Introduction

The impact of the COVID-19 pandemic on child mental health is a significant public health issue. Widely recognized as ‘unprecedented’ in nature, the COVID-19 pandemic has resulted in social disruptions, including community, school and childcare closures that far exceeded anticipated durations (Qualls et al., 2017). Although critical in reducing disease spread, community mitigation strategies also have impacts in the form of job losses, financial insecurity, social isolation and confinement-related stresses, that are expected to impact family and child well-being (Prime, Wade, & Browne, 2020); these ideas stem largely from other studies of acute stress, such as family deterioration during the great recession (e.g. Brooks-Gunn, Schneider, & Waldfogel, 2013). However, it is critical to understand the risk processes driving impacts on families and children during COVID-19; in the absence of such research, practitioners must rely on developmental research conducted under typical circumstances that may not be applicable to the unique context of the COVID-19 pandemic.

Emerging data confirm fears of the negative mental health impact of the pandemic. Early reports suggest increased prevalence of children’s depression, anxiety and behaviour problems during COVID-19 (Liu et al., 2021; Marques de Miranda, da Silva Athanasio, Sena Oliveira, & Simoes-e-Silva, 2020). Findings regarding individuals age 16 and older from the UK Household Longitudinal Study document an increase in mental health problems relative to data collected prior to COVID-19 to April 2020 (Pierce et al., 2020); corresponding results have emerged in cross-national data from 59 countries (Alzueta et al., 2020). Regarding children’s mental health, 14% of parents in a June 2020 US national survey reported that their children’s mental health had worsened during the pandemic (Patrick et al., 2020). Recent longitudinal data indicated large effect sizes regarding increases in children’s externalizing (d = 1.59) and internalizing (d = 1.31) problems (Feinberg et al., 2021). Thus, a critical next step is to identify key pathways of risk that can guide interventionists seeking to minimize the immediate impact and forestall a long-term increase in children’s mental health problems.

In this paper, we draw on ideas that social disruptions incurred during the COVID-19 pandemic and caregiver well-being may lead to disruptions in family-level and parent–child relations, all of which may account for child maladjustment (Prime et al., 2020). We conceptualize family-level functioning (cohesion, conflict and routines) and parenting quality (harshness, laxness and warmth) as proximal risk or protective factors for child mental health, each of which are amenable to existing evidence-based interventions (e.g. Van Ryzin, Kumpfer, Fosco, & Greenberg, 2016). We further expand on these conceptualizations to examine whether pre-existing difficulties or pandemic-related disruptions in family and parenting factors predict declines in child adjustment with the onset of the pandemic.

Family cohesion, conflict and routines all have well-established implications for child adjustment. Family cohesion refers to the quality of emotional bonds among family members (Olson, Waldvogel, & Schlieff, 2019) and is associated with reduced risk for youth externalizing problems (Lucia & Breslau, 2006; McKelvey, Conners-Burrow, Mesman, Pemberton, & Casey, 2015; Richmond & Stocker, 2006) and internalizing problems (Deng et al., 2006; McKeown et al., 1997). Family conflict, including disagreements, anger and hostility among family members, is a robust risk factor for child maladjustment. Namely family conflict is associated with both internalizing and externalizing problems (Benson & Buehler, 2012; Formoso, Gonzales, & a, & Aiken, L. S., 2000; Jaycox & Repetti, 1993). Other work documents family cohesion and conflict as distinct constructs with unique implications for youth well-being (Forgatch & DeGarmo, 1999; Fosco, Caruthers, & Dishion, 2012; Fosco & Lydon-Staley, 2020). Family routines refer to regular practices in family life that are thought to promote predictability and organization (Fiese et al., 2002; Harrist, Henry, Liu, & Sheffield Morris, 2019), and in this study conceptualized as having regular family dinners, organized family activities, and regular wake and bedtimes for children. Developmental evidence from the Add Health study point to family connectedness (inclusive of the participation in daily routines around waking, regular meals together, and regular bedtimes) as highly protective against a number of long-term risks in adolescence (Resnick et al., 1997). Maintaining family routines is protective for child developmental outcomes, even in contexts with elevated risk (Fiese et al., 2002; Kiser, Bennett, Heston, & Paavola, 2005). We propose that the degree to which families maintain routines during the COVID-19 pandemic may thus may be a salient predictor of child well-being (Harrist et al., 2019; Masten & Motti-Stefanidi, 2020; Prime et al., 2020).

Parenting quality also impacts child adjustment; harsh and lax discipline conferring risk and parental warmth operating as a protective factor for children’s maladjustment. Harsh discipline refers to angry, coercive, over-reactive responses to children’s misbehaviour (Arnold, O’Leary, Wolff, & Acker, 1993; Dishion & Snyder, 2015); whereas lax discipline refers to parenting that is overly permissive, failing to apply corrective feedback or consequences to misbehaviour (Arnold et al., 1993; Maccoby & Martin, 1983). Finally, parental warmth refers to supportive, responsive and affectionate parenting practices (Easterbrooks & Goldberg, 1984; Maccoby & Martin, 1983). Decades of empirical evidence, across well over 1000 published studies document the robust implications of each of these parenting dimensions for children’s adjustment, in both cross-sectional and longitudinal studies (Pinquart, 2017a, 2017b).

The current study: Process models of risk for child maladjustment during COVID-19

This study evaluated whether the above domains of family functioning reflect processes by which risk is conferred to children’s internalizing and externalizing problems in the unique context of the COVID-19 pandemic. Through the study of process-focused models, it is possible to identify key risk pathways to guide the selection and delivery of evidence-based, family-focused interventions with optimal effect. We evaluated two hypothesized pathways of risk for children’s maladjustment. First, we considered a pre-existing vulnerabilities pathway, in which families already challenged with poor family relationship quality and/or lower-quality parenting prior to COVID-19 would be at elevated risk for child maladjustment during COVID-19. Second, we tested a family disruption pathway, which posits that the degree to which family functioning and parenting quality deteriorate (e.g. decreases in cohesion, increases in harsh/lax parenting) from pre- to post-COVID-19 may predict subsequent child maladjustment during the COVID-19 pandemic. Our analyses accounted for parents’ emotional distress and family financial strain prior to COVID-19 as potentially important factors in children’s adjustment during the pandemic (Prime et al., 2020). Finally, we controlled for pre-pandemic levels of children’s internalizing and externalizing problems to assess the role of family risk pathways in impacting children’s mental and behavioural health.

Method Participants

We analysed a sample of families from a larger, intergenerational study already in progress prior to COVID-19 onset in the United States. Of the 244 families who had participated in the larger study, 204 agreed to complete surveys during COVID-19. No differences were found between these two samples on any demographic or study variables. Children (45.1% girls) in this sample were an average of 4.17 years old (SDAge = 2.17) at the time of the original study. Participating caregivers (MAge = 27.43, SDAge = 1.67) identified as the child’s mother (70.6%), father (22.5%), stepmother (1.5%), stepfather (2.5%) or other caregiver (1.0%); their racial background was White/Caucasian (90.7%), Black/African American (4.4%), American Indian, Eskimo, or Aleut (0.5%), or Other (4.4%); 10.8% reported that they were of Hispanic origin. Caregivers reported their child’s race as: White/Caucasian (91.7%), Black/African American (7.4%), American Indian, Eskimo, or Aleut (1.0%), Asian or Pacific Islander (0.5%), or Other (4.4%); 12.3% of children were of Hispanic origin. Most caregivers (79.4%; n = 162) reported that they lived with a second caregiving adult for the child (of these, 76.5% were the child’s other biological parent). Of those living with another caregiver, 68.5% were married, 24.1% were in a romantic relationship, 14.9% were cohabiting, and 6.2% were other family members. Of those not living with another caregiver, 45.2% were single, 16.7% were divorced or separated, 14.3% were in a romantic relationship but not living together, and 16.7% reported having another arrangement. Annual family income ranged from ‘0–$9,999’ to ‘$100,000 or more’ (Median: $50,000–59,999).

Procedure

The current sample was recruited from a larger project (Pathways to Health [P2H]; HD092439), an ongoing study evaluating the intergenerational transmission of parenting and family relationships. P2H was an extension of a community-randomized trial of the PROmoting School-community-university Partnerships to Enhance Resilience intervention delivery system (PROSPER; Spoth et al., 2007), which recruited students (N = 10,845) in 28 rural and semi-rural communities in 6th grade. PROSPER followed students through high school and then continued to follow a subset (N = 1,984) of the original sample in young adulthood. Young adults who were parents of children between the ages of 1.5–10 years old were invited to participate in home-based data collection. In March of 2020, recruitment and data collection for P2H were paused due to the COVID-19 pandemic.

Following COVID-19 onset, all families already enrolled in P2H (N = 244) were invited to participate in biweekly surveys beginning 8 May 2020 to assess coping during the pandemic. Participating parents completed web-based surveys. The first surveys were deployed during national stay-at-home orders. A second survey was sent out two weeks later, resulting in one pre-pandemic assessment (T1) and two assessments conducted during the pandemic (T2, T3). To reduce participant burden, scales were abbreviated at T2 and T3. For repeated measures in the current study, only items from the shortened versions of each scale were used. Items for measures can be found in the Appendix S1.

Measures

All variables were scored so that higher values reflected higher levels of each construct.

Family-level functioning (T1, T2)

Family cohesion (αT1 = .80, αT2 = .83) and family conflict (αT1 = .84, αT2 = .84) were each measured using 3 items from the shortened Family Environment Scale (Bloom, 1985). Family routines were measured using 4 items (αT1 = .63, αT2 = .71) from the Child Routines Inventory (Sytsma, Kelley, & Wymer, 2001). At T1, shortened scales were highly correlated with original scales (r’s = .84–.96), suggesting good validity.

Parenting quality (T1, T2)

Harsh discipline (αT1 = .67, αT2 = .71) and lax discipline (αT1 = .60, αT2 = .78) were each measured using 3 from the Parenting Scale (Arnold et al., 1993). Parental warmth was measured with 5 items (αT1 = .59, αT2 = .79) from the Parental Attitudes towards Child Rearing (PACR) warmth scale (Goldberg & Easterbrooks, 1984). At T1, shortened scales were highly correlated with original scales (r’s = .82–.88), suggesting good validity.

Child adjustment

At T1, children’s internalizing and externalizing problems were measured using the Child Behavior Checklist (Achenbach & Rescorla, 2000, 2001). Given the range in age, different versions were used (1½–5 and 6–18). These internalizing (1½–5 α = .87, 6–18 α = .76) and externalizing (1½–5 α = .90, 6–18 α = .86) scales demonstrated good reliability. Scores were converted into internalizing and externalizing T-scores to allow for comparable scores across all children for analysis. To minimize participant burden brief, 5-item assessments of child adjustment were developed for the T3 COVID-19 survey using items adapted from the Strengths and Difficulties Questionnaire (SDQ; Goodman, 1997) and the Child Adjustment and Parent Efficacy Scale (CAPES; Morawska, Sanders, Haslam, Filus, & Fletcher, 2014), which yielded reliable, measures of internalizing (α = .82) and externalizing problems (α = .72).

Covariates

Parent emotional distress was measured at T1 using the 20-item Center for Epidemiological Studies-Depression (Radloff, 1977; αT1 = .91) and the 7-item Penn State Worry Questionnaire (Meyer, Miller, Metzger, & Borkovec, 1990; αT1 = .96). These were highly correlated (r = .62, p < .01), and thus were standardized and averaged to represent parent emotional distress. Financial strain was measured at T1 using four items (αT1 = .78) from the Financial Strain Index (Vinokur, Price, & Caplan, 1996). Time lapse was a measure of the number of weeks between T1 and T2 to capture spacing of measurements assessed in latent change scores (M = 37.9, Range = 8.3–70.7).

Results

Descriptive statistics and bivariate correlations among study variables are presented in Table 1. Variables exhibited acceptable levels of skew (<2.2). Correlations among variables were in the expected directions, with associations of small to moderate magnitude among family-level and parenting quality variables, suggesting that these measures captured distinct facets of family life. We examined whether intervention condition (in P2H) was associated with any study variables; no significant intervention-control differences were found. Thus, structural models were computed as planned and condition was omitted from analyses.

Table 1. Correlations, means, and standard deviations 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 1 Fin T1 2 P Distress T1 .34 3 Time Lapse −.07 .09 4 COH T1 −.27 −.22 .08 5 CON T1 .27 .38 .03 −.35 6 ROUT T1 −.15 −.06 .08 .23 −.21 7 HARSH T1 .13 .19 .03 −.18 .24 −.05 8 LAX T1 .15 .19 −.19 −.18 .18 −.16 .26 9 WARM T1 −.15 −.03 .09 .18 −.01 .16 −.20 −.20 10 INT T1 .20 .39 −.03 −.15 .21 −.13 .17 .17 −.18 11 EXT T1 .17 .32 .06 −.19 .29 −.14 .26 .22 −.02 .61 12 COH T2 −.23 −.30 .01 .32 −.30 .13 −.21 −.24 .10 −.29 −.25 13 CON T2 .22 .23 .15 −.26 .32 −.14 .20 .12 −.06 .30 .30 −.22 14 ROUT T2 −.21 −.21 .02 .36 −.27 .57 −.14 −.27 .14 −.14 −.18 .35 −.22 15 HARSH T2 .20 .19 .17 −.08 .21 −.05 .60 .12 −.01 .21 .28 −.29 .38 −.13 16 LAX T2 .20 .05 −.21 −.18 .12 −.18 .19 .60 −.19 .08 .13 −.17 .08 −.28 .12 17 WARM T2 −.15 −.05 .07 .21 −.25 .12 −.13 −.11 .27 −.11 −.13 .29 −.05 .24 −.07 −.13 18 INT T3 .24 .38 .04 −.14 .17 −.07 .14 −.06 −.05 .43 .30 −.33 .25 −.08 .31 .11 −.06 19 EXT T3 .23 .28 .01 −.11 .13 −.13 .13 .15 −.03 .31 .40 −.23 .29 −.12 .28 .28 .00 .60 N 200 199 200 199 198 199 197 197 199 185 187 201 199 200 201 201 202 181 179 M 1.40 .00 37.90 7.96 2.72 3.44