Design and setting
COVID-19 Young is a longitudinal study of young people aged 11-19 attending secondary and high schools within the municipality of Bergen, Norway (20). The data collection so far comprises two waves. The first data collection (T1) started 27th of April 2020, during the 7th week of the national lockdown, and closed on the 11th of May. The second wave of data was collected between the 16th of December 2020 and 10th of January 2021, during local restrictions implying partly closed schools and sports- and leisure activities put on hold.
The study comprises two subsamples: Cohort 1 were young people aged 12-15 years whose parents participated as part of the study Bergen in Change (21) and consented to the participation by their child(ren) in the present study. Upon consent, parents provided contact information for the youth. A total of 1565 youth were contacted in cohort 1 in wave 1. The consenting parents were more often females (Cramérs V: 0.069, p < 0.001), older (Cramérs V: 0.092, p < 0.001), had higher educational attainment (Cramérs V: 0.155, p < 0.001) and household income (Cohen’s D: 0.19, p < 0.001), and had less often shared residence for the child (Cramérs V: -0.054, p = 0.006) when compared to non-consenting parents (20). These differences were in the range between very small and small effect sizes. Cohort 2 were young people aged 16-19 years, attending high school. For this cohort, the county provided phone numbers from their school contact registers. A total of 5947 youth was contacted in Cohort 2 in wave 1.
The invitation procedures were the same for cohorts 1 and 2 in both waves of data collection. Youth were recruited via SMS and a link to a secure online platform containing an information letter and a 15–30-minute survey (depending on skip-rules). Two SMS reminders were sent. In both waves participants were included in a lottery for a new cellphone.
Characteristics of participants
In wave 1, a total of 7512 youth was invited to participate. Of these, 843 (54%) in cohort 1 and 2154 (36%) in cohort 2 responded, yielding a total of 2997 (40%) youths completing the T1 survey. The mean age was 17 years (SD 1.7), 57.7% were females, and most participants reported living with both parents (77.5%), being born in Norway (93%), and living with siblings (71%). All participants from wave one was invited to answer the second survey. A total of 1598 (53.3%) young people completed the second survey.
Predictor. The predictor in the present study was a variable differentiating between t1 and t2.
Covariates measured at baseline. The included demographic covariates were self-reported age, gender, and country of birth. Age was reported in whole years, and gender differentiated between “boy” and “girl”. We differentiated between being born in Norway and being born in another country. Additional covariates included questions about experiencing loneliness, home school learning, perceived situation at home with one’s family, sleep-problems, and nightmares.
Loneliness was gauged using the single question “Have you felt lonely?” with five possible response options Never (=0), Seldom (=1), Somewhat often (=2), Very often (=3) and Always (=4). For the purposes of the present study, we differentiated between Seldom or less (=0) and Pretty often or more (=1).
Regarding home school learning the participants were asked “During the weeks of homeschooling, do you feel that you have learned…” where the response options were Less (=2), About the same (=1) and More (=0). For situation at home, the participants were asked “How are you getting along with your family during this period after school closing?”, where the response options were A lot better (=1), A little better (=2), As before the school closed (=3), A little worse (=4) and A lot worse (=5). In the present study, we differentiate between Better (=1), The same (=2), and Worse (=3).
For sleep problems, we assessed problems with initiating and maintaining sleep (DIMS) with the following question “During the last couple of weeks, after the closing of your school: Have you had problems sleeping or do you wake up frequently during the night?” with three response options “not true” (=1), “somewhat true” (=2), “true” (=3). Nightmares were assessed using the question “After the schools closed, have you had nightmares or unpleasant dreams?” with three options “Yes, more often than before” (=1), “Not more than before” (=2), “I do not have nightmares or unpleasant dreams” (=3). The phrasing of all items may be found in supplementary Table 1 in Lehmann et al 2021 (20),.
Mental health. Mental health was measured by the 25-item Strengths and Difficulties Questionnaire (SDQ, Goodman 1998). The SDQ is a mental health questionnaire for 3- to 17-year-olds. A self-report version of the SDQ is available for young people aged 11-17 years. SDQ comprises five subscales: emotional problems; conduct problems; hyperactivity; peer problems; and prosocial. Each subscale contains five items rated on a three-point-scale (0-1-2), yielding a subscale-score in the range of 0–10. The scores (excluding the scale for prosocial behaviors) can be combined into a total difficulties score, with scores ranging between 0 and 40. In addition, the emotional problems and peer problems subscales can be combined to comprise the intermediate subscale “internalizing problems”, while conduct problems and hyperactivity subscales can be combined to comprise “externalizing problems”. For the purposes of the present study, we present the total scale scores as well as the individual subscales scores and the intermediate subscale scores Internalizing and Externalizing, at the two time points. In analyses including covariates we only present internalizing and externalizing problems. Cut-off points are available for self-completed SDQ for the total scale and the subscales (SDQEnglishUK4-17scoring-1.PDF (ehcap.co.uk). In the present study we used recommended cut-points to differentiate between “close to average” and “slightly raised (/slightly lowered)” versus “high (/low)” and “very high (/very low)”.
First, baseline characteristics for those with valid scores on at least one SDQ subscale at baseline were calculated. Next, the valid observations, mean scores, and standard deviations across time points of the SDQ total scale and subscales were computed.
Comparison of SDQ scores across time was approached using inferential statistics: Initially, a series of paired t-tests were computed for SDQ total and all subscales, followed by estimating the same associations using mixed models. In longitudinal analyses, mixed models hold the advantage that they make use of all available data points. For the continuous SDQ-measures mixed linear models with maximum likelihood estimation were employed, while mixed effects logistic regression models were used for the dichotomous SDQ-measures. The impact of the included covariates was investigated using the SDQ internalizing and the SDQ externalizing subscale only. Separate adjustment for each covariate is presented across the two subscales, as well as the fully adjusted models. Finally, we investigated the potential moderating role of the covariates in a series of interaction models, and statistically significant interactions were presented in stratified analyses.
Post-hoc analysis. To avoid tautological interpretation of the apparent moderating effect of loneliness on the SDQ internalizing subscale, we also estimated interaction models for SDQ Emotional Problems and SDQ Peer Problems separately as a post-hoc analysis.