Skip to main content
Download PDF

The role of breastfeeding self-efficacy in the relationship between perinatal depressive symptoms and exclusive breastfeeding: a longitudinal mediation analysis

BMC Pregnancy and Childbirth volume 25, Article number: 356 (2025) Cite this article

Abstract

Background

Perinatal depressive symptoms are associated with exclusive breastfeeding; however, the longitudinal mediating process of this relationship remains unclear. Breastfeeding self-efficacy may be an important variable in understanding the complex process involved in their co-occurrence. Therefore, we aimed to explore the role of breastfeeding self-efficacy in the relationship between perinatal depressive symptoms and exclusive breastfeeding using both between- and within-person approaches.

Methods

A prospective longitudinal study was conducted from October 2021 to January 2024 at a tertiary hospital in Hunan, China. Depressive symptoms were measured at 36 gestational weeks and 1 week, 6 weeks, 3 months, and 6 months postpartum. Exclusive breastfeeding and breastfeeding self-efficacy were evaluated at the same postpartum intervals. Cross-lagged panel models, random intercepts cross-lagged panel models, and longitudinal mediation models were used to analyze their relationships.

Results

A total of 334 participants were included. Longitudinal mediation models revealed that breastfeeding self-efficacy mediated the prospective negative effect of perinatal depressive symptoms on exclusive breastfeeding at the between-person level (b = − 0.017, SE = 0.008, 95% CI (− 0.032, − 0.001), P = 0.036), and suppressed the positive effect of exclusive breastfeeding on depressive symptoms at the within-person level (b = − 0.044, SE = 0.022, 95% CI (− 0.087, 0.000), P = 0.047).

Conclusions

Mothers with perinatal depressive symptoms may face challenges in exclusive breastfeeding due to reduced breastfeeding self-efficacy. While increasing exclusive breastfeeding might help reduce depressive symptoms over time, this positive effect can be hindered if breastfeeding self-efficacy remains low. Our findings highlight breastfeeding self-efficacy as a critical target for future interventions.

Peer Review reports

Background

Exclusive breastfeeding (EBF) is recommended as the most optimal feeding practice because of its numerous advantages for infants and mothers [1,2,3]. The World Health Organization (WHO) advises mothers to exclusively breastfeed their babies for at least 6 months [4]. However, few women are able to follow this recommendation [5]. The global rate of EBF practices for infants at 6 months is 43% [4], which is higher than the 29.2% in China [6]. Identifying the key modifiable factors is necessary to effectively address the suboptimal EBF situation.

Perinatal depressive symptoms (PNDs) are a common perinatal psychological disorder whose prevalence reaches 11.9% globally and 16.3% in China [7, 8]. PNDs are a serious threat to the health of mothers and infants [9]. Some researchers have reported the negative impact of PNDs on EBF [10,11,12,13]. A meta-analysis by Kim et al. (2021) [12] revealed that mothers with postpartum depressive symptoms had a 33% reduction in EBF within 3 months postpartum. Others suggested that EBF preceded and predicted subsequent depressive symptoms [14, 15]. The meta-analysis by Alimi et al. (2022) [14] showed that women who did not exclusively breastfeed had 89% higher odds of postpartum depressive symptoms. Previous studies suggested a complex relationship between PNDs and EBF; however, the temporal directionality of this relationship remains ambiguous [10,11,12, 14, 15]. Haga et al. (2018) [16] and Zhu et al. (2023) [17] attempted to use an emerging analytical approach (i.e., cross-lagged panel model, CLPM) to analyze the longitudinal and bidirectional connection among PNDs and EBF and elucidate the temporal causality between them. Zhu et al. (2023) [17] conducted a study in central China and suggested that depressive symptoms at 3 and 6 months postpartum predicted EBF at 6 and 12 months postpartum, and EBF were not found to predict PNDs.

How PNDs and EBF may interact remains unclear, and the possible mediating effect of breastfeeding self-efficacy (BSE) between them has not been examined. BSE reflects the degree of confidence that perinatal women perceive in their ability to breastfeed infants [18]. The strong relationship of BSE with both EBF and PNDs has been widely explored. BSE is a pivotal protective factor for EBF [19,20,21]. Tuthill et al. (2020) [21] showed that the rate of early EBF cessation was predicted to decrease by 2% for every point increase in BSE score. Meanwhile, a negative correlation was found between BSE scores and depressive symptom levels at 42 days postpartum [22] or 2–12 weeks postpartum [23]. PNDs have been reported as a risk factor for BSE [24,25,26,27]. A recent review of the relationships and potential mechanisms of breastfeeding and postpartum depression identified BSE as an important variable in understanding the relationship between breastfeeding and depressive symptoms and suggested that it might mediate this relationship [28]. Therefore, understanding the potential mediating role of BSE in the relationship between PNDs and EBF could aid in uncovering the intricate mechanisms of their co-occurrence and identifying prospective intervention targets for PNDs that impact EBF.

Although Haga et al. (2018) [16] and Zhu et al. (2023) [17] offered insights into the temporal causality linking PNDs and EBF, their statistical approaches (e.g., CLPM) were unable to disaggregate intraindividual differences from interindividual differences [29, 30]. Bandura’s self-efficacy theory posits that individuals’ beliefs in their capabilities (i.e., self-efficacy) vary across different contexts (intraindividual) and exhibit stability between different individuals (interindividual) [31]. Intraindividual fluctuations may reflect moment-to-moment adaptations to situational challenges (e.g., postpartum physiological changes), while interindividual differences represent enduring cognitive schemas shaped by cumulative life experiences [31, 32]. In the context of breastfeeding, this suggests that BSE’s mediating role between PNDs and EBF may operate through distinct mechanisms at different analytical levels: transient efficacy appraisals mediating acute mood-behavior interactions (within-person), versus stable efficacy beliefs moderating long-term depression-parenting pathways (between-person). Previous research also confirmed the necessity of considering both intraindividual and interindividual variations in studying the trajectories of PNDs [33, 34]. Therefore, comparing intraindividual and interindividual variations allowed for more substantial conclusions than using either method alone.

Taken together, this study aimed to explore the possible mediating role of BSE in the relationship between PNDs and EBF by considering both intraindividual and interindividual differences.

Methods

Study design and samples

A prospective longitudinal study was conducted at a tertiary hospital in Hunan Province, China, spanning from October 2021 to January 2024. All eligible women who take routine checks were invited to participate in our study between October 2021 and December 2022. A total of 751 pregnant women volunteered to participate. The eligibility criteria were as follows: (1) age above 18 years, (2) being in the first trimester (7–13 weeks and 6 days of gestation), (3) first-time mothers with a singleton pregnancy, and (4) natural conception. The exclusion criteria were as follows: (1) experiencing abortion, stillbirth, or embryo damage during pregnancy; (2) having suicidal tendencies or severe mental illness diagnosed by a psychiatrist; (3) having an illness that would limit breastfeeding after delivery; and (4) not undergoing routine checks or delivering at the study hospital.

Sample size was estimated using a web-frontend for “semPower 2” (an R-package providing a collection of functions to perform a-priori, post-hoc, and compromise power analyses for structural equation models (SEMs)) [35]. The calculations determined that a sample of 112 participants was required to achieve sufficient power (1-β = 0.80) to detect an effect size of 0.05 for the effect measure, root mean square error of approximation (RMSEA) [36, 37]. The alpha was 0.05. In consideration of a potential 20% non-response rate, the minimum sample size required was set as 140.

Procedures

The participants were recruited during their first trimester and followed up for 6 months postpartum. The five follow-up time points were around 36 gestational weeks (T0), 1 week postpartum (T1), 6 weeks postpartum (T2), 3 months postpartum (T3), and 6 months postpartum (T4). Measures of PNDs were collected at each of these time points (T0–T4), and those of BSE and EBF were collected at four time points (T1–T4). The collection of participants’ characteristics and covariates was distributed across three time points (at inclusion, T0, and T1) to minimize the burden on the participants.

In the longitudinal survey, the participants were approached by researchers at the obstetric clinic. Once informed consent was obtained from the pregnant women, online questionnaires were sent to them via Wenjuanxing (a professional online survey platform) through WeChat messages. A week before the participants’ scheduled routine checks (T0, T2) or their babies’ routine checks (T3, T4), a QR code for the questionnaire was sent to the participants’ WeChat via Wenjuanxing. If the participants did not complete the questionnaire before their routine clinical visit, then a face-to-face reminder was given to them during the visit. For T1, the participants enrolled at T0 were retrieved from the post-delivery discharge roster in the medical record and sent a QR code for the questionnaire. A telephone reminder was also given to those who did not complete the questionnaire on time. The entire recruitment and follow-up was conducted by our research team. All team members received training before the survey and were familiarized with the study’s purpose, content, and implementation.

Measures

Sociodemographic variables

The participants were invited to complete two self-administered questionnaires (one for antenatal information and one for post-delivery) to report their sociodemographic information. The questionnaires were developed based on literature review [16, 17, 27]. To ensure content validity, a panel of experts in obstetrics, epidemiology, and survey methodology reviewed the questionnaires. Feedback from experts led to minor modifications before the final implementation. Given that the questionnaires were used to collect sociodemographic and general variables that have been widely verified and has a high degree of consistency [38], we did not conduct additional reliability and validity tests. The following general characteristics were obtained: age, pre-pregnancy BMI, residence, educational status, occupation, marital status, spousal educational status, household income, and passive smoking during pregnancy, number of classes at maternity school, pregnancy complications, adverse maternal history, mode of delivery, gestational age, infant birthweight, gender, and type of hospitalization.

Perinatal depressive symptoms

The Edinburgh Postnatal Depression Scale (EPDS) was used to assess PNDs [39]. This scale is a self-report questionnaire comprising 10 items. Higher scores indicate more severe depressive symptoms. A cut point of 10 is used in Chinese women to obtain maximum specificity [40]. The EPDS Chinese version has demonstrated sufficient validity and reliability (Cronbach’s alpha was 0.78) [41]. The Cronbach’s α values of EPDS were 0.83, 0.85, 0.84, 0.88, and 0.89 across the five time points (T0–T4).

Breastfeeding self-efficacy

The Breastfeeding Self-Efficacy Scale-Short Form (BSES-SF) was used to measure BSE. This scale comprises 14 positively-worded items regarding mothers’ self-efficacy in their ability to breastfeed [42]. Higher scores indicate better BSE. The Chinese version of BSES-SF has demonstrated sufficient validity and reliability (Cronbach’s alpha was 0.93) [43]. The Cronbach’s α values of BSES-SF at T1 to T4 were 0.96, 0.96, 0.97, and 0.98.

Exclusive breastfeeding

The participants’ exclusive breastfeeding degree index was scored with a single item: “Which of the following types of breastfeeding have you practiced for your child in the past 24 h?” (5 = EBF, 4 = almost EBF, 3 = partial breastfeeding, 2 = token breastfeeding, and 1 = exclusive artificial feeding) [44]. The participants were provided with a detailed explanation of each conceptual option before they made their choice. The definitions of breastfeeding used in our study were based on the WHO’s set of breastfeeding definitions [45]. This variable was handled as a continuous variable [46] because our aim was to test the temporal relationship among the EBF degree index, the level of BSE, and the level of PNDs, rather than to investigate the current status of the population in terms of EBF.

Covariates

The following covariates selected a priori based on the existing literature were obtained: breastfeeding intention, knowledge, and attitudes. Herein, breastfeeding knowledge was evaluated using the Chinese Breastfeeding Knowledge Questionnaire [47], and breastfeeding attitude was evaluated by the Iowa Infant Feeding Attitude Scale [48]. Both scales consist of 17 items each and have been implemented across various countries and regions and applied to diverse populations. Their Cronbach’s α coefficients were 0.87 [49] and 0.62 [50]. In this study, the Cronbach’s α values of them were 0.88 and 0.63.

Statistical analysis

Statistical methods

This study explained further knowledge on the possible mediating role of BSE in the relationship between PNDs and EBF using both between- and within-person approaches [51]. Derived from interindividual score differences, between-person associations have traditionally been estimated using the CLPM. In contrast, within-person associations, reflecting intraindividual score changes over time, are estimated using a random intercepts cross-lagged panel model (RI-CLPM) [29, 30]. Although the CLPM is well-suited for analyzing between-person associations, it does not capture within-person associations. Conversely, the RI-CLPM is designed to estimate within-person associations but is not applicable for between-person associations [52]. The decision to use CLPM or RI-CLPM should align with the specific research objectives. When both between- and within-person associations are expected, the application of both models is recommended for a comprehensive analysis [51].

Data analysis

First, preliminary analyses were conducted using IBM SPSS Statistics for Windows, Version 22.0 (IBM Corp., Armonk, NY, USA). Participant characteristics were detailed through categorical variables presented as frequencies and percentages, and continuous variables were summarized by reporting the mean and standard deviation. Bivariate correlations and descriptive statistics of key variables (PNDs, BSE, and EBF) within each time points were then conducted. Furthermore, the Missing Completely at Random (MCAR) testwas conducted [53]. The result was χ2 (236) = 262.299, P = 0.115, indicating that sample attrition throughout the study was random. Hence, full information maximum likelihood (FIML) was employed for model estimations to address missing data. This approach leverages all available information to provide accurate parameter estimates [54].

Second, the EPDS items associated with anhedonia, anxiety, and depression were averaged to form three parcels [54, 55]. Invariance tests were then performed to assess the stability of the measurement model over time [56]. These tests involved applying increasingly strict equality constraints to model parameters and enabling comparisons among the configural, metric, and scalar models. Invariance can be assumed when the change in RMSEA (ΔRMSEA) between subsequent models remains below 0.015 or the change in standardized root mean square residual (ΔSRMR) remains below 0.010 [57].

Finally, CLPMs and RI-CLPMs were run in Mplus 8.7 to test the temporal associations of PNDs, BSE and EBF and the mediating role of BSE. Our models included latent variable (PNDs) and manifest variables (BSE and EBF). Given that the BSES-SF is a single-dimensional scale [42], BSE was used as manifest variables in our models. For reasons of parsimony and considering the lack of specific hypotheses regarding nonstationarity of the underlying within-person processes in our study, the auto-regressive paths, cross-lagged paths, and within-wave covariances across time points were constrained to be equal. Model fit was compared with the freely estimated models and assessed using chi-square statistics, RMSEA, comparative fit index (CFI), Tucker-Lewis index (TLI), and SRMR. RMSEA and SRMR values below 0.080, CFI and TLI values above 0.900 were considered acceptable model fit [58]. Chi-square difference testing was conducted to examine differences in model fit. The following three criteria were used to assess whether the model fits differ significantly among the different models: statistically significant changes in χ2 (Δχ2) at P < 0.05, changes in CFI (ΔCFI) with values lower than 0.010, and changes in RMSEA (ΔRMSEA) with values higher than 0.015 [59]. Lastly, mediation analyses were run with both CLPMs and RI-CLPMs to examine whether BSE mediates the longitudinal relationship between PNDs and EBF. Breastfeeding intention, knowledge, and attitudes were included as covariates for all models. Statistical significance was set at two-sided P < 0.05.

Results

Participant characteristics

As shown in Fig. 1, we enrolled 751 eligible women in their first trimester between October 2021 and December 2022. As of August 2023, 427 (56.9%) pregnant women delivered at the study hospital. After 93 participants lost connection (88), refused to participate (2), and had duplicated data (3) were excluded, 334 participants who reported both PNDs and postpartum breastfeeding information at least once were included in the final analysis. As of January 2024, the questionnaire response rates of these 334 participants were 74.9%, 91.0%, 84.1%, 71.0%, and 65.9% at 36 gestational weeks, 1 week, 6 weeks, 3 months, and 6 months postpartum, respectively.

Fig. 1
figure 1

Inclusion and exclusion process of participants

Full size image

The participant characteristics are detailed in Table 1. Among the 334 participants, the mean age was 28.50 ± 3.19 years, the mean pre-pregnancy BMI was 21.21 ± 3.02 kg/m2, 93.7% were living in the urban area, 67.4% had a bachelor’s degree or higher, 88.9% were employed, and 96.1% were married. During pregnancy, the passive smoking rate was 31.7%, 10.5% attended maternity school sessions four times or more, and pregnancy complications occurred in 24.8%. At delivery, 55.8% were delivered vaginally. The average birth weight of infants was 3203.92 ± 436.33 g, 53% were male babies, and 20.0% were separated from their mothers after birth. The mean scores for breastfeeding intention, knowledge, and attitude were 3.71 ± 0.98, 9.87 ± 4.61, and 55.46 ± 5.60, respectively.

Table 1 Sample characteristics
Full size table

Preliminary analyses

The bivariate correlations, means, standard deviations, and number of participants are presented in Table 2. The correlations between PNDs and BSE and between BSE and EBF were statistically significant at all time points. The correlations between PNDs and EBF were not statistically significant at all time points.

Table 2 Bivariate correlations and descriptive statistics for key variables
Full size table

Measurement invariance and model comparisons

Table 3 indicates that the measures of PNDs demonstrated scalar invariance, implying that the observed changes in these constructs over time were meaningful and not due to measurement artifacts or item biases.

Table 3 Fit indices for the measurement invariance of perinatal depressive symptoms
Full size table

The fit indices and model comparisons for constrained and unconstrained CLPMs and RI-CLPMs are presented in Additional Table 1 in the Supplemental Information. The findings indicated that setting the cross-lagged paths and all T2–T4 correlated changes between all variables (i.e., M1f, M2f) to be equal over time did not significantly decline the fit of the unconstrained model (M1a, M2a). For reasons of parsimony, Models 1f and 2f were selected as the final CLPM and RI-CLPM.

Between-person effects: Cross-lagged panel model

The temporal relations among PNDs, BSE, and EBF were examined by running a CLPM with PNDs at T0 predicting PNDs at T1 (autoregressive path), BSE at T1 (cross-lagged path), EBF at T1 (cross-lagged path), and EBF at T2 (cross-lagged path). We used BSE at T1 to predict BSE (autoregressive path), PNDs at T2 (cross-lagged path), and EBF (cross-lagged path) at T2. Similarly, we used EBF at T1 to predict EBF at T2 (autoregressive path), BSE at T2 (cross-lagged path), PNDs at T2 (cross-lagged path), and PNDs at T3 (cross-lagged path). The predicted paths of these three variables between subsequent time points were determined. Cross-sectional correlations among PNDs, BSE, and EBF were computed for all measurement points.

The model fit indices indicated an acceptable fit of the model, with χ2 (229) = 336.445, P < 0.001, RMSEA = 0.043, CFI = 0.961, TLI = 0.946, and SRMR = 0.064. As illustrated in Fig. 2, the autoregressive paths of PNDs (Tn) on PNDs (Tn + 1), BSE (Tn) on BSE (Tn + 1) and EBF (Tn) on EBF (Tn + 1) suggested the strong stability of the three constructs over the 6 months postpartum (P < 0.001). PNDs (T0) predicted lower BSE (T1), β = − 0.08, SE = 0.04, P = 0.030; did not predict EBF (T1), β = 0.06, SE = 0.05, P = 0.200; and did not predict EBF (T2), β = − 0.07, SE = 0.05, P = 0.136. Meanwhile, BSE (T1) predicted higher EBF (T2), β = 0.18, SE = 0.05, P < 0.001; did not predict PNDs (T2), β = − 0.08, SE = 0.05, P = 0.076. In turn, EBF (T1) predicted higher BSE (T2), β = 0.16, SE = 0.04, P < 0.001; did not predict PNDs (T2), β = 0.01, SE = 0.05, P = 0.851; and did not predict PNDs (T3), β = 0.04, SE = 0.05, P = 0.400. The standardized coefficients, standard errors, and P-values for the predicted paths of these three variables between subsequent time points were similar to those described above.

Fig. 2
figure 2

CLPM depicting the longitudinal associations among perinatal depressive symptoms, breastfeeding self-efficacy, and exclusive breastfeeding. PNDs = Perinatal depressive symptoms; BSE = Breastfeeding self-efficacy; EBF = Exclusive breastfeeding. Gray dashed lines and gray numbers represent the non-significant paths. For clarity, the effects of controlled variables (breastfeeding intention, knowledge, and attitudes) were estimated but not shown in the figure. Standardized path coefficients and (SE) were reported. T0: 36 weeks of gestation; T1: 1 week postpartum; T2: 6 weeks postpartum; T3: 3 months postpartum; T4: 6 months postpartum. *P < 0.05, **P < 0.01, ***P < 0.001

Full size image

The mediating effect was confirmed by the observed significant indirect effects of PNDs on EBF through BSE (b = − 0.017, SE = 0.008, 95% CI (− 0.032, − 0.001), P = 0.036). No indirect effects of EBF on PNDs through BSE (b = − 0.013, SE = 0.008, 95% CI (− 0.028, 0.002), P = 0.090) were observed. Table 4 lists the direct and indirect effects of the mediation models.

Table 4 Direct and indirect effects from cross-lagged panel models (CLPMs) and random intercepts cross-lagged panel models (RI-CLPMs)
Full size table

Within-person effects: Random intercepts cross-lagged panel model

As presented in Fig. 3, we ran a RI-CLPM to test the longitudinal associations among PNDs, BSE, and EBF at the within-person level. The model showed a good fit, χ2 (223) = 296.257, P < 0.001, RMSEA = 0.036, CFI = 0.974, TLI = 0.962, SRMR = 0.053. The autoregressive effects were almost similar to the results obtained using CLPM. However, PNDs at T0 did not predicted BSE at T1 (β = − 0.02, SE = 0.05, P = 0.754), EBF at T1 (β = 0.04, SE = 0.05, P = 0.469), and EBF at T2 (β = − 0.08, SE = 0.05, P = 0.102). In turn, BSE at T1 predicted lower PNDs at T2 (β = − 0.27, SE = 0.12, P = 0.030) and higher EBF at T2 (β = 0.27, SE = 0.08, P = 0.001). Therefore, individuals with increased BSE (relative to their trait level) experienced a subsequent decrease in PNDs. Meanwhile, EBF at T1 predicted higher BSE at T2 (β = 0.28, SE = 0.08, P < 0.001), did not predict PNDs at T2 (β = 0.09, SE = 0.12, P = 0.469), and did not predict PNDs at T3 (β = 0.08, SE = 0.07, P = 0.270). The standardized coefficients, standard errors, and P-values for the predicted paths of these three variables between subsequent time points were similar to those described above.

Fig. 3
figure 3

Simplified illustration of RI-CLPM depicting the longitudinal associations among perinatal depressive symptoms, breastfeeding self-efficacy, and exclusive breastfeeding. PNDs = Perinatal depressive symptoms; BSE = Breastfeeding self-efficacy; EBF = Exclusive breastfeeding. cPNDs, cBSE, cEBF = within-person level variables; RI-PNDs, RI-BSE, RI-EBF = between-person level factors (random intercepts). Gray dashed lines and gray numbers represent the non-significant paths. For clarity, the effects of controlled variables (breastfeeding intention, knowledge, and attitudes) were estimated but not shown in the figure. Standardized path coefficients and (SE) were reported. T0: 36 weeks of gestation; T1: 1 week postpartum; T2: 6 weeks postpartum; T3: 3 months postpartum; T4: 6 months postpartum. *P < 0.05, **P < 0.01, ***P < 0.001

Full size image

In contrast to the findings obtained using CLPM, the mediating effects of EBF on PNDs through BSE (b = − 0.044, SE = 0.022, 95% CI (− 0.087, 0.000), P = 0.047) were observed. Nevertheless, no indirect effects of PNDs on EBF through BSE (b = − 0.009, SE = 0.028, 95% CI (− 0.064, 0.046), P = 0.750) were found. Given that the mediating effect of BSE was not in the same direction as the direct effect (b = 0.059, SE = 0.054, 95% CI (− 0.046, 0.164), P = 0.271), this finding was interpreted as a suppressing effect [60, 61].

Discussion

To our knowledge, this is the first study that researchers extended prior studies and filled important methodological gaps in the literature by examining the longitudinal mediating role of BSE in the relationship between PNDs and EBF using both between- and within-person approaches (i.e., CLPM and RI-CLPM). The methodological approaches revealed new and unexpected findings. Longitudinal mediation models using CLPM showed that BSE mediated the prospective negative effect of PNDs on EBF at the between-person level, and RI-CLPM showed that BSE suppressed the positive effect of EBF on PNDs at the within-person level. No direct longitudinal association between PNDs and EBF was observed either at the between- or within-person level.

Our study showed that BSE mediated the longitudinal relationship between PNDs and EBF, elucidating how PNDs and EBF interact. At the interindividual level, the populations with higher levels of PNDs might have lower BSE levels, thereby exhibit lower EBF scores. At the intraindividual level, in the group with low EBF scores, if an individual had an increased score of EBF but her BSE was not elevated at the next time point, the levels of PNDs would not decrease. Therefore, the benefit of increased EBF scores on PNDs was suppressed by the low levels of BSE, resulting in no reduction in PNDs and vice-versa. Although no study has explored the mediating role of BSE in the relationship between PNDs and EBF, our findings share some similarities to those of Rosenbaum et al. (2020) [62], who reported that an interaction of self-compassion and negative feelings toward the body mediated the association between early unwanted discontinuation of breastfeeding and depressive symptoms. Self-compassion, also a dimension of individual self‐concept, is strongly related to self-efficacy [63]. Therefore, our findings provide empirical support for Henshaw’s theoretical belief that BSE is a vital element in understanding breastfeeding and depressive symptoms relationship [28]. This concept points toward future directions of conceptualizing interventions and strategies that address common factors for PNDs and EBF.

Our results initially verified the between-person mediating role of BSE, contributing to the existing literature on the mechanisms by which PNDs affects EBF. In concordance with the findings summarized in the introduction, earlier studies supported that depressed women lack confidence and self-efficacy in their ability to overcome barriers to breastfeeding, thereby leading to their reduced EBF [24,25,26, 64, 65]. Papinczak et al. (2000) [64] suggested that a longer duration of EBF was most significantly linked to increased breastfeeding self-confidence and lower scores of postpartum depressive symptoms. Flores-Quijano et al. (2008) [65] also suggested a significant relation between women’s depressive symptoms and lack of confidence in breastfeeding between 2 and 12 weeks postpartum, and both of these variables were associated with EBF. All of the above studies used general statistical analyses such as logistic regression models, linear regression models, and path analysis, illustrating research questions mainly at the between-person level [29, 30]. Thus, between individuals, PNDs might indirectly negatively affect EBF scores by decreasing BSE levels. Furthermore, the between-person level can indicate “who is at high risk for heightened X” and “who should get an intervention to reduce Y” [66]. Our findings based on the between-person level suggested that in a general population, the individuals with higher levels of PNDs are at a high risk for lower EBF scores.

Our study also revealed the within-person suppressing effect of BSE, indicating that the effect of EBF on PNDs within individuals might be completely mediated by BSE [61]. In particular, the within-person suppressing effect of BSE meant that if a woman had a higher EBF score at a certain time point, but her subsequent BSE remained low, it would prevent any reduction in her PND levels. Conversely, if her EBF score was lower but her subsequent BSE remained high, she could still maintain lower levels of PNDs. Individuals with high BSE have high confidence and ability to face breastfeeding challenges [19, 20], thereby reducing the impact of changes in EBF on PNDs. Some past studies partially supported this result [67, 68]. For example, Haga et al.’s (2012) [67] multilevel growth curve analyses revealed that BSE predicted the rate of postpartum depression and the changes in depressive symptoms over time. Jiang et al. (2022) [68] reported that postpartum depressive symptoms were significantly associated with BSE but not with EBF, implicitly suggesting the potential suppressing effect of BSE. A suppression effect refers to a situation where the inclusion of a variable (suppressor) weakens or hides the direct effect of the independent variable on the dependent variable [60]. After the suppressor variable is controlled, the direct effect of the independent variable on the dependent variable becomes apparent or even stronger [60, 61, 69]. The findings of this study were consistent with our previous results that the negative effect of PNDs on EBF became statistically insignificant when the effect of BSE was not controlled in the generalized linear mixed model [70]. According to previous interpretation [60, 61] of the “suppressing effect,” the main effect of EBF on PNDs was positive but the indirect effect through BSE was negative. This finding suggested a broader positive mechanism between EBF and PNDs that was not included in the present study, providing a novel direction for subsequent research on the relationship between EBF and PNDs. Moreover, within-person associations may help identify modifiable targets for intervention [66]. Our findings at the within-person level also suggested that future studies could consider BSE as an intervention target to construct effective intervention strategies in PNDs.

In contrast to our expectations, no direct longitudinal association was found between PNDs and EBF. This finding was consistent with the CLPM results of Haga et al.’s (2018) [16] population-based study, which also reported no significant concurrent associations between EBF and depressive symptoms at 4, 6, or 12 months postpartum, and contradicted the previous findings of Zhu et al. (2023) [17]. This variation could be partially attributed to the differences in measurement timing and confounding factors, particularly with the inclusion of our third variable, BSE. It also underscored the complexity of the relationship between PNDs and EBF. On the one hand, a longitudinal relationship might exists between PNDs and EBF but is masked by the mediating variable BSE, resulting in the lack of statistical significance in the results [49, 71, 72]. In addition, PNDs and EBF have many other potential influences not addressed in this study [73, 74], and these factors might modify or mediate their relationship, resulting in statistical insignificance. On the other hand, there might be no direct effect of PNDs and EBF but only an indirect effect through BSE. This study validated both of these possibilities through the between- and within-person results.

Strengths and limitations

This study is the first to verify the longitudinal mediating or suppressing role of BSE on the longitudinal association between PNDs and EBF. Given that the longitudinal processes mediating the association between PNDs and EBF remain largely unexplored, this study extended existing literature on the temporal relationship between PNDs and EBF by assessing the potential mediating mechanisms. By using a combination of CLPMs and RI-CLPMs, we found robust empirical evidence supporting the temporal relationships between PNDs and EBF. This methodology has advantages over cross-sectional and longitudinal designs that do not account for between-person variance or control for the prior level of outcomes. Furthermore, the differences between the results at the between- and within-person levels illustrated the importance of distinguishing among between- and within-person associations [75]. Lastly, we quantitatively clarified the longitudinal reciprocal associations between EBF and BSE. Bidirectional associations between BSE and EBF were observed at both the between- and within-person levels. Although the strong relationship between BSE and EBF has been widely reported [21, 76], our study was the first to validate their temporal relationship. Moreover, we found that the cross-lagged effects of EBF and BSE on each other were similar, and the effect sizes were large effect (β > 0.12) [77].

Our study also has some limitations. First, for ethical considerations, all women with EPDS scores ≥ 10 were informed of their screening results and prompted on how to self-regulate their emotions, and those with EPDS scores ≥ 13 were advised to visit the psychiatric department of the survey hospital. This phenomenon might have had an impact on the natural progression of PNDs. Second, even though we observed the interesting and significant results on the longitudinal associations among PNDs, EBF, and BSE, the sample size of the study was small. Third, although participants were provided with detailed explanations of EBF criteria in the questionnaire, the EBF degree index was based on a single self-reported question about breastfeeding types. This might miss subtle practices like occasional water or liquid supplementation, risking misclassification. Future research should use objective feeding logs or multi-item assessments for greater accuracy. Fourth, we examined mother-infant hospitalization type (rooming-in vs. separation) rather than directly assessing skin-to-skin contact, early breastfeeding initiation, or prelacteal feeding. In China, these practices are standardized under Baby-Friendly Hospital Initiative (BFHI) accreditation for hospitals implementing rooming-in [78, 79]. Although the surveyed hospital was BFHI-certified, indirect assessment may limit the generalizability of findings. Lastly, cultural differences influence the association between PNDs and EBF [16, 17], and the flexibility of individuals’ levels of PNDs over time vary across cultures [33]. Given that the samples were all drawn from Chinese women, readers should exercise caution when generalizing the findings to other contexts.

Relevance for clinical practice

This study has implications for clinical work in the improvement of EBF and PNDs. By understanding the mediating effect of BSE at the between-person level and its suppressing effect at the within-person level, health professionals can develop more effective intervention strategies targeting different dimensions. First, health professionals must identify the subpopulation within the general population that has higher PND levels and therefore at a high risk for lower EBF scores. Next, for individuals in this high-risk population, BSE can be targeted for intervention. Our within-person results indicated that BSE suppresses the relations from EBF to PNDs, and there is a bidirectional reciprocal relationship between BSE and EBF. Therefore, improving the BSE levels of a woman in the high-risk group can increase her subsequent EBF and PNDs. Enhancing BSE levels also can reveal the benefits of increased EBF in subsequently reducing PNDs, thereby decreasing long-term PND levels. When developing intervention strategies, an interim evaluation of BSE must be designed. This is because even if EBF increases shortly after the intervention, long-term PND reduction will not be achieved unless the increase in BSE is sustained.

Conclusions

Our study demonstrates that BSE mediates the relationship between PNDs and EBF. Specifically, BSE explains why mothers with depressive symptoms are less likely to exclusively breastfeed. However, when examining individual changes over time, we observed that the mental health benefits of breastfeeding are likely to be lost unless BSE improves. These findings emphasize the importance of considering both population-level patterns and individual-level changes. Improving BSE could simultaneously address both low breastfeeding rates and depressive symptoms. Interventions targeting BSE may offer dual benefits for both breastfeeding outcomes and mental health, but progress should be regularly monitored during implementation. Future studies should test practical strategies to boost BSE in mothers experiencing depression.

Data availability

The datasets generated during this study are not publicly available due to the desire to protect patient confidentiality in this small longitudinal study but de-identified data may be made available from the corresponding author on reasonable request and with institutional permission regarding data sharing.

Abbreviations

EBF:

Exclusive breastfeeding

PNDs:

Perinatal depressive symptoms

BSE:

Breastfeeding self-efficacy

CLPM:

Cross-lagged panel model

RI-CLPM:

Random intercept cross-lagged panel model

SEM:

Structural equation models

EPDS:

Edinburgh Postnatal Depression Scale

BSES-SF:

Breastfeeding Self-Efficacy Scale-Short Form

RMSEA:

Root mean square error of approximation

CFI:

Comparative fit index

TLI:

Tucker-Lewis index

SRMR:

Standardized root mean square residual

SD:

Standard deviation

MCAR:

Missing completely at random

References

  1. Meek JY, Noble L. Policy statement: breastfeeding and the use of human milk. Pediatrics. 2022;150(1): e2022057988.

    Article  PubMed  Google Scholar 

  2. WHO. Breastfeeding: World Health Organization; 2024. Available from: https://www.who.int/health-topics/breastfeeding#tab=tab_1.

  3. Yalçın SS, Kuşkonmaz BB. Relationship of lower breastfeeding score and problems in infancy. Breastfeed Med. 2011;6(4):205–8.

    Article  PubMed  Google Scholar 

  4. WHO. Infant and young child feeding: World Health Organization; 2021. Available from: https://www.who.int/en/news-room/fact-sheets/detail/infant-and-young-child-feeding. Cited 2022 08/27.

  5. Victora CG, Bahl R, Barros AJ, França GV, Horton S, Krasevec J, et al. Breastfeeding in the 21st century: epidemiology, mechanisms, and lifelong effect. Lancet (London, England). 2016;387(10017):475–90.

    Article  PubMed  Google Scholar 

  6. CDRF. Survey report on factors affecting breastfeeding in China: China Development Research Foundation; 2019. Available from: https://www.cdrf.org.cn/jjh/pdf/mu.pdf.

  7. Woody CA, Ferrari AJ, Siskind DJ, Whiteford HA, Harris MG. A systematic review and meta-regression of the prevalence and incidence of perinatal depression. J Affect Disord. 2017;219:86–92.

    Article  CAS  PubMed  Google Scholar 

  8. Nisar A, Yin J, Waqas A, Bai X, Wang D, Rahman A, et al. Prevalence of perinatal depression and its determinants in Mainland China: a systematic review and meta-analysis. J Affect Disord. 2020;277:1022–37.

    Article  PubMed  Google Scholar 

  9. Becker M, Weinberger T, Chandy A, Schmukler S. Depression during pregnancy and postpartum. Curr Psychiatry Rep. 2016;18(3):32.

    Article  PubMed  Google Scholar 

  10. Woolhouse H, James J, Gartland D, McDonald E, Brown SJ. Maternal depressive symptoms at three months postpartum and breastfeeding rates at six months postpartum: implications for primary care in a prospective cohort study of primiparous women in Australia. Women Birth. 2016;29(4):381–7.

    PubMed  Google Scholar 

  11. Cato K, Sylvén SM, Georgakis MK, Kollia N, Rubertsson C, Skalkidou A. Antenatal depressive symptoms and early initiation of breastfeeding in association with exclusive breastfeeding six weeks postpartum: a longitudinal population-based study. BMC Pregnancy Childbirth. 2019;19:1–1.

    CAS  Google Scholar 

  12. Kim S, Park M, Ahn S. The impact of antepartum depression and postpartum depression on exclusive breastfeeding: a systematic review and meta-analysis. Clin Nurs Res. 2021;31(5):866–80.

    PubMed  Google Scholar 

  13. Yalçin SS, Orün E. Breastfeeding status and maternal psychopathologies: in a longitudinal study. Arch Dis Child. 2011;96(9):900.

    PubMed  Google Scholar 

  14. Alimi R, Azmoude E, Moradi M, Zamani M. The association of breastfeeding with a reduced risk of postpartum depression: a systematic review and meta-analysis. Breastfeed Med. 2022;17(4):290–6.

    PubMed  Google Scholar 

  15. Nam JY, Choi Y, Kim J, Cho KH, Park EC. The synergistic effect of breastfeeding discontinuation and cesarean section delivery on postpartum depression: a nationwide population-based cohort study in Korea. J Affect Disord. 2017;218:53–8.

    PubMed  Google Scholar 

  16. Haga SM, Lisøy C, Drozd F, Valla L, Slinning K. A population-based study of the relationship between perinatal depressive symptoms and breastfeeding: a cross-lagged panel study. Arch Womens Ment Health. 2018;21(2):235–42.

    PubMed  Google Scholar 

  17. Zhu Y, Yin X, Qiu L, Sun N, An R, Gong Y. Association between breastfeeding and perinatal depressive symptoms: a 13-months cross-lagged analysis in China. Asian J Psychiatr. 2023;82: 103474.

    PubMed  Google Scholar 

  18. Dennis CL. Theoretical underpinnings of breastfeeding confidence: a self-efficacy framework. J Hum Lact. 1999;15(3):195–201.

    CAS  PubMed  Google Scholar 

  19. Lau CYK, Lok KYW, Tarrant M. Breastfeeding duration and the theory of planned behavior and breastfeeding self-efficacy framework: a systematic review of observational studies. Matern Child Health J. 2018;22(3):327–42.

    PubMed  Google Scholar 

  20. Kronborg H, Foverskov E. Multifactorial influence on duration of exclusive breastfeeding; a Danish cohort study. PLoS ONE. 2020;15(9): e0238363.

    CAS  PubMed  PubMed Central  Google Scholar 

  21. Tuthill EL, Miller JD, Collins SM, Widen EM, Onono M, Young SL. HIV infection, hunger, breastfeeding self-efficacy, and depressive symptoms are associated with exclusive breastfeeding to six months among women in western Kenya: a longitudinal observational study. Int Breastfeed J. 2020;15(1):4.

    PubMed  PubMed Central  Google Scholar 

  22. Mercan Y, Tari SK. Association between postpartum depression level, social support level and breastfeeding attitude and breastfeeding self-efficacy in early postpartum women. PLoS ONE. 2021;16(4): e0249538.

    CAS  PubMed  PubMed Central  Google Scholar 

  23. Zubaran C, Foresti K. The correlation between breastfeeding self-efficacy and maternal postpartum depression in southern Brazil. Sexual & Reproductive Healthcare. 2013;4(1):9–15.

    Google Scholar 

  24. Ngo LTH, Chou HF, Gau ML, Liu CY. Breastfeeding self-efficacy and related factors in postpartum Vietnamese women. Midwifery. 2019;70:84–91.

    PubMed  Google Scholar 

  25. Tsaras K, Sorokina T, Papathanasiou IV, Fradelos EC, Papagiannis D, Koulierakis G. Breastfeeding self-efficacy and related socio-demographic, perinatal and psychological factors: a cross-sectional study among postpartum Greek women. Materia socio-medica. 2021;33(3):206–12.

    PubMed  PubMed Central  Google Scholar 

  26. Ahmad Zadeh Beheshti M, Alimoradi Z, Bahrami N, Allen KA, Lissack K. Predictors of breastfeeding self-efficacy during the covid-19 pandemic. J Neonatal Nurs. 2022;28(5):349–55.

    PubMed  Google Scholar 

  27. Karaburun IEG, Yalçın SS. Breast refusal: an analysis of frequency, onset timing, recovery status, and their interplay with breastfeeding self-efficacy and maternal depression. BMC Public Health. 2024;24(1):3568.

    PubMed  PubMed Central  Google Scholar 

  28. Henshaw EJ. Breastfeeding and postpartum depression: a review of relationships and potential mechanisms. Curr Psychiatry Rep. 2023;25(12):803–8.

    PubMed  Google Scholar 

  29. Hamaker EL, Kuiper RM, Grasman RP. A critique of the cross-lagged panel model. Psychol Methods. 2015;20(1):102–16.

    PubMed  Google Scholar 

  30. Lucas RE. Why the cross-lagged panel model is almost never the right choice. Adv Methods Pract Psycholog Sci. 2023;6(1):25152459231158376.

    Article  Google Scholar 

  31. Bandura A, Freeman WH, Lightsey R. Self-efficacy: the exercise of control. J Cogn Psychother. 1997;13:158–66.

    Article  Google Scholar 

  32. Ram N, Gerstorf D. Time-structured and net intraindividual variability: tools for examining the development of dynamic characteristics and processes. Psychol Aging. 2009;24(4):778–91.

    Article  PubMed  PubMed Central  Google Scholar 

  33. Baron E, Bass J, Murray SM, Schneider M, Lund C. A systematic review of growth curve mixture modelling literature investigating trajectories of perinatal depressive symptoms and associated risk factors. J Affect Disord. 2017;223:194–208.

    Article  PubMed  PubMed Central  Google Scholar 

  34. Fredriksen E, von Soest T, Smith L, Moe V. Patterns of pregnancy and postpartum depressive symptoms: latent class trajectories and predictors. J Abnorm Psychol. 2017;126(2):173–83.

    Article  PubMed  Google Scholar 

  35. Moshagen M, Bader M. semPower: general power analysis for structural equation models. Behav Res Methods. 2024;56(4):2901–22.

    Article  PubMed  Google Scholar 

  36. MacCallum RC, Browne MW, Sugawara HM. Power analysis and determination of sample size for covariance structure modeling. Psychol Methods. 1996;1(2):130–49.

    Article  Google Scholar 

  37. Jak S, Jorgensen TD, Verdam MGE, Oort FJ, Elffers L. Analytical power calculations for structural equation modeling: a tutorial and Shiny app. Behav Res Methods. 2021;53(4):1385–406.

    Article  PubMed  Google Scholar 

  38. Krosnick JA. Questionnaire Design. In: Vannette DL, Krosnick JA, editors. The palgrave handbook of survey research. Cham: Springer International Publishing; 2018. p. 439–55.

    Chapter  Google Scholar 

  39. Cox JL, Holden JM, Sagovsky R. Detection of postnatal depression. Development of the 10-item Edinburgh Postnatal Depression Scale. Br J Psychiatry. 1987;150(6):782–6.

    Article  CAS  PubMed  Google Scholar 

  40. Lee DT, Yip SK, Chiu HF, Leung TY, Chan KP, Chau IO, et al. Detecting postnatal depression in Chinese women. Validation of the Chinese version of the Edinburgh Postnatal Depression Scale. Br J Psychiatry. 1998;172:433–7.

    CAS  PubMed  Google Scholar 

  41. Wang Y, Guo X, Lau Y, Chan KS, Yin L, Chen J. Psychometric evaluation of the Mainland Chinese version of the Edinburgh Postnatal Depression Scale. Int J Nurs Stud. 2009;46(6):813–23.

    PubMed  Google Scholar 

  42. Dennis CL. The breastfeeding self-efficacy scale: psychometric assessment of the short form. J Obstet Gynecol Neonatal Nurs. 2003;32(6):734–44.

    PubMed  Google Scholar 

  43. Dai X, Dennis CL. Translation and validation of the breastfeeding self-efficacy scale into Chinese. J Midwifery Womens Health. 2003;48(5):350–6.

    PubMed  Google Scholar 

  44. Barnes J, Stein A, Smith T, Pollock JI. Extreme attitudes to body shape, social and psychological factors and a reluctance to breast feed. ALSPAC Study Team. Avon longitudinal study of pregnancy and childhood. J R Soc Med. 1997;90(10):551–9.

    CAS  PubMed  PubMed Central  Google Scholar 

  45. WHO. Indicators for assessing infant and young child feeding practices (Part 1: Definitions): conclusions of a consensus meeting held 6–8 November 2007 in Washington D.C., USA. Geneva, Switzerland: WHO; 2008.

  46. Rhemtulla M, Brosseau-Liard P, Savalei V. When can categorical variables be treated as continuous? A comparison of robust continuous and categorical SEM estimation methods under suboptimal conditions. Psychol Methods. 2012;17(3):354–73.

    PubMed  Google Scholar 

  47. Zhao M. Factors associated with breastfeeding self-efficacy among Beijing new mothers. Beijing: Peking Union Medical College; 2008.

    Google Scholar 

  48. De la Mora A, Russell DW. The Iowa Infant feeding attitude scale: analysis of reliability and validity. J Appl Soc Psychol. 1999;29(11):2362–80.

    Google Scholar 

  49. Liu L, Xiao G, Zhang T, Zhou M, Li X, Zhang Y, et al. Levels and determinants of antenatal breastfeeding attitudes among pregnant women: a cross-sectional study. Children (Basel). 2023;10(2):275.

    PubMed  PubMed Central  Google Scholar 

  50. Dai HX, Guan XD, Li XM, You LM, Lau Y. Psychometric properties of a mainland Chinese version of the Iowa Infant Feeding Attitude Scale among postpartum women in China. Contemp Nurse. 2013;44(1):11–20.

    PubMed  Google Scholar 

  51. Orth U, Clark DA, Donnellan MB, Robins RW. Testing prospective effects in longitudinal research: comparing seven competing cross-lagged models. J Pers Soc Psychol. 2021;120(4):1013–34.

    PubMed  Google Scholar 

  52. Hihara S, Umemura T, Iwasa Y, Saiga S, Sugimura K. Identity processes and identity content valences: examining bidirectionality. Dev Psychol. 2021;57(12):2265–80.

    PubMed  Google Scholar 

  53. Little RJA. A test of missing completely at random for multivariate data with missing values. J Am Statal Assoc. 1988;83(404):1198–202.

    Google Scholar 

  54. Little TD, Cunningham WA, Shahar G, Widaman KF. To parcel or not to parcel: exploring the question, weighing the merits. Struct Equ Model. 2002;9(2):151–73.

    Google Scholar 

  55. Landis RS, Beal DJ, Tesluk PE. A comparison of approaches to forming composite measures in structural equation models. Organ Res Methods. 2000;3(2):186–207.

    Google Scholar 

  56. Little TD. Longitudinal structural equation modeling. New York: Guilford Press; 2013.

    Google Scholar 

  57. Stenling A, Ivarsson A, Lindwall M, Gucciardi DF. Exploring longitudinal measurement invariance and the continuum hypothesis in the Swedish version of the Behavioral Regulation in Sport Questionnaire (BRSQ): an exploratory structural equation modeling approach. Psychol Sport Exerc. 2018;36:187–96.

    Google Scholar 

  58. Hooper D, Mullen J, Hooper D, Coughlan J, Mullen MR. Structural equation modeling: guidelines for determining model fit. Electronic Journal on Business Research Methods. 2008;6(1):53–60.

    Google Scholar 

  59. Chen FF. Sensitivity of goodness of fit indexes to lack of measurement invariance. Struct Equ Modeling. 2007;14(3):464–504.

    Google Scholar 

  60. MacKinnon DP, Krull JL, Lockwood CM. Equivalence of the mediation, confounding and suppression effect. Prevent Sci. 2000;1(4):173–81.

    CAS  Google Scholar 

  61. Wen Z, Ye B. Analyses of mediating effects: the development of methods and models. Adv Psycholog Sci. 2014;22(5):731.

    Google Scholar 

  62. Rosenbaum DL, Gillen MM, Markey CH. Feeling let down: an investigation of breastfeeding expectations, appreciation of body functionality, self-compassion, and depression symptoms. Appetite. 2020;154: 104756.

    PubMed  Google Scholar 

  63. Liao KY-H, Stead GB, Liao C-Y. A meta-analysis of the relation between self-compassion and self-efficacy. Mindfulness. 2021;12(8):1878–91.

    Google Scholar 

  64. Papinczak TA, Turner CT. An analysis of personal and social factors influencing initiation and duration of breastfeeding in a large Queensland maternity hospital. Breastfeed Rev. 2000;8(1):25–33.

    CAS  PubMed  Google Scholar 

  65. Flores-Quijano ME, Córdova A, Contreras-Ramírez V, Farias-Hernández L, Cruz Tolentino M, Casanueva E. Risk for postpartum depression, breastfeeding practices, and mammary gland permeability. J Hum Lact. 2008;24(1):50–7.

    PubMed  Google Scholar 

  66. Masselink M, Van Roekel E, Hankin BL, Keijsers L, Lodder GMA, Vanhalst J, et al. The longitudinal association between self-esteem and depressive symptoms in adolescents: separating between-person effects from within-person effects. Eur J Pers. 2018;32(6):653–71.

    CAS  PubMed  PubMed Central  Google Scholar 

  67. Haga SM, Ulleberg P, Slinning K, Kraft P, Steen TB, Staff A. A longitudinal study of postpartum depressive symptoms: multilevel growth curve analyses of emotion regulation strategies, breastfeeding self-efficacy, and social support. Arch Womens Ment Health. 2012;15(3):175–84.

    PubMed  Google Scholar 

  68. Jiang Q, Zhang E, Cohen N, Ohtori M, Zhu S, Guo Y, et al. Postnatal mental health, breastfeeding beliefs, and breastfeeding practices in rural China. Int Breastfeed J. 2022;17(1):60.

    PubMed  PubMed Central  Google Scholar 

  69. Agbaje AO, Perng W, Tuomainen TP. Effects of accelerometer-based sedentary time and physical activity on DEXA-measured fat mass in 6059 children. Nat Commun. 2023;14(1):8232.

    CAS  PubMed  PubMed Central  Google Scholar 

  70. Liu L. The impact of perinatal depression on exclusive breastfeeding. China: Central South University; 2023.

    Google Scholar 

  71. Liu L, Zhou M, Xiao G, Zhang T, Li X, Owusua T, et al. The impact of antenatal depressive symptoms on exclusive breastfeeding intention: a moderating effect analysis. Midwifery. 2023;116: 103551.

    Article  PubMed  Google Scholar 

  72. Feng G, Peng J, Tu D, Zheng JZ, Feng C. Two paradoxes in linear regression analysis. Shanghai Arch Psychiatry. 2016;28(6):355–60.

    PubMed  PubMed Central  Google Scholar 

  73. Yin X, Sun N, Jiang N, Xu X, Gan Y, Zhang J, et al. Prevalence and associated factors of antenatal depression: Systematic reviews and meta-analyses. Clin Psychol Rev. 2021;83: 101932.

    Article  PubMed  Google Scholar 

  74. Rollins NC, Bhandari N, Hajeebhoy N, Horton S, Lutter CK, Martines JC, et al. Why invest, and what it will take to improve breastfeeding practices? Lancet. 2016;387(10017):491–504.

    Article  PubMed  Google Scholar 

  75. Nelemans SA, Keijsers L, Colpin H, van Leeuwen K, Bijttebier P, Verschueren K, et al. Transactional links between social anxiety symptoms and parenting across adolescence: between- and within-person associations. Child Dev. 2020;91(3):814–28.

    Article  PubMed  Google Scholar 

  76. De Roza JG, Fong MK, Ang BL, Sadon RB, Koh EYL, Teo SSH. Exclusive breastfeeding, breastfeeding self-efficacy and perception of milk supply among mothers in Singapore: a longitudinal study. Midwifery. 2019;79:102532.

    Article  PubMed  Google Scholar 

  77. Orth U, Meier LL, Bühler JL, Dapp LC, Krauss S, Messerli D, et al. Effect size guidelines for cross-lagged effects. Psychol Methods. 2024;29(2):421–33.

    Article  PubMed  Google Scholar 

  78. UNICEF/WHO. Baby-friendly hospital initiative: revised, updated, and expanded for integrated care. Geneva: WHO; 2009.

  79. NHC. Guidelines for the Management and Supervision of Baby-Friendly Hospitals. Beijing: National Health Commission of China (NHC); 2004 Available from: http://www.nhc.gov.cn/wjw/gfxwj/201304/b77d9959d2cd404498c22e2e4caa978f.shtml.

Download references

Acknowledgements

The authors would like to express their sincere gratitude to the instructors, Qirong Chen, Jiarui Chen, Jinnan Xiao, and Ci Zhang, from the English Medical Paper Editing and Writing course. Their excellent teaching and guidance, especially the constructive feedback from Associate Professor Qirong Chen, significantly contributed to the writing and revision of this manuscript. The authors also acknowledge the support of the Obstetric Departments of the Third Xiangya Hospital of Central South University and the participants, nurses, midwives, and the hard work of the team members of the study.

Funding

This work was supported by the National Natural Science Foundation of China (grant number 72074225), the Key Research and Development Program of Hunan Province of China (grant number 2021SK2024), and the Fundamental Research Funds for the Central Universities of Central South University (grant number 2024ZZTS0184). The funding sources played no role in conducting the research and preparing the article.

Author information

Authors and Affiliations

  1. Health Management Medicine Center and Nursing Department, The Third Xiangya Hospital, Central South University, Changsha, 410013, China

    Li Liu & Chunxiang Qin

  2. Xiangya School of Nursing, Central South University, Changsha, 410013, China

    Li Liu, Shuya Feng, Gui Xiao & Mengjia Zhou

  3. School of Medicine, Jishou University, Jishou, 416000, China

    Yu Zhang & Xingxing Li

  4. Department of Pediatrics, The Third Xiangya Hospital, Central South University, Changsha, 410013, China

    Ying Li

Authors
  1. Li Liu
    View author publications

    You can also search for this author inPubMed Google Scholar

  2. Shuya Feng
    View author publications

    You can also search for this author inPubMed Google Scholar

  3. Yu Zhang
    View author publications

    You can also search for this author inPubMed Google Scholar

  4. Gui Xiao
    View author publications

    You can also search for this author inPubMed Google Scholar

  5. Mengjia Zhou
    View author publications

    You can also search for this author inPubMed Google Scholar

  6. Xingxing Li
    View author publications

    You can also search for this author inPubMed Google Scholar

  7. Ying Li
    View author publications

    You can also search for this author inPubMed Google Scholar

  8. Chunxiang Qin
    View author publications

    You can also search for this author inPubMed Google Scholar

Contributions

LL: Conceptualization, formal analysis, funding acquisition, investigation, methodology, project administration, writing – original draft. SF: Investigation, methodology, project administration, validation, writing – review & editing. YZ: Investigation, methodology, validation, writing – review & editing. GX: Data curation, investigation, validation, writing – review & editing. MZ: Investigation, methodology, writing – review & editing. XL: Investigation, project administration, writing – review & editing. YL: Methodology, project administration, resources, supervision, writing – review & editing. CQ: Funding acquisition, project administration, resources, supervision, writing – review & editing.

Corresponding author

Correspondence to Chunxiang Qin.

Ethics declarations

Ethics approval and consent to participate

Ethical approval was obtained from the Ethics Review Committee of Xiangya Nursing School of Central South University (approval number: E202159). Informed written consent was explained in detail to all the participants to ensure that they were fully aware of the project and agree completely voluntarily. All procedures were conducted in strict compliance with the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

12884_2025_7481_MOESM1_ESM.docx

Additional file 1. Table 1. Model fits and comparisons for cross-lagged panel models (CLPMs) and random intercepts cross-lagged panel models (RI-CLPMs).

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Liu, L., Feng, S., Zhang, Y. et al. The role of breastfeeding self-efficacy in the relationship between perinatal depressive symptoms and exclusive breastfeeding: a longitudinal mediation analysis. BMC Pregnancy Childbirth 25, 356 (2025). https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12884-025-07481-8

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12884-025-07481-8

Keywords

BMC Pregnancy and Childbirth

ISSN: 1471-2393

Contact us