1. IntroductionSub-optimal breastfeeding is estimated to be responsible for 1.4 million child deaths and 43.5 million Disability Adjusted Life Years (DALYs) with non-exclusive breastfeeding accounting for 77 percent (1.06 million) of the deaths and 85 percent of the DALYs [1] . About one-fourth to one-half of all infant deaths in developing countries occur in the first week of life. Breastfeeding helps prevent hypothermia and hypoglycaemia in newborn babies, which are contributory causes of early neonatal deaths especially among low birth weight and premature babies. Immediate breastfeeding within the first hour, followed by early exclusive breastfeeding, improves the health and survival status of newborns. During the late neonatal period, most deaths in developing countries are due to infections such as sepsis, acute respiratory tract infection, meningitis, omphalitis and diarrhoea. Feeding colostrum and breastfeeding, especially exclusive breastfeeding, protect against such deaths. Among 23 interventions considered by the Bellagio Child Survival Study Group, scale-up of exclusive breast-feeding of infants for 6 months and continued feeding until 1 year could prevent an estimated 1.3 million child deaths per year [2] .
Breastfeeding provides young infants with the nutrients they need for healthy growth and development. Apart from being a natural source of nourishment, human milk contains glycans and secretory immunoglobulin, which offer protection against infectious disease [3] [4] reducing the morbidity and mortality due to infectious diseases in childhood [5] . During the first six months of life, when digestive systems are not yet mature, exclusive breastfeeding offers additional protection from illness by limiting exposure to contaminated foods and liquids. This is particularly important in poor environments where early introduction of formula/animal milk is of particular concern because of the risk of pathogens, contamination and over dilution of milk leading to increased risks of morbidity and under-nutrition. A pooled analysis of studies carried out in middle/low income countries showed that breastfeeding substantially lowered the risk of death from infectious diseases in the first two years of life [6] and optimal breastfeeding practices could prevent a substantial proportion of hospital admissions due to diarrhea and lower respiratory tract infection [7] . A systematic review by Kramer et al. confirmed that exclusive breastfeeding in the first 6 months decreased morbidity from gastrointestinal and allergic diseases, without any negative effects on growth [8] . A pooled analysis of data from 3 countries indicated that either predominately or exclusively breastfed infants were at substantially lower risk for infant mortality than non-breastfed infants [9] . Over the past decades, available evidence for the health advantages of breastfeeding has continued to increase and these health benefits are seen to extend into adulthood [10] . Breastfeeding has protective roles against obesity, hypertension, dyslipidemia, and type II diabetes mellitus during adulthood that would have long-term beneficial health effects at individual and population levels [11] . Given such evidences, the WHO has recommended that every child should be exclusively breastfed for the first six months of life with partial breastfeeding continued until two years of age [12] .
India is committed to achieving its National and Millennium Development Goal 4 of reduction in infant mortality. Recognition of breastfeeding as a core component of infant survival, growth and development has led to efforts that promote sound infant and young child nutrition. Guidelines towards this have been given by the Government of India [13] and initiatives to promote breastfeeding have been implemented through the national health programs. In India there are considerable differences in breastfeeding practices, access and utilization of health services and health outcomes across states/provinces and socio-economic groups [14] - [16] . Many studies have been carried out on determinants of infant mortality but very few have examined the trend and relationship between breastfeeding and infant death utilizing survey data of specific population [17] - [19] . This information would be important for evaluating the success of ongoing programs and various other activities that promote breastfeeding. This paper attempts to study the effect of breastfeeding duration on infant survival along with other demographic, socioeconomic and service related factors in India. It is based on the National Family Health Survey (NFHS) conducted with an aim to help the planners and policy makers for informed decision making, besides keeping a track of the health trends in the country. Furthermore, overall development and improved programme coverage in India have led to improvements in child survival. Hence, it may be desirable to examine changes related to breastfeeding and infant mortality over time. The study adopts a multivariate hazard model to analyze the determinants of infant mortality with duration of breastfeeding as a time dependent covariate using two rounds of the National Family Health Survey as well as by pooling the data to capture differences between surveys.
2. Materials and Methods
2.1. Data SourceIndia’s National Family Health Survey (NFHS) provides state and national level information on fertility, family planning, infant and child morbidity and mortality, maternal and reproductive health, nutritional status of women and children, and on the quality of health services. The survey is similar to Demographic Health Survey (DHS) in other countries. It has adopted multistage stratified sampling methods to create a sample representing individuals from all 29 Indian states. So far three rounds of NFHS have been conducted in the year 1992-1993 (NFHS-1), 1998-1999 (NFHS-2) and 2005-2006 (NFHS-3). The present study uses data from the NFHS-1 and NFHS-3 keeping pace for the changes in the study variable as well as explanatory factors.
To make the breastfeeding estimates representative and comparable across the two survey rounds, we used appropriate sample weights in the analysis. Sampling design, sample size and response rate details are published in the round-specific survey reports [20] [21] . All eligible women aged 15 - 49 years were asked to provide information on complete birth history including sex, month and year of birth, and survival status for each live birth. The information on age at death was recorded in days for children who had died in the first month of life; in months for children who had died after the first month but before completion of their second birthday, and in years for children who had died at later ages. For children who had died after their second birthday, age at death in months was computed. Detailed information on antenatal, delivery and postnatal care for births which occurred to eligible women four and five years preceding the survey in NFHS-1 and NFHS-3 respectively were collected. The present study was restricted to all live births that occurred three years preceding survey for analysis of infant mortality. For the purpose of this study, the north-eastern states were excluded and only 18 states and the National capital region of Delhi were considered for the analysis. A total of 36754 births in NFHS-1 and 26782 births in NFHS-3 were included in the analysis.
2.2. Variable DefinitionDependent variable: Dependent variable is infant mortality which is defined as the probability of dying before the first birthday.
Independent variables: The following socio-economic, demographic and health services access variables were considered: breastfeeding (no/yes); place of residence (rural/urban); mother’s age at the time of child birth (15 - 19, 20 - 34, 35+ years); mothers education (illiterate, literate but below middle class, middle class and above); mother’s occupation (non gainful occupation, gainful occupation); preceding birth interval (<2 years, ≥2 years); antenatal and natal care (no/yes); size of the child at birth (small, medium/large), standard of living index based on ownership of household goods (low, medium/high). In the absence of direct information on use of health care services proxy measures such as number of antenatal check-ups, 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 check-ups during pregnancy or received two doses of TT immunization 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”. States were stratified on the basis of post neonatal mortality rate (low mortality states, medium/high mortality states) as observed in Sample Registration System (SRS) [22] .
2.3. Statistical AnalysisExtended Cox regression model has been used to examine the effect of socio-economic, demographic and health service access factors on risk of infant mortality. An extended Cox Regression model, which allows us to specify time-dependent covariates, can be used in situations where the proportional hazards assumption of the Cox Regression model does not hold [23] , i.e., hazard ratios change across time; the values of one (or more) of the covariates are different at different time points. The general hazard ratio formula for extended Cox model is shown below:
![]()
where,
are fixed variables and are time-dependent variables, and β and δ denote vector of coefficients of the respective explanatory variables.
For this model, we have defined dependent variable as time (age of the child) and survival status of child (1 if event is occurred, i.e., child died and 0 if child is alive) during period of infant (<12 months). Results of the multivariate analyses are presented as hazard ratios (risk of dying). The hazard ratio (HR) is the relative increase or decrease in the risk of dying associated with the specific category of predictor variable when compared to the reference category for that variable. If hazard ratio > 1 for a predictor variable, it means that the hazard is greater, i.e., increased risk of death and if hazard ratio < 1, implies a decreased risk of death.
The statistical analysis was conducted using SPSS version 19 using complex survey design. All tests were two-sided and p < 0.05 was considered statistically significant.
2.4. Ethical ConsiderationsAll study procedures, consent forms and tools were approved by the International Institute of Population Science Ethics Review Board (IIPS, 1992-1993, IIPS 2005-2006).
3. ResultsUnadjusted hazard ratio depicting associations between infant mortality and socio-economic and demographic factors are presented in Table 1. Breastfeeding, place of residence, mothers age at birth, education, mother in gainful occupation, birth interval, antenatal/natal care, size of the baby and standard of living Index and stratification of states by level of PNMR (post neonatal mortality rate) were found to be significantly associated with infant death during 1992-1993 (NFHS-1), 2005-2006 (NFHS-3) as well as during 1992-2006.
The unadjusted hazard ratio for infant deaths in NFHS-1 (1992-1993) divulged that the risk of infant death was 97 percent [HR = 0.030, 95% CI = 0.028 - 0.033] less among those children who were breastfed as compared to those who were not breastfed. In contrast to rural area, babies belonging to urban area had a 34 percent [95% CI = 0.612 - 0.707] less hazard of death during the infancy period. Hazard of infant death was higher among births to mothers aged 15 - 19 years (HR = 1.41) and aged 35+ (HR = 1.47) as compared with births to the reference mother’s age at child birth of 20 - 34 years. Children born to mothers having less than 8 years of education or 8 years complete and above had 27 percent and 52 percent less hazard of infant death than children born to illiterate women respectively. Babies born to mothers who were in gainful occupation had a 6 percent less hazard of infant death than babies born to mothers who were not gainfully occupied. The hazard of infant death was 36 percent higher when the birth interval was less than 24 months as compared to a longer birth interval of 24 months or more. Children’ mothers who received antenatal/natal care, the hazard of infant mortality was estimated at 0.49 [95% CI = 0.461 - 0.528] which were substantially lower than that seen in women who did not receive it. Small size of baby at birth had an increased infant death risk of 2.5 times when compared with a large or medium size baby [p value = 0.000]. The risk of infant dying was 35 percent less among those with a high standard of living index. A higher hazard of infant mortality was observed in babies born to women belonging to medium PNMR (48%) and high PNMR (2 times) states in contrast to those born to women belonging to states with low PNMR.
The unadjusted hazard ratio for infant deaths in NFHS-3 (2005-2006) showed that the likelihood of infant death was lower among the children who were breastfed as compared to those who were not breastfeed [HR = 0.023; 95% CI = 0.021 - 0.026]. The risk of infant dying was 33 percent less among urban children than the rural children [p value = 0.00]. Babies born to mothers aged 15 - 19 years and aged 35+ were 46 percent [95% CI = 1.327 - 1.595] and 28 percent [95% CI = 1.056 - 1.552] higher hazard of infant death than babies born to mothers 20 - 34 years of age respectively. Children with mothers having 8 years or more than 8 years of education had an decreases infant death risk of 48 percent when compared with children with illiterate mothers [p value = 0.000]. However children with mother having less than 8 years of education were 14 percent [95% CI = 0.778 - 0.948]