Parenting and Youth Onset of Depression Across Three Years: Examining the Influence of Observed Parenting on Child and Adolescent Depressive Outcomes

Abstract

Concurrent associations between parenting behaviors and youth depression are well established. A smaller body of work has demonstrated longitudinal associations between aspects of parenting and youth risk for depression; however, this limited longitudinal work has predominantly relied upon self- and parent-report questionnaire measures and is thus affected by biases related to retrospective recall and common method variance. The present study used behavioral observation measures of parenting and clinical interview measures of youth depression to examine prospective relationships between observed parental support, responsiveness, criticism, and conflict and youths’ onset of a depressive episode in a 3-year longitudinal design. Participants included 585 community youth age 8-16 (M = 11.92, SD = 2.39, 56.6% female) and a participating caregiver. Parental behavior was coded by trained observers in the context of a 5-minute conflict resolution discussion at the baseline assessment. Youth onset of depression was subsequently assessed every 6 months for a period of 3 years using the Schedule for Affective Disorders and Schizophrenia for School Aged Children (KSADS) to ascertain whether youth experienced onset of depressive episode over the follow-up. Logistic regression analyses indicated that greater parental conflict at baseline predicted higher odds of youth experiencing a depressive onset across the 3-year follow up period, even after controlling for youth and caregiver history of depression at baseline. Findings suggest that parental conflict is particularly influential in youth vulnerability to depression.

Epidemiological research indicates marked increases in the prevalence of depression as youth transition from childhood through adolescence. Results of large, nationally representative epidemiological surveys indicate that rates of depression nearly double between ages 13 and 18, with lifetime prevalence estimates reaching adult rates by age 18 (Merikangas et al., 2010). In addition to significant concurrent distress and psychosocial impairment (Avenevoli, Swendsen, He, Burstein, & Merikangas, 2015), child and adolescent depressive disorders has been associated with adverse outcomes across the life course, including poor health, educational underachievement, unemployment, and mental health difficulties persisting into adulthood (e.g., Fergusson & Woodward, 2002; Keenan-Miller, Hammen, & Brennan, 2007; Johnson, Dupuis, Pinche, Clayborne, & Colman, 2018).

Among those factors implicated in youth risk for depression are a host of parenting behaviors (Birmaher et al., 1996; Sander & McCarty, 2005; Schwartz, Sheeber, Dudgeon, & Allen, 2012; Schwartz et al., 2017). Conceptual models describing relations between parenting and youth depression propose that parenting characterized by lack of sensitivity and support or excessive conflict and criticism contribute to youth vulnerability to depressive outcomes through additive and synergistic effects on children’s social, cognitive, and affective functioning (Sheeber, Hops, & Davis, 2001). Critical and conflictual behavior, as well as sensitive/responsive and supportive behavior, are theorized to be particularly influential to youth depressive outcomes, as these dimensions of parenting may be especially salient for youth socioemotional development. Specifically, theoretical and empirical work suggest that parenting characterized by high levels of hostility (encompassing dimensions of criticism and conflict) and/or low levels of warmth (encompassing dimensions of support and responsiveness) are related to maladaptation involving youth emotion regulation, cognitive processing, and social skill acquisition, and may thus be instrumental in contributing to youth risk for depressive onset (Sheeber et al., 2001; Schleider & Weisz, 2017).

Consistent with this theoretical framework, a rich body of literature has emerged over the past 20 years implicating aberrant parenting processes in youth risk for depression. Results of both cross-sectional and longitudinal studies converge to suggest that both positive (e.g., warm, supportive) and negative (e.g., hostile, rejecting) parenting factors are influential in youth risk for depression (see Yap et al., 2014). For example, families of depressed youth have been observed to express more anger during a laboratory-based problem-solving task relative to families of non-depressed youth (Bodner, Kuppens, Allen, Sheeber, & Ceulemans, 2018), and self-reported family conflict has been found to predict increases in self-reported depressive symptoms among adolescent youth across a three-year follow-up period (Kelly et al., 2016). Moreover, mothers of depressed adolescents have been observed to demonstrate less supportive behavior during laboratory-based interaction tasks relative to mothers of non-depressed adolescents (Pineda, Cole, & Bruce, 2007), and parent-reported parental warmth and support among parents of youth ages 12 to 18 have been found to predict lower levels of offspring self-reported depressive symptoms 5 to 6 years later (Aquilino & Supple, 2001). Indeed, meta-analytic findings support moderate concurrent associations, and small to moderate prospective associations, between such parenting factors as parental warmth and hostility, and adolescent depression (McLeod, Weisz, & Wood, 2007; Yap, Pilkington, Ryan, & Jorm, 2014).

However, the extant body of literature examining relationships between parenting and youth depression is marked by a number of methodological limitations. Much of the work examining associations between parenting and youth depression has been cross-sectional in nature; of studies analyzed by Yap et al. (2014), fewer than 20% of studies specifically examining parental aversiveness, for example, and fewer than 15% of studies examining parental warmth were characterized by longitudinal designs. Cross-sectional work is unable to clarify prospective relationships between parenting processes and youth depression and thus cannot address whether parenting functions as a risk factor for youth depressive onset; cross-sectional findings may instead reflect variance in parenting accounted for by youth depression. Rigorous, longitudinal research is needed to identify parenting factors that prospectively predict youths’ likelihood of developing depressive episodes over time.

Moreover, existing longitudinal research has predominantly relied upon self-report questionnaire measures of parenting variables and youth depression symptoms (e.g., Aquilino & Supple, 2001; Hamza & Willoughby, 2011; Kelly et al., 2016; Vaughan, Foshee, Ennett, 2010) and is thus susceptible to biases associated with retrospective recall and common method variance (McLeod et al., 2007; Podsakoff, Mackenzie, Lee, & Podsakoff, 2003). Indeed, meta-analytic findings suggest that studies examining relationships between parenting and youth depression using questionnaire measures overestimate effect sizes relative to studies incorporating behavioral observation measures of parenting (McLeod et al., 2007).

Behavioral observation measures represent a gold standard method in the study of parenting behaviors, yielding estimates of parenting unaffected by biases characteristic of self- and parent- report measures. Indeed, given evidence of effect size inflation among studies relying only on questionnaire measures of parenting (McLeod et al., 2007), behavioral observation research represents a particularly rigorous method for investigating prospective relationships between parenting and youth depressive onset. Some prospective behavioral observation research has examined observed parenting factors contributing to increases in youth depressive symptoms over time (e.g., Schwartz, Dudgeon, Sheeber, Yap, & Simmons, 2012; Sheeber, Hops, Alpert, Davis, & Andrews, 1997); however, research is needed to clarify parenting behaviors contributing to youth onset of depressive disorder, given high levels of distress and impairment associated with an onset of clinical depressive disorder during childhood or adolescence (Avenevoli et al., 2015; Fergusson & Woodward, 2002; Keenan-Miller et al., 2007; Johnson et al., 2018). Findings from cross-sectional, case-control studies using behavioral observation measures to compare parenting behaviors and processes in families with clinically depressed youth to families with non-depressed youth suggest that observed parenting behaviors are associated concurrently with depressive disorder (e.g., Bodner et al., 2018; Sheeber, Davis, Leve, Hops, & Tildesley, 2007); however, little research has used a longitudinal design to examine prospective relationships between observed parenting and youth depressive disorder across adolescence.

A notable exception to the preponderance of cross-sectional and monomethod self-report questionnaire studies of parenting and youth depression is work by Schwartz and colleagues (2014) conducted in association with the Orygen Adolescent Development Study (OADS). Using a prospective, repeated measures design, Schwartz et al. (2014) found that greater levels of observed parental aggression and reduced levels of observed parental positivity, measured in terms of average frequency of parental aggressive (e.g., contemptuous, angry, belligerent) or positive (e.g., happy, caring, approving) behavior per minute of interaction, predicted youth first onset of major depressive disorder (MDD) across a 6-year follow-up period spanning approximately age 12 to age 18. As the first rigorous, multimethod longitudinal study to probe prospective relationships between observed parenting and youth depression onset, Schwartz et al. (2014) provide important insight into predictive relationships between parenting behavior and youth depressive disorder.

Notably, however, parenting is a multifaceted construct characterized by a wealth of dimensions potentially salient to understanding youth risk for psychopathology. Findings from Schwartz et al. (2014) suggest that parental aggression and parental positivity predict youth depressive outcomes; however, no work has been conducted to date examining prospective relationships between other potentially influential dimensions of observed parenting and youth onset of depression. Parental support, responsiveness, criticism, and conflict, for example, represent commonly assessed and theoretically relevant aspects of parenting influential to youth development (e.g., Fosco, Caruthers, & Dishion, 2012; Johnston, Murray, Hinshaw, Pelham, & Hoza, 2002; Chaplin et al., 2012), yet it is not yet understood how these dimensions of observed parenting behavior relate to youth risk for depressive onset. Meta-analytic evidence suggests that specific subdimensions of positive and negative parenting, such as parental support and parental conflict, demonstrate differential concurrent relations to youth depression (McLeod et al., 2007), indicating the need to study unique relationships between these different facets of parenting behavior and youth liability to experience a depressive episode over time. Identifying prospective associations between parenting behaviors at the facet-level and youth risk for depressive episode may be of particular translational value; namely, enhanced specificity in our knowledge of parenting-depression associations may be critical to informing maximally effective parenting-based interventions to circumvent youth risk for depression.

In summary, decades of work have implicated parenting processes in youth risk for depression, and this rich body of research has made indelible contributions to knowledge of social factors contributing to youths’ vulnerability to psychopathology. Of note, however, much of the work leveraging gold standard behavioral observation measures of parenting has been cross-sectional in nature, precluding analyses of observed parenting processes related to prospective change in youth depression. Further, the longitudinal research that does exist has largely examined relations between observed parenting and youth depressive symptoms (see Schwartz et al., 2014 for an exception), which, while providing important insight into the ways in which parenting may contribute to individual differences in youth emotional experience, does not evaluate if parenting processes uniquely predict the onset of clinically significant depressive disorders. Thus, research is needed to clarify specific aspects of parenting that prospectively predict youth onset of depressive disorder in order to advance knowledge of parenting processes that contribute to youth risk for depression associated with clinically significant distress and impairment.

The Present Study

To address these limitations in the extant body of literature, the present study used behavioral observation measures of parenting to examine prospective relationships between parenting behaviors and onset of youth depressive episodes across a three-year period in a large community sample of youth. Specifically, dimensions of observed positive (supportive, responsive) and negative (critical, conflictual) parenting were examined as predictors of youth onset of depressive episode across three years among children and adolescents. Parental criticism and parental conflict were chosen to represent the content and affective tone of parents’ negative behaviors toward their adolescent children, respectively; parental criticism reflected parenting characterized by blaming or disapproval, whereas parental conflict reflected parenting characterized by affectively charged disagreement or displays of hostility. Similarly, parental support and responsiveness were chosen to represent two related but distinct aspects of positive parenting; parental support reflected parenting characterized by adolescent-directed praise and admiration, whereas parental responsiveness reflected parenting characterized by high levels of sensitivity and attunement with child emotions. Youth diagnoses of depression were assessed by trained clinical interviewers every six months across a 36-month follow up period, yielding up to 7 diagnostic assessment points per participant, providing more precise and reliable estimates of youth experience of clinically significant depression across 3 years (Little, Rhemtulla, Gibson, & Schoemann, 2013).

We hypothesized that negative aspects of parenting, specifically parental criticism and conflict, would predict greater odds of youth experiencing an onset of depression across the 36-month follow up period, and that positive parenting, specifically parental support and responsiveness, would predict lower odds of youth experiencing an onset of depression across this period. We did not make a priori hypotheses regarding the form of differential patterns of associations between specific aspects of parenting and youth risk for depression. Additional exploratory analyses were conducted to examine developmental stage and gender as potential moderators of the prospective relationships between parenting and youth depressive onset. Adolescent development is characterized by normative changes in the parent-child relationship (Longmore, Manning, & Giordano, 2013; Steinberg & Morris, 2001), as well as risk for depression (Hankin et al., 2015; Merikangas et al., 2010); thus, it is possible that the strength of associations between parenting and youth depression may vary across different stages of adolescent development. Gender was also explored as a moderator given gender differences in rates of adolescent depression (e.g., Hankin et al., 1998; Hankin et al., 2015; Wade, Cairney, & Pevalin, 2002) and interpersonal sensitivity among adolescent youth (e.g., Crawford, Cohen, Midlarsky, & Brook, 2003; Hankin, Mermelstein, & Roesch, 2007; Rudolph, 2002). Finally, given evidence that caregiver history of depression predicts both parenting behaviors and youth risk for psychopathology (Goodman, 2007; Goodman & Gotlib, 1999; Goodman et al., 2011; Lovejoy, Graczyk, O’Hare, & Neuman, 2000), secondary sensitivity analyses were conducted with caregiver history of depression added as a covariate in order to assess if parenting behaviors predicted unique variance in youth odds of experiencing a depressive episode, controlling for caregiver depression.

Method

Participants and Procedures

Participants included 585 youth from the community recruited in 3^rd, 6^th, and 9^th grade cohorts (age 8-16 at baseline, approximate M_age = 11.92, SD = 2.39, 56.6% female) and a participating caregiver recruited from the greater Denver and central New Jersey areas in association with the Gene, Environment, and Mood (GEM) Study (Hankin et al., 2015). Participants were recruited using brief information letters sent to families in participating school districts. Of families to whom these letters were sent, 1,108 responded and requested further information. Inclusion criteria included English language fluency, absence of autism or psychotic disorder diagnosis, and IQ > 70 (i.e., lack of intellectual disability) as assessed via parent report. Of those families that contacted the laboratory, 665 (60%) qualified as study participants, as they met inclusion criteria and completed an in-person laboratory assessment. The present studies include those 585 youth for whom complete data from the baseline parent-child behavioral observation task and at least one diagnostic interview follow up assessment were available. Sample demographics were approximately representative of the ethnic and racial characteristics of the United States population (69.1% Caucasian, 10.9% African American, 9.1% Asian/Pacific Islander, 5.5% Multi-racial, 5.5% other racial background; 12.1% Latinx). Participating caregivers included predominantly mothers (89.4%), with smaller numbers of fathers (6.0%) and other caregivers (e.g., grandparents, aunts; 1.2%).¹ Further details regarding sampling procedures and participant characteristics are described in Hankin et al. (2015).

All procedures were approved by the Institutional Review Board.² Informed consent and assent were obtained from all participants prior to administration of study procedures. Participating youth were invited to the laboratory to complete a baseline assessment, during which youth and their participating caregiver completed a semi-structured diagnostic interview (Schedule for Affective Disorders and Schizophrenia for School Age Children [K-SADS-PL]; Kaufman et al., 1997), as well as a standardized 5-minute videotaped interaction task during which the dyad was instructed to discuss a source of disagreement in their relationship (e.g., completing homework; Feng et al., 2009; Kim Park, Garber, Ciesla, & Ellis, 2008). Videotapes were coded by trained research staff to yield observational measures of parenting behaviors, as described below. Youth and their parents subsequently completed diagnostic interview measures every six months for a period of three years (36 months), yielding up to seven total assessment points per participant. Of the total sample, 85.1% (N = 498) was retained through the 36-month assessment. Youth who completed the study did not differ from youth who did not complete the study on any parenting measure at baseline (all ps > .05). Youth who did and did not complete the study were similarly equivalent in age, gender, and history of depression at baseline (all ps > .05).

Measures

Parenting behaviors.

Reliable, independent raters coded parental support, responsiveness, conflict, and criticism during the baseline parent-child interaction task. Global codes for each parenting construct (support, responsiveness, conflict and criticism) were assigned on a scale of 1 to 5 (1 = not at all characteristic of the parenting behavior during the interaction and 5 = highly characteristic of the parenting behavior during the interaction). Codes were based on validated parent-child coding systems and reflect theoretically grounded dimensions of parenting (Melnick & Hinshaw, 2000; NICHD Early Child Care Research Network, 1999), and represent dimensions of parenting described by leading conceptual models describing relations between parenting and youth depression (Sheeber et al., 2001). Parents rated as high in support demonstrated engaged and affirming behavior, providing validating comments (e.g., “I can see that”) as well as praise or recognition of their child. High levels of responsiveness were demonstrated by parents who followed their child’s lead and matched their child’s affect throughout the discussion, remaining undistracted and displaying appropriately responsive social gestures (e.g., nodding). Parental conflict was indicated by displays of hostility or anger. Parental criticism was indicated by such behaviors as statements of disapproval or insults directed toward the child, as well as parental blaming or inappropriately critical behavior. Codes are consistent with prior work assessing positive and negative parenting (Chi & Hinshaw, 2002; Corona et al., 2005; Davidov & Grusec, 2006). Approximately 20% of videotaped observations were double coded. Intraclass correlations between coders ranged from .71 to .85 on all subscales in this study, indicating good interrater reliability.

Youth Depression.

Youth depression diagnoses were assessed at each time point by trained clinical interviewers using the Mood and Psychosis sections of the K-SADS-PL (Kaufman et al., 1997). The K-SADS is a well-validated and widely used measure of youth affective psychopathology. All interviewers were trained and supervised by Ph.D. level, licensed clinical psychologists in administering the K-SADS and assigning clinical diagnoses according to criteria described in the Diagnostic and Statistical Manual of Mental Disorders – Fourth Edition (DSM-IV; American Psychiatric Association, 1994). Consistent with recommendations by K-SADS authors and developers, interviewers applied “skip out” procedures, such that the interview was discontinued if youth did not endorse a criterial symptom of depression (Kaufman et al., 1997). Interviewer training comprised approximately 40 hours didactic instruction, listening to audiotaped interviews, and conducting practice interviews. Interviewers additionally attended weekly supervision sessions facilitated by the principal investigators (PIs) at each site to ensure ongoing diagnostic fidelity. Study Pis also reviewed interviewers’ notes and tapes to confirm diagnostic decisions.

Both youth report and parent report on the K-SADS were considered in determining youths’ diagnostic status using best estimate diagnostic procedures (Klein, Dougherty, & Olino, 2005). Reliability estimates based on 20% of interviews indicated good reliability (κ = .91). For the purposes of the present study, youth were determined to have experienced a depressive episode if they met DSM-IV criteria for major depressive disorder (MDD) definite, MDD probable (four threshold depressive symptoms lasting at least 2 weeks), or minor depressive disorder (mDD) definite (two or three threshold depressive symptoms lasting at least 2 weeks). Diagnoses were collapsed into a single variable given evidence that depression is distributed dimensionally at the latent level (Hankin, Fraley, Lahey, & Waldman, 2005) and that all diagnoses included in the present study (MDD definite, MDD probable, and mDD definite) are associated with significant distress and impairment (Avenevoli et al., 2015; Gotlib, Lewinsohn, & Seeley, 1995).

At baseline, youth were assessed for lifetime history of depression prior to enrollment in the study. At subsequent 6-month follow-ups, youth and a caretaker were interviewed to ascertain if the child met DSM-IV diagnostic criteria for a depressive episode within the preceding 6 months. For the purposes of the present analyses, youth depression diagnoses across all assessment points were collapsed into a single binary variable indicating presence (1) or absence (0) of onset of depression across the 36-month follow up period. Youth history of depression was similarly represented by a binary variable indicating presence (1) or absence (0) of a depression diagnosis prior to enrollment in the study. Youth meeting criteria for a diagnosis of a depressive disorder were asked to report on whether or not they had received mental health treatment since their last assessment. In total, 19.1% (n = 112) youth endorsed receiving some form of mental health treatment over the course of their participation in the study.

Caregiver Depression.

Caregiver history of depression was assessed using the Structured Clinical Interview for DSM-IV (SCID; First et al., 2002) assessing caregivers’ lifetime and current depression. The SCID is a well-validated and widely used semi-structured interview measure of adult psychopathology. Clinical interviewers comprised extensively trained and supervised research staff, as described above. For the purposes of the present analyses, caregiver history of depression comprised a dichotomous score indicating presence (1) or absence (0) of caregiver history of Major Depressive Disorder (MDD) definite, MDD Probable, Minor Depression definite, or Dysthymia following criteria described in the Diagnostic and Statistical Manual of Mental Disorders IV-TR (American Psychiatric Association, 2000) prior to enrolling in the study.

Data Analytic Plan

Logistic regression analysis was used to evaluate prospective relationships between observed parenting variables and youth onset of depression. Youth history of depression at baseline was included as a covariate in all analyses to control for prior history of depression before study entry in order to more robustly evaluate prospective relationships between parenting at baseline and youth onset of depression diagnosis across the 36-month follow up period. To address exploratory aims, separate Grade Cohort × Parenting, and Gender × Parenting, interaction terms were created to probe grade cohort and gender, respectively, as moderators of prospective relationships between parenting and youth depressive onset, controlling for youth prior history of depression. Grade cohort was chosen a potential moderator rather than age given that youth were recruited in discrete developmental cohorts rather than along a continuum of chronological age.

Results

Preliminary Analyses

Descriptive statistics for primary variables of interest in the sample overall, as well as by gender, are reported in Table 1. Mean levels of parental support, responsiveness, criticism, and conflict were not significantly different between boys and girls. As reported in Table 1, 26.0% of participants experienced a depressive episode over the course of the study. Gender differences in depression in this sample have been previously reported (see Hankin et al., 2015 for review).

Table 1.

Descriptive Statistics and Gender Differences in Primary Variables of Interest

	Overall M (SD)	Boys M (SD)	Girls M (SD)	t(df)	p	Cohen’s d
Parental support	2.85 (1.02)	2.81 (1.05)	2.88 (1.00)	−.77 (583)	.441	.07
Parental responsiveness	4.09 (.83)	4.09 (.87)	4.10 (.80)	−.14 (583)	.890	.01
Parental criticism	2.16 (1.04)	2.09 (1.01)	2.21 (1.06)	−1.30 (583)	.194	.12
Parental conflict	1.82 (1.00)	1.76 (1.02)	1.87 (1.00)	−1.30 (582)	.195	.11
	Both Genders N	Boys N	Girls N	OR	p
Youth with past depression history before study start	114 (19.5%)	44 (17.3%)	70 (21.1%)	1.34	.247
Onset of depression over 3-year study follow-up	152 (26.0%)	49 (19.3%)	103 (31.1%)	1.89	.001

Note. OR = odds ratio

Correlations between primary variables of interest are reported in Table 2. All parenting variables were moderately correlated with each other (r’s = ∣.15∣ to ∣.57∣). Parental conflict demonstrated a significant bivariate relationship with youth onset of depression over the 3 years of the study follow-up (r = .14); bivariate relationships between youth onset of depression and all other parenting variables were not significantly different from zero.

Table 2.

Bivariate Correlations between Primary Variables of Interest

	2	3	4	5	6	7
1. Parental support	.30	−.39	−.42	−.01	.02	−.03
2. Parental responsiveness	---	−.15	−.25	.03	−.07	−.07
3. Parental criticism		---	.57	.04	.03	.06
4. Parental conflict			---	.05	.14	.15
5. Past depression				---	.24	.23
6. Onset of depression during study					---	.24
7. Grade						---

Note. All r values greater than or equal to ∣.14∣ are significantly different from 0 at p < .05. Correlation coefficients for binary variables reflect point-biserial correlations. Correlation coefficients for continuous variables reflect Pearson product-moment correlations.

Prospective Relationships between Parenting and Adolescent Depression

Results of logistic regression analysis predicting youth onset of depression from observed parenting behaviors are reported in Table 3. Controlling for adolescent history of depression, parental responsiveness (b = −.23, Wald = 4.00, p = .045, OR = .79) and parental conflict prospectively predicted adolescent depression (b = .30, Wald = 10.24, p = .001, OR = 1.35). Lower observed parental responsiveness at baseline predicted increased odds of youth experiencing an onset of depression across the 36-month follow up period. Similarly, greater observed parental conflict at baseline predicted increased odds of youth experiencing a depressive onset across the follow-up period, even after controlling for youths’ past depression history before study entry. Parental support and parental criticism did not predict youth odds of experiencing an onset of depression.

Table 3.

Results of Logistic Regression Analyses Predicting Adolescent Depressive Onset During Study

Predictor	b	SE	Wald	p	OR	95% CI
Model 1: Parental Support
Constant	−1.47	.30	24.48	<.001	.230
Past depression at baseline	1.23	.22	31.60	<.001	3.43	[2.23, 5.28]
Parental support	.05	.10	.24	.622	1.05	[.87, 1.26]
Model 2: Parental Responsiveness
Constant	−.40	.48	.70	.403	.67
Past depression at baseline	1.25	.22	32.25	<.001	3.50	[2.27, 5.40]
Parental responsiveness	−.23	.12	4.00	.045	.79	[.63, 1.00]
Model 3: Parental Criticism
Constant	−1.42	.23	37.86	<.001	.24
Past depression at baseline	1.23	.22	31.32	<.001	3.41	[2.22, 5.25]
Parental criticism	.04	.09	.18	.668	1.04	[.87, 1.25]
Model 4: Parental Conflict
Constant	−1.89	.22	77.34	<.001	.15
Past depression at baseline	1.22	.22	30.06	<.001	3.37	[2.18, 5.20]
Parental conflict	.30	.09	10.24	.001	1.35	[1.12, 1.61]

Note. OR = odds ratio; 95% CI = 95% confidence interval for the odds ratio

Exploratory Analyses: Moderation by Grade Cohort and Gender

Grade Cohort.

Results of moderation by grade cohort analyses indicated that grade cohort did not moderate prospective associations between parental responsiveness (b = .04, Wald = .51, p = .470, OR = 1.04) or parental conflict (b = .61, Wald = 4.03, p = .045, OR = 1.84) and youth onset of depressive disorder over the three year follow up period.

Gender.

Similarly, results of moderation by gender analyses indicated no significant interaction effects between gender and parental responsiveness (b = −.31, Wald = 1.61, p = .204, OR =.74) or parental conflict (b = .21, Wald = 1.19, p = .275, OR = 1.24) on youth odds of depressive onset, indicating that results were not moderated by gender.

Sensitivity Analyses

Given significant covariation between parental conflict and parental responsiveness (r = −.25), these dimensions of parenting were entered simultaneously into a logistic regression model to assess if observed associations reflected unique relations between these discrete aspects of parenting and youth odds of experiencing a depressive episode, or if these associations were instead driven by shared variance between parental conflict and responsiveness. Results are reported in Table 4. Findings indicated that, when controlling for parental responsiveness, parental conflict remained a significant predictor of youth onset of depression (b = .27, Wald = 7.77, p = .005, OR = 1.31); when controlling for parental conflict, relations between parental responsiveness and youth onset of depression were not significant (b = −.15, Wald = 1.59, p = .207, OR = .86).

Table 4.

Results of Sensitivity Analyses: Controlling for Overlapping Parenting Predictors

Predictor	b	SE	Wald	p	OR	95% CI
Model 1: Parental responsiveness and Parental conflict
Constant	−1.22	.57	4.60	.032	.30
Past depression	1.23	.22	30.65	<.001	3.43	[2.22, 5.30]
Parental responsiveness	−.15	.12	1.60	.207	.86	[.68, 1.09]
Parental conflict	.27	.10	7.77	.005	1.31	[1.08, 1.58]
Model 2: Parental conflict and Caregiver history of depression
Constant	−2.13	.25	75.99	<.001	.12
Past depression	1.36	,24	31.53	<.001	3.89	[2.42, 6.26]
Caregiver depression	.75	.21	12.46	<.001	2.12	[1.40, 3.23]
Parental conflict	.29	.10	8.65	.003	1.33	[1.10, 1.61]
Model 3: Parental conflict and youth receipt of mental health treatment
Constant	−2.40	.25	90.64	<.001	.09
Past depression	.59	.26	5.02	.025	1.81	[1.08, 3.04]
Mental health treatment	2.54	.26	97.47	<.001	12.63	[7.64, 20.90]
Parental conflict	.30	.10	8.21	.004	1.35	[1.10, 1.66]

Note. OR = odds ratio; 95% CI = 95% confidence interval for the odds ratio

As caregiver history of depression is a potent predictor of both parenting behavior and youth onset of depressive disorder (Goodman, 2007; Goodman & Gotlib, 1999; Goodman et al., 2011; Lovejoy et al., 2000), additional sensitivity analyses were conducted to examine if parental conflict predicted youth onset of depression above and beyond the effects of caregiver depression. Results indicated that controlling for caregiver history of depression, parental conflict remained a significant predictor of youth onset of depression (b = .29, Wald = 8.65, p = .003, OR = 1.33; see Table 4).

To further evaluate robustness of results, additional sensitivity analyses were conducted to evaluate if youth receipt of mental health treatment influenced results. Controlling for receipt of mental health treatment, parental conflict remained a significant predictor of youth odds of experiencing an onset of depression (b = .30, Wald = 8.21, p = .004, OR = 1.35). Further, youth receipt of mental health treatment did not interact with parental conflict to predict depressive outcomes (b = .31, Wald = .25, p = .616, OR = 1.14), indicating the receipt of mental health treatment did not moderate results.

Finally, analyses were conducted to evaluate whether patterns of associations between observed parenting behaviors and youth onset of depression would replicate when examining youth onset of MDD, specifically. Of the total sample of youth who experience an onset of depression over the 36-month follow up period, 30.9% (n = 47) experienced an episode of MDD. Results of logistic regression analyses predicting youth onset of MDD are reported in Table 5. Controlling for youth history of MDD at baseline, parental conflict remained a significant predictor of youth odds of experiencing an onset of MDD (b = .34, Wald = 6.28, p = .012, OR = 1.40). Parental support, responsiveness, and criticism did not significantly predict youth odds of MDD (all p’s > .05).

Table 5.

Results of Sensitivity Analyses: Predicting Adolescent Onset of MDD

Predictor	b	SE	Wald	p	OR	95% CI
Model 1: Parental Support
Constant	−2.37	.45	27.96	<.001	.09
Past MDD	1.01	.45	5.09	.024	2.75	[1.14, 6.63]
Parental support	−.06	.15	.16	.689	.942	[.70, 1.26]
Model 2: Parental Responsiveness
Constant	−1.22	.71	2.91	.088	.30
Past MDD	1.00	.45	4.92	.027	2.72	[1.12, 6.57]
Parental responsiveness	−.33	.18	3.44	.064	.72	[.51, 1.02]
Model 3: Parental Criticism
Constant	−2.53	.35	51.04	<.001	.08
Past MDD	1.02	.45	5.06	.025	2.77	[1.14, 6.75]
Parental criticism	−.01	.15	.001	.974	1.00	[.75, 1.33]
Model 4: Parental Conflict
Constant	−3.19	.33	95.44	<.001	.04
Past MDD	.85	.46	3.44	.064	2.34	[.95, 5.76]
Parental conflict	.34	.13	6.28	.012	1.40	[1.07, 1.82]

Note. MDD = major depressive disorder; OR = odds ratio; 95% CI = 95% confidence interval for the odds ratio

Discussion

Concurrent associations between aspects of parenting behavior and youth depression are well-established in the extant literature (McLeod et al., 2007), and a small body of longitudinal work supports prospective associations between parenting and change in youth depressive symptoms over time (Yap et al., 2014). Of note, however, much of this limited longitudinal work relied on self- or parent-report questionnaire measures of both parenting and depressive symptoms and is thus affected by biases associated with retrospective recall and common method variance (McLeod et al., 2007; Podsakoff et al., 2003). Moreover, few studies have examined parenting factors contributing to youth onset of depressive disorder during childhood and adolescence (see Schwartz et al., 2014 for exception). The present study used rigorous, objective behavioral observation measures of parenting and clinical interview measures of youth depressive episodes in a three-year repeated-measures design to examine longitudinal associations between parenting and youth risk for depressive onset. Present results indicate that parental conflict during childhood and adolescence predicts youth likelihood of experiencing a depressive episode over the course of three years, controlling for both caregiver and youth history of depression. No prospective relationships between parental criticism and parental support and youth onset of depression were observed.

The present results demonstrating significant associations between observed parental conflict and youth depressive onset are generally consistent with results from Schwartz et al. (2014), who found that rates of observed parental aggressive behavior predicted youth first onset of depression across a 6-year follow-up period. Findings are also consistent with a large body of literature indicating that hostile, rejecting, and aversive parenting are particularly influential in youth risk for depression (see McLeod et al., 2007; Yap et al., 2014 for reviews). Interestingly, however, the present study did not indicate a significant relationship between parental criticism, another facet of aggressive parenting, and youth depression. While related, parental conflict and parental criticism capture distinct subdimensions of negative parenting behavior, and the differential relationships with youth depression detected in the current study are consistent with work indicating that subdimensions of rejecting and controlling parenting (e.g., parental withdrawal, hostility) demonstrate unique relationships with youth depression (McLeod et al., 2007). Notably, parental criticism captures parental disapproval and blaming behaviors, whereas parental conflict uniquely includes parenting characterized by high levels of hostility and negative affectivity. Thus, it may be that the affective tone, rather than the specific content of parenting behaviors, contributes to youth risk for depressive episode onset.

Similarly, whereas Schwartz et al. (2014) found that rates of observed positive parenting predicted youth onset of depression, the present study indicated no relationship between observed parental support and youth depressive onset across the 36-month follow-up period. Observed associations between parental responsiveness and youth depression became nonsignificant when controlling for parental conflict, suggesting that this initial finding was likely driven by shared variance between parental responsiveness and parental conflict. It is possible that these unexpected findings reflect the complexity of etiological pathways linking positive dimensions of parenting to youth depressive outcomes. It is plausible, for example, that such factors as family functioning or availability of other supportive caregivers in the home may moderate effects. Indeed, recent research using behavioral observation measures emphasizes the need to consider the family system more holistically (e.g., at the triadic level) in order to optimally characterize youths’ lived experience of parenting (Bodner et al., 2018; Hollenstein, Allen, & Sheeber, 2016). Future behavioral observation work should aim to include multiple caregivers to better clarify prospective relationships between parenting processes and youth risk for depression. Additionally, it is possible that youth are more resilient to deficiencies in positive parenting relative to exposure to hostile, conflictual parenting; future work should evaluate the ways in which youth resilience may moderate effects.

Differences in how parenting was conceptualized (aggressive versus critical, conflictual) and assigned codes (rates of behavior versus global codes) may also account for difference in findings. Indeed, as indicated by McLeod and colleagues (2007), research is needed to assess associations between parenting and youth depressive outcomes with greater precision, and it is likely that heterogeneity in the ways in which parenting is measured and conceptualized contributes to inconsistent findings across the literature. Future work is needed to probe prospective relationships between facets of observed aggressive and positive parenting and youth onset of clinically-significant depression across development.

The prospective association between parental conflict and youth depressive onset observed in the present study was not moderated by grade cohort or gender, indicating that prospective relations between observed parenting behavior and youth odds of experiencing a depressive episode across a 3-year period do not significantly differ between boys or girls or across stages of child and adolescent development. Gender results are consistent with previous work indicating no gender differences in the relationship between observed parenting and youth risk for experiencing an onset of MDD (Schwartz et al., 2014). Together with these previous findings, present results suggest that boys and girls are not differentially sensitive to the influence of parenting behavior on subsequent risk for depressive pathology. Also, the relationship between parenting behavior and youth risk for depression did not vary linearly across development, consistent with research indicating that parenting remains influential for child outcomes through late adolescence (Hair, Moore, Garrett, Ling, & Cleveland, 2008).

The present study demonstrated a number of strengths that represent notable contributions to the extant literature examining relationships between parenting behavior and youth depression. Notably, the present study used behavioral observation measures of parenting and repeated semi-structured clinical interview assessments of youth depression to demonstrate prospective relationships between parenting and youth onset of depression in a moderately large community sample in a 36-month longitudinal design. Unlike cross-sectional studies, which are limited in their ability to address directional relationships between parenting and youth depressive outcomes, and monomethod questionnaire studies, which are susceptible to biases associated with retrospective recall and common method variance (McLeod et al., 2007; Podsakoff et al., 2003), the present study represents a rigorous test of the influence of parental support, responsiveness, criticism, and conflict on youth risk for experiencing a clinically-significant depressive episode. Further, the present study is one of few investigations of parenting factors contributing to youth depressive disorder rather than youth depressive symptoms across late childhood and adolescence. Given both concurrent and potentially life-course persistent correlates of onsets of clinical depression in youth (e.g., Avenevoli et al., 2015; Fergusson & Woodward, 2002; Keenan-Miller et al., 2007; Johnson et al., 2018), the present findings provide important insight into parenting behaviors predictive of particularly high levels of youth distress and impairment. Importantly, the present analyses controlled for youth history of depression prior to enrollment in the study, as well as their caregiver’s history of depression, lending further confidence that observed relationships reflect contributions of parenting on youth odds of experiencing a depressive onset across 3 years.

Despite these strengths, the present results should be interpreted in light of several limitations that represent important areas for future work. The behavioral observation measures included in the present study represent a methodological strength; however, the period of observation included in the present study was relatively brief (i.e., 5 minutes) and limited to a single parenting context (i.e., a conflict discussion). Studies incorporating behavioral observation measures of parenting in multiple, natural settings are needed to better understand how youths’ lived experience of parenting across a variety of contexts is related to their risk for depressive onset. Observation of parenting in natural, unstructured settings rather than in the context of a structured laboratory-based task paradigm may yield insight into parenting behaviors typically encountered by youth during normal daily experiences, such as eating a meal or navigating chores and activities (Gardner, 2000). Moreover, parenting was only assessed at one time point in the present study; however, contemporary models of parenting posit dynamic and transactional relationships between parenting behavior and youth outcomes (Pettit & Arsiwalla, 2008). Further longitudinal research is needed to assess bidirectional relationships between observed parenting and youth onset of depression. Notably, there is no research, to our knowledge, in which observed parental behavior and youths’ clinical depression are repeatedly assessed to enable dynamic longitudinal analysis. Additionally, risk for depressive onset was modeled in the present study by controlling for youth history of depression at baseline. This method of predicting prospective change in depression as an outcome over time by controlling for initial levels at baseline represents a standard practice in the field, yet this approach is not without concerns as it is subject to misspecification (e.g., Lord, 1967). Moreover, as skip out procedures in the semi-structured diagnostic interview were applied in the present study, consistent with the developers’ recommendation (Kaufman et al., 1997) and standard practice, we were unable to evaluate effects of parenting on youth dimensional depression symptom severity. It will be important for future research to evaluate the ways in which parenting behaviors may contribute to dimensional depressive disorder severity. Finally, it is possible that the present findings are influenced by unmeasured genetic factors, including heritability of depressive traits and passive gene-environment correlations contributing to youth vulnerability (e.g., Harold, Rice, Hay, & Bolvin, 2011; Reiss et al., 1995; Silberg, May, & Eaves, 2010). Genetically informed designs are needed to disentangle genetic versus uniquely environmental factors contributing to youth risk for depression.

The present study advances the literature on parenting and youth depression by demonstrating prospective relationships between observed parental conflict and youth onset of depressive episodes, controlling for youth prior history of depression. Findings indicate that for both boys and girls, the experience of parenting marked by high levels of conflict functions as a risk factor for depressive onset across the developmental period spanning late childhood to middle adolescence. In contrast, parental criticism and parental support were not related to youth odds of experiencing a depressive onset, and parental responsiveness was similarly unrelated to youth odds of depression after accounting for parental conflict. Findings suggest that these aspects of parenting are less influential in youths’ risk for depressive disorder. Together, results suggest that parental conflict may be a particularly influential aspect of parenting behavior in understanding youth vulnerability to depression.