Data
For this study, the data was used from the National Family Health Survey 2005–06 and 2015–16, which is a nationally representative cross-sectional survey conducted under the stewardship of the Ministry of Health and Family Welfare, Government of India [31]. The survey provides information on population, health, nutrition, and various demographic indicators at households as well as individual level. With the use of a stratified two-stage sampling procedure, NFHS gave the estimates for India as a whole, as well as each state and union territory level [31]. The detailed methodology, with complete information on the survey design and data collection, was published in the survey report [31]. The effective sample size for the present study was 14,422 and 74,132 children aged 6–23 months for 2005–06 and 2015–16, respectively.
Reliability and validity of data
NFHS data has maintained the quality of data and it’s reliability and validity. The NFHS-4 sample is a stratified two-stage sample. The 2011 census served as the sampling frame for the selection of PSUs (primary sampling units). PSUs were villages in rural areas and Census Enumeration Blocks (CEBs) in urban areas. Within each rural stratum, villages were selected from the sampling frame with probability proportional to size (PPS). In every selected rural and urban PSU, a complete household mapping and listing operation was conducted prior to the main survey [31]. Selected PSUs with an estimated number of at least 300 households were segmented into segments of approximately 100–150 households. Two of the segments were randomly selected for the survey using systematic sampling with probability proportional to segment size. Therefore, an NFHS-4 cluster is either a PSU or a segment of a PSU. In the second stage, in every selected rural and urban cluster, 22 households were randomly selected with systematic sampling [31].
Variable description
Outcome variable
The outcome variable was adequately diversified dietary intake (ADDI) among children aged 6–23 months in India. In the NFHS-4, mothers were asked to provide a 24-h recall of foods and food groups given to their children. Children who received foods from four or more of the following food groups were defined to receive minimum dietary diversity: juice; tinned powdered/fresh milk; formula milk; fortified baby food; soup/clear broth; other liquids; chicken, duck, or other birds; bread, noodles, other grains; potatoes, cassava, tubers; eggs; pumpkin, carrots, squash; dark green leafy vegetables; mangoes, papayas, Vit A fruits; any other fruits; liver, heart, other organ meat; fish, shellfish; beans, peas, or lentils; cheese, yogurt, other milk products; other solid/semi-solid food; any other meat; and yogurt.
These food items were categorized into seven food groups followed by the WHO IYCF guidelines [32]: (1) "dairy products" (comprised of formula milk OR tinned powdered/fresh milk; OR cheese, yogurt, other milk products OR yogurt); (2) "legumes and nuts" (comprised of beans, peas, or lentils); (3) " roots, grains, and tubers" (comprised of soup/clear broth OR bread, noodles, other grains OR fortified baby food OR potatoes, tubers, cassava,); (4) "flesh foods" (comprised of heart, liver, other organ meat OR shellfish, fish OR chicken, duck, or other birds); (5) vitamin A rich fruits and vegetables" (comprised of carrots, pumpkin, squash OR dark green leafy vegetables OR mangoes, papayas, Vitamin A fruits); (6) "eggs" (comprised of eggs) "; and (7) "other fruits and vegetables" (comprised of any other fruits). The child was considered to take ADDI if he consumed four or more of the seven groups [13].
Explanatory variable
The other explanatory variables were divided into three categories: 1. Mother's characteristics 2. Child characteristics, 3. Household characteristics.
Mother’s characteristics
The mother’s age was coded as 15–24, 25–34 and 35 + years. The mother's educational status was coded as not educated, primary, secondary and higher. Media exposure was coded as exposed if the respondent was either watching television or reading a newspaper or listening to the radio and otherwise not exposed.
Child characteristics
The child's age was coded as 6–11, 12–17 and 18–23 months. The sex of the child was coded as male and female. Birth order was coded as 1, 2, and 3 + .
Household characteristics
The main explanatory variable was wealth status which represents the socio-economic status of a particular household. The variable wealth status was created using the information given in the NFHS survey. Scores were assigned to households based on the amount and types of consumer items they own, which range from a television to a vehicle or bicycle, as well as home features such as bathroom facilities, drinking water supply, and flooring materials. Principal component analysis was used to calculate these scores (PCA). The national wealth quintiles are calculated by assigning a score to each typical (de jure) household member, rating each individual in the household population according to their score, and then dividing the distribution into five equal groups, each having 20% of the population. The wealth index was categorized as poorest, poorer, middle, richer and richest [31]. Religion was coded as Hindu, Muslim, Christian and others. Other’s included Sikh, Buddhist, and Jain etc. Caste was coded as Scheduled Tribe, Scheduled Caste, Other Backward Class and others [33]. The Scheduled Caste includes a group of people who are socially separated and financially/economically disadvantaged as a result of their low caste position in Hindu society. The Scheduled Castes (SCs) and Scheduled Tribes (STs) are among India’s most economically disadvantaged groups. The OBC are people who have been labelled as "educationally, economically, and socially backwards." The OBCs are considered lower castes in the old caste system, although they are not untouchables [34]. The place of residence was coded as urban and rural. Regions of India were coded as North, Central, East, North-East, West and South.
Statistical analysis
Descriptive analysis, along with bivariate analysis, was used to find the differences in ADDI by background variables in 2005–06 and 2015–16. For the comparison of the prevalence of ADDI from 2005–06 to 2015–16, the study used a proportion test. The test of the normality and multicollinearity was conducted as pre-regression analysis. The model fit indices were tested using chi-square which was p < 0.001. STATA 14 was used for analysis.
The test statistic for comparing two proportions is defined as:
$$Z=\frac{{\widehat p}_1-{\widehat p}_2}{\sqrt{p\dot\ast\left(1-p\dot\ast\right)\left(\frac1{n_1}+\frac1{n_2}\right)}}\sim N\left(0,1\right)$$
where \(p\dot{*}=\frac{{n}_{1}{\widehat{p}}_{1}+{n}_{2}{p}_{2}}{{n}_{1}+{n}_{2}}\); \({\widehat{p}}_{1}\) and \({\widehat{p}}_{2}\) are respectively the proportions of ADDI in the two periods (2005–06 and 2015–16). Similarly, \({n}_{1}\) and \({n}_{2}\) are the respective sample sizes in the two rounds of surveys [35].
Further, logistic regression analysis [36] was used to carve out the significant factors contributing to ADDI among children aged 6–23 months in India. To identify the underlying causes of the decadal difference in the prevalence of ADDI, the technique of decomposition had been used, which is now a day the most common approach used to identify and quantify inter-group differences. That is to compute the group difference (2005–06 to 2015–16) in the prevalence of ADDI among children and to decompose these differences into the major contributing factors, Fairlie's decomposition method [37, 38]. The method is commonly attributed to Blinder (1973) and Oaxaca (1973) [38]. This technique, however, is not appropriate if the outcome variable is dichotomous, such as ADDI, which is coded 0 "no" and 1 "yes". Hence, we used the extension of the Blinder-Oaxaca technique that is Fairlie decomposition which is appropriate for binary models to decompose the decadal change in the prevalence of ADDI into contributions that can be attributed to different factors [37, 38].
$${Y}^{t1}-{Y}^{t2}=\left[\sum_{i=1}^{{N}^{t1}}\frac{F\left({X}_{i}^{t1}{\beta }^{t2}\right)}{{N}^{t1}}-\sum_{i=1}^{{N}^{t2}}\frac{F\left({X}_{i}^{t2}{\beta }^{t2}\right)}{{N}^{t2}}\right]+\left[\sum_{i=1}^{{N}^{t1}}\frac{F\left({X}_{i}^{t1}{\beta }^{t1}\right)}{{N}^{t1}}-\sum_{i=1}^{{N}^{t1}}\frac{F\left({X}_{i}^{t1}{\beta }^{t2}\right)}{{N}^{t1}}\right]$$
where Y is the dependent variable (ADDI) at time t1 (2005–06) and t2 (2015–16), \({N}^{J}\) is the sample size for time t, \({X}^{J}\) is the row vector of average values of the independent variable, and \({B}^{J}\) is the vector of coefficient estimates for time t. This method of decomposition allows us to measure the absolute contribution of factors explaining the decadal variation (2005–06 to 2015–16) in the probability of ADDI among children aged 6–23 months in India. Stata14 [39] was used to carry out the analysis. The authors used svyset in Stata 14 command to control the analysis for complex survey design of National Family Health Survey. Additionally, survey weights were used to make the estimates nationally representative.
Patient and public involvement
No patient involved.