Parent and child adjustment dual trajectories at the beginning of the COVID-19 syndemic.

Children and their families have been significantly impacted by the unfolding of the COVID-19 syndemic. We sought to identify (1) groups of families with distinct profiles of joint trajectories of parental anxiety and child emotional distress and (2) protective and risk factors associated with these dual-trajectory profiles. A sample of 488 parents (65% White; 77% mothers) with 3- to 8-year-old children (MAge  = 5.04, SDAge  = 1.59) was followed from late March to early July in 2020. Survey data on parent (i.e., anxiety symptoms) and child (i.e., emotional distress) adjustment were collected at three time points. Using multivariate growth mixture modeling, we identified one group with low parental anxiety and child emotional distress (42.7%) and three other distinct groups with varying risk levels among parents and/or children. We also identified protective (e.g., positive parenting) and risk (e.g., child negative affect, negative parenting, perceived stress with racism) factors in predicting parent and child adjustment. It can be concluded that, overall, our sample (mostly middle- and high-socioeconomic status families) demonstrated family resilience amid COVID-19, consistent with prior disaster coping literature. At the same time, our findings also indicated the need to identify at-risk families and modifiable factors for post-disaster public health interventions.

INTRODUCTION

The novel coronavirus disease 2019 (COVID‐19) outbreak and the social distancing policies adopted by many countries have significantly altered the daily lives of children and their families (Fong & Iarocci, 2020). By March 25th, 2020, mandated or recommended school closures had been issued in almost all U.S. states (Education Week, 2020), bringing forth unique challenges to families with young and school‐age children. In addition, the long‐standing institutional racism in the U.S. has exacerbated the health disparities following COVID‐19, collectively giving rise to what has been described as a syndemic (Mendenhall, 2020). Using a short‐term longitudinal design, we investigated important protective and risk factors, including perceived stress with racism, associated with parents’ and children's adjustment for 3 months from late March to early July 2020.

Children's adjustment during COVID‐19 syndemic

Children's adjustment post‐disaster (i.e., natural or human‐made disasters) has been historically understudied in the literature, and longitudinal evidence is especially scant, even though children are especially vulnerable (Lee, 2020; Wang et al., 2013). Relative to older children, younger children may have more difficulty encoding or recalling important aspects of the events due to their early stage of cognitive development (Bahrick et al., 1998; Grasso et al., 2016). The psychological and emotional needs of children may be further neglected following disasters when parents or other significant adults in children's lives are affected by the disaster and have problems coping with the event (Cobham et al., 2016). Nonetheless, children may respond to disaster‐related distress in ways that are unique to their developmental stage (Grasso et al., 2016; Saylor et al., 1999). Young children have been observed to respond to disasters with behaviors, such as temper tantrums, refusing to sleep alone, and incontinence (Grasso et al., 2016; Saylor et al., 1999).

Children and parents’ adjustment needs to be understood from a family systems perspective that emphasizes the interconnections between child adjustment and caregiver well‐being and the assumption that risk factors impeding parents’ function can lead to changes in the function of children, and vice versa (Carr, 2015; Fong & Iarocci, 2020; Prime et al., 2020). Accumulating research has documented the impact of COVID‐19 on children and their caregivers (e.g., Feinberg et al., 2021; Fong & Iarocci, 2020; Gassman‐Pines et al., 2020; Marchetti et al., 2020; Wamser‐Nanney et al., 2021). Many caregivers have experienced crisis‐induced hardships not limited to illness, financial insecurity, confinement, and increased caregiving burdens (Fong & Iarocci, 2020; Gassman‐Pines et al., 2020). When affected parents encounter difficulties coping with disaster, the psychological and emotional needs of children may be further neglected (Cobham et al., 2016). Given the uncertainty surrounding the duration and impact of COVID‐19, the increase in anxiety symptoms (more so than depression) among U.S. adults was perhaps unsurprising (Fitzpatrick et al., 2020; Marchetti et al., 2020). Typically, parents serve as primary sources of support for young children and initiate help when children are having problems. However, the impact of disasters may threaten parents’ caregiving capacity, rendering them unavailable to help their children to cope (Feinberg et al., 2021; Hafstad et al., 2012; Prime et al., 2020; Russell et al., 2021). Compared to some prior disastrous events (e.g., September 11, Hurricane Katrina), COVID‐19 and its ensuing social distancing policies and school closures presented unique challenges for parents to provide consistent childcare and to coordinate distance learning at home. With many children quarantined with parents, it is especially important to investigate their adjustment from a family systems perspective using longitudinal methods (Bonanno et al., 2011; Wang et al., 2013).

Need for longitudinal research

A growing body of literature has informed the unfolding process of adults’ psychological adjustment following disastrous events. Longitudinal studies using growth mixture modeling (GMM) have been especially critical in identifying the inter‐individual differences in post‐disaster adjustment trajectories (for a review, see Bonanno et al., 2011). For example, after the outbreak of a highly contagious severe acute respiratory syndrome, Bonanno et al. (2008) followed survivors over 2 years and identified four distinct adjustment trajectories––resilience (35%), chronic dysfunction (42%), psychological recovery (10%), and delayed psychological dysfunction (13%). Trajectories studies such as this one have been typically followed by further exploration of risk factors to identify questions of who is at higher risk, when, and what the precipitating risk factors are that will critically inform targeted prevention and intervention (Bonanno et al., 2011).

Research applying GMM to children and adolescents’ post‐disaster adjustment remains scant (Wang et al., 2013). In the few known studies that have been published, the trajectory profiles that emerged for most children and adolescents was that of adjusting well over time, which similarly mirrored findings for adults (e.g., Kronenberg et al., 2010; Osofsky et al., 2015; Weems & Graham, 2013). The posttraumatic stress and depressive symptoms in one sample of children between 9 and 18 years old, for example, decreased in the 3‐year period after Hurricane Katrina (Kronenberg et al., 2010). However, much evidence is needed to replicate these findings, especially among younger age children.

Although existing literature has mostly examined post‐disaster adjustment trajectories spanning 1–2 years, a drawback of these studies is that they may overlook more nuanced changes happening within a shorter time frame (Silverman & La Greca, 2002). For example, increasing followed by decreasing anxiety symptoms within 3 months post‐disaster may appear to be continuously decreasing if the longitudinal interval was to be 3 months or longer (Silverman & La Greca, 2002). Taken together, past studies have provided little information about the ripple effects of disaster on the family units (Lengua et al., 2005; Prime et al., 2020). The first objective of this current study is to understand the immediate psychological sequelae (i.e., parental anxiety and child emotional distress) following the COVID‐19 outbreak. We hypothesized three to six profiles would emerge based upon past trajectory studies (Lai et al., 2015; Weems & Graham, 2013).

Protective and risk factors

In line with past trajectory studies (Bonanno et al., 2011; Kronenberg et al., 2010; Osofsky et al., 2015) and the systems model and family functioning (Prime et al., 2020), our secondary objective was to identify protective and risk factors that may distinguish trajectory profiles. Emerging data indicated that the mental health of women, ethnic–racial minorities, individuals with children and mothers in particular, and low‐income individuals were disproportionately impacted by COVID‐19 (Feinberg et al., 2021; Fitzpatrick et al., 2020; Marchetti et al., 2020; Russell et al., 2021; Wamser‐Nanney et al., 2021). Post‐disaster coping literature has extensively discussed the role of parenting and child temperament on child adjustment and family functioning (e.g., Cobham et al., 2016; Wilson et al., 2010). Whereas healthy/positive parenting practices and parental support may buffer the impact of external stressors, maladaptive/negative parenting practices can adversely affect child well‐being (Cobham et al., 2016; Prime et al., 2020). In addition, effortful control, a dimension of temperament that concerns children's attention regulation and inhibitory control skills, has been found to be a protective factor from traumatic events (Lengua et al., 2005; Wilson et al., 2010). Children with high effortful control may be better able to regulate their emotions and behaviors with abrupt changes in routines following COVID‐19. Negative emotionality, a child's tendency to experience negative affect such as fear and irritability, has also been shown to increase children's risk by promoting higher levels of negative arousal in response to external environmental stressors (Lengua et al., 2005; Wilson et al., 2010). Surgency, another important dimension of child temperament that is characterized by high activity level and high‐intensity pleasure seeking, has not been studied and identified as an important prospective pre‐trauma predictor (Lengua et al., 2005; Wilson et al., 2010). As some emergent data suggested that adults with higher scores on extraversion trait (akin to child surgency temperament) may be more impacted by quarantine restrictions during COVID‐19 (Wijngaards et al., 2020), we plan to conduct exploratory analysis with surgency as a predictor of trajectory membership as well.

We included an additional factor—parental perceived stress to racism—that was not included in past disaster coping literature in order to understand the ongoing syndemic, in which the impact of COVID‐19 was amplified by biological and social factors rooted in systemic racism (Mendenhall, 2020). This was particularly evident for communities of color, who continue to bear a substantially higher burden of COVID‐19 than Whites (Centers for Disease Control & Prevention, 2021). For example, the rise of COVID‐19 brought a resurgence of hate crimes directed toward Asian Americans (Gover et al., 2020). In the unfolding process of COVID‐19, the murder of George Floyd on May 25th, 2020, along with other high‐profile cases covered by media during quarantine, further exacerbated collective wounds in American society, especially among Black communities (Silver et al., 2020). Emerging research has linked the deleterious effect of exposure to racism events occurring prior to and during COVID‐19 on the physical and psychosocial health of ethnic–racial minorities (Cheah et al., 2020; Heard‐Garris et al., 2018; Martz et al., 2019; Zhou et al., in press). Considering indirect exposure to racism events is different from directly experiencing racism events and that White individuals may experience this as a secondary trauma (Menakem, 2017), some scholars have described the combination of the COVID‐19 syndemic and vicarious exposure to anti‐Blackness racism as a collective trauma (Silver et al., 2020). To explore this phenomenon, we assessed the stress experienced by parents from secondary exposure to racial discrimination experienced by others. As vicarious racism has been consistently linked to increased internalizing symptoms (Heard‐Garris et al., 2018), we hypothesized that perceived stress of racism as a risk factor would be associated with maladjusted trajectories.

METHOD

Participants and procedure

Participants were drawn from a prospective pre‐registered longitudinal study (Purdue IRB 2020‐476). Three waves of online survey data were collected, with about 6 weeks between each survey, from March 30 to July 14, 2020. Each wave was collected in an approximately 2‐week window (i.e., March 30th to April 14th for Wave 1; May 13th to June 1st for Wave 2; and June 30th to July 14th for Wave 3). Participants were recruited from online parenting forums and parenting listservs, as well as through snowball sampling. To qualify for the study, participants must identify as a parent with at least one child between the ages of 3–8 and reside in the U.S. To reduce the burden of study participation during COVID‐19, each survey was designed to be 10–15 min in length. Parents were raffled to receive gift cards of $20, $40, and $40 at Waves 1, 2, and 3, respectively. We excluded parents who reported that either they or their children had contracted COVID‐19 at Wave 3 (n = 7) due to the salience of their trauma exposure. In the two cases where multiple parents within a household responded, one parent was excluded to avoid data clustering. Our final sample included 488 parents at Wave 1, 346 parents at Wave 2, and 332 parents at Wave 3.

Parents were on average 36.73 years (SD = 5.92), with mostly mothers (77%, n = 376). They were from 41 states and Washington D.C., with the most frequently reported states being New York (n = 99), California (n = 50), and Texas (n = 30). The sample included 65% (n = 306) non‐Hispanic White or European Americans; 13% (n = 62) Asian Americans or Pacific Islanders; 8% (n = 35) Black/African Americans; 6% (n = 27) Hispanic/Latinx Americans; 2% (n = 9) Middle Eastern and North Africans; 1% (n = 5) Native Americans; and 5% (n = 24) Multiracial individuals. Most parents were married (80%, n = 374). The majority of parents were working full time or part time before the pandemic (n = 397), with most working from home (n = 282). Participants were biological parents (n = 459, 94%), adoptive parents (n = 19), foster parent (n = 2), stepparent (n = 6), or biological parent's partner (n = 2). Fifty‐two parents reported reduced hours or being furloughed and 126 reported loss in family income as a result of COVID‐19. Most parents received college or above level education (78%, n = 362). Annual family income was measured in eight categories, with 52% (n = 232) reporting lower than $100,000, and 73% (n = 256) reporting eligibility for a stimulus check. Parents reported having an average of 1.96 children (SD = 1.04, range 1–6). Parents with multiple children between the ages of 3–8 (n = 156) were asked to randomly select one as their target child. The mean age of the target child was 5.04 years (SD = 1.59), with 44% (n = 214) girls. Among those who reported their children's schooling prior to COVID‐19 (n = 349), most of the target children were enrolled in daycare/pre‐school (47%; n = 164) or kindergarten/elementary school (43%; n = 151), among which 62% (n = 216) had been provided with online distance learning. The majority of the parents (n = 257) perceived the school closure as “negative,” whereas others reported it as “neutral” (n = 82) or “positive” (n = 61) for their target child.

Measures

Outcome variables

Child emotional distress

Child adjustment was measured by the Pediatric Emotional Distress Scale (PEDS; Saylor et al., 1999), developed to assess 2‐ to 10‐year‐old children's behaviors after experiencing traumatic events. At each wave, parents rated on a 1 (almost never) to 4 (very often) Likert‐type scale, the frequency of their target children's behaviors in the last 30 days. The PEDS consists of four subscales of anxious/withdrawn (6 items, e.g., “cries without good reason”), fearful (5 items, e.g., “refuses to sleep alone”), acting out (6 items, e.g., “gets frustrated too easily”), and trauma‐specific responses (4 items). The trauma‐specific subscale was adapted to coronavirus‐related responses (e.g., “avoids talking about coronavirus even when asked”). The PEDS has demonstrated good internal consistency, interrater reliability, construct validity with child externalizing behaviors, and preliminary evidence for trauma screening (e.g., Bhushan & Sathya Kumar, 2009; Saylor et al., 1999). Mean scores of all 21 items were used in the current analysis, with internal consistency ranging from 0.86 to 90. Prior research has indicated a cutoff of 1.62 across all items for screening trauma exposure (Saylor et al., 1999). Using this cutoff, 52%, 55%, and 53% of children were screened positive at each wave.

Parent anxiety symptoms

Parents’ adjustment was measured by the Generalized Anxiety Disorder 7‐item scale (GAD‐7; Spitzer et al., 2006). The GAD‐7 is a widely used and reliable self‐report instrument. Items were rated on their frequency over the last 2 weeks on a 0 (not at all) to 3 (nearly every day) Likert‐type scale. The internal consistency of GAD‐7 ranged from 0.89 to 0.91 across waves in the current study. A clinical cutoff of 10 or above as a total score (1.43 as mean score) or above has been recommended to indicate a likely case of generalized anxiety disorder (Spitzer et al., 2006). Using this cutoff, 37%, 23%, and 19% of parents were screened positive at each wave. Mean scores were used for better illustration in the multi‐trajectory modeling.

Protective and risk factors

Child temperament

Child temperament was assessed at Wave 2 by the Children's Behavior Questionnaire (CBQ; Putnam & Rothbart, 2006). Measures were staggered at different waves to reduce research burdens on parents. This 36‐item questionnaire is a widely used instrument that captures temperament of Surgency, Effortful Control, and Negative Affect, with strong reliability and validity (Putnam & Rothbart, 2006). Parents were instructed to evaluate descriptions of the target child “during the six months prior to COVID‐19” on a 1 (extremely untrue of your child) to 7 (extremely true of your child) Likert‐type scale. Of the three subscales, surgency (12 items) assesses children's positive emotionality, activity level, impulsivity, and risk‐taking (α = 0.70); effortful control (12 items) measures attention shifting and focusing, perceptual sensitivity, and inhibitory and activational control (α = 0.84); and negative affectivity (12 items) assesses children's fear, anger, sadness, and discomfort (α = 0.82).

Parenting practices

Positive and negative parenting practices were measured at Wave 2 by the Multidimensional Assessment of Parenting Scale (MAPS; Parent & Forehand, 2017). Thirty‐four parenting practices were rated by parents for their frequency over the last 30 days on a 1 (never) to 5 (always) Likert‐type scale. The Broadband Positive Parenting factor of the MAPS includes four narrowband subscales: Proactive Parenting, Positive Reinforcement, Warmth, and Supportiveness. The Broadband Negative Parenting factor includes three narrowband subscales: Hostility, Physical Control, and Lax Control. The MAPS has demonstrated excellent reliability and longitudinal support for the validity of MAPS subscale scores (Parent & Forehand, 2017). Cronbach's alpha in the current sample was 0.86 for the Positive Parenting domain and 0.87 for the Negative Parenting domain.

Perceived stress of racism

To assess parents’ perceived stress in response to the indirect exposure to racism during COVID‐19, a four‐item vicarious discrimination scale (Martz et al., 2019) was administered at Wave 3. These items included “hearing people being the victims of racism in the news,” “hearing about family members or friends who experienced racism,” “seeing other people in public being treated unfairly because of their race,” and “seeing racism depicted in movies or television shows.” Participants were asked to rate their distress levels in the last 3 months on a 4‐point scale from 0 (not at all) to 3 (very much) in response to the four situations above. These items have demonstrated good reliability and incremental validity to personal experiences of discrimination (Martz et al., 2019). Cronbach's alpha was 0.87 in the current study.

Longitudinal measurement invariance

Multigroup confirmatory factor analysis was conducted to examine the longitudinal measurement invariance of the PEDS and GAD‐7, which is an important requirement for longitudinal modeling (Widaman et al., 2010). Following the protocol of Widaman et al. (2010), a series of models with increasing invariance were fitted, and configural invariance (i.e., the same pattern of fixed and free factor loadings across time), weak factorial invariance (i.e., invariant factor loadings across time), and strong factorial invariance (i.e., invariant factor loadings and intercepts across time) were tested for each subscale and wave. A model is determined to have an acceptable fit if it meets the cutoff scores indicating an acceptable model fit for the descriptive fit indices proposed by Hu and Bentler (1999). As indicated in Table S1, PEDS and GAD‐7 demonstrated configural and weak factorial invariance, respectively, warranting comparisons with composite means across time (Widaman et al., 2010).

Missing data analysis

The GAD‐7 and PEDS data were missing at Wave 2 (29%) and Wave 3 (32%). Approximately 61% participants completed all three waves, 17% completed two waves, and 23% completed only one wave of PEDS and GAD‐7. The Little's MCAR test was conducted with all studied variables, which supported a missing completely at random pattern, χ 2(352) = 358.82, p = 0.45. Thus, maximum likelihood (ML) estimation was employed to retain all participants in trajectory analyses (Graham, 2009). In addition, the missingness for demographic variables was <5% except for annual household income (7%). The missingness for parenting practices, child temperament, and stress with racism was around 30%. Thus, we also performed multiple imputation (MI) to account for missingness to reduce parameter bias due to longitudinal attrition (Asendorpf et al., 2014). Therefore, we conducted the analysis with protective and risk factors using 40 imputed datasets in our three‐step approach.

Analytic plan

Multivariate GMM was conducted with Mplus 8.3 (Muthén & Muthén, 1998–2017) to jointly estimate trajectories of parental anxiety and child emotional distress. As an extension of the univariate GMM, multivariate GMM has been utilized to examine symptom comorbidity within the same individual over time (Nagin et al., 2018) or co‐occurring symptoms within the same family systems over time (Zhou & Sun, 2021). Given that there are three time points in our study, only linear slope factors were included. We estimated models with one to five classes, and the fit of these models was assessed with (Nagin et al., 2018; Nylund et al., 2007): (a) Bayesian information criterion (BIC), (b) sample‐size adjusted BIC (SABIC), and (c) bootstrapped likelihood ratio test (BLRT). Lower values of BIC and SABIC indicate a better fitting model. A significant BLRT χ 2 value (p < 0.05) indicates that the specified model fits the data better than the specified model with one less class (Nylund et al., 2007). Classification quality was further considered using entropy values (range from 0 to 1 with higher values indicating that individuals are better discriminated between classes), class sizes, and interpretability of classes (Nagin et al., 2018; Nylund et al., 2007).

After fitting the multi‐trajectory model, we examined potential protective and risk factors associated with trajectory memberships using the three‐step approach, which takes account of the measurement error in the most likely class in the model estimation (Asparouhov & Muthén, 2014). These factors included parent and child gender (male = 1, female = 0), age, race (minority parents = 1, White parents = 0), parental education (below college = 1, college or above = 0), family income (below $25K = 1, $25K–$50K = 2, $50K–$75K = 3, $75K–$100K = 4, $100K–$125K = 5, $125K–$150K = 6, $150K–$200K = 7, above $200K = 8), number of children, racism stress, positive and negative parenting, and surgency, effortful control, and negative affect.

RESULTS

Overall parental anxiety scores decreased from T1 to T3, whereas children's emotional distress scores were stable, and they were moderately correlated across times (rs ranged from 0.37 to 0.55, ps < 0.01, see Table 1). Children's emotional distress across times was highly correlated (rs ranged from 0.66 to 0.70, ps < 0.01), as were parental anxiety scores (rs ranged from 0.55 to 0.71, ps < 0.01).

TABLE 1

Mean, standard deviations, and bivariate zero‐order correlations

	1	2	3	4	5	6	7	8	9	10	11	12
1. GAD‐7 (T1)	–
2. GAD‐7 (T2)	0.66**	–
3. GAD‐7 (T3)	0.55**	0.71**	–
4. PEDS (T1)	0.45**	0.39**	0.41**	–
5. PEDS (T2)	0.37**	0.54**	0.47**	0.66**	–
6. PEDS (T3)	0.43**	0.49**	0.55**	0.66**	0.70**	–
7. Vicarious racism stress	0.34**	0.35**	0.34**	0.17**	0.29**	0.33**	–
8. Positive parenting	−0.07	−0.13*	−0.14*	−0.15**	−0.10	−0.09	−0.08	–
9. Negative parenting	0.19**	0.27**	0.18**	0.38**	0.36**	0.21**	−0.12*	0.04	–
10 Surgency	0.08	0.04	0.07	0.02	0.03	0.05	0.04	0.07	0.02	–
11. Effortful control	0.04	0.05	0.01	−0.10	0.09	0.07	0.12*	0.21**	−0.13	−0.06	–
12. Negative affect	0.16**	0.29**	0.25**	0.34**	0.44**	0.38**	0.23**	−0.13*	0.36**	−0.08	−0.04	–
Mean	1.20	0.95	0.84	1.73	1.74	1.73	3.04	4.28	2.03	4.47	5.22	3.81
SD	0.77	0.71	0.67	0.45	0.46	0.46	0.91	0.42	0.50	0.98	0.98	1.09
Skewness	0.37	0.62	0.93	0.89	0.73	0.72	−0.85	−0.45	0.55	0.01	−0.70	−0.18
Kurtosis	−0.67	−0.15	0.67	0.82	0.67	0.79	−0.38	0.03	1.23	−0.26	0.80	−0.41

Abbreviations: GAD‐7, Generalized Anxiety Disorder 7‐item; PEDS, Pediatric Emotional Distress Scale.

*p < 0.05; **p < 0.01.

Fit statistics of the five latent profile models can be seen in Table 2. The entropy was highest at the 4‐profile solution. The BIC and SABIC kept decreasing from 1 to 5 profile solutions; however, the 5‐profile solution included a small class size (1%). The BLRT never became non‐significant and, thus, was not informative in choosing a preferred model. Taken together, the 4‐profile solution was selected (Figure 1). The first and the largest profile was estimated to account for 42.7% of the sampled population. In Profile 1, parents’ GAD scores slightly decreased over time, whereas children's PEDS scores remained stable. Using 1.43 and 1.62 as clinical cutoffs for GAD and PEDS scores, respectively (Saylor et al., 1999; Spitzer et al., 2006), neither trajectory was clinically significant. The second and the smallest profile (9.0%) presented high and clinically significant symptoms. Both parents and children started with high, clinically significant scores and remained stable across time. In Profile 3 (13.8%), parents’ GAD scores were clinically significant, which started high and decreased over time. Children's PEDS scores were not clinically significant and remained stable. In Profile 4 (34.5%), parents’ GAD scores were not clinically significant, which slightly decreased over time. Children's PEDS scores remained stable across three timepoints and above the clinical cutoff. Additional model parameter estimates are displayed in Table S2.

TABLE 2

GMM fit indices

K	BIC	SABIC	BLRT	Entropy	Sample size (%) per class based on most likely class membership
1	3659.823	3618.562	–	–	100
2	3372.967	3315.838	p < 0.001	0.724	45/55
3	3243.979	3170.980	p < 0.001	0.750	35/13/51
4	3091.109	3002.240	p < 0.001	0.793	43/9/14/35
5	3053.829	2949.091	p < 0.001	0.778	24/16/45/1/11

Number of random starts increased to 500 with 20 optimizations phases due to convergence issues; variances of the intercepts and slopes fixed to zero.

Abbreviations: BIC, Bayesian information criterion; BLRT, bootstrap likelihood ratio test; SABIC, sample‐size adjusted Bayesian information criterion.

FIGURE 1

Parent and child adjustment dual trajectories

We also examined protective and risk factors among the emergent profiles using the three‐step method (Asparouhov & Muthén, 2014). As indicated in Table 3, using Profile 1 as the reference group, being a father decreased the relative odds ratio of membership in Profile 3; having an older child decreased the relative odds ratio of membership in Profile 4; a higher annual household income decreased the relative odds ratio of membership in Profile 3; high racism stress increased the relative odds ratio of membership in Profiles 2 and 3; higher positive parenting decreased the relative odds ratio of membership in Profile 2, whereas higher negative parenting increased the relative odds ratio of membership in Profiles 3 and 4; and lastly, higher negative affect increased the relative odds ratio of membership in Profiles 2, 3, and 4.

TABLE 3

Protective and risk factors associated with GMM profiles

	Profile 2		Profile 3		Profile 4
	OR	SE	OR	SE	OR	SE
Parent gender	1.79	0.94	0.35**	0.23	0.69	0.35
Child gender	0.75	0.42	1.17	0.45	0.65	0.22
Child age	1.15	0.17	0.99	0.13	0.77*	0.09
Race	0.90	0.47	0.91	0.11	0.85	0.15
Parental education	0.57	0.34	1.60	0.83	0.68	0.35
Annual household income	0.76	0.11	0.77*	0.09	0.97	0.09
Number of children	1.29	0.24	0.91	0.18	0.83	0.16
Vicarious racism stress	1.67*	0.59	1.80*	0.56	1.06	0.25
Positive parenting	0.55*	0.36	0.90	0.48	0.86	0.42
Negative parenting	3.03	1.79	2.22*	0.91	2.19*	0.93
Surgency	1.28	0.40	1.15	0.25	1.15	0.22
Effortful control	1.00	0.30	1.25	0.31	0.96	0.18
Negative affect	2.02*	0.68	1.36*	0.32	1.72*	0.35

Profile 1 as the reference group.

Abbreviation: OR, odds ratios.

*p < 0.05; **p < 0.01.

DISCUSSION

To the best of our knowledge, this is the first study that used a multivariate GMM approach to model both young children and their parents’ adjustment at the beginning of the COVID‐19 syndemic in the U.S. This approach allows us to visually represent distinct groups of families that exhibit unique patterns of both change and continuity. Our overarching objectives were to identify any heterogeneity in the co‐occurrence of adjustment patterns between 3‐ and 8‐year‐old children and their parents using this multivariate GMM approach, and to identify protective and risk factors specific to these trajectories that may inform policy and intervention.

Parent and child adjustment trajectories

We identified four family profiles of adjustment trajectories over 3 months from late March to early July 2020. In all groups except Group 2, parents’ anxiety symptoms decreased over time, whereas children's emotional distress remained stable in all four groups. These four profiles were relatively similar in their linear slopes but differed in their intercept. In other words, parental anxiety and children's emotional distress were heterogeneous across families in March, but they were similar in their rate of change across the ensuing 3 months. Around 57.3% of the sample presented with clinically significant symptoms––among which 13.8% (Profile 3) of parents endorsed clinically significant GAD symptoms; 34.5% (Profile 4) of children were reported to have clinically significant PEDS symptoms; and 9.0% (Profile 2) of parents and children both presented with clinically significant symptoms. The GAD scores at T1 (M = 1.20) and T2 (M = 0.95) from our study are on par with the GAD scores (M = 1.08) from a national sample of parents collected from April 27 to 28, 2020 (Russell et al., 2020).

In the few known studies that have been published, the trajectory profile that emerged for most children was that of adjusting well over time, which similarly mirrored findings for adults (e.g., Kronenberg et al., 2010; Osofsky et al., 2015; Weems & Graham, 2013). The emergence of these four profiles was overall consistent with our hypothesis that three to six profiles would exist based upon past trajectory studies (Lai et al., 2015; Weems & Graham, 2013). Our results also converge with past findings that most children and adults make up the low symptom group, suggesting similar resilience observed amid the COVID‐19 pandemic compared to prior disastrous events (Bonanno et al., 2008; Kronenberg et al., 2010; Osofsky et al., 2015; Weems & Graham, 2013). However, diverging from a few studies in which posttraumatic symptoms consistently increased or increased first before decreasing (Bonanno et al., 2008; Janssen et al., 2020; Kronenberg et al., 2010; Osofsky et al., 2015), we did not find an increasing trajectory among parents or children. This may be due to the specific, short‐term timeframe we used in our study design, which differed from the longer intervals used in prior research (Silverman & La Greca, 2002). Another short‐term longitudinal study with U.S. adults from February 2020 to May 2020 also found half of the sample exhibited no/minimal depression or anxiety symptoms without an increasing pattern (Davis et al., 2021). It is also possible that parental anxiety may have already reached its highest in late March when parents needed to quickly adjust to abrupt changes in routines in response to nationwide school closures (Education Week, 2020). Lastly, these 3 months may also not capture a delayed breakdown pattern sometimes observed in children and adults, in which posttraumatic symptoms may not occur until 1–4 years after exposure (Bonanno et al., 2008; Kronenberg et al., 2010; Osofsky et al., 2015). Taken together, our findings add new longitudinal evidence to the growing literature on children's post‐disaster adjustment, especially by elucidating the short‐term change from a family systems perspective.

For our second objective, we conducted explorative analyses to investigate potential protective and risk factors that predicted the dual‐trajectory memberships. We found the odds ratios of group membership compared to the low stable profile (Profile 1) could be accounted for by parent gender, child age, annual household income, racism stress, positive and negative parenting practices, and negative affect temperament. Aligned with emerging COVID‐19 research (Feinberg et al., 2021; Fitzpatrick et al., 2020; Marchetti et al., 2020; Russell et al., 2021), our data also indicated that low‐income mothers were likely to have higher anxiety symptom trajectories. We also found having an older child decreased the relative odds ratio of membership in Profile 4 compared to Profile 1. In other words, even within the 3‐ to 8‐year‐old range, younger age children (but not their parents) were more likely to have deteriorating well‐being. This finding highlights the vulnerability of younger age children amid disasters that warrants special attention from researchers, clinicians, and policymakers (Bahrick et al., 1998; Saylor et al., 1999; Wang et al., 2013).

As expected, we identified positive parenting practices (i.e., proactive parenting, positive reinforcement, warmth, and supportiveness) as protective factors and negative parenting practices (i.e., hostility, physical control, and lax control) as risk factors associated with the adjustment dual trajectories. The caveat is that the current GMM approach cannot address directionality among parenting practice and parental and child well‐being. The role of parental well‐being in supporting healthy parenting practices has been well researched (Abidin, 1992; Roos et al., 2021). Interventions to enhance parental well‐being will likely increase their positive parenting practices, which further protects children from the ramifications of COVID‐19 and the ensuing changes to daily routines.

Because negative affect temperament is a stable trait‐like individual differences among children (Putnam & Rothbart, 2006)—it can be viewed as a risk factor that existed prior to the COVID‐19 outbreak. In accordance with the extant literature (Lengua et al., 2005; Wilson et al., 2010), children who scored higher in negative affect may be more sensitive to sudden changes brought on by environmental stressors and experience more fearfulness or irritability as a result. We did not find the impact of surgency and effortful control in our study (Lengua et al., 2005; Wilson et al., 2010).

Lastly, stress with indirect exposure to racism is one of the strongest predictors in our study. This suggests the impact we found on family well‐being may not solely stem from COVID‐19‐related stressors, but also from exposure to ongoing racism. Our study contributes to the literature linking racism experienced by parents to child health (Heard‐Garris et al., 2018). Future research should continue to study this possible secondary influence from a family system perspective, given the continued national exposure to racially charged events (Carr, 2015; Heard‐Garris et al., 2018). A population‐based survey conducted in August 2020 found that the majority of adults (59%), regardless of race, reported that police violence toward minorities was a significant source of stress in their life (American Psychological Association, 2020). However, exposure and impact from racism have only been marginally conceptualized as a collective trauma that (differentially and inequitably) impacts both people of color and White people, and its measurement has not been established across racial groups (Authors, in press; Heard‐Garris et al., 2018; Martz et al., 2019; Menakem, 2017; Silver et al., 2020). The current measure would not be able to address, for instance, Whites’ perceived stress from indirect exposure to racism was due to cross‐group empathy, an unprecedented racial reckoning and White guilt, or a perceived threat to a White America (Jardina, 2019). More research is needed to unpack the potential different phenomenological experiences of racism stress endorsed by Whites (and people of color) (McKinney, 2005).

Limitations and conclusions

Although this study provides critical evidence for understanding inter‐family differences in adjustment following COVID‐19, there are some important limitations to consider when interpreting the findings. First, despite some retrospective reports on pre‐COVID conditions (e.g., child temperament), we were not able to collect any prospective data prior to COVID‐19. Thus, the “recovered” pattern may be a function of regression to the mean. This also limited our ability to make attributions to the adjustment trajectories. For example, there will likely be families who already presented with high parental anxiety and child emotional distress prior to the COVID‐19 outbreak. Nonetheless, the decreasing trajectory patterns suggest that the majority of the families were impacted by collective COVID‐19‐related stressors. Secondly, parents reported their young children's post‐disaster reactions. The shared methodological variances between parent and child data may, therefore, inflate the relationship between parent and child adjustment. A third limitation is the convenience sampling method used in accounting for the time‐sensitive nature of the project. Individuals who were disproportionately impacted by COVID‐19 especially communities of color were not adequately included in this study. Thus, our findings need to be interpreted with the demographics of the current sample (i.e., higher socioeconomic status, not racially representative) in mind.

Notwithstanding these limitations, this prospective longitudinal study based upon three waves of empirical data collection from March to July 2020 reveals the dynamic interplay between parental anxiety, child emotional distress, and key protective and risk factors. Current findings allow us to conclude that significant interfamily differences in adjustment following the COVID‐19 outbreak in the United States exist, with most families demonstrating resilience as well as some families with high persistent risk. One of the most significant findings to emerge from this study is that parents’ experiences of racism stress were related to the family well‐being, which further underscores the impact of racism during this ongoing syndemic. These important protective (e.g., positive parenting) and risk (e.g., child negative affective, negative parenting, racism stress) factors can guide practitioners and policymakers in identifying at‐risk sub‐populations and designing targeted interventions.

Supplementary Material

Click here for additional data file.