Nutritional Status of Female Children in Comparison to Their Male Siblings in India– A Secondary Analysis of National Family Health Survey (NFHS-5) Data

male siblings within same household. Methods : Individual data of 19083 pairs of siblings was analysed from the NFHS -5 data. Differences in the Z-scores of the siblings was calculated as male’s Z-score minus female’s Z-score and factors influencing them were studied. Results : The mean HAZ and WAZ scores were less than -1.4 Standard deviations (SD) for both genders – while the means of WHZ and BAZ scores were ranging between -0.67 to -0.83. The differences in Z scores were more or less distributed equally on both sides of zero, implying no disadvantage to any of the genders. The mean differences in the Z scores were minimal (-0.07 in HAZ, -0.04 in WAZ, -0.06 in WHZ and 0.01 in BAZ). Conclusion : There was little or no difference in the mean z scores of females and male siblings and did not indicate any gender advantage or disadvantage.


INTRODUCTION
Undernutrition of under-5-year children is a major health concern in India. It not only affects the individual productivity, but impacts the national growth and development as well. 1 Wasting (low weight for height), stunting (low height for age), underweight (low weight for age) and micronutrient deficiencies are the four forms of manifestation of undernutrition. 2 Several intrinsic (age, gender, genetics, diseases) and extrinsic (household characteristics, sociodemographic, economic and political environment) factors have been linked to undernutrition. During the periods of 2006 to 2021, while there has been tremendous decline in the percentages of stunted and wasted children, substantial improvement in the national efforts is required to reach the global targets for malnutrition indicators. 3 Gender disparity and boy preference have been associated since ages with the Indian family systems which deeply impact the growth and development of the children. 4 Gender differential aspects of malnutrition have been widely studied by many researchers linking them to family preferences for birth order as well as disproportionate allocation of household resources. 5,6 Reports from studies conducted in Northern India have stated that the birth order of the index child and the sex composition of the older sibling are important contributors to the gender bias in childhood feeding. 7 However, there exists a discrepancy in the acceptance of these theories of gender disparities. While some researchers have reported female disadvantage in the prevalence rates of stunting, a consequence of chronic malnutrition 8 , several others have contradicted these reports 9,10 .
There is a plausible gap in the literature on substantial data that evaluates the gender disparities in the manifestation of undernutrition within the Indian households, negating the underlying family sociodemographic factors which may skewer the results. In this context, the present analysis was carried out on the National Family Health Survey -5 (NFHS-5) data with the objective of evaluating the differences in the nutritional status of the female Indian children under-5 years of age with their male siblings, quantitatively in terms of distances between their Z scores, viz., Weight for age Z score (WAZ), Height for age Z score (HAZ), Weight for height Z score (WHZ) and BMI for age Z score (BAZ).

METHODOLOGY
Study population and sampling design: The 5 th round of NFHS (NFHS-5) was conducted from June 2019 to April 2021 by the Indian Ministry of Health and Family Welfare with the International Institute for Population Sciences as the nodal agency. The complete methodological details have been elaborated in the India Report of  Concisely, the study subjects were women aged 15 to 49 years, men aged 15 to 54 years and children aged 0 to 59 months. With 2011 census as the sampling frame, the study employed stratified two stage sampling. Villages in rural areas and Census Enumeration Blocks in urban areas were selected as Primary Sampling Units (PSUs) using probability proportional to size (PPS) systematic sampling. In the second stage, 22 households were randomly selected with systematic sampling from the PSUs. The ethics approval for the survey was obtained from ethics review board of International Institute of Population Sciences, Mumbai, India. 12 Data collection: The data was collected through 4 survey schedules -Household, Woman, Man, and Biomarker -using Computer Assisted Personal Interviewing. Informed written consent for participation was obtained from the respondent before initiating the survey. It included eliciting the household demographic and health related details, and included measurements of height, weight, and haemoglobin levels for children, apart from an array of biomarkers and anthropometric measurements for the adults. Seca 874 Digital Scale was used for measurement of Weight, Seca 213 Stadiometer for height and Seca 417 Infantometer for length of the children if less than 2 years old or less than 85 cm. Multiple levels of monitoring and supervision of the fieldwork was done internally and by external agencies for quality assurance.
Statistical analysis: Individual level dataset of NFHS-5 was obtained from The Demographic and Health Surveys -the DHS program and prior permissions were sought for performing this analysis (Authorisation letter no.175759, supplementary file no.1). The file with child recodes IAKR7DFL.dta was used to fetch the data of the children under 5 years. World Health Organization (WHO) Anthro software 13 was used to derive the Weight-for-age z-scores (WAZ), Height-for-age z-score (HAZ), Weight-forheight z-scores (WHZ)and BMI-for-age z-score (BAZ). As compared to the WHO 2006 growth standards, if children had WAZ, HAZ, and WHZ less than −2, then they were classified as underweight, stunting, and wasting respectively. 14 Calculation of wealth Index: Detailed description of arriving at the wealth index is given in the NFHS-5 report. The scoring is given based on the quantity and quality of goods that the househods own, that includes television, bicycle, car, and housing characteristics such as source of drinking water, toilet facilities, and flooring materials. Principal component analysis is then performed to derive the scores. Subsequently, wealth quintiles are compiled at national level by assigning the household score to each household member, ranking each person in the household population by their score, and then dividing the distribution into five equal categories, each with 20 percent of the population. 11 From the main dataset, households were identified which had pairs of children with at least one male and at least one female child. When there were more than 2 children within the household, the youngest pair was selected.
From the dataset, a total of 70304 households were drawn that had any children. Out of them, 18133 had a pair of children -1 male and 1 female. 926 pairs were drawn from the youngest 2 children in a household that had 3 children, 23 pairs from those households that had 4 children and 1 pair from a household that had 5 children (Figure -1).
STATA v14 was used for statistical analysis. A secondary dataset was generated by juxtaposing the details of the male sibling with that of the female sibling of the same household into one row so that the Z score differences of WAZ (dWAZ), HAZ (dHAZ), WHZ (dWHZ) and BAZ (dBAZ) could be calculated. All differences were calculated as the male value minus female value. So, a positive value in the difference would imply the male child being in a better nutritional status as compared to the female sibling and a negative value in the difference would imply the female child being in a better nutritional status as compared to the male sibling. Paired t-test was used to look for the statistical significance of the differences in mean Z scores between the genders and One way Analysis of variation (ANOVA) test was used for factors affecting these differences. A p-value <0.05 was considered statistically significant.

RESULTS
Sample Characteristics: The study sample had 19,083 pairs of siblings with a male-female combination. The sample pairs had relatively older female siblings -with nearly 24% of them above 48 months of age whereas the pairs with older male siblings had 19% of them above 48 months of age. Similarly, of the relatively younger male siblings amongst the pairs, nearly 44% of the boys and 35% of the girls were under 23 months of age. The mean age of the sample was 29.7 months, which was 27.9 months in the boys to 31.5 in the girls (Table 1).
Undernutrition in the sample: In all its forms, undernutrition was more prevalent among the male siblings than the female siblings. The difference was more pronounced in wasting, where the prevalence was 19% in boys as compared to 17% in girls. Overall, around 54% of the children had some form of the undernutrition (Table 1).    The differences in Z scores were more or less distributed equally on both sides of zero, which implies that the differences which were towards the female siblings were as much towards the male siblings. Majority of the dHAZ (72.5%), dWAZ (88%), dWHZ (79.1%) and dBAZ (78.2%) scores ranged between -2 to +2, meaning that distance between the Z scores of the siblings were upto 2 SDs with each other (Table 2).
The mean Z scores were low for both male and females of the sample. They were more negative for the HAZ and WAZ scores -which were less than -1.4 standard deviations (SD) for both genders -while the means of WHZ and BAZ scores were ranging between -0.67 to -0.83. The differences between the Z scores of the siblings were very small and less than 0.1 SD. An attempt was made to analyse the variations in these differences as a factor of age difference between the pair when considered as a categorical variable. This was carried out for two reasons -one, because the sample had older females, and two, since age has a bearing on the nutritional status (Table 3).
When the female child of the pair was younger, the HAZ score on an average was nearly 0.5 SDs better in the female child as compared to the male child, while this difference was nearly 0.2 SDs for the WAZ score. There was a very small difference in WHZ scores, while the BAZ scores were nearly 0.2 SDs better among the male children. When the male child among the pair was younger, the HAZ score was around 0.2 SDs better in them as compared to the female child, whereas the WHZ and BAZ scores were nearly 0.15 SDs better in the female child (Table 3).
There were around 350 pairs of twins among the samples, where the age factor also could be negated to see the differences in the Z scores between the genders. Among these pairs, the mean Z scores were better among the females, with the HAZ, WAZ, WHZ and BAZ score differences being nearly 0.2, 0.15, 0.1 and 0.05 SDs respectively (Table 3).

Difference in Z scores among the categories
The differences in HAZ, WAZ, WHZ and BAZ scores did not vary much among the various categories studied i.e., place of residence, caste, religion, wealth index and maternal education. The dHAZ (-0.182) and dWAZ (-0.11) of scheduled tribes varied from the other communities where the girl child had better nutrition status among the siblings. The dHAZ was found to be higher when the mothers had primary education. Female children had better HAZ scores than their male siblings in Muslim (-0.132) and Christian families (-0.137) as compared to those belonging to Hindu (-0.056) or other religions (-0.068). Male children had better WAZ, WHZ and BAZ compared to females when they belonged to religions other than Hinduism, Islam or Christianity. The HAZ scores were better among females as compared to their male siblings when they belonged to lower wealth quintiles and moving up the quintiles, this difference kept diminishing. The differences in BAZ, WAZ and WHZ did not show any such gradient ( Table 4).

DISCUSSION
The prevalence of undernutrition in under-5 years children in India, despite a phenomenal improvement in our economy, is a matter of concern. Existence of gender bias against the girl child casts a very strong burden on the society in achieving a zeromalnutrition state. However, presence of female disadvantage in the intra-household nutritional status has not been definitively reported so far. Tarozzi have reported gender neutral situation in terms of undernutrition and anaemia in children under-5 years of age, upon analysis of nationally representative data. 15 On the other hand, female disadvantage in feeding practices have been reported by several researchers, where infant girls have been fed for shorter period of times as compared to infant boys. 5,16,17 The present secondary analysis is an attempt at understanding the current situation of malnutrition in the parlance of gender differentials in malnutrition of more than 19,000 male-female sibling pairs drawn from NFHS-5 dataset. An intra-household comparison of nutritional status amongst the siblings has two major objectives-firstly, it negates the effects of socio-demographic environment on nutritional status, and secondly and more importantly, it brings out the possible cases of parental preferences of gender, and associated differences in the allocation of resources if any. The present analysis is probably the first such attempt at quantitative assessment of disparities in the nutritional status between the siblings from different genders of the same household.
The current secondary analysis has observed minimal differences in the nutritional statuses between the genders who hail from the household. The nutritional status of the sample drawn with the sibling criteria to include in the analysis was comparable to that of the full survey sample of NFHS-5. As in the full survey sample (NFHS-5), the prevalence of malnutrition was marginally higher among the males than females. The prevalence of undernutrition was similar to the levels reported by  The mean difference in all the Z scores were more or less close to zero, meaning no appreciable gender disadvantage. The reason for this can be explained by the distributions of these scores as in table 2. All the scores were distributed nearly symmetrically on the either side of zero.
The WHZ and WAZ scores of both males and females were similar to that reported in the NFHS-5 full sample, but the HAZ scores of this sub sample was more negative in both males as well as females. The dHAZ and dWAZ scores tilted slightly towards the male or the female depending on which of the siblings were younger. It is known that the Z scores tend to be less negative among younger children and the distance from normal curves deepens with age 18 , which explains the findings of the current study as well. The analysis was further extended to twins which negates even the age factor between the siblings. Though there were only 350 such pairs of twins for analysis, it had clearly shown that the z scores were better among females compared to males. Similar results have been observed earlier by Alderman et al 19  Socio-demographic factors like caste, religion, wealth quintile, place of residence and maternal education did not seem to influence these differences. However, relatively substantial differences were found among Scheduled tribes in dHAZ and dWAZ scores which showed a female advantage. HAZ among females was also found to be better than their male siblings in Muslim and Christian communities, and male children were better with the dBAZ scores when they were from religions other than the three major religions of India. We did not find such a difference being reported in literature earlier. An interesting observation was that there was a decreasing female advantage in HAZ scores as they moved up on the wealth quintiles. Or rather, it may be seen as a decreasing gender disparity with increasing position in the wealth quintiles. The Asian enigma was a concept put forth in the 90s which ascribed higher levels of malnutrition among women and children due to the discriminations that the women face especially in Southeast Asia. 21 The present study finds results contrasting to this concept.
Improved sex ratio at birth can provide a plausible explanation to this minimised gender disparity in the nutritional status of the siblings. As reported in the Sample Registration System (SRS) bulletin (2020), for every 1,000 boys born in the country in the period 2018-20, 907 girls were born, which is higher compared to earlier data in 2005 (876 girls per 1000 boys) and 2014-16 (898 girls per 1000 boys). 22 A strong and steady efforts on behalf of the governments to curb sex selection and pro-girl child measures such as 'sukanya samridhi' or 'beti bachao beti padhao' coupled with improved education and wealth index have also added to bridging the gender gap. 1,23 While this may not be substantial enough to reach the sustainable development goal no. 2, there seems to be a gradual shift pointing towards a demographic transition in the Indian family set up where children are becoming the focal point of the family. The presence of a large cohort of young women of childbearing age is also a contributory factor to the gender-neutral results.
While the current study provides a positive and hopeful outlook towards achieving a gender egalitarian society, there are also some limitations in the study. It included only one pair of siblings (the youngest pair) when there were multiple pairs within household, which could have affected the results. Also, scope of the study could be further expanded at pinpointing the particular age at which the differences were observed, in order to facilitate the identification of the causes.

CONCLUSION
There was little or no differences in the mean z scores of females and male siblings of the same household and the differences whatsoever observed were equal on both sides, without any gender advantages or disadvantages. Females tend to have a slightly better higher HAZ than their male siblings when they belonged to scheduled tribes, Muslim, Christian or poorer households. The Asian Enigma remains an enigma.

ACKNOWLEDGEMENT
The authors acknowledge the team of IIPS and other organisations which have conducted and generated the data for the present analysis. They also acknowledge the DHS program for granting access to the data.