Breastfeeding difficulties and maternal mental health: role of social support in rural northwestern China | BMC Pregnancy and Childbirth

Data and sample

This study utilized baseline and endline data from a randomized controlled trial (RCT) conducted by Shaanxi Normal University in rural Shaanxi Province, which focused on maternal and infant health. The study employed a multilevel cluster random sampling strategy to identify participants through a three-step procedure. First, ten counties with relatively low levels of economic development were randomly selected from five prefecture-level cities in the province. Second, within each county, ten townships were randomly chosen from a list provided by the local Health Bureau, excluding the township that housed the county seat due to its relatively higher urbanization and affluence. In counties with fewer than ten townships, all townships were included, resulting in a final sample of 79 townships. Third, within each selected township, a list of pregnant women between 8 and 32 weeks of gestation was obtained from the local Township Health Center, from which ten women were randomly selected. In townships with fewer than ten eligible women, all were included. Eligible participants met the following criteria: aged between 18 and 45 years, had resided locally for at least one year, and were within the first six months of pregnancy.

Two rounds of data collection were conducted using the same structured questionnaire in the same sample sites. The baseline survey was administered in March 2021, and the endline survey followed in December 2021, approximately eight months later. Data were collected through face-to-face interviews conducted by trained enumerators recruited from Xi’an university students. Enumerators underwent extensive training and completed a pilot survey with 20 participants prior to formal data collection. Interviews were conducted privately, with efforts made to avoid interruptions from other family members. The survey gathered information on infant and household characteristics, maternal demographics, breastfeeding practices, and maternal mental health. A total of 597 eligible mothers were enrolled at baseline, and 527 completed the endline survey (response rate: 88.3%). After excluding 30 cases with missing outcome or key covariate data, the final Analytic sample included 497 mothers, representing 83.2% of the baseline cohort.

The intervention was delivered from May to December 2021. It consisted of monthly home visits and telephone follow-ups delivered by trained community health workers (CHWs), who were recruited from among women and child health specialists at township health centers. A total of 39 CHWs participated in the study, with An average of 8.7 Years of work experience. On average, CHWs were 33 Years old, and 80% of them held a college degree or higher. Each home visits lasted 40–60 min and focused on maternal nutrition, breastfeeding techniques, and early childhood development. In the meantime, participants in the control group families received no additional intervention.

The study was approved by the Medical Ethics Committee of Shaanxi Normal University and Xi’an Jiaotong University of China (No: 2020 − 1240). Each eligible participant received a consent form with information regarding program objectives, procedures, potential risks, benefits, and an explanation of privacy protection. Participants provided informed consent for inclusion in the study before engaging in a face-to-face interview with a single enumerator.

Measures

Maternal mental health

Maternal mental health (including depression, stress, and anxiety) was assessed using the Depression Anxiety Stress Scales-21 (DASS-21), which is a quantitative measure of depression, anxiety, and stress symptomatology (7 statements each) during the past week. Participants were asked to decide how Much the statements apply to them using a scale from 0 to 3 where 0 refers to “did not apply to me at all”, 1 refers to “applied to me to some degree”, 2 refers to “applied to me to a considerable degree”, and 3 refers to “applied to me very much”. The score of each subscale is Multiplied by 2 to lie within a 0 to 42 scale where higher scores indicate worse outcomes [32]. Each subscale score is Multiplied by 2 to obtain the final score that corresponds to the grading criteria of DASS. Based on the final scores of each subscale, the severity of negative emotions is classified into five levels: normal, mild, moderate, severe, and extremely severe [32, 33]. In this study, the levels were further consolidated, with the “normal” level classified as having no corresponding risk of negative emotions, while the other four levels were classified as having corresponding risks. Specifically, a depression subscale score > 9 indicates a risk of depression; an anxiety subscale score > 7 indicates a risk of anxiety; and a stress subscale score > 14 indicates a risk of stress [33]. The Chinese version of the DASS-21 was validated in a previous study and Cronbach’s alpha for its subscales was 0.78, 0.66, and 0.83, respectively [34].

Breastfeeding difficulty

Breastfeeding difficulties were measured based on maternal reports of 17 challenges experienced within the first two weeks postpartum (e.g., baby not latching, nipple pain, insufficient milk, cesarean-related issues; full list in Supplementary Table A1). The item pool was adapted from the validated Breastfeeding Experience Scale (BES), with modifications to better reflect the rural Chinese postpartum context. Emotionally oriented items (e.g., feeling tired or embarrassed) were excluded to avoid conceptual overlap with psychological outcomes, and culturally specific barriers (e.g., cesarean/episiotomy-related issues, time constraints) were added.

Guided by prior research, we grouped the items into three theoretically defined domains: infant-related difficulties, physical feeding difficulties, and maternal difficulties [24, 35, 36]. To construct composite indicators, we applied exploratory factor analysis using polychoric correlation matrices and principal-axis factoring, extracting one factor per domain. Standardized factor scores were computed using the regression method and used as continuous predictors in subsequent models. In addition, we extracted a total breastfeeding difficulty score to capture the overall level of challenges experienced. As the purpose was to generate summary scores rather than test latent structures, factor loadings are not reported.

Breastfeeding family support

Breastfeeding family support was measured using a 9-item scale developed by Zhu Xiu [37]. The scale demonstrated good internal consistency, with a Cronbach’s alpha of 0.886. Participants rated their level of agreement with each item on a 5-point Likert scale ranging from “strongly disagree” to “strongly agree.” The scale consists of two dimensions: practical support, comprising 2 items (e.g., “My family often prepares food that is good for lactation”), and emotional support, comprising 7 items (e.g., “I think my family wants me to exclusively breastfeed my child”). Practical support in this study refers to tangible help provided by family members, such as caring for the baby or assisting with housework, while emotional support reflects the recognition and encouragement received from family members regarding breastfeeding.

For the analysis, we calculated the mean scores for each support dimension and constructed binary subgroup variables to capture variation in support levels. Participants were classified into “higher” and “lower” family practical support and emotional support groups based on the distribution of scores. To enhance the distinction between groups, those with moderate but not consistently high support levels were grouped into the “lower” category.

Home visits by community health workers

Home visits by community health workers (CHWs) were similarly used to construct a binary subgroup variable. This variable captures whether a participant received the childbirth support home-visit intervention delivered by CHWs. Participants were classified into two groups: “received home visits” (coded as 1) and “did not receive home visits” (coded as 0).

Covariates

In all multivariate models, we controlled for potential confounding factors based on prior literature. These included maternal characteristics (age, education, self-rated health), infant characteristics (gender, birth order, preterm status, underweight), and family context (husband co-residence, conjugal relationship quality, family asset level, and grandparental caregiving) [38,39,40,41]. We also included a binary indicator for receipt of home visits and risk of maternal mental health at baseline to adjust for potential intervention exposure. Variance inflation factors (VIFs) were below 5 for all covariates, indicating no multicollinearity concerns.

Statistical analyses

The data processing and statistical Analyses in this study were conducted using STATA 17.0. Frequency and percentage were used to describe maternal characteristics, infant characteristics, family characteristics, and specific maternal breastfeeding difficulties. Additionally, we performed descriptive analyses of postpartum mental health indicators using frequency, percentage, mean, and standard deviation, as appropriate.

Multiple linear regression models were employed to examine the associations between breastfeeding difficulties and maternal mental health outcomes, including depression, anxiety, and stress. To further assess whether these associations differed by levels of family support, we conducted stratified regressions by family emotional support and family practical support. Similarly, to examine differences by exposure to community support, we conducted stratified analyses based on receipt of home visits from community health workers (CHWs). Multicollinearity was assessed, and all variance inflation factors (VIFs) were below 5, indicating no multicollinearity concerns.