Advertisement
  • Loading metrics

Postpartum depressive symptoms following implementation of the 10 steps to successful breastfeeding program in Kinshasa, Democratic Republic of Congo: A cohort study

  • Robert A. Agler,

    Roles Conceptualization, Data curation, Formal analysis, Writing – original draft, Writing – review & editing

    Affiliations Division of Epidemiology, College of Public Health, The Ohio State University, Ohio, United States of America, Department of Psychology, The Ohio State University, Ohio, United States of America

  • Paul N. Zivich,

    Roles Data curation, Formal analysis, Writing – review & editing

    Affiliations Department of Epidemiology, University of North Carolina at Chapel Hill, North Carolina, United States of America, Carolina Population Center, University of North Carolina at Chapel Hill, North Carolina, United States of America

  • Bienvenu Kawende,

    Roles Data curation, Investigation, Supervision, Writing – review & editing

    Affiliation School of Public Health, The University of Kinshasa, Kinshasa, Democratic Republic of Congo

  • Frieda Behets,

    Roles Funding acquisition, Investigation, Supervision, Writing – review & editing

    Affiliation Department of Epidemiology, University of North Carolina at Chapel Hill, North Carolina, United States of America

  • Marcel Yotebieng

    Roles Conceptualization, Data curation, Funding acquisition, Investigation, Project administration, Supervision, Writing – original draft, Writing – review & editing

    Marcel.yotebieng@einsteinmed.org, myotebieng@gmail.com

    Affiliations Division of Epidemiology, College of Public Health, The Ohio State University, Ohio, United States of America, Division of General Internal Medicine, Albert Einstein College of Medicine, New York, United States of America

Postpartum depressive symptoms following implementation of the 10 steps to successful breastfeeding program in Kinshasa, Democratic Republic of Congo: A cohort study

  • Robert A. Agler, 
  • Paul N. Zivich, 
  • Bienvenu Kawende, 
  • Frieda Behets, 
  • Marcel Yotebieng
PLOS
x

Abstract

Background

Social support and relevant skills training can reduce the risk of postpartum depression (PPD) by reducing the impact of stressors. The 10-step program to encourage exclusive breastfeeding that forms the basis of the Baby-Friendly Hospital Initiative (BFHI) provides both, suggesting it may lessen depressive symptoms directly or by reducing difficulties associated with infant feeding. Our objective was to quantify the association of implementing Steps 1–9 or Steps 1–10 on postpartum depressive symptoms and test whether this association was mediated by breastfeeding difficulties.

Methods and findings

We used data from a breastfeeding promotion trial of all women who gave birth to a healthy singleton between May 24 and August 25, 2012 in 1 of the 6 facilities comparing different BFHI implementations (Steps 1–9, Steps 1–10) to the standard of care (SOC) randomized by facility in Kinshasa, Democratic Republic of Congo. Depressive symptoms, a non-registered trial outcome, was assessed at 14 weeks via the Edinburgh Postnatal Depression Scale (EPDS). Inverse probability weighting (IPW) was used to estimate the association of BFHI implementations on depressive symptoms and the controlled direct association through breastfeeding difficulties at 10 weeks postpartum.

A total of 903 mother–infant pairs were included in the analysis. Most women enrolled had previously given birth (76%) and exclusively breastfed at 10 weeks (55%). The median age was 27 (interquartile range (IQR): 23, 32 years). The proportion of women reporting breastfeeding difficulties at week 10 was higher in both Steps 1–9 (75%) and Steps 1–10 (91%) relative to the SOC (67%). However, the number of reported difficulties was similar between Steps 1–9 (median: 2; IQR: 0, 3) and SOC (2; IQR: 0, 3), with slightly more in Steps 1–10 (2; IQR: 1, 3). The prevalence of symptoms consistent with probable depression (EPDS score >13) was 18% for SOC, 11% for Steps 1–9 (prevalence difference [PD] = −0.08; 95% confidence interval (CI): −0.14 to −0.01, p = 0.019), and 8% for Steps 1–10 (PD = −0.11, −0.16 to −0.05; p < 0.001). We found mediation by breastfeeding difficulties. In the presence of any difficulties, the PD was reduced for both Steps 1–9 (−0.15; 95% confidence level (CL): −0.25, −0.06; p < 0.01) and Steps 1–10 (−0.16; 95% CL: −0.25, −0.06; p < 0.01). If no breastfeeding difficulties occurred in the population, there was no difference in the prevalence of probable depression for Steps 1–9 (0.21; 95% CL: −0.24, 0.66; p = 0.365) and Steps 1–10 (−0.03; 95% CL: −0.19, 0.13; p = 0.735). However, a limitation of the study is that the results are based on 2 hospitals randomized to each group.

Conclusions

In conclusion, in this cohort, the implementation of the BFHI steps was associated with a reduction in depressive symptoms in the groups implementing BFHI Steps 1–9 or 1–10 relative to the SOC, with the implementation of Steps 1–10 associated with the largest decrease. Specifically, the reduction in depressive symptoms was observed for women reporting breastfeeding difficulties. PPD has a negative impact on the mother, her partner, and the baby, with long-lasting consequences. This additional benefit of BFHI steps suggests that renewed effort to scale its implementation globally may be beneficial to mitigate the negative impacts of PPD on the mother, her partner, and the baby.

Trial registration

ClinicalTrials.gov NCT01428232

Author summary

Why was this study done?

  • Depression is common in the postpartum period. Untreated, postpartum depression (PPD) has substantial negative impact on the mother, her partner, and the baby, with long-lasting consequences.
  • Social support and relevant skills training can reduce the risk of PPD by reducing the impact of stressors.
  • The 10-step program to encourage exclusive breastfeeding that forms the basis of the Baby-Friendly Hospital Initiative (BFHI) provides both, but its effect on depression symptoms in postpartum period has not been evaluated.

What did the researchers do and find?

  • We found that depressive symptoms were reduced in the groups implementing BFHI Steps 1–9 or 1–10 relative to the standard of care (SOC), with the implementation of Steps 1–10 associated with the largest decrease.
  • The reduction in depressive symptoms associated with the implementation of BHFI Steps 1–9 or Steps 1–10 was observed for women reporting breastfeeding difficulties, but not for women reporting no difficulties.

What do these findings mean?

  • Our results suggest that the implementation of the BFHI was associated with a reduction in maternal depressive symptoms and that breastfeeding difficulties play a role in this relationship.

Introduction

The postpartum period is associated with numerous physical, emotional, social, and financial stressors, many of which increase the risk of postpartum depression (PPD) [1,2]. Infant feeding difficulties are a particularly common source of stress, with some 60% to 90% of women reporting difficulties with breastfeeding (e.g., pain, proper infant positioning, etc.), leading to an increased risk of PPD, and uncertainty and ambivalence about different methods of feeding [3,4]. The 10 steps to successful breastfeeding, i.e., the basis of the Baby-Friendly Hospital Initiative (BFHI; Table 1), helps to address many issues tied to breastfeeding difficulties, as recent mothers are given a combination of knowledge, skills training, encouragement, and social support to facilitate optimal feeding practices [5]. This reduction in stress provides a pathway through which the BFHI may reduce symptoms of PPD, but this possibility has not been assessed.

To further elaborate, parallels can be drawn between the BFHI and psychotherapeutic interventions, although we do not claim that the BFHI can replace therapy. Both identify stressors, reframe them in a less stressful way, teach how to manage the stressors, and provide encouragement and social support to motivate people to effectively deal with stressors [5,6]. For psychotherapy, the result is that patients express greater motivation, successful goal achievement, and are less discouraged by stressors. Mirroring this pattern, mothers that receive breastfeeding interventions express greater intention to breastfeed, higher rates of exclusive breastfeeding, and are more likely to persist when encountering breastfeeding difficulties [7,8]. Both also provide an opportunity to openly discuss and address potential sources of shame and uncertainty. Psychiatric illnesses and infant feeding practices are also both potentially stigmatizing, as both are often heavily moralized [9,10]. Shame and threat of social rejection are risk factors for the development of psychiatric disorders and a barrier to their management [9]. Mothers regularly report worries of judgment and shaming, as well as instances of outright shaming, that can make mothers feel isolated or inadequate [10]. Open, honest, and judgment-free discussions of shameful topics helps reduce any feelings of shame they may cause [11]. Additionally, simply being uncertain and afraid of the unknown (e.g., concerns about harm to the child based on feeding method) is a risk factor for the development of psychiatric disorders [12].

The community breastfeeding support groups (Step 10) provide an ongoing source of support and knowledge beyond what hospital staff provides at follow-up visits. Support group meetings also last far longer than the brief interactions possible at a hospital, allowing for deeper discussion of ongoing or recent issues and for further provision of both objective (e.g., skills and knowledge) and subjective social support (e.g., encouragement and conveying that the mothers’ problems are neither unique nor indicative of moral character). Additionally, while social support from relative strangers, such as hospital staff, is helpful, social support provided by familiar, closer people provides stronger benefits [13,14]. Further, the opportunity to help others may also reduce stress, as prosocial behaviors result in psychological benefits for the giver [15]. In sum, the BFHI provides mothers with a variety of sources and aid to help with common postpartum stressors that have been shown to increase the risk of PPD. Further, by increasing early initiation (within 1 hour of birth) and prolonged exclusive breastfeeding [16], BFHI interventions are likely associated with increased levels of oxytocin among mothers. Higher levels of oxytocin are associated with lower reported breastfeeding difficulties and improved maternal mood [17,18].

The 10 steps can thus be broken down into 3 groups. Steps 1 and 2 are meant to provide the overall structure of breastfeeding support and care, establishing facility-level policies and guidelines, as well as their implementation. Steps 3–9 provide in-hospital care, with hospital staff encouraging the initiation of breastfeeding, as well as providing information about its benefits to mothers. Step 10 is about community-related care, with hospital staff providing referrals to community-based breastfeeding support groups that can serve as sources of additional support and care for breastfeeding. Combined with the effects of exclusive breastfeeding, which also reduces the risk of PPD [19], the implementation of the 10 steps is expected to reduce depressive symptoms. However, to the authors’ knowledge, the effects of the BFHI on postpartum depressive symptoms have not been previously investigated.

Our objective was to quantify the association of implementing Steps 1–9 or Steps 1–10 on postpartum depressive symptoms. Specifically, we expected that Steps 1–9 (in-hospital care) would be associated with a decrease in depressive symptoms and that when combined with Step 10 (community support groups), the decrease would be larger. Reflecting the hypothesized reduction in stress, we also expected that breastfeeding difficulties would mediate the association of the BFHI on depressive symptoms.

Materials and methods

Data come from a cluster randomized trial of breastfeeding promotion in Kinshasa, Democratic Republic of Congo (NCT01428232) [16]. Depression was not included in registered trial outcomes, and although we have anticipated looking at the effect of the interventions on the prevalence of depressive symptoms, the plan for this analysis was not included in the study protocol (S1 Protocol) and was developed after the trial was funded. The primary purpose of the original study was to investigate whether a partial implementation without Step 10 would be effective in promoting exclusive breastfeeding and whether training nurses to provide additional support to breastfeeding mothers could be used as an alternative implementation of Step 10. Analyses of the primary and secondary outcomes of the trial showed that the implementation of basic training in BFHI Steps 1–9 had no additional effect on initiation of breastfeeding but significantly increased exclusive breastfeeding at 6 months of age and significantly reduced incidence of mild and severe episodes of diarrhea and respiratory infection in the first 6 months of life. Additional support based on the same training materials and locally available breastfeeding support materials, offered during well-child visits (i.e., Step 10), appeared to lessen this effect [16,20].

Six clinics grouped in pair based on their similar workload, number of deliveries, and proportion of mothers returning for their 1 week postpartum visit were randomized to either standard of care (SOC), Steps 1–9 group, or Steps 1–10 group (Fig 1). To inform the selection and pairing, study staff visited 44 facilities in Kinshasa between October and November 2011 and collected information about factors that might affect the quality of care provided in each facility. This selection process was intended to account for differences in important variables, such as utilization of services and socioeconomic status (SES), and to minimize the overall heterogeneity of clinic characteristics. Clinics were randomized rather than mother–infant pairs to minimize cross-group contamination and to better reflect the implementation of the BFHI in practice [16]. Randomization was performed by a statistician not involved in field activities. Staff providing the intervention could not be blinded, but independent interviewers and participants were blinded. The implementation of the BFHI was done by healthcare staff trained with WHO/UNICEF BFHI course. Culturally appropriate informational flyers in official (French) and local (Lingala) languages were provided at facilities randomized to Steps 1–10. The implementation of Steps 1–9 was assessed at the end of the study using the hospital self-appraisal questionnaire, and each of the clinics randomized to intervention groups met at least 80% of the global criteria for each step. In facilities randomized to SOC, besides study briefing prior to randomization, nothing else was provided aside from routine care. The SOC is described in detail elsewhere [21]. The average total number of deliveries in the year preceding the study was comparable across groups: 1,375 in the SOC compared to 1,520 and 1,291 in the Steps 1–9 and 1–10 groups, respectively. However, the average total number of healthcare personnel (physicians, nurses, and nurse assistant) was lower in the Steps 1–9 group (12) as compared to SOC (16) and Steps 1–10 (19) groups. Consequently, the average workload total (deliveries/total personnel) was substantially higher in Steps 1–9 group (144) relative to SOC (90) and Steps 1–10 (83) groups.

thumbnail
Fig 1. Flow diagram of follow-up of mother–infant pairs.

Mothers who missed the week 10 visit were considered censored at week 14 to ensure that censoring was monotonic. One reviewer in the Steps 1–9 group was believed to have incorrectly recorded the EPDS scores for participants. EPDS scores of participants assessed by this 1 staff member were set as missing. * Reported breastfeeding difficulties were assessed at week 10. ** Depressive symptoms were assessed as week 14 via the EPDS. EPDS, Edinburgh Postnatal Depression Scale; LTFU, lost to follow-up.

https://doi.org/10.1371/journal.pmed.1003465.g001

All women who gave birth to a healthy singleton in selected facilities between May 24 and August 25, 2012 were eligible to be enrolled in the study if they intended to attend well-baby clinic visits in the respective facilities. Women were enrolled within 2 days of delivery and follow-up at 6, 10, 14, 18, and 24 weeks postpartum.

Ethics statement

The original study design and implementation was approved by the University of North Carolina at Chapel Hill Institutional Review Board (IRB) and the Kinshasa School of Public Health Ethical Committee (see attached funded proposal in the Supporting information: S1 Protocol). All participants provided written informed consent. This analysis was approved by the Ohio State University IRB as secondary analysis of de-identified data.

Measures

At enrollment, demographics and obstetrical history were collected. Demographic information included age (continuous), education (primary; secondary or higher), marital status (married/live-in boyfriend; never married/separated/divorced), previous children (yes; no), prior miscarriage/abortion/stillbirth (yes; no), not wanting additional children (yes; no), at least 4 antenatal clinic visits before delivery (yes; no), and prior experiences of domestic violence (yes; no). Home ownership (yes; no) and a flush toilet facility (yes; no) were used as indicators of relative SES. The woman’s experience at the clinic was measured as being given their baby right after delivery (yes; no). Exclusive breastfeeding (yes; no) was assessed with retrospective recall for the previous week and defined as no food or water sources other than breastmilk at each follow-up visit. Proportion of feedings (continuous) that were breastfeeding was calculated as the reported number of breastfeeding divided by the total number of reported feedings for the prior 7 days. In predefined sensitivity analysis, the proportions of feedings that were breastfeeding within the past week at week 10 were considered instead of exclusive breastfeeding up to week 10. At each study visit, difficulties experienced with breastfeeding were defined as any self-reported problems with breastfeeding since the week 6 visit. The number of difficulties breastfeeding was calculated as a sum of all reported problems. For this analysis, only exclusive breastfeeding and difficulties breastfeeding at week 10, the visit prior to week 14 when depression symptoms were assessed are considered.

Depression symptoms were assessed at 14 weeks postpartum with the Edinburgh Postnatal Depression Scale (EPDS) [22]. The EPDS is a short 10-item screening tool to assess PPD symptoms that has been shown to have good sensitivity and specificity across many settings [23]. The original English scale was translated to both French and Lingala by translators fluent in all 3 languages and back translated to ensure fidelity. Scores of 13 or higher were considered to reflect probable depression [22]. To ensure the monotonicity of loss to follow-up, women who missed their week 10 visit were considered as censored and had their EPDS score set to missing. One study staff member had systematically different EPDS scores and was believed to have incorrectly recorded the EPDS scores for participants. Specifically, the items that were reverse-scored had systematically lower values than other reviewers. The participants at a clinic the Steps 1–9 arm assessed by this 1 staff member (n = 72) had their EPDS scores set to missing.

Statistical analysis

To estimate the 10-week prevalence difference (PD) in any breastfeeding difficulties, 10-week difference in number of breastfeeding difficulties, 14-week PD of postpartum probable depression (EPDS scores ≥13), and 14-week difference in EPDS scores, we used inverse probability of treatment weights (IPTWs) to account for potential baseline confounders. IPTWs account for potential confounders of the intervention and depression symptoms by creating a weighted population where observed confounders are no longer imbalanced between arms of the intervention [24]. Stabilized IPTWs were calculated as the unconditional probability of the intervention arm divided by the probability of the intervention arm conditional on observed confounders (see Supporting information, S1 IPW Calculation for details on stabilized IPTW calculation). Observed confounders included age, marital status, education, previous miscarriage, parity, whether the pregnancy was wanted, indicators of SES, antenatal care visits, and previous domestic violence. Probabilities for the intervention arms were generated from a multinomial logistic regression model. Age was modeled using restricted quadratic splines with knots at 20, 30, and 40.

To relax the assumption that loss to follow-up was non-informative, we used inverse probability of censoring weights (IPCWs) to account for potentially informative censoring by observed variables [25]. Stabilized IPCWs were calculated conditional on factors believed to be related to loss to follow-up. IPCWs for weeks 10 and 14 were constructed using a logistic regression model predicting censoring as a function of trial intervention arm, number of breastfeeding difficulties, demographics, and clinic experience.

Inverse probability weights (IPWs) were additionally used to estimate the controlled direct associations with probable depression at 14 weeks postpartum, where the controlled direct association corresponds to the change in probable depression if breastfeeding difficulties had uniformly been set to the same level in the population and BFHI steps implementations were compared [26]. Stabilized IPWs for mediation were calculated as the probability of the mediator conditional on the intervention arm divided by the probability of the mediator conditional on the intervention arm and observed confounders. For mediation by number of breastfeeding difficulties, a Poisson model was used to calculate probabilities. Since no women in the SOC group reported more than 4 difficulties, weights and models were estimated only for women reporting 4 or fewer difficulties. A logistic model was used for any breastfeeding difficulties as the mediator. IPWs were conditional on the same confounders as the intervention IPW and further included exclusive breastfeeding at 10 weeks and whether their baby was given right after delivery. Further details on construction of all IPW are provided in the Supporting information (S1 IPW Calculation).

Estimated IPWs were used to fit corresponding marginal structural models. For count variables (EPDS scores and number of breastfeeding difficulties), a linear Poisson marginal structural model was used. A linear binomial model was used for binary outcomes (postpartum probable depression and any difficulties breastfeeding). Saturated marginal structural models were fit for the average total association by intervention arms and the controlled direct association for mediation by any difficulties breastfeeding at 10 weeks. For mediation by the number of breastfeeding difficulties at week 10, the number of breastfeeding difficulties was modeled using a quadratic term. Moreover, 95% confidence intervals (CIs) and p-values for estimated PD were calculated with robust variance estimators, which provides conservative coverage [27]. All analyses were conducted with SAS 9.4 (Cary, North Carolina, United States of America). This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines (S1 STROBE Checklist).

Results

Fig 1 presents the flow diagram of mother–infant pairs through the eligible follow-up period for the presented analyses. In total, 975 mother–infant pairs enrolled in the study, and 718 were considered as uncensored at week 14 postpartum (Table 2). The prevalence of probable depression at 14 weeks postpartum was 18% for SOC, 11% for Steps 1–9, and 8% for Steps 1–10. Median EPDS scores were 8 (interquartile range (IQR): 3, 11) for the SOC, 6 (IQR: 2, 9) for Steps 1–9, and 3 (IQR: 0, 8) for Steps 1–10. More women in the Steps 1–9 (75%) and Steps 1–10 (91%) groups reported difficulties breastfeeding at week 10 compared to SOC (67%). The most commonly reported difficulties of breastfeeding at week 10 were the following: breasts were overfull, the baby nursed too often, and breasts leaked too much (Table 3).

thumbnail
Table 2. Characteristics of women at study enrollment by intervention group (n = 975).

https://doi.org/10.1371/journal.pmed.1003465.t002

thumbnail
Table 3. Reported breastfeeding difficulties at 10 weeks postpartum by intervention group (n = 851).

https://doi.org/10.1371/journal.pmed.1003465.t003

BFHI intervention implementations were associated with reduced EPDS scores and the prevalence of depressive symptoms when compared to SOC (Table 4). Relative to the SOC, both Steps 1–9 (0.42; 95% confidence level (CL): 0.15, 0.70; p < 0.01) and Steps 1–10 (0.48; 95% CL: 0.22, 0.74; p < 0.01) had had a greater number of reported breastfeeding difficulties for IPTW with IPCW. Additionally, the prevalence of any reported difficulties was higher for Steps 1–9 (0.08; 95% CL: −0.01, 0.17; p = 0.06) and Steps 1–10 (0.22; 95% CL: 0.14, 0.30; p < 0.01) as well.

thumbnail
Table 4. Intervention group association with EPDS and breastfeeding difficulties.

https://doi.org/10.1371/journal.pmed.1003465.t004

Among both intervention groups, the associations of the BFHI with probable depression were mediated through the prevalence of any breastfeeding difficulties (Table 5). In the presence of difficulties, EPDS scores were decreased in Step 1–9 (−3.65; 95% CL: −4.80, −2.50; p < 0.01) and Steps 1–10 (−5.21; 95% CL: −6.33, −4.10; p < 0.01) compared to SOC. A similar relationship occurred for probable depression prevalence for Steps 1–9 (−0.15; 95% CL: −0.25, −0.06; p < 0.01) and Steps 1–10 (−0.16; 95% CL: −0.25, −0.06; p < 0.01). When no breastfeeding difficulties occurred, there was a positive but imprecise association of both interventions with EPDS scores or probable depression, 2.92; 95% CL: −7.04, 12.89; p = 0.57 for Steps 1–9 and 1.46; 95% CL: −1.32, 4.23; p = 0.30 for Steps 1–10 (Table 5). There is insufficient indication of increased EPDS scores for either intervention if no breastfeeding difficulties was held constant. Mediation remained apparent when considering the number of breastfeeding difficulties instead. At 0 reported difficulties, Steps 1–9 and Steps 1–10 had little to no effect (Supporting information: S1 Sensitivity Analysis, Table 3.1). In general, as the number of difficulties increased, the strength of the protective association between BFHI and EPDS score or probable depression increased until 3 difficulties. At 4 reported difficulties, the strength of the protective association flattened or decreased (Figs 2 and 3).

thumbnail
Fig 2. Estimated controlled direct associations between BFHI and EPDS score at 14 weeks postpartum mediated through number of difficulties breastfeeding at week 10.

Solid line and solid shaded area indicated Steps 1–9 point estimate and CIs, respectively. Dashed line and hatched region indicated Steps 1–10 point estimate and CIs, respectively. BFHI, Baby-Friendly Hospital Initiative; CI, confidence interval; EPDS, Edinburgh Postnatal Depression Scale; IPW, inverse probability weight; IPCW, inverse probability of censoring weight; IPTW, inverse probability of treatment weight. The marginal structural model for EPDS and number of difficulties modeled was modeled as linear Poisson, where point estimates correspond to differences in counts.

https://doi.org/10.1371/journal.pmed.1003465.g002

thumbnail
Fig 3. Estimated controlled direct associations between BFHI and probable depression at 14 weeks postpartum mediated through number of difficulties breastfeeding at week 10.

Solid line and solid shaded area indicated Steps 1–9 point estimate and CIs, respectively. Dashed line and hatched region indicated Steps 1–10 point estimate and CIs, respectively. BFHI, Baby-Friendly Hospital Initiative; CI, confidence interval; IPW, inverse probability weight; IPCW, inverse probability of censoring weight; IPTW, inverse probability of treatment weight. The model for number of difficulties modeled was modeled as linear Poisson. The marginal structural model for probable depression was linear binomial, with point estimates corresponding to PDs. Probable depression was defined as an EPDS score of at least 13 at 14 weeks postpartum.

https://doi.org/10.1371/journal.pmed.1003465.g003

thumbnail
Table 5. Estimated controlled direct associations between BFHI mediated through any difficulty breastfeeding at week 10.

https://doi.org/10.1371/journal.pmed.1003465.t005

Sensitivity analysis

To assess the sensitivity of the finding to how breastfeeding was defined in the IPW mediation model, we repeated the analysis but with the proportions of feedings that were breastfeeding within the past week at week 10; instead of exclusive breastfeeding up to week 10. As shown in the Supporting information (S1 Sensitivity Analysis), results were similar to the main results. The controlled direct association for both EPDS scores and probable depression reversed directions under no breastfeeding difficulties (Supporting information: S1 Sensitivity Analysis, Table 3.1). These results for mediation for Steps 1–9 are in contrast to Table 5 where increases in EPDS scores and probable depression were seen for Steps 1–9 under no difficulties breastfeeding. However, all these results were imprecise and near the null value. The estimated controlled direct association for Steps 1–9 under no difficulties breastfeeding were dependent on how breastfeeding at week 10 was defined. Together, these results indicate that the association between Steps 1–9 and EPDS is null or relatively small if the population had been held constant at no breastfeeding difficulties at week 10. The remainder of mediation results were consistent across the different approaches to operationalizing breastfeeding at week 10.

Discussion

We hypothesized that the BFHI steps may be understood as an intervention that helps mothers deal with common postpartum stressors, reducing depressive symptoms. Our results provide evidence that the BFHI is associated with reduce postpartum depressive symptoms compared to SOC, with both Steps 1–9 and the modified Step 10 associated with a decrease in the EPDS score. We also showed that this association with depressive symptoms at 14 weeks was mediated by breastfeeding difficulties, with no evidence of an association of the BFHI steps with depressive symptoms in the absence of breastfeeding difficulties. A potential interpretation of these results is that when no breastfeeding difficulties were present, there was little to no associated stress encountered, and so nothing to “treat.” In contrast, when too many difficulties were present, the BFHI steps may have been insufficient to prevent the development of depressive symptoms.

To our knowledge, this is the first study to look at the impact of the implementation of the 10 steps to successful breastfeeding that form the basis of BFHI on PPD symptoms. BFHI interventions, increase early initiation (within 1 hour of birth) of and prolonged exclusive breastfeeding [13], and thus are likely to be associated with increased levels of oxytocin among mothers. Higher levels of oxytocin are associated with lower reported breastfeeding difficulties and improved maternal mood [14,15]. However, our finding of no association between BFHI steps and depressive symptoms when breastfeeding difficulties was not present suggests that at least in settings like DRC where virtually all women initiate and continue breastfeeding for at least 6 months, other mechanisms must be at play. Social support and stress management skills reducing depressive symptoms are both general phenomena. BFHI helps identify stressors, reframe them in a less stressful way, teach how to manage the stressors, and provide encouragement and social support to motivate people to effectively deal with stressors [5,6]. The importance of the identification of stressors is seen in the increased number of difficulties among women in the Steps 1–9 and Steps 1–10 arms. The increased identification of difficulties and the controlled direct association results indicate that the total effect of BFHI operates through the recognition of these stressors and coping with them.

PPD has serious impacts on the woman, her partner, and her baby. Women with PPD experience lower quality of life and an inability to care of themselves, their partner, and infants [2830]. Consequently, infants of mothers with PPD are at increased risk of chronic and acute illness, death, and poorer physical, social, emotional, and intellectual development [31], resulting in lasting psychiatric and cognitive effects [32]. Our finding that the BFHI was associated with reduced depressive symptoms independently through a different path than that associated with increased breastfeeding provides additional evidence to support renewed efforts to scale it up, particularly in low-resource settings where access to mental health services is acutely limited. Coupled with the proven efficacy of the BFHI steps in increasing rates of exclusive breastfeeding, even partial implementation (Steps 1–9) is likely to have wide-reaching and long-lasting effects, both physical and psychiatric.

This study had a number of limitations. The small number of clusters precluded accounting for the clustering in the variance. Therefore, results may be optimistic in terms of precision. In addition, PPD was not a registered trial outcome nor was it included in the funded proposal (see S1 Protocol) as a secondary outcome. Further, some enrolled participants were lost to follow-up, thus a potential for selection bias. However, the consistency of the results from our unadjusted analyses, analyses adjusted for baseline characteristics and for baseline characteristics and censoring, and the magnitude of the estimates suggests that the observed association cannot all be explained by residual confounding or other potential bias.

With only 1 assessment of depressive symptoms (at week 14), we were unable to consider time-varying effects. These include the rate of reduction in symptoms, persistence of associations after the cessation of breastfeeding, and whether breastfeeding difficulties mediate the relationship at other time points. Most psychotherapeutic interventions are generally conducted for 12 to 20 weeks [33], suggesting that the timing of the EPDS administration here was well suited to detecting differences in depressive symptoms. However, our mediation results suggest that benefits will occur potentially as soon as breastfeeding difficulties are encountered, with the small caveat that the first few difficulties will still be relatively more stressful than later difficulties due to their novelty. With respect to persistence of associations, the mediated association through breastfeeding difficulties will of course no longer occur after the cessation of breastfeeding. However, a lighter symptom load will make future stressors easier to manage, reducing the risk of PPD symptoms. Whether the protective associations continue for a matter of weeks or a matter of months is unknown. Similarly, only 1 assessment of breastfeeding difficulties (at week 10) was considered, even though this information was collected at each of the study visits (6, 10, 14, 18, and 24 weeks postpartum). In addition to the fact that earlier difficulties might be relatively more stressful, the increasing protective benefit of BFHI on depressive symptoms with increasing difficulties found in our mediation analysis suggests a potential cumulative effect of the number difficulties over time including a differential impact of transient versus persistent difficulties. Breastfeeding assessments at 6 and 10 weeks can be incorporated into the mediation analysis for multiple mediators [34]. Such relationships might fruitfully be explored in future analyses.

In our analysis, we were unable to explore mechanisms by which BFHI steps exert their influence on postpartum depressive symptoms (i.e., social support quality and quantity, breastfeeding knowledge, and perceived stress of breastfeeding). As such, we are unable to identify what aspects of BFHI steps can be further improved to reduce PPD as well as increase exclusive breastfeeding. Similarly, it is unclear what approaches are best for improving social support (e.g., offering compassion and encouragement versus listening to issues) and breastfeeding knowledge (e.g., maximizing knowledge of major issues versus maximizing knowledge of all issues) without resulting in a decrease in exclusive breastfeeding prevalence as observed for Steps 1–10 in the original trial [16]. It is important for future research to address these questions to provide more actionable information.

For the identification of the average total association and the controlled direct association, we relied on several assumptions. We assumed exchangeability, causal consistency, positivity, and correctly specified parametric models [35]. For exchangeability, we assumed that there is no residual confounding of the treatment–outcome relationship and no residual confounding of the mediator–outcome relationship. Here, our IPW models controlled for common risk factors for depressive symptoms but were unable to directly control for preexisting depression symptoms. Prior symptom severity could affect treatment response, as well as the number and severity of breastfeeding difficulties. We were additionally unable to control for preexisting breastfeeding knowledge, beliefs, and social support, each of which could confound the mediator–outcome estimate. Causal consistency is the assumption that our treatment and mediator are sufficiently well defined such that any additional variations of the treatment do not affect the outcome. We did not distinguish between types of breastfeeding difficulties. Some difficulties are clearly more severe and frustrating than others (e.g., pain versus leak), and the associations we discuss here may not apply equally to more stressful difficulties. Finally, positivity is the assumption that there are treated individuals at every level of confounders. Parametric regression models, and the corresponding assumption of no model misspecification, are necessary to weaken the positivity assumption for the estimation of IPW in the presence of high-dimensional data.

We further assumed no measurement error and non-informative missing data. Measurement error is a concern for 3 variables in particular: SES, difficulties breastfeeding, and EPDS scores. SES was measured through 2 durable goods. Difficulties with breastfeeding were more commonly reported in the Steps 1–9 and Steps 1–10 groups, which likely indicates a greater awareness of the difficulties and a willingness to report them, rather than an actual difference in the prevalence of difficulties breastfeeding. As indicated by our sensitivity analysis, the controlled direct association for no breastfeeding difficulties for Steps 1–9 should be interpreted cautiously. While we assume that other data are missing completely at random, IPCW results instead assume that missing depression data are non-informative conditional on demographics and clinic experience.

Conclusions

In conclusion, we have shown that the implementation of the BFHI steps is associated with a decrease in depression symptoms at 14 weeks postpartum. The findings also suggest that this association is mediated by difficulties with breastfeeding. These results were robust across analytical approaches and different definitions of breastfeeding difficulties and probable depression. Given the numerous health consequences of PPD, the BFHI may then provide a general benefit to both maternal and infant health owing to the simple routine discussions about breastfeeding that the BFHI mandates, with this effect occurring in addition to those afforded by exclusive breastfeeding.

Supporting information

S1 IPW Calculation. This file contains details on the calculation of weight used to implement the IPW. IPW, inverse probability weight.

https://doi.org/10.1371/journal.pmed.1003465.s002

(DOCX)

S1 STROBE Checklist. The completed STROBE Checklist. STROBE, Strengthening the Reporting of Observational Studies in Epidemiology.

https://doi.org/10.1371/journal.pmed.1003465.s003

(DOC)

S1 Sensitivity analyses. This file contains results of the sensitivity analyses.

https://doi.org/10.1371/journal.pmed.1003465.s004

(DOCX)

S1 Analytic Code (SAS). This is the SAS code used to produce the results.

https://doi.org/10.1371/journal.pmed.1003465.s005

(SAS)

S1 Data. This is the dataset used to produce the results.

https://doi.org/10.1371/journal.pmed.1003465.s006

(XLSX)

Acknowledgments

We thank the study mothers and infants for their participation and time and personnel of the participating clinics and study staff for their time and effort.

References

  1. 1. Righetti-Veltema M, Conne-Perréard E, Bousquet A, Manzano J. Risk factors and predictive signs of postpartum depression. J Affect Disord. 1998;49(3):167–80. pmid:9629946
  2. 2. Robertson E, Grace S, Wallington T, Stewart DE. Antenatal risk factors for postpartum depression: a synthesis of recent literature. Gen Hosp Psychiatry. 2004;26(4):289–95. pmid:15234824
  3. 3. Brown A, Rance J, Bennett P. Understanding the relationship between breastfeeding and postnatal depression: the role of pain and physical difficulties. J Adv Nurs. 2016;72(2):273–82. pmid:26494433
  4. 4. Chaput KH, Nettel-Aguirre A, Musto R, Adair CE, Tough SC. Breastfeeding difficulties and supports and risk of postpartum depression in a cohort of womenwho have given birth in Calgary: a prospective cohort study. CMAJ Open. 2016;4(1):E103. pmid:27280109
  5. 5. WHO/UNICEF. Baby-friendly hospital initiative: revised, updated, and expanded for integrated care. Geneva, Switzerland: WHO/UNICEF, 2009.
  6. 6. Meichenbaum D. Cognitive behaviour modification. Cogn Behav Ther. 1977;6(4):185–92.
  7. 7. Munn AC, Newman SD, Mueller M, Phillips SM, Taylor SN. The impact in the United States of the baby-friendly hospital initiative on early infant health and breastfeeding outcomes. Breastfeed Med. 2016;11(5):222–30. pmid:27082284
  8. 8. Hannula L, Kaunonen M, Tarkka MT. A systematic review of professional support interventions for breastfeeding. J Clin Nurs. 2008;17(9):1132–43. pmid:18416790
  9. 9. Wahl OF. Stigma as a barrier to recovery from mental illness. Trends Cogn Sci. 2012;16(1):9–10. pmid:22153582
  10. 10. Thomson G, Ebisch-Burton K, Flacking R. Shame if you do–shame if you don’t: women’s experiences of infant feeding. Matern Child Nutr. 2015;11(1):33–46. pmid:25138617
  11. 11. Brown B. Shame resilience theory: A grounded theory study on women and shame. Fam Soc. 2006;87(1):43–52.
  12. 12. Carleton RN. Into the unknown: A review and synthesis of contemporary models involving uncertainty. J Anxiety Disord. 2016;39:30–43. pmid:26945765
  13. 13. House JS, Landis KR, Umberson D. Social relationships and health. Science. 1988;241(4865):540–5. pmid:3399889
  14. 14. Holt-Lunstad J. Why social relationships are important for physical health: A systems approach to understanding and modifying risk and protection. Annu Rev Psychol. 2018;69:437–58. pmid:29035688
  15. 15. Aknin LB, Barrington-Leigh CP, Dunn EW, Helliwell JF, Burns J, Biswas-Diener R, et al. Prosocial spending and well-being: Cross-cultural evidence for a psychological universal. J Pers Soc Psychol. 2013;104(4):635. pmid:23421360
  16. 16. Yotebieng M, Labbok M, Soeters HM, Chalachala JL, Lapika B, Vitta BS, et al. Ten Steps to Successful Breastfeeding programme to promote early initiation and exclusive breastfeeding in DR Congo: a cluster-randomised controlled trial. Lancet Glob Health. 2015;3(9):e546–55. Epub 2015 Aug 8. pmid:26246225.
  17. 17. Stuebe AM, Grewen K, Pedersen CA, Propper C, Meltzer-Brody S. Failed lactation and perinatal depression: common problems with shared neuroendocrine mechanisms? J Womens Health (Larchmt). 2012;21(3):264–72. Epub 2011 Dec 28. pmid:22204416.
  18. 18. Stuebe AM, Grewen K, Meltzer-Brody S. Association between maternal mood and oxytocin response to breastfeeding. J Womens Health (Larchmt). 2013;22(4):352–61. Epub 2013 Apr 17. pmid:23586800; PubMed Central PMCID: PMC3627433.
  19. 19. Figueiredo B, Canário C, Field T. Breastfeeding is negatively affected by prenatal depression and reduces postpartum depression. Psychol Med. 2014;44(5):927–36. pmid:23822932
  20. 20. Zivich P, Lapika B, Behets F, Yotebieng M. Implementation of Steps 1–9 to Successful Breastfeeding Reduces the Frequency of Mild and Severe Episodes of Diarrhea and Respiratory Tract Infection Among 0–6 Month Infants in Democratic Republic of Congo. Matern Child Health J. 2018;22(5):762–71. Epub 2018 Feb 9. pmid:29417366; PubMed Central PMCID: PMC6645680.
  21. 21. Yotebieng M, Chalachala JL, Labbok M, Behets F. Infant feeding practices and determinants of poor breastfeeding behavior in Kinshasa, Democratic Republic of Congo: a descriptive study. Int Breastfeed J. 2013;8(1):11. Epub 2013 Oct 3. pmid:24083882; PubMed Central PMCID: PMC3850507.
  22. 22. 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. pmid:3651732
  23. 23. Gibson J, McKenzie-McHarg K, Shakespeare J, Price J, Gray R. A systematic review of studies validating the Edinburgh Postnatal Depression Scale in antepartum and postpartum women. Acta Psychiatr Scand. 2009;119(5):350–64. pmid:19298573
  24. 24. Robins JM, Hernan MA, Brumback B. Marginal structural models and causal inference in epidemiology. Epidemiology. 2000;11(5):550–60. Epub 2000 Aug 24. pmid:10955408.
  25. 25. Robins JM, Finkelstein DM. Correcting for noncompliance and dependent censoring in an AIDS Clinical Trial with inverse probability of censoring weighted (IPCW) log-rank tests. Biometrics. 2000;56(3):779–88. Epub 2000 Sep 14. pmid:10985216.
  26. 26. Robins JM, Greenland S. Identifiability and exchangeability for direct and indirect effects. Epidemiology. 1992:143–55. pmid:1576220
  27. 27. Hernán MA, Brumback BA, Robins JM. Estimating the causal effect of zidovudine on CD4 count with a marginal structural model for repeated measures. Stat Med. 2002;21(12):1689–709. pmid:12111906
  28. 28. Rahman A, Iqbal Z, Bunn J, Lovel H, Harrington R. Impact of maternal depression on infant nutritional status and illness: a cohort study. Arch Gen Psychiatry. 2004;61(9):946–52. Epub 2004 Sep 8. pmid:15351773.
  29. 29. Howard LM, Flach C, Mehay A, Sharp D, Tylee A. The prevalence of suicidal ideation identified by the Edinburgh Postnatal Depression Scale in postpartum women in primary care: findings from the RESPOND trial. BMC Pregnancy Childbirth. 2011;11:57. Epub 2011 Aug 5. pmid:21812968; PubMed Central PMCID: PMC3161003.
  30. 30. Sadat Z, Abedzadeh-Kalahroudi M, Kafaei Atrian M, Karimian Z, Sooki Z. The Impact of Postpartum Depression on Quality of Life in Women After Child's Birth. Iran Red Crescent Med J. 2014;16(2):e14995. Epub 2014 Feb 5. pmid:24719747.
  31. 31. Field T. Postpartum depression effects on early interactions, parenting, and safety practices: a review. Infant Behav Dev. 2010;33(1):1–6. pmid:19962196
  32. 32. Halligan SL, Murray L, Martins C, Cooper PJ. Maternal depression and psychiatric outcomes in adolescent offspring: a 13-year longitudinal study. J Affect Disord. 2007;97(1–3):145–54. pmid:16863660
  33. 33. Weissman MM, Markowitz JC, Klerman G. Comprehensive guide to interpersonal psychotherapy: Basic Books; 2008.
  34. 34. VanderWeele TJ, Vansteelandt S. Mediation Analysis with Multiple Mediators. Epidemiol Method. 2014;2(1):95–115. pmid:25580377.
  35. 35. Hernán MA, Robins JM. Estimating causal effects from epidemiological data. J Epidemiol Community Health. 2006;60(7):578–86. pmid:16790829