Mental sequelae of the COVID-19 pandemic in children with and without complex medical histories and their parents: well-being prior to the outbreak and at four time-points throughout 2020 and 2021

Families were recruited from two ongoing prospective cohort studies at the University Children’s Hospital Zurich, Switzerland: The EpoKids study [42] investigates the potential long-term neuroprotective effect of erythropoietin on executive functions in children born very preterm (VPT). Children born VPT were eligible for EpoKids if they had been enrolled in the trial ‘Does erythropoietin improve outcome in very preterm infants?’ (NCT00413946) at birth and had participated in the 2-year follow-up assessment [43]. They were recruited for EpoKids when they were between 7 and 12 years old. Typically developing, term-born children of the same age were recruited as siblings, friends, and through flyers at local schools and hospitals, and included into a control group [42]. The Research and Child Health Outcome (REACHOUT) study longitudinally follows children with congenital heart disease (CHD) who underwent cardiopulmonary bypass surgery at the University Children’s Hospital Zurich, Switzerland before 6 years of age between 2004 and 2009. At 10 years of age, only children without genetic or syndromal disorders were assessed. The REACHOUT study focuses on the neurodevelopmental outcome of these children [44, 45]. Families were eligible for the current study on well-being during the COVID-19 pandemic if the child had participated in the neurodevelopmental assessment and the parents had completed a set of questionnaires on child and parent well-being between January 2013 and mid-March 2020 as part of the EpoKids or the REACHOUT study (T0: prior to the implementation of measures to reduce COVID-19 in Switzerland). Parents of eligible families were invited to complete an online survey once during the first wave of the pandemic (T1: between April 17 and May 10, 2020), while lockdown measures, including school closure, were in place in Switzerland [13]. Families who had participated in the T1 assessment were approached again once during the second wave (T2: between October 30 and November 22, 2020), once during the third wave (T3: between April 23 and May 23, 2021), and once during the fourth wave (T4: between October 29 and November 21, 2021), when governmental restrictions were less severe: schools were open, but public and private assemblies were restricted. The parents could either fill out the questionnaire online or in a paper–pencil format (sent by mail with postage-paid return envelope). The vast majority of families chose the online format. Supplementary Fig. 1 details the assessment procedure for the current study. Supplementary Table 1 lists the restriction measures in Switzerland at each assessment time-point.

Families who participated in the COVID-19 survey at T1 did not differ in parental education or parent well-being prior to the pandemic (T0) from those who did not participate at T1 [13]. Furthermore, parents who participated at T2 (P = 0.192) and T3 (P = 0.671) did not differ in parent well-being at T1 from those who did not participate at T2 and T3, respectively. Parents who participated at T4 had lower parental well-being at T1 in comparison to parents who did not participate at T4 (P = 0.048) with small effect size (Median difference = 2.86, Cohen’s d = 0.287). The length of the time interval between T0 and T1 was not associated with changes in child well-being [13]. The study was approved by the local ethics committee, and all parents gave written informed consent.

Standardized questionnaires assessing quality of life were selected from the protocols of the two cohort studies (T0), and were included in the online surveys at T1, T2, T3, and T4.

For children, the psychological well-being subscale of Kidscreen-27 [46] was used to assess well-being. Parents completed the proxy-report of the scale at all time-points. Children completed the self-report of the scale at T0, T2, T3, and T4. At T0, children completed the questionnaires during the on-site assessment for the prospective cohort studies. At T2, T3, and T4, the questionnaires were sent to the children by mail after the parents had completed the online survey. The psychological well-being subscale includes seven items that assess the child’s positive emotions and satisfaction and the absence of feelings of loneliness and sadness. Raw subscale scores were transformed into T values with Swiss norms (n = 1672, adjusted for age and sex). Low values indicate poor well-being [46].

For parents, the mental subscale of the Short Form Health questionnaire was used to assess self-reported well-being. The 36-item version (SF-36) was used at T0, and the 12-item short form (SF-12) was used at T1, T2, T3, and T4 [47]. For the analysis at T0, only the 12 items overlapping with the short form were considered. The mental subscale of the SF-12 assesses four dimensions: vitality, social function, role limitations due to emotional problems, and mental health. Raw scores were transformed into T values based on German norms (n = 2524) [48]. Both the Kidscreen-27 and the SF-12 have acceptable to good internal consistency [46, 47].

Several predictors of child and parent well-being were assessed: maternal and paternal education were assessed as an indicator of socio-economic status. Higher scores indicate higher education (range 2 to 12). The 14-item short form of the Social Support Questionnaire (F-SozU K14 [49]) was assessed as part of both original cohort studies, and, thus, data on the perceived extent of support from the social network that is accessible if needed were available for all families at T0. Three dimensions, emotional support, practical support, and social integration, were summed according to the manual. Higher scores indicate more social support. The F-SozU K14 has excellent internal consistency [49]. At T1, the quality of family functioning was assessed with the 27-item Family Relationship Index (FRI [50]). Three dimensions, cohesion, expressiveness, and conflict, were summed according to the manual. Higher scores indicate better quality of family functioning. The FRI has good internal consistency [50]. Previously, family functioning during the first wave of the pandemic (i.e., T1) has been reported to be impaired in these families [13], and it was, thus, investigated whether this initial impairment continued to impact child and parent well-being as the pandemic further evolved. Familial COVID-19 risk status was assessed at T1 by asking parents whether a family member was at risk for a severe disease course in case of an infection with SARS-CoV-2 due to a pre-existing health condition. A dichotomous variable differentiated families with a member at increased risk from those without.

Parent and child characteristics are expressed as numbers and proportions of totals (dichotomous data), median and interquartile range (ordinal data), and mean and standard deviation (continuous data). Child and parent well-being of the total sample were compared to normative data (as provided by the respective manuals [46, 48]) at each time-point using one-sample t tests for normally distributed data and Mann–Whitney U tests for skewed data. Effect sizes were estimated with Cohen’s d (small effect > 0.2, moderate effect > 0.5, and strong effect > 0.8 [51]). The proportion of children and parents scoring below the normal range (> 1 SD below the normative mean or median) were reported separately for each time-point to illustrate clinical relevance of low well-being.

Longitudinal changes of child and parent well-being were investigated with mixed-effect models: Three models were calculated with the following outcomes: (1) proxy-reported child well-being, (2) self-reported child well-being, and (3) parent well-being. As fixed effects, the models included assessment time (categorical: T0 = prior to the pandemic; T1 = first wave, spring 2020; T2 = second wave, fall 2020; T3 = third wave, spring 2021, T4 = fourth wave, fall 2021), group (categorical: typically developing, CHD, and VPT), and age at assessment and sex of the child or the parent, respectively. T0 (factor ‘time’) and typically developing children (factor ‘group’) were defined as reference categories. As random effect, family-specific intercepts were included to take respondent-specific variability and shared variance between siblings into account (pairs of siblings: n = 21). Children’s individual intercepts were nested within families. The additional predictors of child and parent well-being [i.e., parental education, (T0), perceived social support (T0), family functioning (T1), and familial COVID-19 risk status (T1)] were then added as fixed factors to those models with a significant time effect.

Unstandardized regression coefficients (B) and generalized semi-partial R² (R²_B), to quantify effect sizes for mixed models, were reported (small: R²_B < 0.01; medium: R²_B > 0.09; large: R²_B > 0.25 [52]). The distribution of residuals was examined to evaluate normality.

All analyses were performed with R version 4.0.3 [53]. P values at a α-level of 0.05 were considered statistically significant. False discovery rate (FDR) control was used to correct for multiple comparison [54].

Before the pandemic (T0), families of 346 children had participated in one of the two cohort studies and thus were eligible for the current study. Families of 200 children participated in the online survey at T1 (follow-up rate = 58%). These children were between 7 and 13 years at T0 (mean age = 10.4 ± 1.2) and between 8 and 17 years at T1 (mean age = 12.8 ± 2.0). Details are presented in Table 1. At T2, families of 138 children (follow-up rate = 70%) completed the online survey again. At each T3 and T4, families of 134 children (follow-up rate = 67%) completed the online survey again. Primarily, mothers reported on the well-being of their children (> 90%). The questionnaires of both parents were available for 27 children; therefore, only the mothers' responses were retained for further analyses. In a number of families, parents completed the survey for more than one child, resulting in 175, 122, 117, and 117 parents participating at T1, T2, T3, and T4. Sample characteristics are displayed in Table 1.

Of the children with CHD, 21% had a univentricular heart defect. Children born VPT were born at a mean gestational age of 28.9 weeks (SD = 1.6). In 36 families, at least one family member had been infected with SARS-CoV-2 until November 2021. In 75 families, at least one family member was reported to be at risk for a severe course of disease in case of a SARS-CoV-2 infection (assessed at T1).

Table 2 displays the comparison to normative data as provided by the respective manuals for child and parent psychological well-being.

Table 3 reports the statistical estimates of the three linear mixed models assessing longitudinal changes in child and parent well-being. Compared to before the pandemic, child proxy-reported well-being was significantly lower during the first but not the second, the third, and the fourth wave of the pandemic (Fig. 1A). The model’s effect size was small [R²_B(CI-95) = 0.044 (0.085 to 0.028)]. The effect size in the model with the additional predictors increased to moderate [R²_B (CI-95) = 0.132 (0.187 to 0.102)]. Well-being was independent of sex and age of the child. Proxy-reported well-being of children born VPT or with CHD did not differ from typically developing children after FDR correction. Sparse social support before the pandemic and poor family functioning during the first wave significantly predicted lower well-being. There was no significant interaction either between assessment time and perceived social support (P = 0.087) or assessment time and family functioning (P = 0.088) on well-being.

For child self-reported well-being, there was no significant change in well-being across time with small effect size [R²_B (CI-95) = 0.022 (0.055 to 0.012, Fig. 1B)]. Self-reported well-being of children born VPT or with CHD did not differ from typically developing children after FDR correction. Self-reported well-being and proxy-reported well-being correlated weakly to moderately at T0 (Spearman’s r = 0.15, P = 0.050), T2 (Spearman’s r = 0.11, P = 0.343), T3 (Spearman’s r = 0.35, P = 0.001), and T4 (Spearman’s r = 0.49, P = 0.001).

Parent well-being dropped significantly during the first wave compared to before the pandemic, and remained significantly lower during the second, third, and fourth waves (Fig. 1C). The model’s effect size was small [R²_B(CI-95) = 0.057 (0.095 to 0.038)]. Adding the predictors increased the model’s effect size to moderate [R²_B(CI-95) = 0.140 (0.189 to 0.112)]. The well-being of parents of children born either VPT or with a CHD was higher than the well-being of parents of typically developing children. Well-being was independent of parent age. Female sex, sparse social support before the pandemic, and poor family functioning during the first wave significantly predicted lower well-being. There was no significant interaction between assessment time and group (P = 0.065), assessment time and sex (P = 0.094), assessment time and social support (P = 0.276), or assessment time and family functioning (P = 0.298) on well-being.

Parents’ self-reported well-being and parents’ proxy-report well-being of their children correlated weakly to moderately at T0 (Spearman’s r = 0.28, P < 0.001), T1 (Spearman’s r = 0.20, P = 0.008), T2 (Spearman’s r = 0.21, P = 0.020), T3 (Spearman’s r = 0.30, P < 0.001), and T4 (Spearman’s r = 0.17, P = 0.077).

The current longitudinal investigation draws on two cohort studies not originally designed to investigate the mental sequelae of COVID-19 for families. Thus, some limitations require consideration: the questionnaires for the assessments during the pandemic were selected from the study protocol of the cohort studies to allow changes to be investigated. These questionnaires assess well-being rather than mental health symptoms; thus, the conclusions that can be drawn about the prevalence of mental health problems during the pandemic are limited. The study sample includes children with and without complex medical histories and is not representative of the general population of children and parents in Switzerland. However, the longitudinal findings presented here are in line with and complement findings from cross-sectional studies conducted during the initial wave of the pandemic with nationally representative samples that report child and parent well-being compromised in comparison to normative data [11, 25].

The sample size of the current study was relatively small compared to previous cross-sectional studies (e.g., [15, 16, 20]). However, its longitudinal design ensured well-powered analyses owing to within-subject correlations [69]. Of the participating families, 71% participated at four or five measurement time-points. Well-being at T1 was not different between those parents who participated at T2 and T3 compared to those who did not participate, but was lower in those who participated at T4 compared to those who did not participate (small effect size; P = 0.048, Cohen’s d = 0.287).

Participating families come from high and rather homogenous socio-economic backgrounds, as is often seen in prospective cohort studies [70]. This may explain the absence of any effect of parental education on well-being, which had been expected from previous findings [1].

Finally, the participation rate in child self-reports was lower compared to the parent-reports. Also, no self-report of child well-being was assessed during the first wave of the pandemic. Thus, no conclusion can be drawn from this study about the potential immediate effects on child self-reported well-being. Correlations between parent- and self-reports of child well-being were weak to moderate. This is in line with other studies reporting modest correlations, likely because parent and self-reports reflect different realities of perceived well-being [71, 72]. Therefore, studying both, child and parental perspectives, is crucial to comprehensively understand child well-being [73]. Importantly, however, parents’ self-reports of their well-being and parents’ proxy-reports of their children’s well-being also correlated only weakly to moderately. This suggests that parents are able to distinguish the perception of their own well-being from that of their children rather than the low parental well-being negatively impacting their report of child well-being.

This study provides evidence of the long-term mental sequelae of the COVID-19 pandemic for parents of school-aged children with and without complex medical histories. One in four parent reports substantially compromised well-being in late 2021—more than 1 year into the crisis. Children’s well-being was specifically affected during the initial wave of the pandemic, as reported by their parents, and subsequently recovered. Importantly, even small psychological impacts of the pandemic have been argued to require careful attention as they may pose a substantial public health problem if reproduced across the whole population [2]. Consequently, it is crucial to provide psychological support to those in need alongside the comprehensive economic measures that have been implemented by governments to recover from the ongoing COVID-19 pandemic.

The findings of the current study provide evidence that parents of school-aged children require attention throughout a pandemic as their psychological well-being remains compromised well beyond the initial wave. This is specifically true for those with limited social support and poor family functioning. Strategies to support vulnerable families during times of crises need to be developed and implemented, and most importantly, they must remain in place even if social distancing measures are necessary to reduce infection rates (e.g., virtual counselling). Further, children’s well-being may be particularly compromised during times of school closures. Keeping schools open whenever possible, or re-opening them early after periods of strict lock-down should, thus, be part of future pandemic strategies to support psychological well-being of children throughout a pandemic.