The interplay of maternal and paternal postpartum depressive symptoms with children's internalizing and externalizing symptoms from childhood to adolescence: does socioeconomic status matter? A longitudinal cohort study

Abstract

Background

Maternal postpartum depression is an important risk factor for internalizing and externalizing problems in children. The role of concurrent paternal depression remains unclear, especially by socioeconomic status. This study examined independent and interactive associations of postpartum maternal and paternal depression with children's internalizing/externalizing symptoms throughout childhood and adolescence (ages 3.5–17 years).

Methods

We used data from the Québec Longitudinal Study of Child Development, a representative birth cohort (1997–1998) in Canada. Data included self‐reported maternal and paternal depressive symptoms at 5 months' postpartum using the Center for Epidemiologic Studies Depression Scale. Internalizing and externalizing symptoms in children were reported by parents, teachers and children/adolescents using the Social Behaviour Questionnaire (ages 3.5–13 years) and the Mental Health and Social Inadaptation Assessment for Adolescents (ages 15–17 years). We used three‐level mixed effects modelling to test associations after adjusting for confounding factors.

Results

With 168 single‐parent families excluded, our sample consisted of 1,700 families with useable data. Of these, 275 (16.2%) families reported maternal depression (clinically elevated symptoms), 135 (7.9%) paternal depression and 39 (2.3%) both. In families with high socioeconomic status, maternal depression was associated with greater child internalizing (β = .34; p < .001) and externalizing symptoms (β = .22; p = .002), regardless of the presence/absence of paternal depression. In families with low socioeconomic status, associations with symptoms were stronger with concurrent paternal depression (internalizing, β = .84, p < .001; externalizing, β = .71, p = .003) than without (internalizing, β = .30, p < .001; externalizing, β = .24, p = .002).

Conclusions

Maternal depression increases the risk for children's internalizing/externalizing problems in all socioeconomic contexts. In families with low socioeconomic status, risks were exacerbated by concurrent paternal depression. Postpartum depression, especially in low socioeconomic environments, should be a primary focus to optimize mental health across generations.

Introduction

The intergenerational transmission of mental health problems is a recognized public health concern (Goodman, 2020; O'Donnell & Meaney, 2017), with mental health problems affecting 13%–22% of children and adolescents around the world (Costello, Copeland, & Angold, 2011; Polanczyk, Salum, Sugaya, Caye, & Rohde, 2015; Vasileva, Graf, Reinelt, Petermann, & Petermann, 2021). Of the various genetic, biological and environmental influences on child socioemotional development, parental depression is a well‐documented modifiable risk factor (Connell & Goodman, 2002; Goodman et al., 2011; Ivanova, Achenbach, & Turner, 2022; Kane & Garber, 2004; Rogers et al., 2020). With 16%–17% of Canadian mothers (Daoud et al., 2019; Lanes, Kuk, & Tamim, 2011; Liu, Wang, & Wang, 2022) and 13% of Canadian fathers (Da Costa et al., 2015; Rao et al., 2020) experiencing clinically elevated depressive symptoms in the first year after the birth of a child, parental mental health problems become a relevant target of intervention.

Observational and intervention research has traditionally focused on maternal, rather than paternal, postpartum depression. However, it is essential to better understand the independent and interactive roles of mental health in both parents. Assortative mating and emotional contagion theories suggest that maternal and paternal depression often coexist (Field, 2018; Thiel, Pittelkow, Wittchen, & Garthus‐Niegel, 2020). Parental depression and marital conflict/dissatisfaction have a bidirectional relationship (Briscoe & Smith, 1973; Johnson & Jacob, 1997; Whisman, 2001) that could explain why couples are prone to concurrently experience depression. Concurrent paternal depression was reported in 20%–31% of families with maternal postpartum depression (Nishigori et al., 2020; Paulson, Dauber, & Leiferman, 2006; Smythe, Petersen, & Schartau, 2022; Takehara, Suto, & Kato, 2020). We can thereby infer that research on the contribution of each parent's depressive symptoms as well as their co‐occurring effect is important to better understand the risk for children's mental health problems. Unfortunately, longitudinal studies on depression in both parents remain scarce (Fisher, Brock, O'Hara, Kopelman, & Stuart, 2015; Fredriksen, von Soest, Smith, & Moe, 2019; Gutierrez‐Galve et al., 2019; Gutierrez‐Galve, Stein, Hanington, Heron, & Ramchandani, 2015; Hanington, Heron, Stein, & Ramchandani, 2012; Huhdanpaa et al., 2020; Letourneau et al., 2019b; Malmberg & Flouri, 2011; Narayanan & Naerde, 2016; Pietikainen et al., 2020; Van Batenburg‐Eddes et al., 2013; Vänskä et al., 2017; Velders et al., 2011).

Whether concurrent maternal and paternal postpartum depression acts cumulatively or synergistically on child socioemotional development has not been determined. Family resilience theory (Patterson, 2002b) suggests that in the presence of a family stressor like depression in one parent, the other parent could compensate by becoming more involved or developing better parenting skills, thus buffering the effect. On the other hand, family stress theory (Patterson, 2002a) posits that accumulated exposure to depression in both parents could aggravate stress levels and social withdrawal up to a threshold that exponentially exacerbates the negative consequences on the child. Goodman & Gotlib's Developmental Model for Understanding Mechanisms of Transmission (Goodman, 2020; Goodman & Gotlib, 1999) proposes that the father's mental health or psychopathology could either attenuate or exacerbate the effects of maternal depression on child functioning, but few empirical studies have supported this hypothesis (Brennan, Hammen, Katz, & Le Brocque, 2002; Feldman, Wilson, & Shaw, 2020; Gere et al., 2013; Jacobs, Talati, Wickramaratne, & Warner, 2015).

Empirical evidence supports various theoretical models (Figure 1). Some studies support the ‘unique maternal model’ for children's internalizing (Letourneau et al., 2019b; Narayanan & Naerde, 2016; Pietikainen et al., 2020) or externalizing (Carro, Grant, Gotlib, & Compas, 1993; Fredriksen et al., 2019; Pietikainen et al., 2020) symptoms. This model posits that maternal depression is independently related to children's mental health outcomes and that paternal depression has neither an independent nor exacerbating effect. Other studies support the ‘independent additive model’ for children's externalizing behaviours, whereby parental effects are cumulative but not interactive (Brennan et al., 2002; Narayanan & Naerde, 2016; Weinfield, Ingerski, & Moreau, 2009). Yet other studies support the ‘synergistic interactive model’ for internalizing (Carro et al., 1993; Feldman et al., 2020; Goodman, Brogan, Lynch, & Fielding, 1993; Jacobs et al., 2015; Vänskä et al., 2017) and externalizing (Feldman et al., 2020; Goodman et al., 1993) problem behaviours, whereby depression in one parent exacerbates the effect of the other parent's depression. Finally, the ‘crossover interactive model’ posits that the association between maternal depression and child outcomes depends on the mental state of the father. In support of this model, a study observed that maternal depression was associated with offspring's depression when the father did not experience depression, but the association disappeared when the father did (Brennan et al., 2002). In a second study, maternal depression was associated with externalizing symptoms in children only when the father also suffered from depression (Hails et al., 2019).

Figure 1.

Competing theoretical models for the interplay of maternal and paternal depression on children's mental health symptoms. Associations between maternal depression and child symptoms are represented by red slopes for families without paternal depression and blue slopes for families with paternal depression. In the unique maternal model, the red and blue slopes are superimposed, meaning that paternal depression is not associated with child symptoms and does not moderate the association between maternal depression and child symptoms. In the independent additive model, the blue slope is higher than the red slope, meaning that paternal depression is associated with child symptoms, but parallel to each other, meaning that paternal depression does not moderate the association of maternal depression. In the synergistic interactive model, the blue slope is higher than the red slope, meaning that paternal depression is associated with child symptoms, but also steeper than the red slope, meaning that paternal depression strengthens the association of maternal depression. In the crossover interactive model, the blue slope is descending and crosses the red slope. This means that when the father is not experiencing depression, maternal depression is associated with increased child symptoms, but when the father is, maternal depression is associated with decreased child symptoms

Understanding the interplay between maternal and paternal depression is important with respect to screening, prevention and treatment of postpartum depression to optimize children's mental health. Recognizing that promoting the mental health of both parents is important, it may be important to allocate limited resources more effectively when it comes to more intensive, costly interventions. While the unique maternal model would favour targeting mothers exclusively, support for the independent additive or the synergistic interactive model would justify targeting both parents. Understanding the interplay between maternal and paternal depression could also contribute to a better comprehension of the mechanisms explaining its association with child socio‐emotional development. If the independent additive model is confirmed, it would suggest that the mechanisms operate more on the individual level and could be different for maternal versus paternal depression. Results in support of an interactive model would indicate that the mechanisms operate more on the familial level, emerging from the relationships between family members. Unfortunately, few studies have tested all models simultaneously (Vänskä et al., 2017; Weinfield et al., 2009). Vänskä et al.'s (2017) study supported the unique maternal model for internalizing symptoms and the independent additive model for externalizing symptoms. The Weinfield et al. (2009) study also supported the independent additive model for externalizing symptoms but reported no associations between either parent's depression and children's internalizing symptoms. While most have suggested an independent association between postpartum maternal depression and children's mental health problems, results are inconsistent for paternal depression and concurrent parental depression (Capron et al., 2015; Dietz, Jennings, Kelley, & Marshal, 2009; Fredriksen et al., 2019; Hails et al., 2019; Hanington et al., 2012; Laurent et al., 2013; Letourneau et al., 2019b; Malmberg & Flouri, 2011; Narayanan & Naerde, 2016; Pietikainen et al., 2020; Van Batenburg‐Eddes et al., 2013; Velders et al., 2011). These studies have heterogenous methodologies, with the outcomes measured at various periods of child development, by various informants, some considering the socioeconomic contexts and others not. These factors may account for the varying results observed; however, the diversity in methodologies employed in previous studies complicates their interpretation. Therefore, it is essential to re‐examine the interplay between maternal and paternal postpartum depression in relation to children's internalizing and externalizing symptoms across childhood and adolescence.

Family socioeconomic context

Low socioeconomic status (SES) is a major risk factor for parental mental health problems (Eastwood, Phung, & Barnett, 2011; Nath et al., 2016) and child socioemotional development (Letourneau, Duffett‐Leger, Levac, Watson, & Young‐Morris, 2013; Reiss, 2013). Our previous work with the Québec Longitudinal Study of Child Development (QLSCD) (Clément et al., 2024) showed that family SES moderated the independent associations of maternal and paternal depression with child mental health. However, the moderating role of SES in the presence of concurrent maternal and paternal depression was not analysed. While one parent may compensate for depression in the other by greater involvement with the child or better parenting skills (Goodman, 2008), families in precarious living situations may lack the resources to do so (Al‐Matalka, 2014; Hoff & Laursen, 2019). Depression in both parents within a low socioeconomic context may thus synergistically exacerbate the risks for child socioemotional development (Kuh, Ben‐Shlomo, Lynch, Hallqvist, & Power, 2003). To our knowledge, no studies to date have yet examined this question.

Multiple perspectives of children's mental health

Our study also addresses a frequent source of bias in intergenerational mental health research: same‐informant bias. Most longitudinal studies rely solely (Gutierrez‐Galve et al., 2015, 2019; Hanington et al., 2012; Letourneau et al., 2019a; Malmberg & Flouri, 2011; Van Batenburg‐Eddes et al., 2013) or mainly (Fredriksen et al., 2019; Huhdanpaa et al., 2020; Pietikainen et al., 2020) on mother reports of children's mental health symptoms. A recent meta‐analysis highlighted that correlations between parental depression and children's outcomes were larger when the same informants rated both parental depression and children's problems, even in longitudinal analyses (Ivanova et al., 2022). Using both mother and father reports is recommended for assessing different perspectives, especially when parental psychopathology is present (Moreno, Silverman, Saavedra, & Phares, 2008). Likewise, the observations of a child's behaviour in various contexts, such as home versus school, seem to provide different but complementary perspectives. Only one prior study examining associations between parental postpartum depression and children mental health has used a multi‐informant design, including mother, father and teacher reports, and analysed data through a structural equation model (Narayanan & Naerde, 2016). Multi‐level longitudinal modelling is another effective method for analysing multi‐informant data over time, even with missing information and adjusting for autocorrelation of observations within informants and families (Kuo, Mohler, Raudenbush, & Earls, 2000).

Current study

In the present study, our goal was to describe the interplay between clinically elevated maternal and paternal depressive symptoms – hereafter referred to as depression – at 5 months postpartum and their associations with internalizing and externalizing symptoms in children and adolescents from ages 3.5 to 17 years, in families with low and high SES. We used multi‐informant longitudinal data from a population‐based birth cohort in Québec, Canada. Using a family‐level perspective, we aimed to identify which theoretical model best fits the data.

Methods

Participants

The QLSCD is a representative birth cohort study conducted by the Institut de la Statistique du Québec (Québec Institute of Statistics), covering all of Québec except for the Northern Territories and Cree, Inuit, and First Nations Reserves (2.2% of births) (Orri et al., 2021). Singletons born at 24–42 weeks' gestation were selected from the 1997 to 1998 Québec birth registry and stratified by geographical area. Children of mothers who spoke neither French nor English, or who were already participating in another longitudinal study, were excluded. For the present study, we further excluded single‐parent families, because the main objective was to investigate the interplay between paternal and maternal depression.

Study design

Data in the QLSCD include parental depressive symptoms and covariates at 5 months after childbirth. Follow‐up assessments on children occurred annually from ages 3.5 to 8 years, and every 2 years thereafter to age 17 years. Data were collected by telephone interviews with mothers, as well as self‐reported questionnaires from mothers, fathers (or biological mother's live‐in partner in 0.5% of cases), teachers and children. The QLSCD protocol was approved by the Institut de la statistique du Québec and the St‐Justine Hospital Research Centre ethics committees, and informed consent was obtained at each data collection.

Measures

Child mental health

The primary outcome measure was internalizing and externalizing symptoms in the child. For children and early adolescents, these were reported by mothers (five timepoints measures, ages 3.5–8 years), fathers (six timepoints, ages 3.5–13 years) and teachers (six timepoints from 6 to 13 years), using the Social Behaviour Questionnaire (SBQ) (Collet, Orri, Tremblay, Boivin, & Côté, 2023). The SBQ integrates items from the Rutter Children's Behaviour Questionnaire (Rutter, 1967), the Child Behaviour Checklist (Achenbach, Edelbrock, & Howell, 1987), the Ontario Child Health Study Scales (Boyle et al., 1993) and the Preschool Behaviour Questionnaire (Behar, 1977). From ages 10–13 years, children self‐reported their own mental health symptoms using the SBQ. At ages 15 and 17 years, they self‐reported using the Mental Health and Social Inadaptation Assessment for Adolescents (MIA) (Côté et al., 2017).

The number of items in the SBQ varied by age. Internalizing symptoms were measured by SBQ subscales for emotional problems (3–6 items) and anxiety (3–4 items). Externalizing symptoms were measured by SBQ subscales for hyperactivity (4–6 items), inattention (2–4 items), aggressivity (12–14 items), opposition (3–5 items) and conduct behaviours (5–7 items). All items were rated on a three‐point Likert scale (never = 0, sometimes = 1, often = 2) and have good psychometric properties for assessing emotional (Cronbach's α = 0.64–0.85) and behavioural (Cronbach's α = 0.78–0.94) problems during childhood (Collet et al., 2023).

Internalizing symptoms in adolescence were measured by the MIA subscales for depression (eight items, α = 0.90), social phobia (eight items, α = 0.90) and generalized anxiety (nine items, α = 0.86). Externalizing symptoms were measured by MIA subscales for attention deficit with or without hyperactivity disorder (16 items, α = 0.89), conduct disorder (16 items, α = 0.95), opposition disorder (10 items, α = 0.84) and aggression (18 items, α = 0.96) (Côté et al., 2017). The total score for each subscale was standardized as 0–10, and internalizing and externalizing symptoms scores were averaged to obtain total mental health symptoms scores for each measure from various informants and waves of data collection.

Parental depression

Postpartum maternal and paternal depressive symptoms were self‐reported by each parent when children were 5 months old, using a 13‐item scale combining a validated short version (Poulin, Hand, & Boudreau, 2005) (α = 0.81) of the Center for Epidemiological Studies Depression (CES‐D) Scale (Radloff, 1977) and one item from the Edinburgh Postpartum Depression Scale (EPDS): ‘I have felt scared or panicky for no very good reason during the past week’ (Cox, Holden, & Sagovsky, 1987). Responses to each item ranged from 0 (none) to 3 (all the time), for a total score of 0–39. Scores were dichotomized using the validated clinical cut‐off of 16/60 on the original CES‐D (Weissman, Sholomskas, Pottenger, Prusoff, & Locke, 1977), equivalent to 10/39 after adding the EPDS item. Many prior studies have used abridged versions of the CES‐D and these are highly correlated with the original CES‐D (Poulin et al., 2005).

Covariates

We selected potential confounding factors associated with either exposure or outcome (Bradshaw, Riddle, Salimgaraev, Zhaunova, & Payne, 2022; Carneiro, Dias, & Soares, 2016; Dol et al., 2021; Eastwood et al., 2011; Farbstein et al., 2010; Helle et al., 2015; Jansen et al., 2010; Tore et al., 2018). These included: child sex, birth weight and term or preterm (<37 weeks' gestation) birth as obtained from obstetrical records; parental immigration status (born in Canada, European immigrant of non‐European immigrant), age group, education level, occupational status, lifetime history of depression and antisocial behaviours during adolescence and adulthood, as self‐reported by each parent on a 0–10 scale adapted from the Diagnostic Interview Schedule III – Revised (Robins, Helzer, Croughan, & Ratcliff, 1981); and household information such as family type (intact vs. reconstituted), number of siblings, household income, all reported by the mother during interviews at baseline.

Socioeconomic status

Socioeconomic status (SES) refers to the relative position of a family or individual in a hierarchical social structure, based on their access to, or control over, wealth, prestige and power (Statistics Canada, 1996). In the QLSCD, family SES at baseline (1997–1998) was created from the following five variables reported by mothers: maternal and paternal years of education, maternal and paternal occupational prestige, and household annual income before taxes, following a method proposed by Willms and Shields (1996), described in Appendix S1. The standardized continuous score was then dichotomized at the median into a low versus high SES variable. According to examples provided by Statistics Canada (1996), a family with a median SES can be characterized by the following characteristics: parents with high school levels of education, a father employed in a semi‐professional occupation, a mother not participating in the labour force, and a household income near the low‐income cut‐off (Statistics Canada, 1996). This cut‐off is defined as the threshold below which a family is likely to devote a larger portion of their after‐tax income on necessities such as food, shelter and clothing compared to the average Canadian family (Statistics Canada, 2024).

Statistical analysis

We used cross‐classified three‐level mixed effects modelling to test for interactive associations of parental depression and family SES with children's mental health symptoms, with observations (Level 1) nested within informants and waves of data collection (Level 2), and informants and waves nested within children (Level 3) (Figure 2). Multilevel mixed‐effects modelling was used to account for the correlation of errors between measures from multiple informants within the same child at a specific wave of data collection and between repeated measures within the same child from a specific informant. Multi‐level models were cross‐classified because there were two nested variables at level 2 (upper and lower parts of Figure 2): waves of data collection, and informants (Claus, Arend, Burk, Kiefer, & Wiese, 2020). Random effects on the intercepts of waves and informants accounted for the within‐child variability of their mean symptom level across waves or informants. Random effects on the slopes accounted for the within‐child variability of their symptom levels' change across time. Waves and informants were also included as fixed effects at level 3, along with parental depression, family SES, and confounding variables, to account for between‐child variability. Various nesting structures, including or not each of the nested variables, with or without random slopes were compared in exploratory analysis to select the best fitting structure according to the Akaike information criterion.

Figure 2.

Flow diagram showing cross‐classified multilevel study population, using data from the QLSCD 1997–1998 birth cohort. Data were compiled from the final master file of the Québec Longitudinal Study of Child Development (1998–2015), © Gouvernement du Québec, Institut de la statistique du Québec. The outcome data provided by mothers, fathers, teachers, children and adolescents, with repeated measures from ages 3.5 to 17 years, form a cross‐classified three‐tiered hierarchical (multilevel) data structure. Level 1 contains all the outcome measures as observed on the questionnaires. These observations are nested in waves of data collection (upper part of the figure), representing variations between informants within each wave, and nested in informants (lower part of the figure), representing longitudinal variations within informants. Waves and informants at level 2 are nested in children at level 3. Level 3 contains waves of data collection, informants, all covariates measures at the family level, representing variations between children. Cross‐classified multilevel modelling was used to account for the correlation of errors between measures from multiple informants within the same child at a specific wave and between repeated outcome measures within the same child by a specific informant. QLSCD, Québec Longitudinal Study of Child Development; obs., observations; Mo., Mother‐reported; Fa., Father‐reported; Te., Teacher‐reported; Ch., Child self‐reported; Ad., Adolescent self‐reported

We used multiple imputations to account for missing covariate data at baseline, with a ‘missing at random’ assumption resulting in 20 imputed datasets. We applied inverse probability weighting to control for potential bias from selective censoring due to missing data on exposure or outcome variables, using the method described by Weuve et al. (2012). For each observation contributing to the analysis, the weights corresponded to the inverse probability of being uncensored, at specific waves by a particular informant. The covariate balance for each weight is presented in Figure S1. Sensitivity analyses were conducted without weighting. All analyses were performed using R v4.1.0 (R Core Team, 2021), and the packages linear mixed‐effects models (lme4) (Bates, Mächler, Bolker, & Walker, 2015), multiple imputations by chained equations (MICE) (van Buuren & Groothuis‐Oudshoorn, 2011) and tools for multiple imputations in multilevel modelling (mitml) (Grund, Lüdtke, & Robitzsch, 2021).

Model selection

The modelling selection steps are presented in Figure S2. Statistical significance was set at p < .05. The first step was to determine if the model selection was performed on the whole sample or separately for families with low or high SES. Besides the conceptual reasons already mentioned, it was decided to proceed with this step before comparing the various theoretical models to avoid diluting an additive or interactive effect of paternal depression that would have been present only in families with low SES, as a previous study observed (Pearson et al., 2013). Thus, a first model tested two‐way interactions between SES and maternal or paternal postpartum depression, and a three‐way interaction between SES, maternal and paternal postpartum depression at 5 months, adjusted for a set of covariates selected using a stepwise algorithm based on the Akaike information criterion (AIC). If any of these interactions with SES were significant, we stratified our analytic sample by family SES.

The second step was to compare three statistical models: the first predicting child symptoms solely by maternal depression, the second predicting them by both maternal and paternal depression, and the third including the interaction between maternal and paternal depression. The AIC values for each model were compared to select the statistical model that best fitted the data.

The third step was to examine the associations' significance level and select the appropriate theoretical model. If the first model best fitted the data and maternal depression was significantly associated with children's symptoms, the unique maternal model was selected. If the second model best fitted the data but only maternal depression was significantly associated with children's symptoms, the unique maternal model was also selected. If the second model best fitted the data and both maternal and paternal depression were significantly associated with children's symptoms, the independent additive model was selected. If the third model best fitted the data and the interaction term between maternal and paternal depression was significant and positive, the synergistic interactive model was selected. If the interaction term was significant and negative, the crossover interactive model was selected. If an interactive model was selected, analyses were further stratified by maternal or paternal depression to examine more closely how the depression of each parent moderates the association of the depression of the other parent with children's symptoms.

Following sensitivity analyses, we tested the moderating effect of time on the significant associations for each theoretical model fitting our data. This enabled us to evaluate if parental postpartum depression played a transient, an enduring, or a delayed role in the emergence of children's symptoms.

Results

Participants

Of the 2,120 families in the QLSCD, we excluded the 168 families with a single parent 5 months. Of the remaining 1952 families, data on depressive symptoms in both parents and at least one measure of children's mental health symptoms from 3.5 to 17 years of age were available for 25,230 observations from 1,700 children (Figure 2).

Parental depression at child's age of 5 months

Mothers experienced depression in 275 (16.2%) families, fathers in 135 (7.9%) and both in 39 (2.3%). Table 1 shows participant characteristics at baseline, by parental depression after multiple imputation. There were fewer boys in families with paternal depression only (42.7%), and more boys in families where both parents suffered from depression (61.5%), although these differences were not statistically significant. However, almost one out of three of the families with dual‐parent depression were immigrants. Many families with dual depression had the following characteristics: greater tendency than others to have three or more children, child born before age 25 or after age 40, lifetime depression, low SES. Antisocial behaviours were present to some extent in all categories of parental depression, but differences were significant only for maternal behaviours and maternal or dual‐parent depression.

Table 1.

Participant characteristics of the study participants in the QLSCD ^a at child's age 5 months, by parental depression

	Parental depression b
	None (n = 1,329)	Mother only (n = 236)	Father c only (n = 96)	Both parents (n = 39)	All (n = 1,700)	p value
Child characteristics
Male, no. (%)	664 (50.0)	124 (52.5)	41 (42.7)	24 (61.5)	853 (50.2)	.19
Birth weight, g, mean (SD)	3,410 (496)	3,420 (487)	3,410 (473)	3,360 (464)	3,410 (493)	.69
Maternal characteristics
Immigration status, no. (%)
Born in Canada	1,223 (92.0)	209 (88.6)	83 (86.5)	26 (66.7)	1,541 (90.6)	<.001
European immigrant	37 (2.8)	5 (2.1)	3 (3.1)	2 (5.1)	47 (2.8)
Non‐European immigrant	69 (5.2)	22 (9.3)	10 (10.4)	11 (28.2)	112 (6.6)
Age group, no. (%)
<25 years	232 (17.5)	67 (28.4)	20 (20.8)	9 (23.1)	328 (19.3)	.01
25–39 years	1,073 (80.7)	166 (70.3)	75 (78.1)	29 (74.4)	1,343 (79.0)
≥40 years	24 (1.8)	3 (1.3)	1 (1.0)	1 (2.6)	29 (1.7)
Antisocial behavior d , mean (SD)	0.952 (1.12)	1.26 (1.33)	1.15 (1.16)	1.36 (1.48)	1.01 (1.17)	.001
Lifetime depression, no. (%)	181 (12.5)	76 (29.6)	17 (17.0)	15 (33.3)	289 (15.6)	<.001
University degree, no. (%)	449 (31.0)	43 (16.7)	36 (36.0)	7 (15.6)	535 (28.9)	<.001
Paternal characteristics
Immigration status, no. (%)
Born in Canada	1,196 (90.0)	210 (89.0)	88 (91.7)	28 (71.8)	1,522 (89.5)	<.001
European immigrant	40 (3.0)	5 (2.1)	0 (0)	0 (0)	45 (2.6)
Non‐European immigrant	93 (7.0)	21 (8.9)	8 (8.3)	11 (28.2)	133 (7.8)
Age group, no. (%)
<25 years	94 (7.1)	25 (10.6)	6 (6.3)	7 (17.9)	132 (7.8)	.005
25–39 years	1,135 (85.4)	199 (84.3)	78 (81.3)	26 (66.7)	1,438 (84.6)
≥40 years	100 (7.5)	12 (5.1)	12 (12.5)	6 (15.4)	130 (7.6)
Antisocial behavior d , mean (SD)	1.37 (1.60)	1.44 (1.58)	1.58 (1.68)	1.83 (1.95)	1.40 (1.61)	.20
Lifetime depression, no. (%)	127 (9.6)	42 (17.8)	30 (31.3)	17 (43.6)	216 (12.7)	<.001
University degree, no. (%)	369 (27.8)	39 (16.5)	26 (27.1)	6 (15.4)	440 (25.9)	.001
Household characteristics, no. (%)
Intact 2‐parent family e	1,178 (88.6)	206 (87.3)	78 (81.3)	35 (89.7)	1,497 (88.1)	.18
Siblings in family
0	580 (43.6)	93 (39.4)	36 (37.5)	18 (46.2)	727 (42.8)	.05
1–2	685 (51.5)	130 (55.1)	56 (58.3)	15 (38.5)	886 (52.1)
3+	64 (4.8)	13 (5.5)	4 (4.2)	6 (15.4)	87 (5.1)
Low socioeconomic status f	607 (45.7)	149 (63.1)	43 (44.8)	27 (69.2)	826 (48.6)	<.001

SD, standard deviation. Boldface type indicates significant associations at p < .05.

^a Data were compiled from the final master file of the Québec Longitudinal Study of Child Development (1998–2015), © Gouvernement du Québec, Institut de la statistique du Québec.

^b Clinical cut‐off 16/60 in original Center for Epidemiological Studies Depression Scale, equivalent to 10/39 in this study.

^c In 11/1,700 families, the mother's partner living in the household was reported as the father.

^d Score on a 0–10 scale based on self‐reported antisocial behaviours from adolescence to adulthood, adapted from the Diagnostic Interview Schedule III‐Revised.

^e Intact versus blended (reconstituted).

^f Dichotomized score composed of five variables: years of education for the parents, occupational status of the parents and household income.

Moderating effect of socioeconomic status

Family SES played a role in children's internalizing and externalizing behaviours. After adjustment for confounding factors, there was a three‐way interaction between maternal depression, paternal depression and low SES in association with children's internalizing symptoms (β = .84 (0.19 to 1.48); p = .01). This three‐way interaction was marginally significant for externalizing symptoms (β = .62 (−0.02 to 1.25); p = .06) (Table 2). Given that SES was a significant moderator for the associations with internalizing symptoms, we decided to stratify all our analyses by SES. Results before inverse probability weighting are shown in Table S1. Estimates and confidence intervals were almost identical.

Table 2.

Interactions between maternal depression, paternal depression and socioeconomic status at 5 months and children's internalizing and externalizing symptoms from 3.5 to 17 years old, using QLSCD data ^a

	Internalizing symptoms b		Externalizing symptoms c
	β (95% CI)	p	β (95% CI)	p
Maternal depression d	.39 (0.23 to 0.55)	<.001	.23 (0.08 to 0.39)	.003
Paternal depression d	.07 (−0.14 to 0.27)	.54	.04 (−0.16 to 0.25)	.68
SES e (Reference: high)	.17 (0.08 to 0.26)	<.001	.12 (0.04 to 0.21)	.006
Interactions
Maternal depression × Paternal depression	−.38 (−0.87 to 0.12)	.14	−.05 (−0.55 to 0.44)	.83
Maternal depression × SES	−.12 (−0.33 to 0.1)	.28	−.06 (−0.27 to 0.16)	.60
Paternal depression × SES	−.04 (−0.35 to 0.28)	.83	−.01 (−0.33 to 0.31)	.95
Maternal dep. × Paternal dep. × SES	.84 (0.19 to 1.48)	.01	.62 (−0.02 to 1.25)	.06

β, Estimate; CI, Confidence interval; SES, socioeconomic status.

All models weighted to account for selective missing data in the exposure or outcome, using propensity scores from 24 predictors of missingness.

^a Data were compiled from the final master file of the Québec Longitudinal Study of Child Development (1998–2015), © Gouvernement du Québec, Institut de la statistique du Québec.

^b Adjusted for child sex, maternal lifetime history of depression, maternal age at childbirth, maternal and paternal antisocial behaviours, family type and number of siblings.

^c Adjusted for child sex, maternal immigration status, maternal age at childbirth, maternal and paternal antisocial behaviours and family type.

^d Clinical cut‐off 16/60 in original Center for Epidemiological Studies Depression Scale, equivalent to 10/39 in this study.

^e Dichotomized score composed of five variables: years of education for the parents, occupational status of the parents and household income.

Synergistic interactive effect of parental depression in families with low SES

Concurrent maternal and paternal postpartum depression in a low socioeconomic context occurred in 27/1,700 families of the sample. In families with low SES, the third model, including an interaction between maternal and paternal depression, best fitted the data for the associations with both internalizing and externalizing symptoms (Table 3). The interactions between maternal and paternal depression on children's internalizing (β = .48; p = .02) and externalizing (β = .59; p = .01) symptoms were significant (Table 3). These results lead to the selection of the synergistic interactive model as best representing the interplay between maternal and paternal postpartum depression on children's internalizing and externalizing symptoms in families with low SES.

Table 3.

Model selection for the interplay of maternal and paternal postpartum depression on children's internalizing and externalizing symptoms at ages 3.5–17 years, in families with low socioeconomic status, using QLSCD data ^a

	Model 1		Model 2		Model 3
	β (95% CI)	p	β (95% CI)	p	β (95% CI)	p
Internalizing symptoms b
Maternal depression c	.34 (0.2 to 0.47)	<.001	.33 (0.19 to 0.47)	<.001	.27 (0.12 to 0.42)	<.001
Paternal depression c			.19 (−0.01 to 0.39)	.06	.02 (−0.23 to 0.27)	.85
Interaction
Maternal × Paternal depression					.48 (0.06 to 0.89)	.02
Fit index
AIC d	42,427,7		38,560,7		38,558,9
Externalizing symptoms e
Maternal depression c	.26 (0.12 to 0.4)	<.001	.25 (0.1 to 0.4)	.001	.18 (0.02 to 0.34)	.03
Paternal depression c			.26 (0.04 to 0.47)	.02	.05 (−0.21 to 0.32)	.69
Interaction
Maternal × Paternal depression					.59 (0.14 to 1.03)	.01
Fit index
AIC d	36,263.2		32,829.3		32,826.0

All models are weighted to account for selective missing data in the exposure or outcome, using propensity scores from 24 predictors of missingness.

β, Estimate; AIC, Akaike Information Criterion; CI, Confidence interval.

^a Data were compiled from the final master file of the Québec Longitudinal Study of Child Development (1998–2015), © Gouvernement du Québec, Institut de la statistique du Québec.

^b Adjusted for child sex, maternal lifetime history of depression, maternal age at birth, maternal and paternal antisocial behaviours, family type and number of siblings.

^c Clinical cut‐off 16/60 in original Center for Epidemiological Studies Depression Scale (CES‐D), equivalent to 10/39 in this study.

^d Estimated from the first imputed dataset.

^e Adjusted for child sex, maternal immigration status, maternal age at birth, maternal and paternal antisocial behaviours and family type.

Analyses stratified by paternal depression in families with low SES revealed that associations between maternal depression and child symptoms were stronger in the presence of concurrent paternal depression than in the absence thereof: internalizing (Figure 3), β = .84 (0.52 to 1.17), p < .001 versus β = .30 (0.17 to 0.43), p < .001; externalizing (Figure 4), β = .71 (0.24 to 1.19), p = .003 versus β = .24 (0.09 to 0.40), p = .002 respectively.

Figure 3.

Associations between maternal postpartum depression and children's internalizing symptoms at ages 3.5–17 years, by paternal depression and socioeconomic status, using QLSCD data. Data were compiled from the final master file of the Québec Longitudinal Study of Child Development (1998–2015), © Gouvernement du Québec, Institut de la statistique du Québec. Associations between maternal depression and children's symptoms are represented by red slopes for families without paternal depression and blue slopes for families with paternal depression. In families with low SES, the blue slope is steeper than the red slope, implying that paternal depression strengthens the association between maternal depression and children's internalizing symptoms, supporting the synergistic interactive model. In families with high SES, the blue and red slopes are confounded, meaning that paternal depression is not associated with children's internalizing symptoms and does not exacerbate the association of maternal depression, supporting the unique maternal model. SES, socioeconomic status

Figure 4.

Associations between maternal postpartum depression and children's externalizing symptoms at ages 3.5–17 years, by paternal depression and socioeconomic status, using QLSCD data. Data were compiled from the final master file of the Québec Longitudinal Study of Child Development (1998–2015), © Gouvernement du Québec, Institut de la statistique du Québec. Associations between maternal depression and children's symptoms are represented by red slopes for families without paternal depression and blue slopes for families with paternal depression. In families with low SES, the blue slope is steeper than the red slope, implying that paternal depression strengthens the association between maternal depression and children's externalizing symptoms, supporting the synergistic interactive model. In families with high SES, the blue and red slopes are confounded, meaning that paternal depression is not associated with children's externalizing symptoms and does not exacerbate the association of maternal depression, supporting the unique maternal model. SES, socioeconomic status

Analyses stratified by maternal depression in families with low SES showed that associations between paternal depression and child symptoms were significant only in the presence of concurrent maternal depression: internalizing, β = .47 (0.15 to 0.8); p = .004 versus 0.18 (−0.03 to 0.39), p = .09; externalizing; β = .67 (0.29 to 1.05); p = .001 versus 0.18 (−0.07 to 0.43), p = .16. In sum, for families with low SES, the mother's absence of depression was enough to compensate for the presence of paternal depression. Concurrent depression in both parents led to worse outcomes than maternal depression alone.

Sensitivity analyses revealed that the interplay between maternal and paternal depression on children's outcomes was moderated by time among low SES families (Table S2). Specifically, the synergistic role of maternal and paternal depression was not observed during preschool age (3.5 to 5 years), emerged during school age (6 to 12 years) and continued into adolescence (13 to 17 years) (Table S3).

Unique effect of maternal depression in families with high SES

Among families with high SES, the second model, including maternal and paternal depression, best fitted the data for the associations with both internalizing and externalizing symptoms (Table 4). However, only maternal depression was associated with increased children's internalizing (β = .34; p < .001) and externalizing (β = .22; p = .002) symptoms (Table 4). These results lead to the selection of the unique maternal model as best representing the interplay between maternal and paternal postpartum depression on children's internalizing and externalizing symptoms in families with high SES. In other words, children from families with high SES exposed to maternal depression had more internalizing and externalizing symptoms than children never exposed to parental depression. Children exposed to depression in both parents had the same level of internalizing and externalizing symptoms as children exposed only to maternal depression. The sensitivity analyses revealed that the unique maternal association did not vary over time (Table S4).

Table 4.

Model selection for the interplay of maternal and paternal postpartum depression on children's internalizing and externalizing symptoms at ages 3.5–17 years, in families with high socioeconomic status, using QLSCD data ^a

	Model 1		Model 2		Model 3
	β (95% CI)	p	β (95% CI)	p	β (95% CI)	p
Internalizing symptoms b
Maternal depression c	.34 (0.19 to 0.48)	<.001	.34 (0.19 to 0.49)	<.001	.38 (0.23 to 0.54)	<.001
Paternal depression c			−.01 (−0.19 to 0.18)	.95	.06 (−0.14 to 0.27)	.54
Interaction
Maternal × Paternal depression					−.39 (−0.88 to 0.1)	.12
Fit index
AIC d	56,348.6		53,930.6		53,930.9
Externalizing symptoms e
Maternal depression c	.23 (0.09 to 0.36)	.001	.22 (0.08 to 0.36)	.002	.22 (0.08 to 0.37)	.003
Paternal depression c			.03 (−0.14 to 0.2)	.75	.03 (−0.16 to 0.22)	.76
Interaction
Maternal × Paternal depression					−.01 (−0.46 to 0.45)	.98
Fit index
AIC d	47,068.3		45,059.7		45,062.7

All models are weighted to account for selective missing data in the exposure or outcome, using propensity scores from 24 predictors of missingness.

β, Estimate; AIC, Akaike Information Criterion; CI, Confidence interval.

^a Data were compiled from the final master file of the Québec Longitudinal Study of Child Development (1998–2015), © Gouvernement du Québec, Institut de la statistique du Québec.

^b Adjusted for child sex, maternal lifetime history of depression, maternal age at birth, maternal and paternal antisocial behaviours, family type and number of siblings.

^c Clinical cut‐off 16/60 in original Center for Epidemiological Studies Depression Scale (CES‐D), equivalent to 10/39 in this study.

^d Estimated from the first imputed dataset.

^e Adjusted for child sex, maternal immigration status, maternal age at birth, maternal and paternal antisocial behaviours and family type.

Results of the model's comparison in the sample before stratification by SES are presented in Table S5. Model selection would have led to the unique maternal model for internalizing symptoms, and the synergistic interactive model for externalizing symptoms.

The results of the first model for internalizing symptoms including all fixed effects is provided in Table S6.

Discussion

Using a multi‐informant, population‐based cohort study, we examined the independent and interactive associations of parental postpartum depression with children's internalizing/externalizing symptoms from early childhood through to adolescence, while considering the family socioeconomic context. In families with low SES, our results supported the ‘synergistic interactive model’. Effects were synergistic because the presence or the absence of paternal depression, respectively, exacerbated or buffered the associations of maternal depression with mental health outcomes. In families with high SES, our results supported the ‘unique maternal model’, whereby maternal depression increased the risk of children's mental health symptoms, whereas paternal depression did not.

The synergistic nature of the interplay between maternal and paternal postpartum depression on children's mental health outcomes in families with low SES suggests an accumulation of risk. First, concurrent depression in both parents often leads to a rupture of ties with their families of origin and their social network through social withdrawal (Porcelli et al., 2019) or increasing conflicts with network members (Seguin, Bouchard, St‐Denis, Loiselle, & Potvin, 1995). Parents may not seek help from the community and support network outside the family unit, especially in non‐western families (Tang, Zhu, & Zhang, 2016) or in low socioeconomic circumstances (Séguin, Potvin, St‐Denis, & Loiselle, 1999). Second, exposure to depression in both parents in the context of low (vs. high) financial and social resources leads to aggravated risk resulting in a cascading effect throughout the life course and threatens worse outcomes than individual risks borne separately (Kuh et al., 2003). This cascading effect could explain why the synergistic interaction between maternal and paternal postpartum depression increased over time. The fact that we observed no differences between children exposed or not exposed to paternal depression may be explained by mothers compensating and protecting their children from the effects of a partner experiencing depression (Belsky, 1984). However, without the resources available to families living in higher socioeconomic conditions, it is possible that mothers with depression could not effectively compensate for paternal depression.

Previous studies testing interactions between maternal and paternal postpartum depression on child mental health outcomes during preschool age found no significant interactions (Fredriksen et al., 2019; Letourneau et al., 2019b; Malmberg & Flouri, 2011; Narayanan & Naerde, 2016). Our findings which cover the preschool to the adolescence periods, add to this literature by revealing that the interaction observed between maternal and paternal postpartum depression and its effect over children's symptoms emerged only during the school age and adolescence periods and was restricted to children in families with low SES. This novel finding will need to be replicated in other cohort studies that include a long‐term follow‐up.

The synergistic nature of the associations in families with low SES, as well as the unique maternal effect in families with high SES, suggest that maternal postpartum depression has stronger associations with children's socioemotional development than paternal, a finding in line with previous studies (Connell & Goodman, 2002; Fredriksen et al., 2019; Gjerde et al., 2017; Gutierrez‐Galve et al., 2019; Hanington et al., 2012; Letourneau et al., 2019b; Malmberg & Flouri, 2011; Narayanan & Naerde, 2016; Pietikainen et al., 2020; Velders et al., 2011). These results underscore the central role of maternal mental health in the socioemotional development of the child, regardless of socioeconomic status and the father's mental health. Mothers typically remain the primary caregiver, even to this day, and will usually spend more time with their children than fathers do (Fredriksen et al., 2019; Gjerde et al., 2017; Narayanan & Naerde, 2016). However, with changing government policies such as the Québec Parental Insurance Plan (introduced in 2006), modern fathers are taking more paternity leave and providing more hands‐on routine care in the postpartum period (Patnaik, 2019). The impact of such policies should be examined in more recent cohorts when considering paternal mental health.

Our results have implications for therapy, social policy and future research. The mechanisms explaining the association between parents' and children's mental health must be further elucidated. The presence of an interaction between maternal and paternal depression in the association with children's mental health suggests synergistic mechanisms at work, that is, an effect greater than the sum of the parts. According to family systems theory, certain phenomena are understandable only by studying the family as a whole, because of interactions between family members (Bavelas & Segal, 1982; Titelman, 2014). It is equally plausible that mechanisms underlying the observed interactive associations operate in the same way. For example, it has been suggested that intergenerational transmission of attachment could be moderated by the quality of the marital relationship (Dickstein, Seifer, St Andre, & Schiller, 2001), which is highly correlated with parental depression (Garthus‐Niegel et al., 2018; Ramchandani et al., 2011). More research on family‐level dynamics could inform systemic interventions aiming to reduce the risk of externalizing behavioural problems in children exposed to depression in both parents, with the goal of improving mental health in this and the next generation, to break the cycle of mental health problems.

A greater understanding of whether maternal and paternal postpartum depression plays a transient, persistent or delayed role in child socioemotional development is critical to better understanding its role in child development and adapt interventions. Examining these patterns fell outside the scope of this study and was therefore only explored in sensitivity analyses. More research is needed to investigate how the interplay between parents' and children's mental health varies across critical periods of child development.

Strengths and limitations

This study has major strengths. Our sample consisted of a large, longitudinal, representative birth cohort. Data included measures for both parents as well as family conditions at the child's aged 5 months, and mental health outcomes across childhood and adolescence up to age 17 years, offering a life course perspective. The stratification of our analyses by SES before comparing the competing models allowed to detect the synergistic interaction between maternal and paternal depression, which was not observable from the analysis performed on the whole sample. This supports the importance of taking into account the socioeconomic context to avoid diluting information. We systematically adjusted for confounding factors about both parents and the household (adding covariates) to facilitate analysis of the family system as a whole. We used a multi‐informant design and multilevel mixed effects modelling, which is a solid methodological approach to reduce potential same‐informant bias. Finally, we used robust methods like multiple imputation and inverse probability weighting to deal with missing data and mitigate attrition bias.

Nonetheless, the study has several limitations. We did not have clinician reports for diagnostic assessments of clinically relevant mental health problems such as conduct disorder, major depression and generalized anxiety. Although observations were not available from every informant for each timepoint, data on internalizing/externalizing behaviours were available through parents and teachers who saw the child daily, as well as by self‐reports in adolescence. Additionally, the observed prevalence of clinically elevated depressive symptoms in both parents in our study (2.3%) is similar to that reported in the literature (2.4%–3.2%) (Smythe et al., 2022), suggesting that self‐reported scales effectively estimated the prevalence of clinical depression. Second, we could not account for continued parental depression after 5 months because it was not measured for fathers. As the influence between parents and children's mental health can be reciprocal, our data could not explore this bi‐directional aspect. Since a prior study with this cohort revealed the importance of the chronicity of maternal depression in predicting children's outcomes (Ahun et al., 2017), more research is needed to investigate if the same applies to paternal depression. Third, the observational design of the study precluded inferences on causality, allowing only for associations. Fourth, baseline data on parental depression and family circumstances were a single timepoint measure, precluding longitudinal data on timing and duration thereof. These would necessarily impact child development and family dynamics over time. However, postpartum paternal depression is an occurrence at a critical transitional period in the infant's life. The association of postpartum depression with child outcomes up to 17 years later suggests that future research should attempt to understand the mechanisms underlying this association. Fifth, baseline data were taken in 1997–1998. Trends and government policy have changed in the past 25 years, such as the introduction of paid paternity leave. Follow‐up research should try to account for fathers' increasing role in childcare. Estimates of paternal associations in the present study should therefore likely be considered understated and conservative, with preventive and therapeutic programs increased accordingly. This would also apply to families with high SES, with maternal professional careers becoming the norm. Sixth, the sample is ethnically homogenous (91% Caucasian), thus limiting the generalizability of results to other ethnic groups. Since ethnicity and SES are known to overlap in Canada (Kazemipur & Halli, 2001), it is possible that poor families were also under‐represented in our sample, which might have under‐estimated the associations observed in our ‘low’ SES subsample. Seventh, this study adopted a hypotheses‐generating approach, interpreting the role of SES in the associations even with a marginally significant moderation for externalizing symptoms. Thus, results need replication. Finally, the low prevalence (12/1,700) of children exposed to both parents' depression in families with high SES may have been insufficient to detect an interaction between maternal and paternal depression in relation to children's symptoms. Consequently, the potential role of paternal depression in families with high SES cannot be ruled out and should be further investigated.

Conclusion

Exposure to maternal postpartum depression increased the risk for children's internalizing and externalizing problems in all socioeconomic contexts. This risk was exacerbated by the co‐occurrence of paternal depression in low socioeconomic contexts. While more research is needed to elucidate the mechanisms associating parental depression with children's mental health problems (within a family systemic perspective), timely screening and intervention programs promoting postpartum mental health should be offered universally to all pre‐, peri‐ and postpartum mothers. Additional screening and interventions to address paternal postpartum depression may be needed, particularly in families struggling with maternal depression and low socioeconomic conditions.

Ethical approval

The QLSCD protocol was approved by the Institut de la statistique du Québec and the St‐Justine Hospital Research Centre ethics committees, and informed consent was obtained at each data collection.

Key points.

• What's known? Maternal postpartum depression is associated with children's increased internalizing and externalizing problems during childhood and adolescence.

• What's new? In families with high socioeconomic status, our results supported the ‘unique maternal model’, whereby maternal depression increased the risk of children's mental health symptoms, whereas paternal depression did not. In families with low socioeconomic status, we found a ‘synergistic interactive model’, whereby the presence of paternal depression exacerbated associations of maternal depression with children's mental health outcomes.

• What's relevant? The models are relevant to future research. Screening and interventions promoting postpartum mental health should be offered universally to all mothers and prioritized for fathers living in families struggling with concurrent maternal depression and low socioeconomic circumstances.

Supporting information

Appendix S1. Construction of the family socioeconomic status (SES) Variable.

Figure S1. Covariate balance between censored and uncensored observations reported by mothers, fathers, teachers, children and adolescents from ages 3.5 to 17 years, before and after inverse probability weighting (IPW).

Figure S2. Decision tree for model selection.

Table S1. Interactions between maternal depression, paternal depression and low socioeconomic status at 5 months and children's internalizing and externalizing symptoms at ages 3.5–17 years, before inverse probability weighting.

Table S2. Moderating effect of time on the synergistic association of maternal and paternal postpartum depression with children's internalizing and externalizing symptoms at ages 3.5–17 years in families with low socioeconomic status.

Table S3. Interplay of maternal and paternal postpartum depression on children's internalizing and externalizing symptoms at ages 3.5–17 years in families with low socioeconomic status by developmental periods.

Table S4. Moderating effect of time on the maternal unique association of postpartum depression on children's internalizing and externalizing symptoms at ages 3.5–17 years in families with high socioeconomic status.

Table S5. Model selection for the interplay of maternal and paternal postpartum depression on children's internalizing and externalizing symptoms at ages 3.5–17 years (sample not stratified by SES), using QLSCD data.

Table S6. Fixed effects predicting children's internalizing symptoms at ages 3.5–17 years, using QLSCD data.

Data availability statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.