Breastfeeding practices and associated factors at the individual, family, health facility and environmental levels in China

Abstract We examined the association between breastfeeding practices and associated factors using cross‐sectional data from face‐to‐face interviews with 9,745 mother–child dyads in China. The study collected information on breastfeeding practices and potential associated factors at the individual, family, health facility and environmental levels in China. We used survey commands in Stata to consider sampling weight and survey design effects. Although breastfeeding was the norm (97.4% ever breastfed), the prevalence of early initiation of breastfeeding (EIBF) in 0–11 months old was 8.2%, exclusive breastfeeding (EBF) in 0–5 months old was 27.8% and breastfeeding on the previous day in 6–11 months old was 77.5%. The prevalence of EIBF was lower for caesarean delivery and among mothers belonging to ethnic minority groups. The prevalence of EBF was higher among mothers who practiced EIBF, received information that encouraged breastfeeding and knew that a baby should be breastfed on demand and exclusively. By contrast, the prevalence of EBF was lower in mothers who received infant formula advice or felt uneasy breastfeeding in public places. The prevalence of breastfeeding on the previous day was higher among mothers whose partners supported breastfeeding and who knew about timing of colostrum production, EBF for 6 months, and to nurse more to stimulate milk production. The prevalence of breastfeeding on the previous day was lower in mothers who received infant formula advice or felt uneasy breastfeeding in public places. In conclusion, we found that the prevalence of EIBF and EBF practices in China was low and associated with factors at individual, family, health facility and environmental levels.


| INTRODUCTION
Despite their established benefits, breastfeeding practices remain suboptimal worldwide: the global prevalence of early initiation of breastfeeding (EIBF), exclusive breastfeeding (EBF) and continued breastfeeding at 2 years is less than 50% (Global Nutrition Report, 2018;Rollins et al., 2016;UNICEF, 2019). The annual increase in the prevalence of EBF was just 5 percentage points in the last 14 years (from 35% in 2005 to 42% in 2018). In contrast to the global trend, the prevalence of EBF in China declined from 27.6% in 2008 (Center for Health Statistics and Information at Ministry of Health of People's Republic of China, 2009) to 20.7% in 2013 (Duan et al., 2018). The annual economic cost of not breastfeeding according to recommendation has been estimated to be US$ 66.1 billion or 0.61% of China's gross national income (Walters, Phan, & Mathisen, 2019).
Various factors are associated with breastfeeding practices in China, such as the level of education, employment status, ethnicity and family income (Duan et al., 2018;Gao et al., 2016;Li, Li, Ali, & Ushijima, 2003;Xu et al., 2007) as well as mothers' nutrition knowledge (Gao et al., 2016;Guldan et al., 1993;Kong & Lee, 2004;Shi, Zhang, Wang, & Guyer, 2008). These studies focused primarily on socio-economic and individual-level factors in specific geographic areas with the exception of a study by Duan et al. (2018). As suggested by Rollins et al. (2016), breastfeeding practices are also associated with factors related to health care systems, families, workplaces, the built environment and the marketing of breastmilk substitutes (BMSs).  (International Labour Organization, 2014). It also does not align with the recommended EBF duration of 6 months.
In addition, China is the largest market for BMSs globally (Rollins et al., 2016)

| Participants and data collection
This study is based on data from 10,408 mother-child dyads who participated in a survey on factors influencing breastfeeding in China conducted in 2017-2018. This study was a part of the Breastfeeding Promotion Initiative at the China Development Research Foundation (CDRF). The main mission of the CDRF is to improve the policy environment and address cultural norms affecting breastfeeding in China.
The study was designed, administered and implemented by an independent research institute-the National Institute for Nutrition and Health at the Chinese Center for Disease Control and Prevention (NINH, China CDC). The sample size calculation was based on the prevalence of EBF rate for infants under 6 months of 20.8% (Yang et al., 2016), precision of 4%, design effect of 2 and an estimated 10% of nonresponse rate for 12 strata.
A multistage, stratified cluster sampling approach was used for the selection of the survey sample. We selected 12 districts/counties from all seven regions and included 12 of 34 provincial-level administrative divisions considering population size, executive capacity and collaboration of the provincial-level CDC in China. These districts/counties represented three strata: large cities (four districts/counties), medium and small cities (four districts/counties) and rural areas (two districts/counties in general rural areas and two in rural poor areas). One county in rural areas was purposely selected to provide baseline data for a CDRF project intervention.

Key Messages
• Few large-scale studies have examined factors associated   with  poor  breastfeeding  practices  in  China comprehensively.
• This study, covering seven regions in China, shows a low prevalence of recommended breastfeeding practices.
• Reported barriers to exclusive breastfeeding (EBF) were feeding the infant formula and giving water, limited breastfeeding knowledge, infant formula promotion, early return to work, limited workplace support and lack of public breastfeeding rooms.
• Timely investments in a comprehensive EBF programme are warranted for China to meet the World Health Assembly Target to increase the rate of EBF up to at least 50% by 2025.
In each selected district/county, the researchers at NINH, China CDC obtained the list of all clusters and their population size using data provided by the provincial-level CDC. Four clusters were randomly selected from the list via PPS. In cities, a cluster was typically equal to a subdistrict. Small-size subdistricts in rural areas were combined to form a cluster. In each selected cluster, the data collection team visited the corresponding immunization clinic and invited mothers who brought their 0-to 11-month-old children to the clinic for immunization and to participate in the study. With written consent of a mother, the interviewer proceeded with the questionnaire. The process continued until the sample size reached 210 mother-child dyads; about 50% were mothers of children

| Outcome variables
Infant and young child feeding practices were assessed using indicators recommended by the World Health Organization (WHO), based mainly on foods and drinks consumed the previous day and night (Daelmans, Dewey, & Arimond, 2009). The following indicators and descriptions were used: (i) Children ever breastfed-Proportion of children born in the last 12 months who were ever breastfed; (ii) EIBF-Proportion of children born in the last 12 months who were put to the breast within 1 h of birth; and (iii) EBF under 6 months-Proportion of infants 0-5 months old who were fed exclusively with breastmilk in the previous 24 h (no foods, no liquids except for medications such as drops and syrups). In this study, we used receiving any breastmilk on the previous day as a proxy indicator for the continuation of breastfeeding among infants 6-11 months old.

| Exposure variables
Correct latching was defined when mothers recalled that they kept their babies latched before their milk came in. Questions related to breastfeeding included five topics: (1) number of days after birth, a mother still produces colostrum (within 7 days); (2) the duration a child should be exclusively breastfed (6 months); (3) the best way to stimulate milk production (nurse more); (4) how often to feed a baby (on demand); and (5) awareness of the benefits of breastfeeding. The benefits of breastfeeding included 10 items: (5.1) helps uterus and other organs to return to normal; (5.2) helps mothers to regain a desired figure and reduce weight; 5.3) lowers the risk of ovarian and breast cancer; (5.4) helps to delay menstruation; (5.5) more economical and safer than infant formula; (5.6) can meet the baby's physiological needs; (5.7) promotes baby's emotional and intellectual development and strengthens the connection between baby and mother; (5.8) helps to reduce the risk of being overweight, obese, and developing chronic diseases; (5.9) helps to build up immunity and reduce the risk of infectious diseases and (5.10) reduces the likelihood of babies having an allergy. For the analysis, we generated a score for the awareness of the benefits of breastfeeding from these 10 items (scale 0-10; mean of 4.7; median of 4), in which one point was added to the score for each benefit mentioned. We plotted the prevalence of EBF by each awareness score to determine a cut-off point for good awareness of the benefits of breastfeeding. The result showed a score of six indicated good awareness. Twenty-two per cent of the women had an awareness score of 6-10.
We also asked mothers about the environment related to breastfeeding practices. The environmental factors included (i) whether her partner agrees that breastmilk is better than BMS, (ii) having received information that encourages breastfeeding, (iii) exposure to infant formula advertisements and free samples and (iv) feelings of inconvenience when breastfeeding in public and the use of infant formula due to this inconvenience. For formally employed women, we asked about the availability of refrigerators at work to store breastmilk and if they have returned to work.

| Covariate variables
We collected data on maternal characteristics such as age (categorized into 18-25, 26-35 and ≥36 years), ethnicity (Han and other ethnicities), education (junior high school or less [<9 years], high school [9-12 years] and college or higher education), employment status (formally employed and not formally employed) and the place of residence (big cities, medium and small cities and rural areas). A mother was classified as formally employed if she responded that she was formally employed and covered by the paid maternity leave policy. We asked mothers if they had had an antenatal visit and if they had experienced gestation diabetes or high blood pressure during their pregnancy. We also collected information about child characteristics and delivery-related factors including gender, age in months, first birth, low birthweight or caesarean delivery. We also asked the mothers if the child had had diarrhoea or a symptom of respiratory infection within 2 weeks prior to the interview.

| Statistical analysis
From 10,408 records of interviewed mothers, we excluded 185 migrant women in rural areas because the study protocol only included locally registered residents in rural areas. We also excluded 12 records with missing values for outcomes (e.g., ever breastfed, EBF and EIBF), 397 records for exposure variables (e.g., knowledge, social norms and the environment relating to breastfeeding) or covariates relating to childbirth (e.g., first birth, caesarean delivery and low birthweight), and 69 records due to missing information on maternal characteristics (e.g., age, ethnicity and employment status). The sample size for the analysis was 9,745, which is about 95% of the original sample. Sensitivity analysis using the complete sample showed very similar findings with those reported in this manuscript.
We used survey commands in Stata 14.

| Ethical considerations
The study design was reviewed and approved by the Medical

| RESULTS
Among 9,745 mothers interviewed, 89.6% belonged to the Han ethnicity, 28.1% had at least a college degree and 16.6% had formal employment (Table 1). About 50.5% of the infants were boys, 38.7% were the first child and 42.0% were born by caesarean delivery.
Mothers had a moderate level of breastfeeding knowledge (Table 1).  Table 4). The prevalence of breastfeeding on the previous day was higher in mothers whose partners supported breastfeeding and who knew about the timing of colostrum production, EBF for 6 months, and to nurse more to stimulate milk production. The prevalence was lower in mothers receiving infant formula advice or feeling uneasy breastfeeding in public places (Table 4).

| DISCUSSION
In this study, we found that breastfeeding practices were inadequate in 12 districts/counties in China. The prevalence of EIBF found in our study (8.2%) is lower than the national prevalence reported in 2013 of 28.7% (UNICEF, 2018). The prevalence we reported is similar to a study in Sichuan province of 9% (Tang et al., 2013) but much lower than a study in central and western China of 59.4% (Guo et al., 2013).
Given that the prevalence of EIBF varied substantially by region in  (Nguyen et al., 2016;Nguyen et al., 2017). Second, at the household level, we found that mothers with support from partners (e.g., proxy through the support of breastfeeding over BMS) were more likely to continue to breastfeed their infants. In China, especially in urban areas, there has been a shift from multiple generation households to two-generation households (Su, Hu, & Peng, 2017). Among migrants, the simultaneous migration of the couple has become the dominant form of migration for young couples (Meng, Zhao, & Liwu, 2016). Although taking care of newborns is considered to be the responsibility of the extended family in China, conflicts often happen among new mothers, their mothers and mothersin-law about how to take care of babies, especially regarding breastfeeding (Tang, Zhu, & Zhang, 2016). To avoid such conflicts, young families might choose to separate from their extended families to have full autonomy over their decisions about child feeding and other concerns (Tang et al., 2016). The shift might reduce the influence of mothers or mothers-in-law in the decision making and F I G U R E 1 Breastfeeding pattern for infants from (a) all mothers (n = 9,745), (b) formally employed mothers (n = 2485) and (c) not formally employed mothers (n = 7260). Data were weighted prevalence; (n) was for nonweighted sample size Values are weighted, adjusted odds ratios (OR) and 95% CI from survey multiple logistic regression, controlled for place of residence and child gender. Significantly different from the null value (OR of 1). (n) was for nonweighted sample size. Third, for factors at the health facility level, we found that the prevalence of caesarean delivery was high and associated with a lower prevalence of EIBF. The prevalence of caesarean deliveries in our study, 42%, was comparable with previous studies (Feng, Xu, Guo, & Ronsmans, 2012;UNICEF, 2019) and much higher than the WHO's suggested prevalence between 10% and 15% (Betran, Torloni, Zhang, & Gülmezoglu, 2016). This finding suggests the need of regulating the indication of caesarean delivery due to nonclinical reasons, which is high and affected by factors from women and family members, health workers, the health system and socio-cultural norms Values are weighted, adjusted odds ratios (OR), and 95% CI from survey multiple logistic regression models, controlled for maternal age, education and ethnicity, and child gender. Significantly different from the null value (OR of 1). (n) was for nonweighted sample size. * P < 0.05. ** P < 0.01. *** P < 0.001. (Feng et al., 2012;Long et al., 2018). Although the WHO issued guidelines for early essential newborn care for both vaginal and caesarean deliveries (Betran et al., 2016), current childbirth and early essential newborn care policy and practice in China do not align with the recommendation from the WHO according to a recent study in four provinces in China (Xu, Yue, Wang, Murray, & Sobel, 2018 Values are weighted, adjusted odds ratios (OR) and 95% CI from survey multiple logistic regression models, controlled for place of residence, maternal age, education and ethnicity, and child gender. Significantly different from the null value (OR of 1). (n) was for nonweighted sample size. * P < 0.05. ** P < 0.01. *** P < 0.001.
Escamilla, 2019). However, the situation is challenging in China, a country with almost 17 million births annually (UNICEF, 2019) and the world's largest market for BMS (Rollins et al., 2016). China pro- Sixth, our study also suggests the need for better maternity protection and workplace breastfeeding support in China, a country with 70% of women participating in the labour force (Ye & Zhao, 2018). In our sample, only 16.6% of the women interviewed were formally employed and eligible for paid maternity leave. We found that the child's age did not predict EBF in formally employed mothers, while it was associated with a lower prevalence of EBF among those mothers who were not formally employed. Because paid maternity leave protects women's employment and income during their pregnancy and after childbirth, the women are more likely to have better breastfeeding practices (Chai, Nandi, & Heymann, 2018;Rollins et al., 2016). Previous studies showed that mothers who returned to work were less likely to practice EBF (Jia, Dong, & Song, 2018;Mirkovic, Perrine, & Scanlon, 2016). Given that the duration of paid maternity leave in China is just 14 weeks, it is challenging for mothers to exclusively breastfeed their children for 6 months. Additionally, mothers need lactation support at work, including nursing breaks and refrigerators for storing breastmilk (International Labour Organization, 2014). In our study, we found that simple interventions, such as the availability of a refrigerator to store breastmilk in the workplace, were potentially associated with higher EBF (OR: 1.33; 95% CI: 0.95, 1.86). These findings suggest that legislation and interventions are needed to make the workplace a more supportive environment for breastfeeding mothers.
Our study examined factors associated with breastfeeding practices at multiple levels in China using a large sample drawn from all seven regions and included 12 of 34 provincial level administrative divisions, covering large cities, medium and small cities and rural areas.
The sample size was large enough to stratify the analyses by the working status of the women. The study has limitations as well.
Because the selection of provinces and districts/counties was not random, our results might only apply to study sites. The recruitment of mothers from vaccination clinics would provide comparable findings with those from households because of a strong immunization registry and the immunization programme's high coverage in China (99%; UNICEF, 2019). Because the respondent was the mother, we excluded children who were not with their mothers at the time of the survey, including 'children left behind' (Tian, Ding, Shen, & Wang, 2017), which might overestimate the prevalence of EBF and breastfeeding on the previous day. Due to the limitations in sampling, we might expect a random variation in the prevalence of ever breastfed and EIBF from the 'true' value. The cross-sectional design cannot be used to conclude causal relationships. However, plausibility of causality was increased in our study by selecting age ranges for which exposure or covariate variables occurred before or at about the same time as the practices. We did not collect information about the intermediate stages between awareness and practices (e.g., intentions, trials, adoption and maintenance) because they were not the focus of this study. Also, although we capture key information relating to BMS and maternity protection, the content of our questionnaire was not as exhaustive as the complete list of the WHO's NetCode Assessment

Module and International Labor Organization's Maternity Protection
Assessment Toolkit. Nonetheless, we captured key individual, health facility and policy components.
In conclusion, we found that the prevalence of EIBF and EBF practices in China were low and associated with factors at the individual, family, health facility and environmental levels. Interventions to improve breastfeeding practices in China should include improved breastfeeding knowledge and skills by mothers and other caregivers, and breastfeeding-friendly environments in health facilities, and public places. In addition, improved maternity protection legislation and national regulations aligned with the Code seem critical to protect breastfeeding and ensure the health and well-being of children and women in China. To achieve the above, China should build on the evidence of successful experiences, tools and platforms for breastfeeding protection, promotion and support in the region and in the world (Rollins et al., 2016;UNICEF, 2019).
designing, administering and implementing the survey, and suggestions on data analysis at the early stage. We are grateful to Jessica Blankenship from UNICEF East Asia and the Pacific Regional Office,

CONFLICTS OF INTEREST
The authors declare that they have no conflicts of interest.

CONTRIBUTIONS
The authors' responsibilities were as follows: JL, TTN, XW and JF