This study used a longitudinal prospective design that included three time points (first and third trimester of pregnancy, and 4 months postpartum). The study adhered to the most stringent ethical guidelines governing human participant research. Specifically, all experimental procedures were meticulously designed and implemented in full compliance with the 1964 Helsinki Declaration and its subsequent amendments, which represent the international gold standard for ethical medical research. It received comprehensive ethical review and obtained formal approval from Fuenlabrada University Hospital Ethics Committee (approval reference: CPMP/ICH/135/95). All women signed a consent form to declare a voluntary agreement with all procedures implicated in this project. According to the Enhancing the Quality and Transparency Of health Research (EQUATOR) checklist, the Strengthening the Reporting of Observational studies in Epidemiology (STROBE) statement was chosen for this study.

This study is part of a larger study that aims to analyze the evolution of maternal psychosocial variables during pregnancy and postpartum. The study was carried out in Fuenlabrada University Hospital in Madrid (Spain), with an allocated population of 216,803, which provides care to about 3000 obstetric patients per year. Women were considered eligible for the study if they were of legal age (over 18 years), had a low-risk pregnancy with a maximum gestational age of 14 weeks, and were mentally and physically healthy without a psychiatric diagnosis during pregnancy. Out of 290 Spanish women evaluated during first-trimester, 120 (41.37%) met the inclusion criteria for the present study, which included being followed up in the third-trimester, immediately postpartum, and 4 months postpartum. Health status was defined by the absence of depression or any other psychiatric disorders. Additionally, the study exclusively included naturally conceived pregnancies, explicitly excluding in vitro fertilization-embryo transfer (IVF-ET) pregnancies.

In the first trimester of pregnancy (1–13 weeks of pregnancy), and according to the established inclusion criteria, a total of 290 pregnant women agreed to participate in this study. All women received their obstetric clinical care at the hospital where the study was carried out. During the first obstetric clinical appointment, coinciding with the first ultrasound (about 12–14 weeks of pregnancy), the midwife, and a member of the research team, informed the women about the study that was being carried out, and if they agreed, they signed the corresponding informed consent and filled out the evaluation protocol (see variables and instrument section), in the presence of the midwife in case any doubts arose. Sociodemographic and clinical data of interest were extracted from the clinical history. Of the total 290 women who completed the questionnaires five were excluded because of a miscarriage (n = 285). In the third trimester (around weeks 30), a new evaluation protocol (see variables and instrument section) was mailed to the women with a prepaid envelope for returning the completed questionnaire to the research team, 122 participants completed this phase. These women were evaluated again in the immediate postpartum period, during their hospital stay (although this measurement is not part of the present study) (n = 122). Finally, 4 months after delivery, the last evaluation protocol (see variables and instrument section) was sent in the post, obtaining a response from 120 women. To understand the sample loss, which is common in prospective studies, we should highlight that, in addition to there being different evaluation modalities used (in person, via mail), around 35% of the pregnant women treated at the hospital where we carried out the study did not end up giving birth there. There were no significant differences in sociodemographic, clinical or pregnancy variables between the first, third trimester, and 4 months after birth. Moreover, when main analyses were conducted no significant results were found in sociodemographic variables.

Data were processed using SPSS version 23 (IBM, Chicago, IL, USA), with a predetermined significance level of p 0.05. A minimum sample was established at 200, following the established criteria for regression analyses [39]. Bivariate Pearson correlations were then performed as a precondition for the mediation analysis [39]. Descriptive analyses were also performed. Next, according to our aim, two models were conducted. First, a multiple mediation model (model 4) was conducted for the outcome (postpartum depression) (Fig. 1) by the SPSS macro-PROCESS (SPSS tool for estimating regression models with mediation and/or moderation effects), allowing us to test the effect of the proposed mediators (depressive symptoms in the third trimester, and lack of confidence in caretaking) to be carried out (i.e., the total indirect effects). Then, a serial model (model 6, Fig. 2) was tested to analyzed if both mediators were working serially. According to our hypothesis, depressive symptoms were included as a first mediator, and mothers’ lack of confidence as a second mediator. These models were tested using regression (to calculate statistics for specific paths) and bootstrapping (to generate a confidence interval [CI] for the mediation’s effect). All variables were centered before the analyses, and statistical significance was set at an alpha level of 0.05. These analyses represent a sophisticated analytical method that transcend traditional mediation analyses by offering a more nuanced understanding of complex psychological mechanisms, providing robust statistical testing of indirect effects, allowing simultaneous examination of multiple mediators and moderators, and enabling more sophisticated exploration of contextual variables. Moreover, they have methodological superiority because they overcome limitations of traditional regression-based approaches, provide bootstrap confidence intervals for indirect effects, handle complex mediation and moderation scenarios, and allow for more precise estimations of psychological processes. Finally, in relation to contextual variable exploration, these models have unique capabilities in examining indirect effects, conditional indirect effects, moderated mediation models and complex interaction patterns [40].

Fig. 1.

Hypothesized multiple model of the interrelationships between pregnancy worries, depressive symptoms (mediator 1), a mothers’ confidence in caring for her infant (mediator 2) and postpartum depression (dependent variable) a1-b1 (indirect effect of mediator 1); a2-b2 (indirect effect of mediator 2); C´: direct effect.

Fig. 2.

Hypothesized serial model of the interrelationships between pregnancy worries, depressive symptoms, a mothers’ confidence in caring for her infant and postpartum depression (dependent variable) a1-b2 (indirect effect of both mediators working in serial; d21 (effect of depressive symptoms on mothers’ lack of confidence); C´: direct effect.

Focusing on theoretical and methodological justification for both integrated mediation models, the integration of serial and parallel mediation models represents a sophisticated approach to understanding complex psychological mechanisms, offering a more comprehensive and nuanced analysis of variable interactions. To explain the rationale for this integrated mediation modeling, psychological phenomena rarely follow linear, single-pathway mechanisms. The integrated approach allows for the comprehensive exploration of multiple mediational pathways, simultaneous examination of direct and indirect effects, and capturing complex interrelationships between variables. The methodological advantages of serial mediation model characteristics are to examine sequential mediational processes, capturing cascading psychological mechanisms and tracing variable influences through multiple intermediary stages. In relation to parallel mediation model characteristics, they look for simultaneous mediational pathways, identifying competing or complementary mediational mechanisms, and provide a comprehensive viewpoint of variable interactions [40].

The mean of age of women included in this study was 31.28 (standard deviation (SD) = 3.96) years ranging between 23 and 42 years old. Descriptive data related to pregnancies planned, multiparous women, previous miscarriage, working status when women got pregnant, educational level, type of delivery, type of anesthesia and breastfeeding are presented in Table 1.

The correlation analysis showed (Table 2) that pregnancy worries were positively associated to depressive symptoms (p 0.01), postpartum depression (p 0.01) and a MLC in caring for her infant (p 0.01). A MLC in caring for her infant was positively correlated to depressive symptoms (p 0.01), and postpartum depression (p 0.01). Depressive symptoms were positively correlated to postpartum depression (p 0.01). Related sociodemographic variables, only age was negatively associated with a mothers’ lack of confidence (–0.25; p 0.01). All of others sociodemographic variables were no significant in their relation to worries, depressive symptoms, postpartum depression and mothers’ lack of confidence.

In relation to sociodemographic and clinical variables, age, planned pregnancy, previous pregnancies, previous abortions, type of delivery, birth weight and anesthesia were analyzed in relation to the variables of interest. Only age correlated negatively with mothers’ lack of confidence in caring for her infant (p = 0.008). Likewise, in relation to the characteristics of pregnancies, while pregnancy planning did not show significant associations with any variable under study, previous pregnancies did show significant differences on a mothers’ lack of confidence in caring for her infant and on pregnancy worries (Table 3). Finally, previous abortions, type of delivery, birth weight and anesthesia also showed no significant differences with any of the variables analyzed.

Firstly, a parallel multiple model was tested (Fig. 1). This model included PW as the independent variable, DS at third trimester and mother’s lack of confidence in caring for her infant as mediator variables, and PPD as the dependent variable, considering the previously described covariates (age, previous pregnancies). The results showed that, the predictive effect of pregnancy worries on depressive symptoms was significant (a1 = 0.29, t = 4.08, p 0.001 [95% CI = 0.1507, 0.4346] (Table 4). Secondly, the prospective effect of worries (evaluated at first trimester) and depressive symptoms (evaluated at third trimester) were analyzed related to the mother’s lack of confidence (evaluated after birth). Results showed that depressive symptoms had a significant effect on a mothers’ lack of confidence in caring for her infant (Coeff = 5.11, t = 3.95, p 0.001 [95% CI = 2.5521, 7.6763]), but pregnancy worries were not significant (a2 = 1.67, t = 1.60, p 0.11 [95% CI = –0.3959, 3.7525]. Moreover, related to covariates, results showed that previous pregnancies also had a significant negative effect on a mothers’ lack of confidence (Coeff = –6.008, t = –3.83, p 0.001 [95% CI = –9.1141, –2.9028]). Third, related to postpartum depression, while pregnancy worries did not have a significant effect on postpartum depression, depressive symptoms (Coeff = 5.0312, t = 9.9753, p 0.001, [95% CI = 4.0345, 6.0352]), and mothers’ lack of confidence (Coeff = 0.1432, t = 4.1132, p 0.001, [95% CI = 0.0742, 0.2123]) were both significant. However, previous pregnancies remained not significant. Focusing on PPD, results showed a significant total effect of pregnancy worries on PPD (c = 1.6438, t = 2.9293, p = 0.004, [95% CI = 0.5318, 2.7558]). Meanwhile, regression coefficient estimates — based on the use of 95% bias-corrected CI as evidence of the mediation of direct effect and indirect effects for pregnancy worries, depressive symptoms and mothers’ lack of confidence in caring for her infant – were calculated as follows: direct effect (PW on PPD) c´ = –0.2843, t = –0.7352, p = 0.4638 [95% CI = –1.0509, 0.4822]; a1b1 (PW by DS on PPD) = 1.4733, [95% CI = 0.7105, 2.2309]; a2b2 (PW by MLC on PPD) = 0.2404, [95% CI = –0.0569, 0.5826], respectively (Fig. 1). The total amount of variance accounted by the overall model was 63% (F = 46.85; p 0.0001).

Looking to test if there was a possible serial effect between variables, a serial model was tested (Fig. 2). This model included PW as an independent variable, DS as a first mediator, and mother’s lack of confidence in caring for her infant as a second mediator (MLC). PPD was included as the dependent variable. In this model, covariates (i.e., age, previous pregnancies) were also added. As Table 4 shows, indirect effects of pregnancy worries on postpartum depression via depressive symptoms on mothers’ lack of confidence in caring for her infant (indirect effect 3) were significant (d21 (PW by MLC on PPD) = 0.2144, [95% CI = 0.0595, 0.4562]. However, when contrast effects were analyzed, results showed that the mediation effect of depressive symptoms at third trimester between pregnancy worries and postpartum depression (indirect effect 1) was higher than the mediated effect of a mothers’ lack of confidence in caring for her infant (indirect effect 2), and showed also larger effects compared to the serial mediation path, as indicated by bootstrap CI that were entirely above zero for both contrast effects (contrast 1: indirect effect 1 $>$indirect effect 2: 1.2329 [95% CI = 0.4004, 2.0903]; contrast 2: indirect effect 1 $>$indirect effect 3: 1.2590 [95% CI = 0.5658, 1.9826], respectively). The third contrast, that compared indirect effect 2 and 3 was not significant (contrast 3: indirect effect 2 $>$indirect effect 3: 0.0260 [95% CI = –0.4118, 0.4314]).

This study has some limitations that should be highlighted. First, the interpretation and extrapolation of the results must be linked to some limitations because the sample was taken from a single hospital and from women with low obstetric risk. Based on the literature reviewed where there is a higher percentage of depressive symptomatology in women with pregnancy complications, the fact that one of the inclusion criteria was low obstetric risk opens the opportunity to generalize our results (specifically in relation to postpartum depression) to a larger number of women. Moreover, the loss of participants in the different phases of the study could be another bias to the interpretation of the results. Thirdly, self-reported measures used to assess variables included may be subject to the usual bias found in these types of assessment tools. Nevertheless, demographic, and clinical variables associated with postpartum depression in pregnancy were considered, but the effects of other physical symptoms such as anxiety related to postpartum depression should also be controlled. Fourthly, we have confirmed the importance of depressive symptoms as significant mediators between pregnancy worries in the prenatal stage and their implication in bonding in the postnatal stage, however, within a broader family context, it would be interesting to consider other variables, such as the role of the father figure in the bond [59], that may influence the possibility of developing an adequate bond with the baby [60]. The father’s role emerges as a critical psychological mediator in maternal mental health and perceived childcare competence, extending far beyond traditional conceptualizations of parental support. The contemporary understanding transcends traditional support conceptualizations, positioning fathers as active, transformative agents in maternal psychological adaptation. For example, researchers have shown that paternal emotional responsiveness significantly modulates maternal self-perception by enhancing maternal resilience, increasing perception of caregiving competence, and reducing vulnerability to postpartum depressive symptoms [59, 60]. Moreover, practical co-parenting engagement impacts maternal psychological adaptation by sharing responsibilities that also reinforces maternal self-efficacy [61]. The fathers’ role should be integrated into future studies as a contextual role trying to understand how it impacts maternal self-efficacy.

The apparent contradiction in the research finding—where worries predict postpartum depression at an individual level but not at a global model level—can be explained through several nuanced psychological and methodological perspectives. First, the lack of direct effects in the global model suggests that worries operate through intermediary psychological processes rather than having a straightforward, linear relationship with postpartum depression. Potential mediating factors might include worries that may activate underlying cognitive schemas and negative thought patterns that indirectly contribute to depressive symptoms. These mechanisms might be more pronounced at an individual level but become diluted when examining broader population dynamics. Furthermore, individual differences in coping mechanisms, social support, and personal resilience can modulate how worries translate into depressive symptoms. This heterogeneity could explain why a direct relationship disappears in aggregate models. Moreover, related to methodological considerations, in complex psychological models, some variables can suppress or mask the direct effects of other variables when additional factors are introduced. Worries might have a statistically suppressed effect when controlling for other psychological and environmental variables.

Future studies should consider these variables to gain in-depth knowledge of all the factors that may influence the mother-baby bond. As we know, psychological phenomena rarely follow simple linear patterns. The global model might reveal more complex, non-linear interactions that obscure the direct relationship visible at the individual level. The research reveals that the relationship of worries with postpartum depression is intricate and context-dependent. While predictive at an individual level, these effects become more complex when examined through broader statistical lenses, reflecting the sophisticated nature of psychological processes.