Introduction

Maternal postpartum depression (PPD) usually starts between childbirth and 4–6 weeks thereafter and has negative consequences for the mother and her infant.1,2,3,4 However, evidence on the effects of PPD on early childhood socioemotional development (SED) is sparse.5,6,7 SED is a continuous process through which humans acquire the skills to express, experience, and regulate emotions as a precursor to develop social competencies and emotional intelligence.8 SED in the first year of life is sensitive to the caregiver’s positive parenting practices that can regulate the infant’s stress and contributes to the infant’s emotional wellbeing.8,9 In the long-term, high parenting quality acts as a buffer enhancing the healthy development of neurobiological systems governing stress reactivity and emotion.10,11

Mothers are usually the primary caregivers, and their infants can be considerably exposed to the symptoms of PPD.3,12 The mechanisms by which PPD can negatively influence infants’ SED are complex and not fully clarified. Insofar research has focused on direct causal mechanisms including genetic inheritance and indirect mechanisms including the effects of PPD on parenting quality and behaviors.7,13,14,15 Besides the uncertainty about the influence of PPD on SED, it is also not clear if the variations in the PPD trajectory, i.e., in the onset, severity, and continuity of PPD are of additional importance for infants’ SED.16,17,18

One of the potential mechanisms by which PPD can indirectly impact infants’ SED is the influence of maternal self-efficacy (MSE).19,20 MSE refers to the mother’s self-confidence in the organization and execution of tasks dealing with the care of her child. Depression could lower MSE by altering her cognitions including perceptions of worthlessness and negative self-attribution.21 Additionally, MSE has been shown to predict the child’s adjustment directly, or indirectly via positive parenting practices and behaviors.22,23 Therefore, we hypothesized that the trajectory of PPD is negatively associated with infants’ SED and this association is mediated by MSE.

Finally, PPD often co-occurs with anxiety. Postpartum anxiety by itself is a risk factor for poor SED of infants.24 However, owing to the comorbid nature of PPD and postpartum anxiety it is difficult to disentangle the distinct impacts of these disorders on infants’ SED.25,26 Evidence lacks on whether anxiety has an independent effect on infants’ SED on top of PPD.

Therefore, the main aim of the study was to assess the effects of maternal PPD trajectories until 6 months on infants’ SED at age 12 months. In addition, we aimed to assess the mediating role of MSE and the additional effect of postpartum anxiety at 12 months of age.

Methods

Study design and participants

We used longitudinal data from the POST-UP trial.27,28 POST-UP had a prospective, quasi-experimental, comparative design to assess if regular screening for PPD in Preventive Child Healthcare (PCHC) followed by advice on treatment options and referral when indicated improved mental health outcomes at the maternal and child level. The participants of the POST-UP trial were mother-child dyads consulting the PCHC between December 1, 2012, and April 1, 2014, during the first month after birth. The participants allotted to the intervention arm (n = 1843) were repeatedly screened for PPD during the PCHC consultation, using the Dutch version of the Edinburgh Postnatal Depression Scale (EPDS),29 whereas the participants allotted to the care as usual arm (n = 1246) were not screened for PPD. Since the PPD trajectory is the main independent variable for the current study only mother-child dyads of the intervention arm were included (n = 1843).

Procedure and measures

During home visits around 2 weeks postpartum informed consent for participation was obtained from the parents by the PCHC nurse. Baseline data were recorded during the first in-person consultation at the PCHC center 4 weeks after birth. The intervention arm received regular screening of PPD at 1, 3, and 6 months after childbirth. More details on the recruitment and blinding procedure can be found elsewhere.27 The POST-UP trial procedures were in accordance with the protocols of the World Medical Association Declaration of Helsinki and were approved by the Medical Ethics Committee, University of Twente (approval number: 80-82470-98-012).

Socioemotional development (SED) of infants at age 12 months was the primary outcome variable. SED was assessed using the Ages and Stages Questionnaire-Social-Emotional 12-month version (ASQ-SE).30 The ASQ-SE consists of 25 items related to developmentally appropriate behavior and is completed by the caregiver. The scores of all items are summed up (ranging from 0 to 145): higher scores represent a worse SED. The original and the Dutch version of ASQ-SE have good validity and reliability.31,32

Postpartum depression (PPD) was measured at 1, 3, and 6 months postpartum using the EPDS. The EPDS is a widely used measure for screening for PPD in clinical practice and research, and the Dutch version of the EPDS exhibits good internal consistency.33 It consists of 10 items and the scores are summed up to get a total score (ranging from 0 to 30), with higher scores indicating a higher level of depressive symptoms. Scores on the EPDS ranging from 9 to 12 are indicative of mild depression and scores above 12 indicate major depression.34 Since both clinical PPD and subclinical symptoms of PPD are associated with infants’ SED, we modeled the EPDS scores as a continuous variable instead of as a categorical variable.16

Maternal self-efficacy (MSE) was assessed using the postnatal version of the Self-Efficacy in the Nurturing Role (SENR) questionnaire at age 12 months of the infant,35 which includes 16 items representing basic parenting competencies. Negatively formulated items were scored inversely so that higher scores reflect higher MSE. The SENR has shown good internal reliability for the postpartum period.22,36

Postpartum anxiety was measured using the short version of the state form of the Spielberger State-Trait Anxiety Inventory (STAI-6) at 12 months after birth.37 The STAI-6 has been widely used for the screening of anxiety in pregnant women and mothers during the perinatal period.38,39 The Dutch version of the STAI-6 has shown good psychometric properties.40

Other covariates regarded maternal education level (high versus low), gestational age (in weeks), lifetime history of depression (yes or no), having experienced stressful life events before or after childbirth, defined by at least 2 or more stressful life events at either time periods i.e., stressful life events including an accident, illness, death of family or friends, losing a job, financial, housing and relationship problems (more details can be found in Appendix 1).41

Statistical analyses

For an optimal estimation of the effects, multiple imputation was carried out to resolve the missingness of data (Table 1 provides details on missing data). The multiple imputation method was based on the method of chained equations, with the assumption of missingness at random over 20 imputed data sets, and 75 iterations to ensure convergence.42 Imputations were based on variables potentially associated with the main variables of the study i.e., PPD, postpartum anxiety, MSE, and SED. These included age of the mother (in years), maternal education level (high versus low), first pregnancy (yes or no), method of delivery (cesarean section versus vaginal delivery), alcohol consumption (yes or no), social support (assessed using the Social Support List-Interactions questionnaire43), quality of life of the mother (assessed using the Medical Outcome Study-Short Form44), stressful life events before and after childbirth, birth weight (in grams), breast-feeding status (yes or no), physical health problems of the child (yes or no), hospitalization of the child (yes or no), and infants’ temperament (assessed using the Infant-Toddler Quality Of Life questionnaire-Short Form-4745). Binary variables were imputed using logistic regression, and continuous variables using predictive mean matching.42,46

Table 1 Characteristics of the mothers and infants in the study; source is POST-UP trial (n = 1843).
Full size table

For the first aim i.e., to assess the effects of the PPD trajectory on infants’ SED, firstly, the individual EPDS trajectories over the three measurement points were estimated using multilevel regression analysis with a random intercept and slope. Given the skewed count-type nature of the EPDS scores, these were modeled using Poisson regression.47 Due to the low variability of the random slopes, the longitudinal trends of the EPDS score were estimated with a random intercept model and a fixed slope. The ASQ-SE at 12 months was regressed on the random intercept of the PPD trajectory, to assess the effects of overall variations in the levels of the EPDS scores on the ASQ-SE scores using structural equation modeling (SEM) (Total effects model).

For the second aim, i.e., to assess the mediating role of MSE, the values of MSE at 12 months were added in the pathway between PPD and SED to test mediation using SEM (Mediation model). To study the additional independent effect of postpartum anxiety, we applied conditional regression analysis to obtain a measure of anxiety independent of PPD48 conditional values of anxiety i.e., the residuals of the linear regression of the STAI-6 scores at 12 months on the EPDS scores at 6 months (hereafter referred to as ‘postpartum anxiety variable’). ASQ-SE scores were adjusted by the postpartum anxiety variable in the mediation models. Finally, all covariates were included in the model (as adjusting variables for the mediator and outcome variables).

The descriptive analyses were carried out in IBM SPSS 27.0. Multiple imputations were done in R (package mice, version 3.13.0).49 Trajectories of PPD and SEM were performed in M-Plus version 8 and estimated using maximum likelihood estimation with robust standard errors.50,51

Results

Background characteristics

The sample (n = 1843) consisted of mothers who were Dutch citizens by birth (95.3%) and were generally highly educated (88.7%). Other characteristics of the mother-child dyads are presented in Table 1.

The effects of PPD trajectories on SED of infants

The initial level of the EPDS scores among the mothers varied, but generally decreased slightly from the first month to the sixth month postpartum (Fig. 1). Higher levels of PPD over time were associated with a lower SED (coefficient for log-EPDS 3.5, 95% confidence interval (95% CI) 2.8; 4.2, e.g., an increase in the EPDS score from 9 to 13 was related to a decrease of 1.3 points on the ASQ-SE (Table 2, model 1). To make this association better interpretable, the predicted values of ASQ-SE based on EPDS scores in the original scale are presented in Appendix 2.

Fig. 1: Trends of the Edinburgh Postnatal Depression Scale trajectories from 1 to 6 months after birth representing the random intercepts and its variance (n = 1843).
figure 1

Areas within the dotted lines include 70% and 90% of the trajectories, respectively.

Full size image
Table 2 The direct, indirect (i.e., mediated), and total effects of postpartum depression trajectory, based on the EPDS values at 1, 3, and 6 months of age on infants’ socioemotional development as assessed with the ASQ-SE score (n = 1843).
Full size table

Mediating role of MSE and additional effect of postpartum anxiety

MSE mediated the relationship between PPD and SED (indirect effect 1.7, 95% CI: 1.2; 2.1) as seen in Table 2, model 2. The addition of the postpartum anxiety variable and other covariates reduced all effects marginally, but these remained statistically significant (Table 2, model 4, and Fig. 2). About 50% of the total effects of the PPD on SED were mediated via MSE (indirect effect 1.4, 95% CI: 1.0; 1.9). Postpartum anxiety had a small but significant additional negative effect on SED (B 0.17, 95% CI: 0.10; 0.24).

Fig. 2: Mediation model demonstrating the associations (B) and their standard errors (SE); “c” is the total effects model, where c is the total effect of postpartum trajectories (log transformed) on infants’ socioemotional development.
figure 2

For the mediation model, a*b is the indirect, i.e., the mediated, effect and c- is the direct, i.e., not-mediated, effect. The effects are adjusted for postpartum anxiety and for other covariates (maternal education, history of depression before childbirth, stressful life events during or after birth of the child, and gestational age). Estimates were pooled over 20 multiple imputed datasets. All paths indicated in the figure were statistically significant at p < 0.001.

Full size image

Discussion

We found a negative association of PPD trajectories with infants’ SED. Moreover, we found that MSE mediated this relationship. Furthermore, postpartum anxiety had a small additional negative effect on infants’ SED.

Higher PPD levels during the first 6 months postpartum were associated with a worse SED of infants at 12 months of age. Our findings are in line with a study by Porter et al. among 282 mother-child dyads which showed negative effects of the trajectories of perinatal depression on infants’ SED at the age of 12 months.52 Porter et al. also assessed the effects of the changes in perinatal depression over time (random slopes). We could not assess this association due to the small variability of the changes in PPD over time, whereas this was much larger in the study by Porter et al. The difference in the variability of PPD between both studies could be due to the differences between the study samples; Porter et al. assessed a clinical sample whereas we assessed a community sample.

We found that MSE mediated the association between PPD and infants’ SED. To our knowledge, this is the first study to examine the role of MSE as a mediator of the relationship between the course of PPD and infants’ SED at age 12 months. A possible explanation is that depressive states could negatively affect the mother’s parenting abilities which in turn might lead to a negative appraisal of her infant’s behavior.53 However, PPD and MSE have been shown to have a bidirectional relationship in which MSE is a predictor as well as a consequence of PPD;54 It is also possible that they have a shared psychological, neurobiological, or genetic background rather than being mutual causes.55 Therefore, our results reiterate the importance of recognizing both PPD and MSE as important targets for preventive interventions to improve infants’ SED.

Finally, our finding that postpartum anxiety had a small additional negative effect on top of PPD on infants’ SED, is in line with the study by Prenoveau et al. that demonstrated moderate effects of postpartum anxiety on infants’ SED at age 2 years.56 However, in that study anxiety did not independently predict SED problems beyond PPD. Postpartum anxiety is associated with intrusive parenting practices which could explain the impact of anxiety on infants’ SED.24

Strengths and limitations

Our study has several strengths. First, our study had a longitudinal design with a representative sample of 1843 mother-child dyads from the general Dutch population. Second, PPD has been assessed repeatedly at three time points (1, 3, and 6 months after birth), which allowed for studying the effects of the course of PPD on infants’ SED precisely. Finally, we used a Poisson-based regression approach to handle the skewed PPD scores that provide robustness to the model and results.

However, some limitations should be mentioned as well. The main limitation of our study regards the use of self-reported maternal mental health and infants’ SED which may have introduced information bias. For instance, mothers who are depressed may be more likely to rate their infants having more socioemotional problems as well as rate their self-efficacy lower. However, all three questionnaires used in the study have been validated among the population of mothers with PPD and anxiety.29,35,57 Second, we did not assess prenatal depression and prenatal anxiety whereas these could be important confounders in the association between PPD and SED of infants.1,24 However, including ‘lifetime depression’ and ‘stressful life events’ as covariates is likely to reduce potential bias here. Third, due to the design of the study, the other direction of the association between PPD and MSE was not tested.

Implications

The strong associations of PPD trajectories with infants’ SED in our data emphasize the need for effective and regular screening of PPD during the postpartum period. In addition, also screening for postpartum anxiety should be considered because of its independent, additional negative effect on infants’ SED. Furthermore, the mothers with PPD, as well as their infants, may benefit from interventions that focus on both decreasing the duration and level of the PPD and improving maternal self-efficacy via interventions such as peer parenting support groups.58

Future studies are needed to assess the effects of the duration of all types of mental health problems before and after giving birth, including repeated measures of other mental health problems including perinatal depression and anxiety. Additionally, more insight is needed into protective factors affecting bonding between mother and child, into the effects of paternal mental health, and into biological mechanisms such as epigenetic changes or changes to the morphology of the brain in order to obtain further evidence enabling to develop and offer effective personalized preventive interventions.59,60

Conclusions

The levels of the PPD trajectory between birth and 6 months after birth are negatively associated with infants’ SED. Moreover, MSE significantly mediated the relationship between PPD and SED, indicating MSE as a potential target for preventive interventions to alleviate the negative effects of PPD on children’s SED. Additionally, our findings show the importance of the (repeated) assessment of PPD and anxiety among mothers in the postpartum period in order to offer preventive interventions in an early phase.