Introduction

The COVID-19 pandemic is a biological natural disaster with a serious impact on both physical and mental health (EM-DAT, 2022; Jones et al., 2021). As many other disasters, pandemics can cause great suffering at the physical, biological, and social level, with dangerous consequences for individuals’ health and wellbeing (World Health Organization, 2022). Generally, children and adolescents exposed to a disaster are considered to be at risk, because of their heightened vulnerability. For this reason, they need special attention compared to adults during emergencies (Peek et al., 2018).

Disasters such as the COVID-19 pandemic have cascading and cumulative effects that pose many challenges to young people and their families (Masten & Motti-Stefanidi, 2020). In terms of mental health, among the traumatic consequences of the COVID-19 pandemic there can be an increase of psychological disorders and/or symptoms, such as depression, anxiety, posttraumatic stress disorder (PTSD), and sleep disorders in children and adolescents (Golberstein et al., 2020). The knowledge on their prevalence is paramount to develop and implement evidence-informed interventions to cope with the traumatic consequences of the pandemic and to foster their resilience, both during and after its occurrence. Therefore, we examined the literature on psychological disorders and/or symptoms, assessed through self-report or other-report instruments, in children and adolescents; we took into account studies published in the first 18 months of the COVID-19 pandemic, using a meta-analytical approach.

Impact of COVID-19 on Children and Adolescents

On March 11, 2020, the World Health Organization (2022) declared that the spreading of the COVID-19 was a global pandemic. Many countries claimed a state of emergency, implementing strict public health measures. The safety measures taken, such as school closures, social distance, and indications on health protection behaviors, have had a strong impact on global mental health for children and adolescents (Ellis et al., 2020; Holmes et al., 2020). Recent literature confirmed that they are particularly exposed to the consequences of the COVID-19 pandemic from a psychological perspective (Cost et al., 2021; Kılınçel et al., 2020; Lavigne-Cerván et al., 2021; Ravens‑Sieberer et al., 2021; Tang et al., 2021, 2022b; Vicentini et al., 2020).

Method

Literature Search and Search Results

We conducted systematic searches in three databases during September 2021: PsycINFO, PubMed, and Scopus. We decided to focus on PsycINFO and PubMed as they are among the most authoritative databases for conducting meta-analyses in mental health research (Cuijpers, 2016). In addition, following guidelines suggesting not to confine reviews to one or two databases (Cheung & Vijayakumar, 2016; Lemeshow et al., 2005), we decided to examine also a citation database like Scopus to find other relevant articles, in line with previous experiences (e.g., Filatova et al., 2017; Hendriks et al., 2018). We used the following search terms: “COVID-19” AND “children and adolescents” AND “mental health”. Concerning inclusion criteria, we considered studies which: (a) involved participants until 20 years of age; (b) included the assessment of at least one measure of psychological disorders and/or symptoms; (d) reported the occurrence of disorders and/or symptoms so that effect sizes (ES) could be calculated; (e) analyzed data that were collected during the COVID-19 pandemic; (f) were written in English. We excluded publications reporting reviews, discussions, single-case studies, and qualitative studies. Moreover, we excluded those studies whose participants suffered from physical or mental illness prior to the COVID-19 pandemic and those studies which did not include the statistical indexes necessary as inputs for a meta-analysis.

The initial search identified a total of 1163 works published between the outbreak of the pandemic and September 30, 2021. Four-hundred and thirty-eight publications were indexed in PsycINFO, 448 were indexed in PubMed, and 277 were indexed in Scopus. As a first step, we removed 432 duplicates from this initial set, i.e., the same publications downloaded in different searches. Then, we screened the 731 selected publications. As a second step, we read all the titles and abstracts and included only the publications pertinent in terms of topic—i.e., respecting the inclusion criteria—for a total of 83. As a third step, we read each article, and this led us to exclude 36 publications because they were off topic, and 21 because they reported reviews, discussions, single-case studies, and qualitative studies. This last step of the selection process was conducted by two independent judges; the reliability was 100%. No publications were excluded after the discussion between judges. Thus, the search identified a selection of 26 publications. We report in the PRISMA diagram (Fig. 1) the results of the selection process (Moher et al., 2009).

Fig. 1
figure 1

PRISMA diagram (Moher et al., 2009)

Full size image

For ethical issues, we adhered to the recommendations of the American Psychological Association.

Coding and Reliability

We reviewed and coded the eligible studies for several variables. We coded measures of psychological disorders and/or symptoms, in terms of occurrence of depression, anxiety, PTSD, and psychological distress. We also coded the type of instrument (distinguishing self-report and other-report instruments) and the continent (North America, Asia, and Europe). The moderating effect of age was not examined, because in most studies children and adolescents were not separated. For an overview of the included studies, see Table 1.

Table 1 Overview of the selected studies
Full size table

Measures

Several authors explored the incidence of psychological disorders and/or symptoms during the COVID-19 pandemic on children and adolescents. We coded their occurrence, distinguishing them in four categories: depression (e.g., Cao et al., 2021; Tang et al., 2008).

Following the suggestion of Van den Noortgate and Onghena (2003), we used a two-step procedure. First, we ran a traditional random-effects meta-analysis: we evaluated the main effects, performed the influence analyses, and checked the publication bias. Second, we ran a multilevel mixed-effects meta-analysis to examine the role of the moderators. We checked whether the results of the traditional random-effects model differed from the results of the multilevel mixed-effects model, and we found that they were not substantially different.

For the multilevel mixed-effects models, we chose to use the restricted maximum-likelihood estimation method, because it seemed more appropriate for considering non-independent sampling errors due to the presence of multiple effects in the studies (Borenstein, 2009). We used Cochran’s heterogeneity statistic (Q) to test if the effect sizes of different studies were similar or not. A significant value of Q means that there is heterogeneity between the effect sizes. We used the I2 statistic to assess the level of heterogeneity. It measures the proportion of total variance due to the variability between the studies. We have low heterogeneity if the value of the statistic is between 1 and 49, medium if the value is between 50 and 74, and high if the value is between 75 and 100. We assessed the role of each moderator with mixed-effect models, considering the dependence of effect sizes through multilevel modelling (level 1 = effect sizes; level 2 = study). We conducted a test to evaluate the possible moderating effect of one or more variables included in the model. In this test, the null hypothesis was that all the regression coefficients were equal to zero, while the alternative hypothesis was that at least one of the regression coefficients was not equal to zero.

We examined the moderating role of the kind of disorder and/or symptom (depression, anxiety, PTSD, and psychological distress), the type of instrument used to assess disorders and/or symptoms (self-report and other-report), and the continent (North America, Asia, and Europe). We excluded those studies that did not have information on each moderator in the corresponding analysis. We assessed the potential publication biases using the trim and fill approach of Duval and Tweedie (2000). This approach estimates the number of studies missing from a meta-analysis by eliminating those studies that create patterns of asymmetry, and adding new data estimated on the initial sample to generate a symmetrical distribution of effect sizes. The output of this analysis is a funnel plot that is designed using the effect size against the standard error for each study.

Results

Prevalence of Psychological Symptoms and/or Disorders in Children and Adolescents

Initially, we analyzed the data concerning the proportions of children and adolescents affected by psychological disorders and/or symptoms reported in the studies included in the meta-analysis. Their incidence ranged from .07 to .67. Then, we transformed the proportions applying the Freeman-Tukey double arcsine transformation and we ran a first random-effects model. This model, k = 47, n = 155.282, estimated a pooled incidence of psychological disorders and/or symptoms equal to .21, 95% CI [.17, .26], SE = .03. The effect sizes were heterogeneous, Q(46) = 22,862.65, p < .001, and the proportion of total variance due to the variability between the studies was very high, I2 = 99.76%.

Considering the forest plot, we identified two potential outlying effect sizes. We examined them further to determine whether they were really influential to the overall effect size. Following Viechtbauer and Cheung’s (2010) suggestions, we analyzed the outlying effect sizes (Chen et al., 2020a, 2020c). However, our results can be viewed focusing on the bright side of the medal: about 80% of the participants did not show the disorders and/or symptoms at issue. Future research should further explore the factors underlying the occurrence of mental health disturbances in certain cases and the absence of their development in many other cases. Such knowledge would be of primary relevance to implement actions to sustain children and adolescents’ resilience, also for possible future traumatic events.