Bidirectional effects between maternal mental health and adolescent internalizing problems across six years in Northern Ireland

Abstract Background Emerging evidence indicates the existence of bidirectional relations between mothers’ mental health and adolescent adjustment, but few studies have examined these relations in contexts of high environmental adversity, including economic deprivation and political violence. Given other empirical connections between political violence and adolescent adjustment problems, the impact of child adjustment problems on maternal mental health may be exacerbated in contexts of sectarian violence. Methods Addressing this gap, latent change score modeling was used to examine interrelations between trajectories of maternal mental health and adolescent internalizing symptoms over time in communities afflicted by political conflict. Over six years, 999 adolescent‐mother dyads participated in a longitudinal study in Belfast, Northern Ireland. Six‐hundred ninety‐five families were originally recruited in year 1, with 304 recruited to supplement the sample in year 3; the largest available sample for a given year was 760 dyads. Models including maternal mental health, adolescent internalizing symptomatology, and political violence (i.e., sectarian antisocial behavior) as a time‐varying covariate were tested. Results Results demonstrated that for both mothers and adolescents in a dyadic pairing, higher rates of symptomology in one member of the dyad were related to symptoms observed in the other member. Results also suggest that political violence and factors related to social deprivation increased symptoms across the dyad. Conclusion This study advances understanding of the bidirectional impact between maternal mental health and adolescent internalizing over time in contexts of political violence.

The impact of maternal mental health on child development has been a major area of research for many years. Empirical work suggests a bidirectional relationship between maternal mental health and child adjustment, indicating that child functioning also impacts maternal functioning (Cummings & Schermerhorn, 2003). However, little is known about factors that may influence these relations in high-risk contexts, such as contexts of socioeconomic deprivation and political violence. Recent research has examined interrelations between maternal and adolescent mental health in contexts of socioeconomic deprivation (Baker et al., 2019;Milan & Carlone, 2018;Shaw et al., 2016), but questions remain about dynamic change in the maternal-adolescent relationship over extended time periods in communities experiencing political violence. Currently, there is limited evidence to confirm that a bidirectional effect persists in adverse contexts. Given other empirical connections between political violence and a range of adolescent adjustment problems (Cummings et al., 2017), the impact of child adjustment problems on maternal mental health may be more pronounced in such conditions.
In the current study, we examined bidirectional relations between maternal mental health and adolescent internalizing symptomatology in the context of socioeconomic deprivation and political violence in Northern Ireland. We hypothesized that maternal mental health and adolescent internalizing symptomatology are predictive of each other's prospective longitudinal trajectories. We also hypothesized that the context of political violence specifically would impact maternal mental health and child internalizing, with higher perceived conflict in the community increasing symptomology for each pair in this reciprocal dyad. Better understanding of these effects over time may lead to improved intervention strategies for mothers and adolescents in high-risk communities (Cummings et al., 2017).

Bidirectionality between maternal and adolescent mental health
Past research has established that maternal depression predicts child internalizing symptomatology and indicates a possible reciprocal relationship in internalizing symptomatology between mothers and children (Goodman & Gotlib, 1999;Goodman et al., 2011). For example, Elgar et al. (2003) examined reciprocal effects of children's adjustment problems and mothers' depressive symptomatology.
Their four-year longitudinal study of 16,581 parent-child dyads found a bidirectional risk of child adjustment problems and maternal depression at clinically significant levels. Specifically, children with depressed mothers were twice as likely to engage in problematic behaviors compared to their peers, and maternal depression also increased with more reported emotional problems in children.
Mothers' and children's distress were thus dynamically interrelated over time, even after prior levels of distress were statistically controlled. Recently, Spiro-Levitt et al. (2019) analyzed the relationship between mother-child depressive symptomatology in 167 dyads undergoing a Depression Prevention Initiative (DPI) program.
A significant positive correlation between child improvement in depressive symptoms and maternal improvement in depressive symptoms was found; as the adolescents improved over the two years, the mothers did as well. Sellers et al. (2016) was one of the first longitudinal studies to establish that adolescent depressive symptoms could trigger a subsequent recurrence of maternal depressive episodes in high-risk mothers who had been previously diagnosed with major depressive disorder.

Bidirectional effects in contexts of socioeconomic inequality and political violence
Recent research has focused on relationships between maternal and adolescent mental health in specific contexts of adversity, including socioeconomic deprivation. For example, Shaw et al. (2016) conducted two longitudinal studies examining the links among child conduct problems, maternal depression, and neighborhood deprivation. In their first study, using mother-son dyads in an economically deprived urban area, longitudinal effects of child conduct problems on maternal depression and maternal depression on childhood conduct problems were indicated, supporting a reciprocal relationship between childhood externalizing symptomatology and maternal depression. Their second study validated the predictive effects of maternal depression on childhood conduct problems. Neighborhood deprivation was predictive of conduct problems in children in both studies. In the current study, we hypothesized that indicators of political violence, commonly associated with socioeconomic deprivation, would factor into maternal-adolescent relations similarly.
Furthermore, Baker et al. (2019) examined the bidirectional relationship between child behavioral problems and maternal depression using longitudinal cross-lagged modeling in mother-child dyads in low-income families. Mothers reported on their own depressive symptomatology, their parenting, and their child's internalizing and externalizing symptomatology. Bidirectional relationships were found between behavioral problems in children and maternal depression across every wave of the study.

Key points
� Further exploring known correlations between maternal mental health and adolescent internalizing psychopathology, this study examined the bidirectionality of maternal and adolescent functioning within communities experiencing sociopolitical adversity � Bidirectional relations were analyzed over six consecutive years in families living in communities impacted by political violence in Belfast, Northern Ireland � Results demonstrated that higher symptomology in one pair of the mother-adolescent dyad predicted an increase of symptoms for the other member in the following year. Politically-related antisocial behavior in the community increased mother's and adolescent's symptomatology, thereby impacting these bidirectional relations � Findings suggest that community programs designed to improve adolescent adjustment in adverse community contexts could benefit from involving the entire family system Milan and Carlone (2018) explored how depression and PTSD symptoms in mothers and adolescent daughters predicted their relational behaviors among mother-daughter dyads from diverse, low-income backgrounds. Adolescents' depressive symptoms significantly impacted their mothers' behaviors towards them after accounting for maternal symptoms. While mothers' relational behaviors predicted changes in adolescents' depressive symptoms, adolescent behaviors were unrelated to subsequent maternal symptoms. Mother and daughter PTSD symptoms both predicted daughters' relational behaviors.
Current work thus supports bidirectional relations between maternal and child mental health in communities with lower socioeconomic statuses. However, bidirectional relations may differ widely depending on the context, and little is known about relations in other contexts adversities, such as ethnopolitical conflict. Given links between political violence and adolescent adjustment problems (Cummings et al., 2017), it is expected that internalizing problems for adolescents will impact maternal mental health symptoms and that reciprocal relations will be found. The current study breaks new ground by examining relations between adolescent internalizing and maternal mental health in a multi-wave longitudinal study of families in Northern Ireland in communities afflicted by protracted political violence.

Northern Ireland
The present study utilizes data from families with adolescents in The current study sampled from families with preadolescent and adolescent youth, as this demographic is more likely than others to be exposed to political conflict and more likely to actively contribute over time to intergenerational transmission of political violence (McEvoy-Levy, 2006). Adolescents in Northern Ireland around the time of data collection were not merely passive victims of sectarianism, but were active participants in protests, rioting, and acts of violence (McEvoy-Levy, 2006;Northern Ireland Riots, 2011). Moreover, as paramilitary organizations in Northern Ireland were disbanded, evidence indicates that adolescents were more likely than other demographics to be involved in sectarian conflicts; young people, unaffiliated with such organizations, demonstrated increased activity in antisocial behavior (e.g., rioting and crime; Haydon & Scraton, 2008;Shirlow & McEvoy, 2008).
Negative outcomes for youth associated with political violence and armed conflict have been identified worldwide (Cummings et al., 2017). Additionally, research has identified links between exposure to political violence, family processes, and psychopathology in adolescence. Family processes are longitudinal predictors of youth adjustment in contexts of political violence (Betancourt et al., 2013;2015;Panter-Brick et al., 2011). Further examination of relations between environmental adversity, family processes, and adolescent adjustment is warranted to improve understanding of family functioning in contexts of political violence.

The present study
In the current study, trajectories of maternal mental health and adolescent internalizing symptoms were modeled in 999 adolescent and mother dyads who participated in a six-year longitudinal survey study in Belfast, Northern Ireland. We hypothesized that symptomology in mothers and adolescents would be reciprocally and positively associated with each other, accounting for political violence in each family's local community. We also hypothesized that political violence would further increase symptoms and reciprocal relations in the dyad.

METHOD Participants
The data come from a 6-year longitudinal survey of motheradolescent dyads in Belfast, Northern Ireland. Mothers and one adolescent were surveyed annually. For households with more than one child, the youngest child eligible for participating was selected to maximize the amount of time families could be followed before the child left home. About half of participating adolescents had older siblings; others were the eldest or only children. At the start of data collection, 695 dyads were surveyed; a supplemental sample was added at time 3 for a total of 999 unique family dyads across the six years. Mothers were surveyed instead of fathers.
Mothers head many households in working-class Belfast, and generally were more available for in-home surveys.  for emotion problems and peer problems (Goodman, 1997). Goodman et al. (2010) recommend combining the emotion and peer subscales to assess internalizing symptomology in non-clinical populations. Adolescents responded to each item with 0 (not true), 1 (somewhat true), or 2 (certainly true). Example items include "I am often unhappy, down-hearted, or tearful" and "Other children or young people pick on me or bully me." On average, internal consistencies for internalizing scores were 0.62.

Maternal mental health.
Mothers completed the 12-item version of the General Health Questionnaire (GHQ-12), which measures mental health indicators (Goldberg et al., 1997). Responders rate their mood and level of distress on a scale from 1 (not at all or much less than usual) to 4 (much more than usual). Sample items include "Have you recently lost much sleep over worry?" and "have you recently felt that you are playing a useful part in things?" Higher scores indicate more mental distress. On average, internal consistencies for maternal mental health were 0.89.

Sectarian Antisocial Behavior. Adolescents and mothers
completed the Sectarian Antisocial Behavior Scale (SAB) to assess awareness of these behaviors within their communities. The SAB is a 12-item scale assessing respondents' awareness in the previous 3 months to a range of sectarian-motivated behavior, from witnessing stones thrown over peace walls to politically-driven murder. Items were answered using a 5-point Likert-type scale, with choices ranging from (1) (not in the last 3 months) to (5) (every day). Average Cronbach's alphas were 0.94 (mother report) and 0.93 (child report).

Statistical analysis
Univariate and bivariate latent change score models ( First, the best LCSM was determined to characterize trajectories of adolescent internalizing symptoms and maternal mental health separately. Second, the trajectories were modeled simultaneously to test for dyadic effects between mothers and adolescents. All latent variable models were fitted using the lavaan package (Rosseel, 2012) in the R statistical computing software (R Core Team, 2021). We assumed the data were missing at random for families that did not respond in a given year and estimated the model using fullinformation maximum likelihood to address missing data.
Univariate longitudinal models. Univariate LCSMs were tested to evaluate the trajectory of adolescent and maternal distress symptoms separately. The LCSM includes a constant change parameter, which indicates that the processes are changing linearly (with intercept and slope parameters) and a proportional change parameter that allows for dynamic, possibly non-linear growth. Proportional change means that the rate of change depends on the initial severity of symptoms. A path diagram representing univariate LCSM is presented in Figure 1. Sequential models were fitted examining linear change only, proportional change only, and the dual-change LCSM with both change processes. SAB was included as a timevarying covariate in all models (i.e., adolescent's GHQ score at time T A B L E 1 Descriptive statistics of the variables used in the study reported as "mean (standard deviation)" Year Note: The age range for each year is reported instead of SD. Attrition is reported as the proportion of lost families for a given year relative to the previous year. The asterisk* for year 3 denotes that a supplemental sample of 304 families was added at this time; attrition is based on the continuation of the original 567 from year 2 to year 3.
1 was regressed on adolescent's SAB score at time 1). The effect of SAB was freely estimated at each time point because the context of political conflict changed year-to-year.
Dyadic longitudinal models. Next, the bivariate LCSM was fitted ( Figure 2). In this model, the best-fitting trajectories for adolescents and mothers from the univariate step were fitted simultaneously, and different patterns of "coupling" were tested across the dyad. Direct sectarian-related events. We also controlled for age by regressing adolescent intercept and slope on each person's age at baseline. 1 Models were compared using a χ 2 difference test, comparing nested models that differed by inclusion or exclusion of key parameter estimates of interest. If a simpler model did not fit significantly worse than its more complex counterpart, then the simpler model was preferred over the fuller model because the additional parameter(s) did not significantly improve fit with the observed data. In these analyses, the full model included dyadic coupling, and reduced models allowed for only maternal impact on adolescents, only adolescent impact on mothers, and no coupling in the dynamic process.  Table 3.

RESULTS
Starting with the fully coupled process model, the model displayed Key parameter estimates are presented in Table 4. Critically, the coupling parameter estimates were statistically significant and positive in value, meaning higher symptoms in one pair for each year contributed to higher symptoms in the other member for the following year. The proportional change coefficients were negative in value and remained significant, but the linear trajectory was no longer significant when accounting for the coupling process. In other words, changing symptoms from year-to-year were a function of the previous year's symptoms (in addition to the coupling effect) and depended upon one's initial level (i.e., intercept). While both mothers and adolescents generally declined in symptoms by the final year of study, higher symptomatology of the other partner in the dyad could slow or even reverse this decline in symptoms. Figure S2 displays model-predicted trajectories of internalizing and GHQ scores over the six years for mothers and adolescents with high, average, and low scores. Adolescent internalizing at time t had a larger impact on maternal mental health at time t + 1 than did the maternal proportional decline coefficient (0.635 (SE = 0.14) vs. −0.343 (SE = 0.08), respectively). Ultimately, mothers with high initial GHQ symptoms and whose adolescents displayed higher levels of internalizing F I G U R E 2 Full bivariate latent change score model. Change processes modeled in Figure 1 are simultaneously modeled for maternal and adolescent symptoms. Dyadic coupling is modeled by γ a and γ m (impact of adolescent internalizing at time t on the change in maternal mental health distress between time t and t + 1, and vice versa), plus correlations between residuals x = adolescent internalizing, y = maternal general health questionnaire. realized an initial increase in GHQ score over time, before returning to initial levels in later years ( Figure S2).
It was also critical to examine the effect of SAB at each time point, allowing for time-varying effects to reflect differences from year to year in sectarian conflict. For both mothers and adolescents, SAB had a significant positive prediction of symptoms at times 2 and 3 (Table 4). SAB also had a significant positive effect on GHQ at time 4, and had a small, surprising negative effect on internalizing at time 1. The correlation between mother and adolescent SAB report was high within each wave of the study ( Figure S1). Based on the model,  Specifically, both mothers and adolescents generally declined in symptomatology temporally; this decline occurred in proportion to the level of symptomatology expressed when they entered the study.
However, the rate of decline was slowed or reversed by the symptomatology of the other partner in the dyad. Adolescent internalizing at time t had a larger impact on maternal mental health at time t + 1 than did the maternal proportional decline coefficient, thus highlighting the critical effect adolescent functioning has on maternal functioning over time. Moreover, it was found that high initial levels of adolescent symptomology resulted in a net increase in mental health problems (i.e., GHQ scores) for mothers over time.
These findings support past research which emphasizes the reciprocal relationship between maternal and child depressive symptomatology (Baker et al., 2019;Elgar et al., 2003;Hughes & Gullone, 2010

Limitations
This study has a few limitations to consider. First, these results are based entirely on self-reported questionnaire assessments. Another limitation is that we operationalized emotional difficulties as continuous traits. Although this methodology yields some strengths (e.g., continuous outcome scores increase statistical power and differentiate the severity of problems among non-clinical samples in the absence of formal diagnosis), these results may not apply to those with clinical diagnoses. Additionally, there were low internal consistencies in the SDQ scale. However, the SDQ is widely used, and relatively low internal consistencies are typically reported in other studies as well. Finally, this study considers only the relations between mothers and adolescents and did not survey fathers. The role of fathers' impact on adolescent adjustment in Northern Ireland is discussed elsewhere (Luningham et al., 2020).

Future directions
Future research should address the impact of bidirectional effects between adolescent internalizing symptomatology and paternal mental health. Past work (Cummings et al., 2016) supports that emotional security in the family or in the community merits consideration in models testing the bidirectional link between maternal and adolescent functioning.

CONCLUSIONS
Our study emphasizes the significance of the bidirectional relationship between maternal functioning and adolescent adjustment, and the exacerbation of these relations in contexts of political violence. This study suggests that family processes, specifically the relationship between mothers' mental health and adolescent internalizing symptomatology are not only linked but influence each other over time.