Study design
This study used data from the Under 5 Child Nutrition and Health Surveillance System (U5CNHSS), a component of the National Maternal and Child Health Surveillance System (NMCHSS) in China [18]. Although the surveillance system was running from October 1, 2011, data quality was erratic until 2013. Therefore, we used data from 2013 to 2018 in the present study. The U5CNHSS contains data on about 140,000 children under 5 years each year, including almost 40,000 infants under 6 months. The U5CNHSS dynamically collected information on growth and development, nutrition and health status, and impact factors of nutrition and health at 1, 3, 6, 8, 12, 18, 24 and 30 months, as well as at 3 and 4 years for each child. After multi-stage stratified cluster random sampling of the NMCHSS, we selected 80 of the 334 districts/counties in mainland China. First, all districts/counties in China were divided into four levels (urban, eastern rural, central rural and western rural areas) and ranked by total population. The reported rate of underweight among children under 5 years in the four levels were used to calculate the sample size in each level. We imposed a requirement of data from at least 2000 children in each district/county in order to compensate for loss of subjects due to mobility and other causes. Then the number of districts/counties in each level were decided. Second, four levels were further divided into subgroups based on the number of districts/counties. Third, 334 districts/counties from NMCHSS were assigned to each subgroup. Then one district/county was randomly selected from each subgroup. Fourth, four townships (communities/sub-districts) in each selected district/county were systematically sampled based on total population. Finally, villages/committees randomly selected until reaching the required minimal sample size.
In one part of this study, trends in the two breastfeeding indicators over time were analyzed based on the cross-sectional survey data. In the second part of this study, prevalence of exclusive breastfeeding over time was analyzed based on longitudinal cohort data.
Study population
Analysis of EIBF trends involved data for infants born between January 1, 2013 and December 31, 2018 by year. Analysis of EBF trends involved data for infants who were surveyed between September 1 and September 30 in each year of the study period and who were no older than 180 days at the surveys. Longitudinal analysis of EBF prevalence involved data for infants who participated in all three visits at 1, 3 and 6 months and who were no older than 180 days at the 6-month visit. The study flow chart was shown in the Supplementary Figure 1.
Data collection
Staff collecting data were trained before the surveillance. Unified protocol and study forms were used for each surveillance site. Briefly, village or committee doctors were required to register all newborns and under-5 children within their areas of responsibility. They conducted family visits of newborns, and notified children for health visits. Doctors at township or community health care centers assembled a health profile for each child and were responsible for health visits. During the health visits, the children’s weight and length/height were measured, and they underwent a basic physical exam and developmental assessment, and their hemoglobin were tested (once every year). During the visits, parents were asked to complete a structured questionnaire collecting demographic information about the household, the child’s diseases, and feeding practices. Data were entered into the network reporting system within one month after each visit, and data were subjected to a quality control process that was audited level-by-level.
EIBF and EBF were defined according to the WHO [19]. We calculated the indicators for each year during the study period. EIBF was defined as the proportion of infants born in the year who were put to the breast within one hour of birth. This was based on recall by infants’ mothers at the first visit, who were asked “How soon after birth was your baby be first put to the breast? [1=never, 2=within 1 hour (immediately), 3=1-23 hours, 4=the second day or later, 5=unknown]”.
EBF was defined as the proportion of infants 0-5 months old who were fed exclusively with breast milk, including breastfeeding by a wet nurse and feeding expressed breast milk, as well as administration of any necessary medications, vitamins, or oral rehydration solution. This indicator was based on recall of the previous day (i.e. 24 hours before) by mothers at each visit, who were asked “Did your infant receive breast milk yesterday (i.e. 24 hours ago)?” If they responded “yes”, then they were asked, “Which kind of liquid food did your infant receive yesterday?” and allowed to choose from preset responses (1=water, 2=sweet water/juice/other liquid food, 3=milk powder/milk/goat’s milk, 4=formula milk powder, 5=none of above, 6=unknown). The women who answered “yes” were also asked, “Did your infant receive solid/semi-solid/soft food yesterday (such as rice flour, paste, cooked rice, steamed bread, etc)?” and allowed to choose from preset responses (1=yes, 2=no, 3=unknown). “Exclusive breastfeeding” was defined as the feeding of infants 0-5 months old with breast milk but neither liquid food listed in the questionnaire nor solid/semi-solid/soft food during the previous 24 hours.
Quality control
The village or community doctors checked the list of newborns and under-5 children monthly to make sure that each child participated in health visits on time. Health workers in the township or community health care centers cross-checked their lists of children who finished health visits with the lists of the village or community doctors, and they also checked the completeness of data collection forms. Health workers at the levels of county/district, prefecture, province, and nation would sample 2-3 surveillance sites semi-annually or annually for quality control concerning work flow, measuring methods, instrument adjustment records and questionnaire data quality.
Statistical analysis
In the analysis, mainland China was divided into three regions (i.e., East, Central and West) based on economic development level as per the surveillance system, and we divided the country into rural and urban areas based on classification by the National Bureau of Statistics of China [20].
The prevalence of the two breastfeeding indicators with 95% confidence intervals (95% CIs) were calculated at the national and subnational levels. The Cochran-Armitage trend test [21, 22] was used to explore trends in the two breastfeeding indicators over time. The annual percent of change (APC) in prevalence of the two breastfeeding indicators at the national and subnational levels were estimated, together with 95% CIs, using log-linear regression [log(δb) = log(δa) + (tb −ta )log(θ)], followed by exp transformation [APC = exp (log(θ))-1) x 100] [23].
All analyses were performed using SAS 9.4 (SAS Institute, Cary, NC, USA). Statistical significance was assessed by two-tailed tests at an alpha level of 0.05.