Trends and predictors of optimal breastfeeding among children 0–23 months, South Asia: Analysis of national survey data

Abstract Optimal breastfeeding practices, including early initiation of breastfeeding (EIBF) within 1 hr of birth, exclusive breastfeeding (EBF) for the first 6 months of age, and continued breastfeeding (CBF) for 2 years of age or beyond with appropriate complementary foods, are essential for child survival, growth, and development. Breastfeeding norms differ within and between countries in South Asia, and evidence is needed to inform actions to protect, promote, and support optimal practices. This study examines time trends and predictors of EIBF, avoidance of prelacteal feeding (APF), EBF, and CBF to 2 years using survey data from Afghanistan, Bangladesh, India, Nepal, and Pakistan since 1990. EIBF, APF, and EBF increased in Bangladesh, India, and Nepal from 1990 to 2016. EIBF and EBF increased in Pakistan from 1990 to 2013, but both EIBF and APF decreased in recent years. In Afghanistan, EIBF, APF, and EBF decreased from 2010 to 2015. CBF remained fairly constant across the region although prevalence varied by country. Significant (p < 0.05) predictors of suboptimal practices included caesarian delivery (4–25%), home delivery, small size at birth, and low women's empowerment. Wealth, ethnic group, and caste had varied associations with breastfeeding. Progress towards optimal breastfeeding practices is uneven across the region and is of particular concern in Afghanistan and Pakistan. There are some common predictors of breastfeeding practices across the region, however country‐specific predictors also exist. Policies, programs, and research should focus on improving breastfeeding in the context of women's low empowerment and strategies to support breastfeeding of infants born small or by caesarian section, in addition to country‐specific actions.

the first 3 days of life also promotes optimal breastfeeding practices.
The benefits of breastfeeding for children up to 2 years of age include lower child morbidity and mortality, higher intelligence scores and academic performance, and lower risk of ovarian and breast cancers among mothers . Breastfeeding could save more than 800,000 child lives and add more than $300 billion to the global economy if scaled up to near universal levels Victora et al., 2016).
South Asia is home to 26% of the world's children under 5 years of age (United Nations Children's Fund, 2016). Although the region is making progress on child nutrition, including infant feeding, there is much heterogeneity among and within countries in the region with regard to breastfeeding practices. In order to inform policies and programs to reach the global target of the World Health Assembly on EBF by 2025 and improve other breastfeeding practices, a deeper understanding of this heterogeneity is needed (World Health Organization, 2014).
Breastfeeding is influenced by multiple environments, that is, individual, family/household, community, workplace, health systems, and policy United Nations Children's Fund, 2016).
Many researchers and practitioners have used a social-ecological perspective to describe how individual breastfeeding behaviours are subject to environmental factors, such as family support, community norms, and health system policies (Bronfenbrenner, 1979). This study describes the epidemiology of optimal breastfeeding between 1990 and 2016 in five countries in South Asia, which account for 99% of the region's population: Afghanistan, Bangladesh, India, Nepal, and Pakistan. The study describes trends in breastfeeding practices in South Asia since 1990 and uses a socioecological approach to identify factors associated with optimal practices in each country.
We discuss the policy and program implications of the findings for these countries. All surveys collected data from women 15-49 years of age.
Breastfeeding data were collected from women with a live birth in the 2 years preceding the interview date. We restricted our sample to include only the last-born children living with their mother at the time of the interview. All surveys used a modified 24-hr recall of infant feeding practices. Further information about national survey methodology and sampling procedures are available in the respective DHS, NFHS, and MICS reports.

| Outcomes
The following breastfeeding outcomes were examined: EIBF, defined as the proportion of children born in the last 24 months who were put to the breast within 1 hr of birth; APF, the proportion of children aged 0-23 months who did not receive any food/drink other than breast milk during the first 3 days following delivery; EBF, the proportion of infants aged 0-5 months who received only breast milk, during the previous day; and CBF at 2 years, the proportion of children aged 20-23 months who received breast milk during the previous day (World Health Organization, 2008).

Key messages
• The generally positive trends in EIBF, APF, and EBF in Bangladesh, India, and Nepal are encouraging, but progress towards improved practices is regressing in Pakistan and Afghanistan.
• Caesarean delivery is a common predictor of suboptimal EIBF across all countries. Home delivery, small size at birth, and low women's empowerment were also associated with suboptimal breastfeeding.
• Creating a supportive breastfeeding environment can improve breastfeeding practices in South Asia. Action is required to address the child-, maternal-, and household-level disparities in breastfeeding practices.

| Data analyses
Breastfeeding trends analyses used data from two surveys in Afghanistan, seven surveys in Bangladesh, four surveys in India, five surveys in Nepal, and three surveys in Pakistan (Table 1).
Breastfeeding outcomes (EIBF, APF, EBF, CBF) were expressed as dichotomous variables. All statistical analyses were conducted using Stata 14.0 (StataCorp. 2015. Stata Statistical Software: Release 14. College Station, TX: StataCorp LP) with the svy command to allow for the cluster sampling design of the surveys. Probability weights were applied to achieve nationally representative samples, and all sample sizes and proportions reported are based on these weights.
The proportion of women reporting the breastfeeding practices was calculated separately for each country and for each survey round. To account for differences across surveys, nonoverlapping confidence intervals (CIs) were used to identify statistically significant differences.
However, this approach is not conclusive for overlapping CIs, and significant differences may be underestimated (Knezevic, 2008). Values presented for the trends analyses are proportions with 95% CI.
Child-, maternal-, and household-level variables were regressed on EIBF, APF, EBF, and CBF at 2 years in separate models for each outcome were restricted to complete cases. Regression models were constructed using a manual stepwise backwards elimination approach to identify factors associated with the outcomes. All variables 2 were entered into the initial model, and independent variables were eliminated in stages beginning with the child-level, maternal-level, and household-level last. Variables were retained in the model based on conceptual relevance and p value <0.15 to ensure we kept any potential meaningful confounders (Bursac, Gauss, Williams, & Hosmer, 2008). We assessed multicollinearity   (Table S1-S4).

| Ethics
More information about the survey designs and data collection methods are available in the survey reports. Permission to use the data was obtained from ICF International (Rockville, Maryland) and UNICEF (New York, New York).
EIBF increased between the first and last survey in all countries except Afghanistan. Between 1990Between -1994Between and 2014Between -2016 Figure S1).  Table 2 summarizes select characteristics of the analytic sample across the five countries. Between 21% and 28% of children were under 6 months of age, the majority (≥99%) of children were singleton births, 12-25% of children were perceived to be smaller than average at birth, and most children did not receive a postnatal check-up within 2 days of birth. Among maternal-level variables, 5-25% of women were adolescents, facility births ranged from 40% to 85%, caesarean section deliveries ranged from 4% to 25%, and 21-55% of women had four or more ANC visits. Household wealth quintiles were distributed between richest (14-20%) and poorest (17-23%).

Continued breastfeeding at 2 years
Few variables were associated with CBF across countries (Table 3b).
Lower decision-making autonomy in Pakistan was a significant barrier to CBF at 2 years. Facilitators to CBF at 2 years included maternal underweight in India, maternal primary education in Pakistan, and      Gender role attitudes Nonconforming             poorer wealth quintiles in India and Pakistan. There were no significant associations with CBF at 2 years in Afghanistan, Bangladesh, or Nepal.

| DISCUSSION
This study examined trends and predictors of EIBF, APF, EBF, and CBF in five countries in South Asia. The evidence shows positive trends for EIBF, APF, and EBF in Bangladesh, India, and Nepal, but Afghanistan and Pakistan lag behind in almost all breastfeeding practices. Predictors of breastfeeding practices varied across countries; however, cesarean delivery and child age were common predictors for EIBF and EBF, respectively. Taken together, the findings summarize progress and identify predictors of EIBF, APF, EBF, and CBF for each country in the region and have implications for policy and programs.

| Early initiation of breastfeeding
Despite steady improvements in EIBF in Bangladesh, India, and Nepal, recent declines in Afghanistan and Pakistan are a concern. Lack of health system support for EIBF and security problems in much of Afghanistan and parts of Pakistan may hinder implementation of interventions designed to support EIBF. Total EIBF prevalence for Afghanistan, Bangladesh, India, and Nepal are comparable with the global average (46%; United Nations Children's Fund, 2016). However, accelerating progress to support child health and achieve global targets is required in all countries.
Among child-level factors, women's perception of small birth size was identified as a risk factor for delayed initiation in Afghanistan, India, and Nepal. Perceptions of small birth size could reflect low birth weight, which has been associated with poor EIBF (Pollitt, Gilmore, & Valcarcel, 1978;Sundaram et al., 2013). Actions to prevent low birth weight, which is prevalent in the region (United Nations Children's Fund, 2014), and to support the establishment of breastfeeding in low birth weight infants could increase EIBF. Postnatal check-up, a proxy for the quality of health services in our analysis, was positively associated with EIBF in Bangladesh and India but negatively associated in Afghanistan and Pakistan. The inconsistent findings suggest that the provision of breastfeeding promotion and support by health providers may vary in coverage or quality across countries.
Caesarean delivery was associated with delayed initiation of breastfeeding in all countries. Additionally, health facility delivery was associated with delayed breastfeeding in Bangladesh but not in Nepal and India, and birth assistance from traditional birth attendants was associated with delayed breastfeeding in Nepal. Whereas in India, four or more ANC visits predicted EIBF. The results highlight the role of health workers and health facilities in supporting EIBF, and as cesarean deliveries rise in South Asia (Betran et al., 2016), it underscores the importance of ensuring policies such as Baby Friendly Hospital Initiative support EIBF in all infants, including those born by caesarian section. In Pakistan, overweight and obesity were risk factors for delayed initiation.
As overweight and obesity rise in the South Asia region, more countries will need to address breastfeeding support for this population (NCD Risk Factor Collaboration, 2017). Low decision-making autonomy was another factor associated with suboptimal EIBF in India and Pakistan, whereas higher education was a predictor of EIBF. The results are similar to a previous findings showing women's decision-making autonomy is a predictor of EIBF in South Asia but not other breastfeeding practices (Smith, Ramakrishnan, Ndiaye, Haddad, & Matrorell, 2003).
Improving the status of women in South Asia may be beneficial for EIBF, but further research is warranted.
At the household level, there were inconsistent findings for wealth in Pakistan, Afghanistan, and India, and caste and ethnicity were associated with EIBF in Nepal and Afghanistan. The results show an influence of these household-level variables on EIBF, but the relationships are likely context dependent. These results suggest research is required to better understand and address the drivers of the behaviours specific to different castes and ethnicities.

| Avoidance of prelacteal feeding
Prelacteal feeding is known to disrupt EIBF and EBF by delaying the onset of breastfeeding and milk arrival (Ahmed, Rahman, & Alam, 1996;Perez-Escamilla, Segura-Millan, Canahuati, & Allen, 1996). APF steadily rose in Bangladesh, India, and Nepal but declined recently in Afghanistan and Pakistan where rates are currently lowest. The declines may be related to sociocultural barriers and marketing of breast milk substitutes. However, in all countries, prelacteal feeding is still a problem, and common prelacteals include honey, water, and other animal milks ( Figure S2). Measures to improve EIBF and EBF must continue to address prelacteal feeding. across cultures, and caregivers and women who perceive their infants as small, may provide prelacteals to compensate for their perceived milk insufficiency (Gatti, 2008;Sundaram et al., 2013). Small newborns may also have anatomical problems with latching and poor suckling patterns for which lactation support is required, but often these services are not available in many parts of South Asia (Gryboski, 1969 Prelacteal feeding is influenced by misconceptions about colostrum and local feeding customs (Ali, Ali, Imam, Ayub, & Billoo, 2011;Patel, Banerjee, & Kaletwad, 2013;Sharma & Byrne, 2016). Therefore, addressing context-specific beliefs and behaviours are crucial.

| Exclusive breastfeeding
EBF rates in Bangladesh, India, and Nepal were slightly above the World Health Assembly 2025 breastfeeding target of 50%; however, the recent decline in Afghanistan and low rate in Pakistan are a concern. Possible reasons for this include geographic and security barriers limiting interventions to support EBF, weak health systems, and marketing of breast milk substitutes.
Younger child age was the only child-level predictor of EBF in all countries, whereas small birth size and female child were negatively associated with EBF in Afghanistan and India, respectively. Another study in India also reported associations between female children and shorter duration of EB (Jayachandran & Kuziemko, 2011). Cesarean delivery in Pakistan, delivery assistance from traditional birth attendants or health professionals in Afghanistan, middle-aged mothers in India, and larger household size in Bangladesh were risk factors for suboptimal EBF. EBF was associated with ethnic groups in Afghanistan and India, illustrating the influence of local norms on EBF in these countries.
The perception of insufficient milk supply and sociocultural norms (Gatti, 2008;Pries et al., 2016) are all reported barriers to EBF (Patel et al., 2015). EBF for 6 months is a challenge for women everywhere, and our results illustrate understanding local context is important for overcoming barriers to EBF. Interventions that are multicomponent, targeting women, families, communities, and health facilities can help women start and sustain EBF for 6 months (Menon et al., 2016).

| Continued breastfeeding
Although CBF at 2 years is higher in all five countries than the global average (46%; United Nations Children's Fund, 2016), recent declines in CBF prevalence are a concern for Afghanistan and Nepal. In regions with low diet diversity, CBF helps provide children with essential nutrients, continues to offer protection from infection, and is associated with child development (United Nations Children's Fund, 2016).

Maternal-and household-level predictors of CBF at 2 years
included women with low BMI in India, no education in Pakistan, and poorer households in both countries. The results suggest socioeconomic disparities play a role in the maintenance of breastfeeding for 2 years with those more disadvantaged breastfeeding longer . Interventions to support CBF must be tailored to target women across the socioeconomic spectrum, and this will be increasingly important as countries in the region experience economic growth.

| Implications for policy and programs
The variation in predictors of optimal breastfeeding between countries illustrate that policy and program interventions need to be based on an understanding of local determinants for suboptimal breastfeeding.
In South Asia, national action for breastfeeding, including national policies and laws protecting, promoting, and supporting optimal breastfeeding, exists (Thow et al., 2017). However, our results indicate more action is required at the national and subnational levels to ensure that these polices are implemented and monitored.
At the health facility level, suboptimal breastfeeding practices among cesarean delivery births is a clear example of where action is required in all countries. All mothers, regardless of the mode of delivery, should be supported to initiate breastfeeding immediately after delivery (Prior et al., 2012). Strategies to improve breastfeeding outcomes for infants delivered by caesarian section include adoption of supportive hospital policies, training of medical staff to support breastfeeding postdelivery, education about caesarian delivery and breastfeeding, and reduction of caesarian deliveries that are not medically required (Kuyper, Vitta, & Dewey, 2014).
Low birth weight infants require special attention. Ensuring facility and community health workers are trained to support mothers to put low birth weight infants who are able to breastfeed to the breast as soon as possible after birth when they are clinically stable or to assist mothers to express breast milk or access human milk for infants that cannot be fed at the breast is essential (World Health Organization, 2011). In South Asia, India has an established network of human milk banks for low birth weight infants (Haiden & Ziegler, 2016). Further, preventing low birth weight by improving maternal nutritional status before and during pregnancy may also improve breastfeeding practices.
The health system also has a crucial role in promoting and supporting breastfeeding practices during antenatal care and throughout the first 2 years of a child's life. Our findings indicate that interventions should be based on a local understanding of social-cultural barriers, including potentially harmful traditional practices and the low autonomy of women. Interventions delivered by health workers and community-based workers to inform, educate, and counsel on breastfeeding in the home/family, community, and health facility environments all demonstrate effectiveness, highlighting the impact of multiple supportive environments . A breastfeeding supportive environment should include not only the immediate family and health system but also the broader community.

| Limitations
The cross-sectional nature of the survey data limits inferences about directionality of associations in this study; however, our analyses do not make any causal claims. There is also the possibility of response bias from mothers due to the recall method used to assess breastfeeding practices, and this could affect accuracy of the estimates. In addition, social desirability bias for all self-reported breastfeeding practices could affect accuracy of the estimates. Sample sizes for some breastfeeding practices such as CBF were small in some countries, limiting inferences. Finally, the comparability across countries may be limited by the different survey years.

| CONCLUSION
Over the last 25 years, there has been a steady increase in EIBF, APF, and EBF in Bangladesh, India, and Nepal. CBF, however, has not shown the same improvement, and, in recent years, all breastfeeding practices in Afghanistan and EIBF and APF in Pakistan have declined. Our study identified child-, maternal-, and householdlevel factors associated with suboptimal breastfeeding practices in South Asia. The most common predictors of suboptimal breastfeeding included caesarian delivery, small size at birth, home delivery, and low women's empowerment. This information can assist policy and program managers to strengthen the design and implementation of actions to protect, promote, and support breastfeeding.