This study was a secondary analysis of the National Family Health Survey (NFHS-4) conducted in 2015–2016. The NFHS is a nationally representative cross-sectional survey using a multistage cluster random sampling. This survey is conducted amongst households, children (age 0–5), women (age 15–49), and men (age 15–54). The main aim of the survey is to collect information on fertility, contraception, infant mortality, maternal and child health, and nutritional assessment at the national level.

The NFHS-4 included 640 districts in India as per the 2011 Census. A total of 572,000 households,  were selected through a complete mapping of both urban and rural areas. The urban sample was selected through a two-stage sampling, with Census Enumeration Blocks (CEB) being selected in the first stage after which a random selection of 22 households in each CEB are selected. The rural sample was also selected through a two-stage sampling. The villages were taken as the Primary Sampling Units (PSU) in the initial stage, after which 22 households were randomly selected in each PSU. The NFHS-4 has collected data on children below five years. After applying the sample weight, 21,145 children aged below six months were there in the data [11].

Computer-Assisted Personal Interview was the data collection mode. The NFHS-4 fieldwork for India was conducted from 20 January 2015 to 4 December 2016 by 14 Field Agencies and gathered information from 601,509 households, 699,686 women, and 112,122 men. Four types of tools were used in the survey to collect information. It includes the household questionnaire, women’s questionnaire, men’s questionnaire, and biomarker assessment. Questionnaires are available in 19 languages using Computer Assisted Personal Interviewing (CAPI) [11].

Exclusive breastfeeding among children was the outcome variable. Infants less than six months who received only breastmilk except for medicines were regarded as exclusively breastfed [20, 21]. This indicator was assessed based on the youngest child living with the mother and the feeding practice within 24 h before the day of data collection.

The population was the youngest living children aged 0–5 months, living with their mothers (women aged 15–49 years).

As per the DHS coding manual, prevalence of EBF was measured as the proportion of infants aged 0–5 months who received breastmilk as the only source of nourishment (but also received prescribed oral rehydration solution, medicines and vitamin syrups or drops) in the last 24 h before the survey based on mothers’ recall [12, 22]. This measurement was consistent with the WHO/UNICEF guidelines for assessing IYCF practices [23]. Missing data on breastfeeding is treated as not currently breastfeeding in the numerator and included in the denominator [22]. Missing and “don’t know” data on foods and liquids given is treated as not given in the numerator and included in the denominator [22].

In this study, four predictor variables (religion, household wealth, maternal education, and place of residence) which were associated with exclusive breastfeeding as per the existing literature and were more likely to determine the social position of the child were combined to generate intersecting categories which was the main predictor variable. Three other variables (maternal age, caste, and gender of child) served as confounders. Firstly, the selected variables for creating intersecting categories were converted into dichotomous variables as follows.

Religion: Hindu – Hindu & (Muslim, Christian, Buddhist, Sikh, Jain, and others) – Others.

Wealth index: Poorest, poorer, and middle - poor & richer and richest – rich.

Education of mother: Illiterate and primary education – primary & secondary and higher education – secondary.

Then we combined these variables and the place of residence (rural/urban) to create intersecting social categories. There were 16 intersecting categories, as depicted in Table 1.

The data were analyzed using the SPSS Version 27 (univariate and multivariate analysis) and Stata Version 17 MP-Parallel Edition (slope inequality index and graphs). Prior to carrying out the univariate and multivariate analyses, the appropriate sample weights were applied to the data. Descriptive statistics (frequency and percentage) were used to explore the distribution of the sociodemographic factors and EBF prevalence in the population. Binary logistic regression was used to assess the relationship between the intersecting categories and EBF prevalence. The model was subsequently adjusted for maternal age, caste, and sex of the child. Odds ratio with a p - value < 0.05 at a 95% confidence interval was considered statistically significant. Model fitness was assessed using the Pearson goodness of fit test.

Inequalities in EBF prevalence across the intersecting groups was assessed using absolute and relative indices. The absolute measure of inequality was derived by computing the difference in EBF prevalence within the most advantaged group and disadvantaged group. The relative disparity in exclusive breastfeeding prevalence was also computed by dividing EBF prevalence among the most advantaged group and disadvantaged group. Slope and relative inequality indices were also obtained via a logistic regression model using the siilogit command [24].

The National Family Health Survey-4 (Demographic and Health Survey) datasets were available in the public domain, after removing the identifying information. To conduct the study, the first author obtained the Ethics clearance exemption from the Institutional Human Ethics Committee (IHEC), Central University of Kerala (Approval number: IHEC/CUK/2021/15).