Youth Well-being During the COVID-19 Pandemic

This study leverages cross-sectional caregiver and adolescent self-report data from the National Institutes of Health (NIH) Environmental influences on Child Health Outcomes (ECHO) research program. Started in 2016, ECHO brings together 69 existing longitudinal observational pediatric cohorts from >200 sites across 44 US states and Puerto Rico to investigate a broad range of environmental exposures on 5 primary areas of child and adolescent health: pre-, peri-, postnatal; airways (eg, asthma); neurodevelopment; obesity; and positive health (eg, life satisfaction).^30,31  The current study included a subset of ECHO cohorts with participants from 34 states who had complete data collection for parent-report (n = 11 cohorts from 27 states) and/or adolescent self-reports (n = 5 cohorts from 25 states) on the primary variables required for analysis. Cohort descriptions, including recruitment strategies, inclusion/exclusion criteria, and participant details (Supplemental Table 3; Supplemental Figs 2 and 3) provide state-level geographic distribution of each multicohort sample that includes families from rural, urban, and suburban communities across the United States. Data were collected between May 1, 2020, and May 31, 2021. Given the diversity of ECHO cohorts’ study schedules and capabilities during the COVID-19 pandemic, different strategies were employed to collect these data, including sending an e-mail to all participants at the same time with a link to the online survey, collecting data at the next scheduled visit (whether remote or in-person), and adding a new study visit to collect COVID-19–related data. Cohorts primarily collected data via self-administered remote online survey, with some data collected via self-administered paper/pencil surveys and researcher-administered phone interviews. Local institutional review boards (IRBs) and/or the central ECHO IRB (Western IRB) reviewed all research methods and procedures.

For analytic purposes, participants were divided into 2 samples on the basis of child age and reporter of the child-level data, given known low correlations between parent-report and child self-report.^32,33  The first sample included data from 977 caregivers from 11 ECHO cohorts reporting on their own COVID-19 pandemic experiences and that of their children aged 2 to 12 years (mean age = 8.3, SD = 2.3). Parent-reports for child-level data were collected because of limited reporter reliability for younger children,^34,35  particularly for the outcome variables of child stress and well-being.¹⁹  The second sample was composed of 669 11- to 17-year-olds (mean age = 16.4, SD = 1) and their caregivers from 5 ECHO cohorts. Adolescents self-reported on their stress, life satisfaction, social connections, and family engagement, and their caregivers contributed data on family-level COVID-19 pandemic-related hardships, their own stress, and family engagement. There were no overlapping participants across samples. Table 1 summarizes the demographic characteristics of each group. Separate analyses were conducted for each participant group. For brevity, the 2- to 12-year-old sample is referred to as the “child” sample and the 11- to 17-year-old sample is referred to as the “adolescent” sample.

Apart from preexisting mental health conditions and sociodemographic variables, all measures came from the ECHO COVID-19 questionnaires, which were developed in April 2020 to capture the physical, mental, and social impact of the COVID-19 pandemic on children and families. A team of ECHO investigators, including measurement scientists, epidemiologists, and clinical psychologists, developed de novo survey items and modified items from existing surveys and source materials (eg, COVID-19 symptom checkers) from leading organizations (eg, Centers for Disease Control and Prevention, World Health Organization). See Supplemental Information for a description of the ECHO COVID-19 Questionnaire development process.

Child life satisfaction was measured via a single item derived from the rigorously developed Patient-Reported Outcome Measurement Information System Life Satisfaction Assessment.¹⁵  For the 2- to 12-year-old sample, parents were asked, “Since becoming aware of the COVID-19 outbreak, how often has your child felt happy and satisfied with his/her life?” For the 11- to 17-year-old adolescent sample, adolescents were asked, “Since becoming aware of the COVID-19 outbreak, how often have you felt happy and satisfied with your life?” Each was scored on a 5-point Likert scale from not at all (1) to very often (5).

COVID-19 pandemic-related stress was derived from the ECHO Acute Stress Disorder Scale,³⁶  which has parallel self-report versions for adults (≥18 years), adolescents (11–17 years), and a parent-report version for children (2–12 years; see Supplemental Table 4 for items by version). This instrument is aligned to the Diagnostic and Statistical Manual of Mental Disorders, fifth edition for acute stress disorder (ie, intrusive thoughts, negative mood, disassociation, and avoidance, arousal).³⁷  All versions use the same item stem (“Since becoming aware of the COVID-19 outbreak…”). The adult and adolescent versions have 9 items, and the parent-report version has the same 9 items with minor wording modifications for developmental appropriateness (eg, “My child seemed in a daze” for “I felt in a daze”), as well as 1 additional item to capture age-inappropriate behaviors resulting from COVID-19 pandemic-related stress (“My child did things that he/she had outgrown or acted younger than current age”). All items were scored on a 5-point Likert scale from not at all (1) to very often (5), and a total average score was computed (range: 1–5).

COVID-19 pandemic-related family hardships included 24 caregiver-reported hardships, such as unemployment, food insecurity, child care difficulties, health concerns, and worries about the child and family. Items were modified from the COVID-19 Perinatal Experiences Impact Survey.³⁸ 

Social connectedness represented the youth’s current level of social connectedness, or feelings of interpersonal closeness and a sense of belonging to a social group,³⁹  compared with before the COVID-19 outbreak (“Compared to before the COVID-19 outbreak, does the child seem/do you feel…”). This single-item measure used a 6-point Likert scale anchored by much less socially connected (1) and much more socially connected (6). Parallel items were used for parent-report and adolescent self-report. This item was modified from the Adolescent Social Connection and Coping During COVID-19 Questionnaire.⁴⁰ 

Family engagement represented how much the family unit interacted as a method of coping with the COVID-19 pandemic. Caregivers were asked whether, to cope with the pandemic, they (a) talked to friends and family, and/or (b) engaged in more family activities, scored as 0 (no) or 1 (yes). The average of both items was used as our measure of family engagement for the 2- to 12-year-old child analyses. Adolescents were also asked item b, and their response, along with their parents’ responses, were included in the family engagement score for the 11- to 17-year-old adolescent analyses. Items were modified from the COVID-19 Perinatal Experiences Impact Survey.³⁸ 

Child preexisting mental health problems were assessed via parent-reports of their child’s previous diagnosis of anxiety, depression, attention-deficit disorder or attention-deficit/hyperactivity disorder, autism spectrum disorder or pervasive developmental disorder, developmental delay, intellectual disability, and learning disability. Given the specific interest in anxiety and depression in the context of the COVID-19 pandemic and the small sample sizes for other conditions, a categorical variable with 3 mutually exclusive categories was constructed: no mental health conditions (reference); anxiety and/or depression, which included children with one or both conditions only, as well as those with anxiety and/or depression and other comorbidities; and other mental health condition (ie, has at least 1 mental health condition but does not have anxiety or depression).

Covariates included child age (continuous), sex assigned at birth (male or female), primary caregiver educational attainment (high school degree or less, some college, college degree, master’s, professional, or doctorate degree), ECHO cohort in which the child was enrolled, and COVID-19 diagnosis, defined as whether a doctor ever told the child and/or caregiver they had COVID-19 and/or a positive result on a nasal, mouth swab, or blood test. Child race and ethnicity are presented in Table 1. These variables were collected via parent-report or adolescent self-report and were required by the NIH, consistent with its Inclusion of Women, Minorities, and Children policy. Racial and ethnic categories were based on the NIH Policy on Reporting Race and Ethnicity Data. Since we did not have specific hypotheses regarding race and ethnicity, and previous research indicates that family and social supports function as protective factors across racial and ethnic groups,^26,29  these variables were not included as covariates.

We used mega-analysis^41–43  by pooling raw data from all cohorts into a larger single sample for each analysis (ie, one for the 2- to 12-year-old child analysis and one for the 11- to 17-year-old adolescent analysis), thus improving power beyond that of any single cohort sample. Also called “meta-analysis of individual participant/patient data,” mega-analysis is often considered superior to traditional meta-analysis, which relies on summary statistics because mega-analysis retains more detailed information (ie, raw data) from the original samples.^41–43  Because this technique uses raw data, mega-analysis requires either standard data collection or a way to harmonize data across individual samples. For the current study, all cohorts collected the same COVID-19–related data via the ECHO COVID-19 questionnaires, and all other data (eg, mental health conditions, sociodemographic variables) were harmonized to create standard variables as described above. We selected participants with data on all predictors, outcomes, and effect moderators, while allowing missing data on covariates. As shown in Table 1, all but 1 covariate had missing data from <10 participants; for the adolescent sample, 44 participants (6.6%) had missing caregiver educational attainment data, which is still well within the suggested 5% to 10% missingness.⁴⁴  We used multiple imputation to account for missing data; specifically, 25 completed data sets were constructed using the mice package⁴⁵  in R,⁴⁶  with imputed values on covariates. Aside from Table 1 and the least absolute shrinkage and selection operator (LASSO) regressions (see below), all descriptive and inferential analyses were conducted separately for each imputation, and the results were combined using Rubin’s⁴⁷  pooling rules.

After initial descriptive frequencies analyses, we used the “glmnet”⁴⁸  package in R to perform 2 separate multivariate LASSO regressions, 1 with child sample and 1 with the adolescent sample, to identify the sets of caregiver-reported hardships that were the strongest predictors of caregiver and child acute stress, conditioned on the other hardships. The LASSO approach removes or “shrinks” the model of the least predictive variables while maximizing prediction precision by retaining the most predictive variables with the least shared covariance. This is achieved by constraining the sum of the absolute values of the model coefficients, which effectively reduces the coefficients for some variables to 0 and results in a subset of predictors with the largest associations.⁴⁹  LASSO thus enabled us to reduce the large set of correlated hardships to the subset that minimized the prediction error of the parent and child acute stress variables. For children aged 2 to 12 years, the LASSO predicted caregiver reports of their own and their children’s COVID-19 pandemic-related acute stress; for adolescents aged 11 to 17 years, the LASSO predicted caregiver self-reported and adolescent self-reported COVID-19 pandemic-related acute stress.

For the primary analyses, we estimated separate models for the child and adolescent samples using structural equation modeling ( ( (Fig 1).⁵⁰  For the 2- to 12-year-old group, parent-report data were used for all variables. For the 11- to 17-year-old group, analyses included adolescent self-reported stress, life satisfaction, and social connectedness, as well as adolescent-reported family engagement, as described above. To aid in interpretability, youth stress, life satisfaction, social connectedness, and family engagement were standardized within each sample.

To evaluate model fit, we used standard fit indices: χ²P > .05; ratio of χ² to degrees of freedom (df) < 5; root mean square error of approximation (RMSEA) < 0.06; comparative fit index (CFI) and Tucker–Lewis Index > 0.95; and standardized root mean squared residual (SRMR) < 0.08.⁵¹  Path coefficients were judged to be significant if the associated P value $≤$ .05. To estimate and test mediated effects, we used a Monte-Carlo–based test.^52,53  Path models were estimated using the lavaan⁵⁴  package in R, with the semTools⁵⁵  package used to aggregate results from multiple imputation and compute mediation statistics.

The child and adolescent samples included approximately equal numbers of males and females (male: child: 51.9%, adolescent: 49.2%). The child sample was predominantly White (77.2%), with 6% American Indian or Alaska Native, 5.9% Black, 2.1% Asian, and 7.7% multiracial. This sample also had 10.8% Hispanic children. The adolescent sample was predominately White (63.5%) and Black (30.3%), with 3.6% multiracial and <10 children from other racial backgrounds; 5.5% of adolescents were Hispanic. Caregiver educational attainment was higher in the 2- to 12-year-old child sample: 64.1% had a college degree or higher compared with 37.1% in the 11- to 17-year-old adolescent sample. Additionally, fewer children aged 2 to 12 years had mental health problems (14.9%) compared with adolescents (38.4%; see Table 1 for sample descriptive statistics). Geographically, the child sample represented 27 states across the continental United States, ranging from <5 participants to 258 participants per state. The adolescent sample included participants from 25 states in the continental United States, with <5 to 211 participants per state. Supplemental Fig 2 (child sample) and Supplemental Fig 3 (adolescent) provide details on the geographic distribution of each multicohort sample.

On average, the child and adolescent groups had similar levels of youth COVID-19 pandemic-related acute stress (child: mean = 1.7, SD = 0.6; adolescent: mean = 1.9, SD = 0.8). Caregivers in the 2- to 12-year-old child sample had slightly higher pandemic-related acute stress (mean = 2.1, SD = 0.7) compared with caregivers in the 11- to 17-year-old sample (mean = 1.8, SD = 0.7). Additionally, child samples had higher life satisfaction (mean = 4.1, SD = 0.8) compared with adolescents (mean = 3.4, SD = 1; see Table 1). See Supplemental Table 5 (child) and Supplemental Table 6 (adolescent) for complete covariance matrices.

Multivariate LASSO regression for the 2- to 12-year-old child sample identified the same 8 hardships predicting caregiver- and parent-reported child COVID-19 pandemic-related acute stress (see Supplemental Table 7 for hardship items and frequencies and Supplemental Table 8 for LASSO results): health care access, health concerns, impact on family members, impact on the child, social distancing/being quarantined, financial concerns, access to personal/household supplies, and difficulties arranging child care. For the 11- to 17-year-old adolescent sample, the same 5 hardships predicted caregiver-reported and adolescent self-reported COVID-19 pandemic-related acute stress: health concerns, impact on family members, impact on the child, social distancing/being quarantined, and food insecurity (Supplemental Table 7). Summary count scores were computed for each set of hardships (child range: 0–8; adolescent range: 0–5).

The structural model with labeled path coefficients is displayed in Fig 1; caregiver stress was predicted by the number of COVID-19 pandemic-related family hardships, youth stress was predicted by number of family hardships and caregiver stress, and youth life satisfaction was predicted by youth stress. These paths resulted in a mediation model with 2 df, the omitted paths being the direct relationships between COVID-19 pandemic-related family hardships and youth well-being, and between caregiver stress and youth well-being. We also examined effect moderation by 2 types of social support, youth social connectedness and family engagement, and by youth preexisting mental health conditions. We controlled for youth and family sociodemographic variables and COVID-19 infection status (covariates) for all outcomes and estimated (residual) covariances between the predictors unique to the youth stress and life satisfaction regression and the 2 upstream predictors (family hardships and caregiver stress), yielding a fully saturated model except for the 2 df.

SEM results indicated excellent fit of the hypothesized model (χ²[2] =1.61, P = .45; RMSEA < 0.001; CFI ≈ 1; TFI ≈ 1; SRMR = 0.002). As summarized in Fig 1, COVID-19 pandemic-related family hardships had a significant direct association with caregiver and children’s acute stress, where each additional hardship equated to a 0.15 (SE = 0.01, P < .001) and 0.04 (SE = 0.01, P < .001) SD increase in stress for caregivers and children, respectively. In turn, child stress had a direct negative association with well-being: a 1 SD increase in acute stress was associated with a 0.36 (SE = 0.04, P < .001) decrease in life satisfaction. Mediation effects were significant for all mediated paths (Table 2). Of note, the largest indirect association was found for caregiver stress on child life satisfaction via child stress (⁠$b^$ = 0.15, 95% confidence interval [CI] = −0.2 to −0.11). COVID-19 pandemic-related family hardships had a total indirect association of −0.04 (95% CI = −0.05 to −0.03), with equal contributions from the hardships/caregiver stress path (⁠$b^$ = −0.02, 95% CI = −0.03 to −0.02) and hardships/child stress path (⁠$b^$ = −0.02, 95% CI = −0.02 to −0.01).

Social connectedness and family engagement both had positive direct associations with children’s life satisfaction: 1 SD increase in each of these social supports equated to a 0.17 (P < .001) and 0.18 (P = .002) SD increase in life satisfaction, respectively. Additionally, family engagement significantly moderated the relationship between children’s stress and well-being, where a 1 SD increase in family engagement was associated with a 0.15 (SE = 0.05, P = .002) decrease in the effect of child stress on child life satisfaction. No direct association or moderation was found for preexisting mental health conditions. Results also suggested older children had slightly lower parent-reported life satisfaction (⁠$b^$ = 0.04, SE = 0.01, P < .001) and slightly higher acute stress (⁠$b^$ = 0.02, SE = 0.01, P = .02) compared with younger children. Compared with children whose caregivers had a high school degree or less, children whose caregivers had a college degree (⁠$b^$ = 0.23, SE = 0.08, P = .001) or a master’s, professional, or doctorate degree (⁠$b^$ = 0.23, SE = 0.09, P = .001) had higher life satisfaction. No other covariate paths were statistically significant.

SEM results showed excellent fit for the hypothesized model (χ²[2] = 0.52, P = .77; RMSEA < 0.001; CFI ≈ 1; TFI ≈1; SRMR < 0.001). In general, results replicated those found for the 2- to 12-year-old child model (see Fig 1 path coefficients), with several notable exceptions. First, although the strength of the association between COVID-19 pandemic-related hardships and adolescent acute stress (b = 0.05, SE = 0.02, P = .02) mirrored that of the child model, the association between caregiver acute stress and adolescent acute stress was weaker and nonsignificant (b = 0.09, SE = 0.05, P = .06). Similarly, the indirect association of caregiver acute stress with adolescent life satisfaction was significant but weaker than that in the 2- to 12-year-old sample (b = −0.03, 95% CI = −0.08 to −0.00) (Table 2). Second, although both social connectedness (b = 0.11, SE = 0.04, P = .004) and family engagement (b = 0.11, SE = 0.05, P = .02) had direct associations with adolescents’ life satisfaction, no effect moderation was observed. Third, preexisting anxiety and/or depression was associated with a 0.24 (SE = 0.11, P = .02) SD decrease in adolescent life satisfaction; no moderation was found. There was a large association between acute stress and adolescent sex, where being female was associated with a 0.4 (SE = 0.6, P < .001) SD increase in acute stress. No significant associations of caregiver educational attainment were found.

Our findings in a diverse multicohort sample of >1600 children and adolescents provide new insights into varied stress-coping adaptations of caregivers, children, and adolescents to the COVID-19 pandemic. We know from previous research that, despite youth having much lower severe acute respiratory syndrome coronavirus 2 infection rates compared with adults during the study period,⁵⁶  social disruptions resulting from the pandemic have adversely affected their mental health.^7,57  Identifying children and adolescents with mental health problems is critical, but a preventative lens toward identifying the underlying mechanisms that promote well-being, and not just the lack of mental health problems, is required to get ahead of the pediatric mental health crisis that is swiftly becoming its own epidemic. The current study addresses this by testing and expanding the FSM^24–26  and Prime and colleagues’ COVID-19 Family Well-Being Risk and Resilience Framework²⁷  to examine malleable factors that promote youth’s psychological well-being.

Whereas the FSM emphasizes economic hardship as the upstream instigator of downstream youth maladjustment, LASSO results presented here suggest that social (eg, concerns about the child or adolescent, social distancing/being quarantined) and health-related (eg, access to health care) hardships are also important, at least in the context of the COVID-19 pandemic. In both the child and adolescent samples, “concerns about the child” was the most predictive factor of both caregiver and youth COVID-19 pandemic-related acute stress, followed by health-related hardships. For the adolescent sample, only one economic hardship (food insecurity) was present in LASSO results, and it had the lowest magnitude compared with the other four hardships. Although interventions focused on alleviating economic hardships may also impact social and health-related concerns (eg, if a family is financially secure, they may have fewer concerns about the child or family or have additional resources to enable better access to personal/household supplies), additional focus on social and health-related hardships is required to fully address the most impactful hardships impacting families.

In addition to identifying specific hardships, results showed the number of COVID-19 pandemic-related family hardships was associated with caregiver and youth’s pandemic-related acute stress, a finding consistent with the FSM and model by Prime et al. Caregiver stress mediated this relationship in the 2- to 12-year-old sample only, whereas adolescents were directly affected by pandemic-related family hardships. The lack of mediation for adolescents may in part be because of teenagers being more acutely aware of pandemic-related disruptions, particularly regarding family-level hardships because teenagers themselves may have become care providers for younger siblings; engaged in remote schooling longer than elementary school children; or became unemployed⁵⁸  or were unable to secure summer employment.⁵⁹ 

For children and adolescents, social relationships were significant contributors to their life satisfaction, independent of COVID-19 pandemic-related acute stress. In addition, greater family engagement was associated with decreased influence of pandemic-related acute stress for 2- to 12-year-olds, supporting research on the critical nature of child-caregiver relationships, particularly for younger children. Therefore, COVID-19 pandemic-related stress may have an acute impact on children’s lives, but it is important to remember that other youth’s lives, such as connecting with peers and family engagement, contribute to their appraisals of life satisfaction, emphasizing the critical nature of relational health to pediatric well-being.⁶⁰  It is also important to consider the COVID-19 pandemic social context, where social distancing and quarantine guidelines made contact with friends challenging, with most social connections likely occurring remotely rather than in-person. Even so, this level of interaction had measurable impact on children’s life satisfaction.

Additionally, having prepandemic anxiety and/or depression placed adolescents, but not children, at increased risk for lower life satisfaction. A significant association between adolescents’ stress and sex indicated that females had higher levels of psychological stress compared with males, a finding corroborating known gender differences in anxiety and depression before⁶¹  and during the COVID-19 pandemic.^57,62–65  Addressing children’s preexisting mental health problems, particularly for adolescent girls, may be another strategy to help decrease the negative impact of the COVID-19 pandemic on youth well-being.

Regardless of age, youth’s COVID-19 pandemic-related acute stress had a large direct impact on their well-being; for every 1 SD increase in acute stress, life satisfaction decreased by more than one-third SD. The moderate magnitude of the stress/life satisfaction association suggests these are related but also separable constructs. Measuring one does not provide a proxy for the other. Future interventions can find ways to target both constructs to decrease psychological stress, as well as improve well-being in and of itself. In some cases, the intervention may be the same. For example, mindfulness interventions can decrease stress, but they can also promote well-being.⁶⁶  Alternately, the 2 LASSO regressions highlighted 9 hardships that were related to caregiver and youth stress and could be targeted by federal or state agencies. Federal assistance to address caregiver’s financial and work concerns may help alleviate stress for both caregivers and youth. Evidence on the effectiveness of the financial benefits from the federal Coronavirus Aid, Relief, and Economic Security (CARES) Act,⁶⁷  the Consolidated Appropriations Act,⁶⁸  and American Rescue Plan Act⁶⁹  suggests government assistance helped stave off increases in poverty during the pandemic,⁷⁰  with individuals using funds to pay for basic needs (food, rent, mortgage, or utilities) and household or personal supplies⁷¹ ; continuing similar programs for the lowest income families could help continue to decrease poverty rates and reduce financial hardship. In turn, programs such as these could have important downstream influences on youth’s psychological well-being.

Social connectedness and family engagement may also be of prime interest to health care professionals, above and beyond evaluating negative risk factors such as stress. Results presented here support the American Academy of Pediatrics’ revised policy statement regarding childhood toxic stress, which proposes shifting from the deficit-based adversity focus to a strengths-based relational health framework that emphasizes safe, supportive, and nurturing relationships as a solution to both buffering adversities and promoting resilience.⁷²  One potential way to address the statement’s call for a public health approach to relational health is through social prescribing, or providing health professionals with nonmedical referral options that support social connection (eg, community arts activities, walking or running clubs, sports teams) and leverage local community assets while promoting social connectedness.^73–75  With concerns over the spread of the coronavirus through indoor activities, evidence from nature-based social prescribing offers unique outdoor opportunities to enhance youth’s social connectedness.⁷⁵  Additional strategies might include:

1. outdoor learning pods, which emerged as ways for small groups of youth to learn together without losing in-person social connection during school closures;

2. formal programs such as school-based social-emotional learning programs,⁷⁶  positive youth development interventions,^77–79  organized sports and extracurricular activities^80,81 ; and

3. informal activities such as recess,⁸²  unstructured play, and leisure activities.

Well-child care visits also provide clinicians an opportunity to discuss family engagement, including evidence-based strategies of talking and working together as a family to solve problems, identifying family strengths, and staying hopeful in difficult times.^60,83  The Well-Visit Planner⁸⁴  tool may be particularly useful to open conversations with families and individuate engagement strategies because the tool places the voice of the family at the center of clinical interactions.^85–87  Accounting for developmental differences and based on findings presented here, encouraging positive family engagement for children and emphasizing peer connections for adolescents can help differentiate intervention strategies.

This study has several limitations. First, these data are not nationally representative and include only youth-caregiver dyads with complete data on the primary variables required for the analysis. This limited which ECHO cohorts were included in the final samples. For the adolescent analysis specifically, ECHO cohorts contributing data were limited to primarily the east coast and western United States, possibly limiting our ability to generalize conclusions to other regions because policy approaches to the pandemic differed across regions of the United States. Compared with the 2- to 12-year-old child sample, the 11- to 17-year-old adolescent sample also had more frequent indicators of socioeconomic disadvantage (mothers without college education and mothers from racial and ethnic minoritized groups). These differences were driven in part by geography (Utah and South Dakota have lower frequencies of the population of racial and ethnic minoritized groups) and inclusion criteria for one of the cohorts (Cohort L only enrolled infants born <28 weeks’ gestation, a condition that is associated with socioeconomic disadvantage). We would anticipate that our model is robust across socioeconomic positions based on the large body of literature on the robustness of the FSM model across diverse samples.²⁶  Nonetheless, higher socioeconomic status very likely lessens COVID-19 pandemic-related stress experienced by caregivers and youth and might moderate the association with caregiver pandemic-related stress and diminish youth well-being. Both models were conditional on caregiver educational attainment and showed similar results despite differences in samples, adding to the robustness of the model findings and providing the basis for future research with nationally representative samples. Second, data include both parent-reported and adolescent self-reported outcomes, making it impossible to distinguish whether differences in models are because of reporter or child age. Third, the family engagement measure evaluated quantity of engagement rather than quality of the youth-caregiver relationship; however, the framing as coping strategies to deal with pandemic stress, specific-item wording suggesting positive engagement (eg, engaging in more family activities [eg, games or sports]) and positive associations with life satisfaction aligns with work on relationship quality. Finally, data were cross-sectional, making it impossible to control for or evaluate changes from pre- to post-COVID-19 pandemic regarding stress and well-being; however, the reference time point for survey questions was, “Since becoming aware of the COVID-19 pandemic…” which accounts for shifts in the primary constructs evaluated here.

As one of the only investigations of child and adolescent positive psychological functioning during the COVID-19 pandemic, this study supports the FSM and Prime and colleagues’ framework during a unique time when the population at large has been “exposed” to a ubiquitous stressor. Findings provide novel insight into upstream mechanisms that can be targets for future interventions to improve pediatric well-being during pandemics and other external stressors. The evidence presented here suggests that youth and caregiver stress have the potential to be reduced and well-being benefited by the systematic creation of routinely available opportunities for family and peer group social engagement, and with programs tailored to community and neighborhood strengths and needs. These findings further emphasize the importance of assessing the presence of positive psychological well-being instead of only the absence of mental health problems as the primary outcome and highlights potential opportunities for future interventions to target both constructs to decrease psychological stress, as well as improve well-being. By measuring COVID-19 pandemic-related family hardships and caregiver stressors that influence youth’s stress and their sense of well-being, we identified varied pathways to outcomes and potential targets for optimizing responses to ongoing COVID-19 pandemic-related challenges.

We thank our ECHO colleagues and the medical, nursing and program staff, as well as the children and families participating in the ECHO cohorts. We also acknowledge the contribution of the following ECHO program collaborators:

ECHO components: Coordinating Center: Duke Clinical Research Institute, Durham, North Carolina: Smith PB, Newby KL, Benjamin DK; Data Analysis Center: Johns Hopkins University Bloomberg School of Public Health, Baltimore, Maryland: Jacobson LP; Research Triangle Institute, Durham, North Carolina: Parker CB; Person-Reported Outcomes Core: Northwestern University, Chicago, Illinois: Gershon R, Cella D.

ECHO awardees and cohorts: Columbia University, New York, New York: Perera FP, Herbstman JB; Dartmouth College, Hanover, North Hampshire: Karagas MR; George Washington University, Washington, D.C.: Ganiban J; Kaiser Permanente, Oakland, California: Ferrara A, Croen LA; Michigan State University, East Lansing, Michigan: Paneth N, Kerver JM, Ruden DM; Memorial Hospital of Rhode Island, Pawtucket, Rhode Island: Deoni S, D’Sa VA; New York University Langone Health, New York, New York: Blair C; Penn State, University Park, Pennsylvania: Neiderhiser JM; University of California, San Francisco, California: LeWinn KZ; Bush NR; University of Colorado Anschutz Medical Campus, Aurora, Colorado: Dabelea D; University of North Carolina, Chapel Hill, North Carolina: O’Shea M, Fry R; University of Oregon, Eugene, Oregon: Leve LD; University of Tennessee Health Sciences Center, Memphis, Tennessee: Mason A; University of Utah, Salt Lake City, Utah: Stanford JB, Giardino A, Porucznik CA; University of Washington, Seattle, Washington: Karr C, Sathyanarayana S; University of Wisconsin, Madison, Wisconsin: Gern J.

CFI

comparative fit index

CI

confidence interval

COVID-19

coronavirus disease 2019

df

degrees of freedom

ECHO

Environmental influences on Child Syndrome Health Outcomes

FSM

Family Stress Model

IRB

institutional review board

LASSO

least absolute shrinkage residual selection operator

NIH

National Institutes of Health

RMSEA

root mean square error of approximation

SRMR

standardized root mean squared residual