Participants
Paediatric clinical teams from 46 UK hospitals referred 2856 CYP receiving care for an immunosuppressive condition, defined as having a medical indication for an annual influenza vaccine, to the ImmunoCOVID-19 team. 1631 (57.1%) of these CYP and their carers consented to participation and were sent weekly online questionnaires assessing weekly COVID symptom presentation, COVID testing and results, COVID vaccinations, NHS attendance, hospital admissions, missing out on school, sport/leisure activities and parental concern. In July 2021, the 1021 CYP and their carers who had not dropped out of the study, were invited to participate in a mental wellbeing extension study, 467 consented (45.7%) and were sent online mental wellbeing surveys to complete in November/December 2021, January/February 2022, and March/April 2022. The CYP and their carers completed informed consent/assent for both the weekly ImmunoCOVID-19 survey and the mental wellbeing survey. Ethical approval for both components of the ImmunoCOVID-19 study was provided by the Leeds NHS Research Ethics Committee (IRAS 281544). The ImmunoCOVID-19 study complied to the latest version of the Helsinki Declaration.
Measures
CYP’s emotional wellbeing was assessed using the self-report child version of the positive and negative affect schedule (PANAS-C) [13]. The PANAS-C contains two 10-item subscales designed to measure positive (i.e., active, alert, attentive, determined, enthusiastic, excited, inspired, interested, proud, and strong) and negative affect (i.e., afraid, ashamed, distressed, guilty, hostile, irritated, jittery, nervous, scared, and upset). Answer categories on the PANAS-C range from 1, ‘very slightly’/‘not at all’, to 5, ‘extremely’/‘very much’, leading to a minimum combined score of 10 and a maximum combined score of 50. CYP’s health related quality of life (HRQOL) was measured using the self-report child and adolescent version of the KIDSCREEN-27 [14]. The 27-items scale measures CYP HRQOL in relation to physical wellbeing, psychological wellbeing, parent relations and autonomy, social support and peers, and school/education. A 5-point Likert-scale is used for the answer categories. For each of the 5 subscales, a scoring algorithm is employed to calculate T-scaled scores with a mean of 50 and a standard deviation of 10, higher scores indicating a better quality of life. Parental mental wellbeing was assessed with the 21-item self-report Depression and Anxiety Stress Scales (DASS—21) [15]. Answer categories range from 0, ‘did not apply to me at all’, to 3, ‘applied to me very much’/‘most of the time’, leading to a minimum subscale score of 0 and a maximum subscale score of 21. The subscale scores for depression and anxiety were taken into account when determining the parent’s mental health status in the currently described project. If a parent scored 10 or higher on the depression subscale they experienced depressive symptoms. If the parent scored 8 or higher on the anxiety subscale they experienced anxiety symptoms. Parents experiencing anxiety and/or depressive symptoms were coded as experiencing parental mental health problems (PMH).
The four previously discovered parental COVID concern patterns (resilient, recovering medium concern, recovering high concern, continuous high concern) [10] were considered potential fixed effects influencing CYP emotional wellbeing and HRQOL. Further factors considered to impact CYP psychosocial wellbeing were CYP’s clinical (diagnosis) and demographic (age, gender) characteristics, which were assessed at baseline (March-July 2020). Parental demographic and household information (gender, age, geographical location, rural/urban, green space, household income, household composition, number of siblings, and employment), collected in September 2021, were also taken into account.
Quantitative analysis
The data has been cleaned, processed, and analysed in SAS9.4 [16]. First the descriptive characteristics of the sample were determined. Then we used repeated-measure analysis of variance to assess unadjusted differences in emotional wellbeing and HRQOL measured in November/December 2021, January/February 2022, and March/April 2022. Next, we visualized unadjusted differences in emotional wellbeing and HRQOL over time for those CYP with and without parents experiencing PMH. Finally, a linear mixed model repeated measure (MMRM) approach was used to explore the predictors of the CYPs HRQOL and emotional wellbeing [17]. Different models were calculated for the separate subscales (positive mood, negative mood, physical wellbeing, psychological wellbeing, parent relations and autonomy, social support and peers, and school functioning). PMH was entered as fixed effect, this independent variable changed over time, as parental mental wellbeing was measured at the same intervals as CYP’s emotional wellbeing and HRQOL.
Time, CYP’s age (0–9 vs 10+), CYPs gender, CYPs diagnosis (rheumatological diagnosis vs other), receipt of COVID vaccine by September 2021, report of SARS-CoV-2 infection by September 2021, parental gender, parental age (27–40, 41–50, 51–62), geographical location (South -, Middle -, North of England, Wales/Scotland/Ireland), urbanization (rural, semi-rural, urban), access to green space (difficult, easy), household income (<£29500, ~ £29500, > £29500), household composition (single parent vs 2 parents), number of siblings (0, 1, 2+) parental employment (full/parttime vs non-working/disabled/retired), and parental concern level were entered into the model as fixed effects. The significance of these independent variables was first tested in univariable models. Only significant variables were entered into multivariable models, with the aim of building similar multivariable models for each of the outcomes. Parental concern, CYP’s diagnosis, report of SARS-CoV-2 infection by September 2021, geographical location, urbanization, and household composition were not significant in predicting CYP’s emotional wellbeing or HRQOL in the univariable models and thus these fixed effects were not entered in the multivariable models.
A heterogeneous Toeplitz covariance matrix was chosen for the MMRM models as this matrix structure had the best fit (AIC = 4263 and BIC = 4283). It represented the simplest model with lowest AIC/BIC combination [17] when compared to compound symmetry (4274/4282), unstructured (4262/4287), autoregressive (4281/4289), heterogenous autoregressive (4270/4287), Toeplitz (4273/4285). The Kenward-Roger correction was applied to reduce bias in estimation of standard errors and F-Statistics [17]. The results are presented as parameter estimates with standard errors and P-values.