Data
This study uses data from two rounds of the National Family Health Survey (NFHS) conducted in 1992–93 (NFHS-1) and 2005–06 (NFHS-3). Data from the NFHS-2 were not considered in the analysis as significant changes in EBF practices were not expected to have occurred in the five to six years between successive surveys.
The NFHS is a nationally representative cross-sectional study using a multistage cluster sampling design with internationally validated instruments. The survey is conducted every five years under the stewardship of the Ministry of Health and Family Welfare, Government of India and coordinated by the International Institute of Population Sciences, Mumbai. The key aim of NFHS is to provide national level data on key indicators like fertility, family planning, infant and child morbidity and mortality, maternal and reproductive health and nutritional status of mothers and infants.
In each of the NFHS, all currently married eligible women were asked to provide information on complete birth history including sex, month and year of birth and survival status for each live birth. Detailed information was obtained on antenatal and delivery care, breastfeeding duration and practices, vaccination and recent illnesses for the recent births which had occurred for each eligible woman during the four years preceding the survey in NFHS-1and five years preceding the survey in NFHS-3. The area covered by each of the surveys accounted for 99 per cent of the country’s population and the survey response rate ranged between 89–100 per cent across different states.
Sampling design, sample size and response rate details are published in the round-specific survey reports [20, 21]. All study procedures, consent forms and tools were approved by the Ethics Review Board of the International Institute for Population Sciences, Mumbai, India (IIPS 1992–93, IIPS 2005–06).
For the purpose of the present study, only 19 bigger states (18 states and the National Capital region of Delhi) excluding the six north-eastern states were considered for the final analysis. In these 19 States, a total of 83,511 and 102,572 eligible women respectively in NFHS-1 and NFHS-3 were interviewed. Our window of observation for analysis in this research is three years of life in both the surveys, i.e., NFHS-1 and NFHS-3, based on retrospective data of children who were aged 0–35 months at the time of survey. The analysis would therefore refer approximately to the period 1990–1993 for NFHS-1 and 2003–2006 for NFHS-3.
The aforesaid 19 states of India were further grouped into high, moderate and low post-neonatal mortality regions in view of their association with the socio-economic differentials and breastfeeding practices. Sample Registration System (SRS) figures on post neonatal mortality rate (PNMR) for the period 1990–92 [22] were used to group the states/provinces in three groups: states with high level of PNMR ≥ 27/1000 live births (Odisha, Uttar Pradesh, Bihar, Assam, Madhya Pradesh, Rajasthan, Haryana); states with medium level of PNMR ≥ 20/1000 live births (West Bengal, Andhra Pradesh, Gujarat, Karnataka, Punjab, New Delhi) and states with low level of PNMR < 19/1000 live births(Tamil Nadu, Himachal Pradesh, Maharashtra, Jammu, Goa, Kerala).
Study variable
In the present study we used exclusive breastfeeding (EBF) among infants aged < 6 months as the outcome variable. This was defined, based on the WHO key infant feeding indicators and the guide to DHS statistics [23, 24] as the infants 0–5 months of age who received only breastmilk in the previous 24 h. From NFHS-1, information on duration of exclusive breastfeeding of currently breastfeeding children and of those who had died was included while data of only currently breastfeeding children was included from NFHS-3. The status of exclusive breastfeeding was ascertained at each month. Since the ages when the child started receiving plain water and other liquid and solid foods were recorded in completed months, for analysis purposes, we have added 0.5 months in computation of present age, or ages when the child started receiving water or other food supplements.
Independent variables
The independent variables included the demographic and socio-economic characteristics of mothers and antenatal and natal care. In order to see the effect of mother’s age at the time of child birth on her breastfeeding practice, it was categorised as woman under 20 years, 20 – 34 years, or 35 years or more. Mother’s place of residence was categorized into rural and urban. For the work status of mother, those engaged in household work or working without wages, were categorized as non-gainful and the rest were considered in gainful occupation. The economic status of the household was measured by a composite index called standard of living index (SLI) by summing up the scores based on different facilities and items present in the household such as type of the house, source of lighting and ownership of various consumer durables [25] and were categorized into three groups as low, medium and high SLI.
To study the variation in the exclusive breastfeeding pattern in the context of educational attainment of mother, we considered three categories: (1) illiteratemothers and those who do not have any formal schooling, (2) mothers with 1–7 years of schooling and (3) mothers with 8 or more years of schooling. This grouping corresponds to the educational levels as illiterate, primary and middle school, and high school and above respectively. Thelength of the preceding birth interval was consideredintwocategoriesandwas child of first birth order or preceding birth interval < 24 months and the other if the preceding birth interval ≥ 24 months. The effect of the size of child at birth as perceived by mother was also considered under the two categories of large/average and small.
In the absence of direct information on use of health care services, proxy measures such as number of antenatal appointments, number of tetanus toxoid injections (TT) received by mother and place of delivery were used as indicators of access to use of health care facilities. A ‘variable antenatal and natal care’ was created. If mother received at least three antenatal appointments or was vaccinated against tetanus (received two doses of TT immunization) during pregnancy or delivered the child in a health facility then antenatal and natal care received by the mother was coded as ‘Yes’, otherwise, coded as ‘No’.
In sequence of above independent variables, children whose mothers were in the age group 20–34 years at last childbirth, from rural area, illiterate, with non-gainful occupation, from low SLI household, having first birth/preceding birth interval ≥ 2 years, not availing antenatal or natal care and having large/average size of the baby, were considered as reference categories of predictor variables.
Statistical model
The rate of EBF was reported at age one, four and six months to identify changes that may occur according to the age of the infant during the surveys. This was examined against various aforesaid demographic, socioeconomic and service related characterizes of the mothers using univariate and bivariate analysis. As the study variable (type of breastfeeding: exclusively breastfeeding, breastfeeding and water only, breastfeeding with supplements and not breastfeeding) had more than two categories, the associations between exclusive breastfeeding and the aforesaid predictors (those found significant in bivariate analysis) were examined using multinomial logistic regression model controlling for potential confounders [26, 27].
Utilizing the data on the EBF status of children at different ages of the infant in NFHS-1 and at the time of survey in NFHS-3 the effects of predictor variables on EBF at the beginning of 1, 4, 6 months were estimated separately. For each variable, the effects of the other variables included in the model were controlled by setting them at their mean values. Three multinomial logistic regressions in conjunction with multiple classification analysis were estimated, one for each of two groups of states (high and medium/low PNMR) and one for all states combined. This analysis accounted for study design and sample weight as the raw data from NFHS provides sampling weights. The statistical analyses were conducted using Stata Version 13 with SVY commands that adjusted for the complex cluster sampling design used in the survey. All tests were two-sided and p < 0.05 was considered statistically significant. An unweighted total of 34,176 and 25,459 births under three years of age in the NFHS-1 and NFHS-3 respectively were assessed for the WHO defined EBF indicator and there were 2278 and 1896 mother-infant pairs (less than six months) in the 1992 and 2005 NFHS survey respectively.