Time trends in emotional well-being and self-esteem in children and adolescents during the COVID-19 pandemic

Study design

We investigated time trends in emotional well-being and self-esteem of children in Japan using the Corona-Codomo surveys, which collected serial cross-sectional data during the COVID-19 pandemic. These surveys explored the mental health and social experiences of youth aged 6–17 years of age and parents of youth aged 0–17 years and were repeated seven times in online format during the course of the pandemic: the first wave from April 30th to May 31st, 2020; the second wave from June 15th to July 31st, 2020; the third wave from September 1st to October 31st, 2020; the fourth wave from November 17th to December 27th, 2020; the fifth wave from February 19th to March 31st, 2021; the sixth wave from September 15th to September 30th, 2021; and the seventh from December 1st to December 31st, 2021. Participants were recruited through multiple platforms including the website of the National Center for Child Health and Development, child health organizations such as Japan Pediatric Society, social media, and other print and digital media platforms. The questionnaire consisted of self-report (age 6–17) and parent-report (age 0–17) questions, with slightly different questions depending on the age of the child. Children were able to participate on their own if their parent gave consent. The details of the study are described elsewhere [11, 12].

Inclusion criteria

We included youth aged 6–17 years old whose self-reported answers to the question on the outcomes were available. We excluded participants with only parent-reported answers. Additionally, we excluded waves four and six of the Corona-Codomo survey since the outcome variables were not collected in these waves, and thus used data from waves 1, 2, 3, 5 and 7.

MeasuresOutcome variables: emotional well-being and self-esteem

Emotional well-being and self-esteem were measured using the self-report KINDL for children and adolescents, which is a well validated tool for measuring health-related quality of life in children and adolescents [13]. KINDL consists of six domains, physical, emotional, self-esteem, family, friends, and school. We focused on the emotional and self-esteem domains in the present study. Each domain is assessed using four questions, in which the participant chooses an answer from five categories (never, seldom, sometimes, often, or always) on quality of life in the past week. Each question is scored from 1 (never) to 5 (always), and a total is calculated for each domain. We transformed the scores into percentage scores, with a score of 100 representing the best quality of life in that domain (corresponding to a score of 5 for all four questions).

Explanatory variable: COVID-19 Stringency Index

As a measure of the stringency of social distancing policies, we utilized the COVID-19 Stringency Index by the Oxford COVID-19 government response tracker, which measures the stringency of containment and closure policy indicators for various countries [14]. The Index is computed using nine metrics of social distancing policy: school closures, workplace closures, cancellation of public events, restrictions on public gatherings, closures of public transport, stay-at-home requirements, public information campaigns, restrictions on internal movements, and international travel controls. The Index is calculated for each country on any given day, by taking the mean score of the nine metrics, taking a value between 0 and 100. Gender (male or female) and age group at baseline (6–11 or 12–17 years old) were used as covariates.

Statistical analysesMain analysis

We first conducted a descriptive analysis of participant characteristics for each wave. Second, we computed adjusted estimates for each outcome in each wave. Outcomes were adjusted with zero–one inflated beta (ZOIB) regression models to account for the non-normality of the outcomes and extreme scores of 0 and 100 [15]. The wave number was used as a categorical variable in all models. Adjusted models included interaction terms between the wave number and the gender or age group variable to take into account the varying sample characteristics in each wave. Then, we re-computed the estimates of the outcomes after weighting observations to reflect the distributions of age and gender in the Japanese population based on 2019 Census data (6–11 year-olds, 48.7%; 12–17 year-olds, 51.3%; males, 51.2%; and females, 48.8%) [16]. The weighted averages of the outcomes were superimposed on a timeline with the 7 day averages of COVID-19 cases [17], dates of school closures and academic breaks obtained from United Nations Educational, Scientific and Cultural Organization’s school closure dashboard [18], and dates of declarations of state of emergency. Finally, we tested the associations between Japan’s COVID-19 Stringency Index and mental health outcomes (emotional well-being and self-esteem). We created a ZOIB model for each outcome and plotted the adjusted estimates of the outcomes and 95% confidence intervals for each value of the Stringency Index in 5-unit increments from 0 to 100. We added additive three-way interaction terms between the Stringency Index and age group and gender, to see whether the associations between the Stringency Index and outcomes differed between subgroups. We created plots of the estimates with 95% confidence intervals (CIs) as well as plots stratified by age group and gender. Cluster-robust standard errors were used to take into account the dependencies between observations within each wave.

We performed all analysis with Stata SE version 15.1 (StataCorp, Texas, USA) and the command “margins.” 95% CIs were computed based on the delta method [19]. Two-sided P < 0.05 were considered statistically significant for all tests.

Missing data

Proportion of missing data for outcomes emotional well-being and self-esteem was 6.2% in wave 1, and zero in other waves (see flow chart in Additional file 1: Figure S1). The participant was allowed in the first wave to stop answering the questionnaire before reaching in the end, which may explain the missing data for the outcomes. Child’s age and gender were missing in 7.3%, 2.5%, 1.9%, 2.0%, and 1.2% of respondents in waves 1, 2, 3, 5, and 7, respectively. Adjusted analyses were performed on complete cases.

Sensitivity analyses

We conducted two sensitivity analyses. First, we excluded participants from three municipalities in the third survey, since data were collected using a different method in these municipalities (flyers with QR codes were provided to students through schools in addition to nationwide online recruitment). Next, we excluded individuals who participated in multiple surveys, who were identified based on their answer to a question in wave 3 (19.6% of participants replied that they had already participated in previous surveys). We performed this sensitivity analysis since individuals participating in multiple surveys could compromise the independence of the data across waves.

Ethical considerations

The study was approved by the institutional review board of the National Center for Child Health and Development (approval number 2020-21) and all parents gave informed consent to participate.

Role of the funding source

The Corona-Codomo survey was funded by the Ministry of Health, Labour and Welfare Grant (Grant Number 20GC1019; wave 1), Japan Science and Technology Agency J-RAPID Collaborative Research/Survey Program for Urgent Research on the Coronavirus Disease 2019 Grant (Grant Number JPMJJR2008; waves 2, 3, and 5), and Japan Science and Technology Agency SICORP Grant (wave 7). Analysis was conducted with funding by the Japan Society for the Promotion of Science JRP-LEAD with UKRI Grant (Grant Number JPJSJRP20211709).

留言 (0)

沒有登入
gif