-score for underweight by approximately -0.0657702 units compared to urban areas. This
indicates that children in rural areas are more likely to be underweight compared to their urban
counterparts, even after considering other factors.
The following figure shows the mean distribution of weight to age(z score) across different
wealth index.
WASTING

According to World Health Organization (WHO), child growth is an internationally accepted

outcome reflecting child nutritional status. Child wasting refers to a child who is too thin for
his or her height and is the result of recent rapid weight loss or the failure to gain weight.
Wasting, based on a child’s weight and height, is a measure of acute undernutrition.

Moderately or severely wasted: Weight-for-height z-score is below -2.0 standard deviations

from the median on the WHO Child Growth Standards.

The model for wasted is given as follows

Wasted = α +β1 breastfeeding + β2 mother’s education + β3 mother’s age+ δ1 rural + δ2 poorer

+ δ3 middle + δ4 richer + δ5 richest + δ6 (“2-3”) +δ7 (“4_5”) + δ8 (“6+”)
The above figure shows the summary statistics for multiple linear regression output for the
given model. The intercept represents the estimated baseline z-score of weight to height ratio
when all independent variables are zero or not applicable. In this model, the estimated baseline
z-score is approximately -0.2643813. There is a statistically significant positive relationship
between the duration of breastfeeding and the z-score of weight to height ratio, i.e., child
wasted. For each additional month of breastfeeding, the z-score increases by approximately
0.000546 units, holding all other variables constant. This shows that as months of breastfeeding
increases the likelihood of child being healthy and not wasted. Similarly, there is a negative
relationship between mother's education level and possibility of child being wasted. For each
additional unit increase in the mother's education level, the z-score increases by approximately
0.0183031 units, holding all other variables constant. This shows that when mothers are more
educated about the nutritional value, it can prevent the child from being too thin for his height.
The ceteris paribus effect of mother’s age on her child wasted possibility is shown by the
respective coefficient value, 0.0048861 units.

Relationship between place of residence and child wasted.

This can be shown by the coefficient of the dummy variable ‘rural’ (base group is ‘urban’)
which shows the differential z scores between the places (0.3814689 units). Living in rural
areas, compared to urban areas, is associated with a statistically significant decrease in the z-
score of weight to height ratio. This means children residing in the rural area tend to be wasted
against those who live in urban area.

Relationship between child wasted and Wealth index

Poorer: Being in the "Poorer" wealth index category, compared to base category ("Poorest"), is
associated with a statistically significant increase in the z-score of weight to height ratio. For
children in the "Poorer" category, the z-score increases by approximately 0.1786977 units
compared to the poorest

Middle: Children in the "Middle" category have a z-score higher by approximately 0.310975
units compared to the base group.

Richer: Children in the "Richer" category have a z-score higher by approximately 0.4307326
units compared to the reference category.

Richest: Finally, being in the "Richest" wealth index category is associated with the highest
increase in the z-score compared to the reference category. Individuals in the "Richest"
category have a z-score higher by approximately 0.6438171 units compared to base group.

We can observe that as compared to poorer children, the other category children are more likely
to have healthy height weight ratio and not being wasted. As they from poorer to richest group,
the likeliness of child being healthy increases at an increasing rate.

Relationship between child wasted and wealth index

There is no statistically significant difference in the z-score of weight to height ratio between
children in the 2-3 birth order category and base group, i.e., first child. Children in the 4-5 birth
order category have a statistically significant decrease in the z-score compared to the reference
category. For children in the 4-5 birth order number, the z-score decreases by approximately
0.0859332 units compared to the reference category. Same case for 6+ birth order number
children. This shows that children born as 4th child or above are more prone to be wasted than
the children born as 1st child.

Testing the effect of interaction of months of breastfeeding with place of residence of the child:
The interaction term shows how the effect of breastfeeding duration on the outcome variable (wasted)
varies between rural and urban areas. The negative sign (-0.0320748) of the coefficient after running
the regression on the following model indicates that the effect of breastfeeding duration on the
outcome variable (wasted) is weaker in rural areas compared to urban areas.

Z-score(wasted)= β0+ β1months breastfed+ β2wealth index + β3 education + β4 age of mother+ β5

birth order+ β6 type of residence+ β7bf_urban

We are to test the null hypothesis that the above-mentioned interaction has no impact against the two-
sided alternative that its impact is significantly different from zero.

H0: β7=0 against H1: β7≠0

Wald Test Result

F (4,151525) = 2.60
Prob > F = 0.0343

Since the p-value (0.0343) is less than the significance level (typically 0.05), we reject the null
hypothesis. Therefore, we have evidence to suggest that there is a significant interaction effect
between type of household and wealth index.

Testing the significance of interaction between type of household and wealth index
Z-score (Wasted)= β0+ β1months breastfed+ β2wealth index + β3 education + β4 age of mother+ β5
birth order+ β6 type of residence+ β7 rural_urban_wealth_interaction
The observed value of coefficient of the interaction is positive. To assess the significance of the
interaction between type of household (e.g., rural or urban) and wealth index in our regression model,
we would perform a Wald test.
H0: β7=0 against H1: β7≠0

The regression analysis below states that the z score difference amongst urban poor and rural and urban
rich is 0.026 (2.6%) once again insinuating the fact that we have iterated before.
Wald test result
F (1,151523) = 39.46

Prob > F = 0

Since the p-value is less than 0.05, we reject the null hypothesis. Therefore, we conclude that
there is evidence of a significant interaction effect between type of household and wealth index.
The above figure shows the mean distribution of weight to height (z score) across different
wealth index.

STUNTING:

Testing the impact factors have on stuntedness of a child between the age of 0 and 5:

In this study we are interested in testing the impact various factors such as months breastfed,
wealth index, education of the parent, age of the mother, birth order, and type of residence have
on stuntedness. Stuntedness in a child is denoted by the z score of height to age.

The regression equation can be written as:

Z-score (Stunted)= β0+ β1months breastfed+ β2wealth index+ β3 education+ β4 age of

mother+ β5 birth order+ β6 type of residence

The table below displays the coefficients of the factors along with many other statistical
information. Information that is important for our analysis is the p-value. According to rejection
rule based on p-value, the null hypothesis is rejected when the p-value is below the alpha or
level of significance.
 Stunting Z Coef. St. Err. t- p- [95% Interval] Sig
score value value Conf

months_breastf -.009 0 -46.82 0 -.01 -.009 ***

ed

wealth_index .058 .002 27.25 0 .054 .062 ***

age_mother .01 .001 16.46 0 .009 .011 ***

birth_order -.046 .003 -18.31 0 -.051 -.041 ***

education .014 .001 22.18 0 .013 .015 ***

type of place of 0 . . . . .
r~s

rural -.078 .007 -11.76 0 -.091 -.065 ***

Constant -.221 .017 -13.09 0 -.255 -.188 ***

Mean dependent var 0.000 SD dependent var 1.000

R-squared 0.045 Number of obs 144477

F-test 1134.260 Prob > F 0.000

Akaike crit. (AIC) 403370.513 Bayesian crit. (BIC) 403439.680

*** p<.01, ** p<.05, * p<.1

To test:

H0: β1=β2= β3= β4= β5= β6= β6 =0 against H1: β1≠β2≠β3≠β4≠β5≠β6≠β6 ≠0

The hypothesis that we are interested in testing above simply means that each factor has no
influence in the stuntedness of a child’s growth.

Upon analyzing the p-value(which is 0 for each factor) we can reject the null hypothesis at
α=0.01 level of significance and conclude that each factor has some influence on determining
the stuntedness of a child’s growth.
We can also infer the relationship between the factors and stuntedness z score. It is useful to
note that wealth index, education level of the parents, age of the mother are positively
correlated with the z score while the other exhibit an inverse relationship.

A unit increase in breastfeeding months causes the z score to decrease by 0.9% (this can be
substantiated with the fact that the data consists of children who are breastfed for a prolonged
period. Birth order also shares the same pattern as duration of breastfeeding i.e for a unit
increase in birth order the z score decreases by 4.6% .This is because the increased birth order
indicates more division of resources which leads to less resource for one child hence less
nutrition for a child. Everything remaining same the difference in z score between rural and
urban household children is -7.8%. Urban households are considered as base and therefore
indicates that children in urban households are more nutritious than rural kids.

A educated parents can provide 1.4% better nutrition than ill-educated set of parents. The same
pattern is shared by wealth index and age of the mother.

Testing the effect interaction of months of breastfeeding with other terms on the stuntedness of
the child:

As seen from the table before that for each point increase in months breastfed the z score
decreases by 0.006 or 0.6%. According to an article by NCBI the prolonged breastfeeding can
cause stunting and sometime lead to severe stunting. It can be concluded that the more months
the child is breastfed the more they are prone to having stunted growth or stunting. So it is of
interest we test the hypothesis:

H0: β1=0 against H1: β1≠0

According to health organizations breastfeeding children in their infant years improves their
nutritional status. Usually, a child must be breast fed for at least six months from their birth.
The months of breastfeeding ranges from 0 months to 96 months (3 years) in the dataset
considered. So, it is essential we convert it into a categorical variable as it would be helpful
when analyzing the interaction with other terms as well. Since it is a categorical variable, a
base group must be assumed which in our case is a group of mothers who breastfeed their child
for less than a month. It can be inferred from the table that there is a 9.7% difference in stunting
between the mothers who breastfed their child more than a month and less than a month. This
once again
 Breastfed -0.097 0.003 -38.40 0 -0.102 -0.092 ***
months
(base
group <1)

proves that more months of breastfeeding increases the risk of stunting. The table below
displays the results of Wald test which tests if a particular parameter in our case duration of
breastfeeding is statistically significant from a certain value (like 0 in this case). The p-value
indicated in the table explains that the evidence from the data explains that the duration of
breastfeeding a child is statistically differing from zero and hence we can reject the null
hypothesis that breastfeeding has no impact on stunting.

Wald test results

F (4, 144465) = 777.91

Prob > F = 0.0000 (p-value)

Having examined the effect breastfeeding duration has on a child’s development, investigating
the interaction between duration of breastfeeding and other variables like type of residence,
education level of parents and birth order is of consequence.

Interaction between type of residence (rural and urban) and duration of breastfeeding:

The duration of breastfeeding is dependent on the place of residence of the family. A mother in
the rural area can have more time off during and after birth of the child to nurture for the child
and hence will be able to breastfeed for a prolonged duration. However, a mother in urban area
might not have the flexibility to take time off to cater to the child (because of limited time off
from work, lack of paid maternity leave etc.).

The bar graph below insinuates that there are some significant inter effects between the type of
residence and months of breastfeeding.
To test the hypothesis, we add the interaction term breastfeeding*rural_urban(bf_urban) and
the new regression looks like:

Z-score (Stunted)= β0+ β1months breastfed+ β2wealth index + β3 education + β4 age of

mother+ β5 birth order+ β6 type of residence+ β7bf_urban

We are to test the null hypothesis that the above-mentioned interaction has no impact against
the two-sided alternative that its impact is significantly different from zero.

H0: β7=0 against H1: β7≠0

The table below shows the results of the Wald test testing the above hypothesis. Despite seeing
some evidence that there might be some significance in the interaction between type of
residence and duration the p-value (0.13>0.01) is greater than alpha hence we accept the null
hypothesis and conclude that this interaction does not influence the stunting significantly.

Wald Test Results

F( 1,153281) = 2.29
 Prob > F = 0.1301

Testing the interaction between duration of breastfeeding and education level of parents:

The education level of parents is a important indicator when it comes to understanding the
nutritional development of a child. The regression table introduced before informs that for a
unit increase in the year of education the z score increases by 0.013(1.3%) implying that the
more years of education the more nutrition the child can get. As a consequence, we examine
its interaction with the months of breastfeeding.

Z-score (Stunted)= β0+ β1months breastfed+ β2wealth index + β3 education + β4 age of

mother+ β5 birth order+ β6 type of residence+ β7bf_educ

To test:

H0: β7=0 against H1: β7≠0

The results after conducting the Wald test are given below. From the p-value (which is less than
alpha) it can be inferred that the effect of the interaction is significant and therefore we can
reject our null hypothesis.

Wald Test Result

F (1,144467) = 34.27

Prob > F = 0.0000

Testing the significance of interaction between type of household and wealth index:

The above diagram displays the z score across different wealth index cases like poorest, poorer,
middle, richer and richest. It is evident that for poorest and poorer households the mean z score
is less than 0 indicating signs of stunting and children from middle and rich households have
better nutrition.

Here we test the null hypothesis that the interaction between type of residence and wealth index
is not significant against the two-sided alternative that their impact is significant.

Z-score (Stunted)= β0+ β1months breastfed+ β2wealth index + β3 education + β4 age of

mother+ β5 birth order+ β6 type of residence+ β7weal_urban

H0: β7=0 against H1: β7≠0

The regression analysis below states that the z score difference amongst urban poor and rural
and urban rich is 0.017 (1.7%) once again insinuating the fact that we have iterated before.
Children from Urban rich and rural rich are able to collectively provide more nutrition than the
poor.

Wealth_urban .0179761 .0044946 4.00 0.000 .0091669 .0267854

Wald test result
F( 1,144469) = 16.00

Prob > F = 0.0001

Conclusion:

Stunting, wasting, and underweight as an indicator of child malnutrition is influenced by the

maternal education, age of the mother, the wealth of the household, place of residence, duration
of breastfeeding and birth order. The significance of each of these variables to each indicator
differs. Breastfeeding months are a positive indicator for wasting and underweight whereas for
stunting prolonged duration of breastfeeding causes stunting and can lead to severe stunting.
Educated parents who breastfeed their children for more than a month have significantly
increased the z score of the indicators which implies that educated parents have more awareness
about the appropriate duration for breastfeeding. Children from poor households are at the risk
of being underweight, stunted, and wasted regardless of their residence (urban or rural).

In conclusion, our regression analysis has shed light on the multifaceted determinants of child
malnutrition. By examining various socio-economic, demographic, and environmental factors,
we have identified significant predictors contributing to malnutrition prevalence. Our findings
underscore the critical importance of addressing not only nutritional intake but also broader
socio-economic disparities and access to healthcare. Moving forward, targeted interventions
aimed at improving maternal education, household income, and sanitation facilities can play a
pivotal role in combating child malnutrition and fostering healthier outcomes for vulnerable
populations.
References:

1. Scherbaum V, Srour ML. The Role of Breastfeeding in the Prevention of Childhood

Malnutrition. World Rev Nutr Diet. 2016;115:82-97. doi: 10.1159/000442075. Epub
2016 May 19. PMID: 27198529.
2. Singh, S.K., Gudakesh & Vishwakarma, D. Malnutrition among children in India:
exploring the contribution of the integrated child development service scheme. SN Soc
Sci 4, 49 (2024). https://doi.org/10.1007/s43545-023-00811-7
3. Katoch OR. Determinants of malnutrition among children: A systematic review.
Nutrition. 2022 Apr;96:111565. doi: 10.1016/j.nut.2021.111565. Epub 2021 Dec 11.
PMID: 35066367.

4. Kangas, S.T., Ouédraogo, C.T., Tounkara, M. et al. Nutritional treatment of children

6–59 months with severely low weight-for-age z-score: a study protocol for a 3-arm
randomized controlled trial. Trials 25, 30 (2024). https://doi.org/10.1186/s13063-023-
07890-0
5. Prentice AM, Ward KA, Goldberg GR, Jarjou LM, Moore SE, Fulford AJ, et al. Critical
windows for nutritional interventions against stunting. Am J Clin Nutr. 2013;97: 911–
918. 10.3945/ajcn.112.052332
6. De Onis M, Blössner M. The World Health Organization global database on child
growth and malnutrition: methodology and applications. Int J Epidemiol. Oxford
University Press; 2003;32: 518–526
7. Yang H, de Onis M. Algorithms for converting estimates of child malnutrition based on
the NCHS reference into estimates based on the WHO Child Growth Standards. BMC
Pediatr. 2008;8: 1–6. 10.1186/1471-2431-8-1
8. WHO Child Growth Standards. 2003.