COVID-19 Induced Environments, Health-Related Quality of Life Outcomes and Problematic Behaviors: Evidence from Children with Syndromic Autism Spectrum Disorders

For this study, 230 families with children diagnosed with PMD, RTT and SYNGAP1-ID completed self-report questionnaires in order to ascertain COVID-19 induced lifestyle changes or changes in their environments between July 2020 and January 2021. The principal caregivers completed a questionnaire for proxy-assessed health status in children. Study participants were distributed as follows: 138 (60%) of the participants were diagnosed with PMD, 50 (22%) with SYNGAP1-ID and 42 with RTT (18%). The study was approved by the Baylor College of Medicine’s Institutional Review Board. A recruitment email provided access to an online Qualtrics survey through the following organizations: Phelan–McDermid Syndrome Foundation, RettSyndrome.org, SYNGAP1 Foundation and SynGAP Research Fund, Inc. Participants had to read and understand English as well as have access to the Internet in order to complete the web-based questionnaire. All respondents were required to give their informed consent online prior to completing the questionnaire.

Our main dependent variables were children’s HRQoL and behavioral outcomes. HRQoL of the study’s participants was ascertained using the \(\hbox {PedQL}^{TM}\) Version 4.0 Generic Core Scales self-report questionnaire. This instrument was developed to measure the main dimensions of health as well as role functioning (school/day care) Varni et al. (1999). The proxy (third person) version of the questionnaire contains 23 items examining physical functioning (8 items), emotional functioning (5 items), social functioning (5 items) and school functioning (5 items). Each item utilized a 5-point response scale (0 = never a problem; 1 = almost never a problem; 2 = sometimes a problem; 3 = often a problem; 4 = almost always a problem). Items were reverse-scored and linearly transformed to a 0-100 scale (0 = 100, 1 = 75, 2 = 50, 3 = 25, 4 = 0). Thus, higher scores indicate better HRQoL. For scale and total scores, the mean was computed as the sum across all items divided by the number of items answered. Furthermore, a psychosocial health summary score was calculated as the mean score over the number of items answered across the emotional, social, and school functioning scales. We considered the following behavioral outcomes: indicator variables which denoted that the child startled easily sometimes or often; that the child had angry outburst sometimes or often and if the child seemed in a daze sometimes or often during the COVID-19 outbreak.¹

We designed a survey to investigate the association between the COVID-19 outbreak and the following domains: overall health, behaviors, school activities, access to healthcare, daily routine, as well as, basic household and medical expenses. Specific exposures of interest were indicator variables for: COVID-19 reduced caregiver’s mental health appointments; COVID-19 reduced access to medical health care services; during COVID-19 some money was left over; the child spent more than 2 hours outdoor each day during the COVID-19 outbreak; the child had access to a yard or other outdoor space; the child had in-person, phone or video contact at least once per week; on average during the COVID-19 pandemic the child spent at least 3 hours per day playing computer games; on average since the COVID-19 pandemic the child spent at least 3 hours per day watching TV; the child has been somewhat or very successful in completing the recommended homework during COVID-19. Furthermore, we used an interaction variable between time outdoors during COVID-19 of more than 2 hours each day and annual family income over \(\$55\) thousand (U.S. dollars) per year.

Covariates included in the multivariable models were informed by previous theoretical or empirical models (Olson et al., 2021; Thomas et al., 2012; Larsson et al., 2005) and included: age at assessment (years), English as the main language (yes (referent), no), caregiver’s highest educational qualification, marital status (married (referent), not married), race (white (referent), not white), and income.

Differences in characteristics between children with PMD, SYNGAP1-ID and RTT were tested using the ANOVA test for continuous variables and Chi-squared test for categorical variables. ANOVA and Chi-squared tests had also been used to assess differences by diagnosis status related to COVID-19 variables and HRQoL outcomes. Differences in \(\hbox {PedQL}^{TM}\) scale, psychosocial and total scores, between children with PMD, SYNGAP1-ID and RTT were estimated using an ANOVA test under standard distributional assumptions. Finally, we performed separate multivariable regressions to explore the adjusted association between COVID-19 related factors on HRQoL measured using \(\hbox {PedQL}^{TM}\) and clinically relevant outcomes. Furthermore, to test for the presence of indirect socio-economic effects moderating the association between time spent outdoors and children’s outcomes we examined the association of interaction variable between time outdoors during COVID-19 of 2 or more hours each day and annual family income over \(\$55\) thousand U.S. dollars per year. Fixed effects multivariate regression analyses were performed where the type of diagnosis was modeled as fixed effects due to condition-specific behavioral effects by diagnosis, which corrected for the fixed unobserved heterogeneity within each diagnosis (PMD, SYNGAP1-ID and RTT). We used the Ordinary Least Squares (OLS) estimator to perform the multivariable regressions. Linear probability models were fitted to investigate the adjusted association between COVID-19 induced covariates and the probability of developing problematic behaviors.

The data was collected between July 2020 and January 2021, and the response rate was not uniformly distributed across each day. However, the median participation rate was achieved on September 25, 2020. For the purpose of our study, we defined the early COVID-19 phase as the time between the start of data collection up to and including September 25, 2020, and the late COVID-19 phase after September 25, 2020 up to January 2021. We employed a 1 to 1 random matching (Rothman, 2012; Stuart, 2010) to compare children’s outcomes during earlier and later phases of the outbreak across all diagnoses.

Furthermore, we estimated the impact of the later COVID-19 phase by randomly matching upon the clinical diagnosis as this type of matching has been shown as an effective method to control for clinical and behavioral profiles of disorders (Rothman, 2012; Stuart, 2010). Differences between the time before or on September 25, 2020, and after this date matched observations were ascertained using the Student t-test for unequal variances. However, due to small sample sizes this analysis was restricted for PMD children only. Differences between early and late groups were ascertained using the Student t-test for unequal variances. Confidence intervals (CI) were computed using bias-corrected and accelerated bootstrap method (BCa). Statistical analyses were conducted using Stata 16.0 (Stata Corp, College Station, TX). P-values of 0.05 or less were considered statistically significant.

Table 1 reports baseline characteristics of the study’s population across each diagnosis. Of the 230 participants there were no differences found across diagnoses (138 with PMD, 50 with SYNGAP1-ID and 42 with RTT) among the following variables: child’s age, age at diagnosis, participation in early behavioral intervention, primary language spoken at home, caregiver’s marital status, race or ethnicity, caregiver’s relationship to the child, and participation in the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) program.² However, statistically significant differences were recorded across diagnoses within the following domains: the child’s gender (p-value < 0.01) as expected because RTT mostly impacts girls, type of health insurance (p-value <0.01), and the estimated annual household income (p-value < 0.01). The overall evidence suggested that the majority of socioeconomic covariates were largely balanced across diagnoses.

Tables 2 and 3 show the distribution of COVID-19 related covariates across each diagnosis. Statistically significant differences were detected across the diagnoses within multiple questions which ascertained the impact of COVID-19 on lifestyle changes and problematic behaviors in children. Results showed statistical significance in relation to a child’s behavior’ when we asked if the child startled easily, had angry outburst, or seemed to be in a daze more often during the pandemic. For example, almost 90% of SYNGAP1-ID parents reported angry outbursts sometimes or often versus the other disorders where it was around 60%. Also, over 45% of SYNGAP1-ID parents reported their child startled easily, sometimes or often compared to PMD and RTT around 35%. Furthermore, almost 75% of SYNGAP1-ID parents reported their child seemed in a daze sometimes or often, followed by RTT around 66% and PMD 44%.

Additionally, we found statistical significance in a child’s behavior with regard to their time spent outdoors with 35% of individuals with RTT rarely going outside compared with 7% and 14% of SYNGAP1-ID and PMD respectively (Table 3). Moreover, we also found significance across genetic diagnoses in variables dealing with caregiver and family support, measuring whether or not a child had expressed fears that they, a parent or another close family member could contact and/or die from COVID-19.

Table 4 presents descriptive statistics for the PedsQL scale, psychosocial and total scores across PMD, SYNGAP1-ID and RTT. The evidence demonstrated that there were no significant differences across the mean scores for school functioning (PMD highest 57.12, followed by RTT (48.80) and SYNGAP1-ID 45.68). However, results showed that mean scores were significantly different for physical functioning (PMD 45.68, SYNGAP1-ID 41.50 followed by RTT 19.95, p-value < 0.01), emotional functioning (PMD 71.22, SYNGAP1-ID 62.03 followed by RTT 61.20, p-value <0.05), psychosocial functioning (PMD 60.65, RTT 53.00 followed by SYNGAP1-ID 51.71, p-value <0.05) and for the overall score (PMD 55.78, SYNGAP1-ID 48.20 followed by RTT 41.67, p-value <0.01).

Table 5 separately reports results from two models: one which assessed the adjusted association of COVID-19 induced lifestyle changes and pre-specified clinical and sociodemographic covariates on total PedsQL score, and the other model which ascertained the association of COVID-19 factors on psychosocial PedsQL scores only. We found that COVID-19 induced constraints related to caregivers’ mental health appointments was associated with a decrement of 14.5 points (95% CI − 27.26, − 1.82) in PedsQL total score and a decrement of 21.25 (95% CI − 35.76, − 7.25) in PedsQL psychosocial score. Empirical models identified the adjusted negative effect of a child watching at least 3 hours of TV per day on his/her HRQoL which was associated with a decrement of 13.76 in PedsQL total score (95% CI − 21.41, − 6.11) and a decrement of 12.88 (95% CI − 20.21, − 5.55) for PedsQL psychosocial score. However, if caregivers reported their child spent more than two hours outdoors per day, there was a positive association with PedsQL total score of 6.63 (95% CI − 0.69, 13.95) adjusted for all other covariates. Furthermore, we found that marital status and level of education were statistically significant predictors of PedsQL during the COVID-19 outbreak. We note here that other variables, while not statistically significant, were of the plausible direction.

Table 6 reports results from OLS models which assessed the association between COVID-19 factors and problematic behaviors. There was a positive association between a higher reported reduction in a caregiver’s mental health appointments due to COVID-19 and all three negative child outcomes (startled easily, angry outburst and daze). Moreover, time spent outdoors was found to decrease the risk of angry outbursts and the child being in a daze. These two findings are consistent with previous results reported in Table 5 which explored the association between COVID-19 induced lifestyle changes on HRQoL outcomes. Furthermore, the evidence showed that the interaction variable between spending more than 2 hours outdoors during COVID-19 and annual family income over \(\$55\) thousand U.S. dollars per year was a statistical predictor for several variables. In particular, this interaction variable had a virtually identical relationship with the indicator for a child easily startling and had the indicator for a child’s angry outburst around − 0.19 (95% CI − 0.35, − 0.02).

Table 7 reports results from comparing HRQoL outcomes and problematic behaviors by date of responses. This study started data collection in June, 2020; however, on September 25th, 2020, a little over half of respondents (55.7%) had completed the survey. The overall evidence in Table 7 shows that generally PedsQL scores were higher for the late phase vs early phase. Additionally, psychosocial functioning scores were 5.97 CI (95% CI − 0.22, 12.15) higher for respondents during the later phase. Furthermore, the proportion of reported problematic behaviors decreased slightly by − 0.14 (95% CI − 0.30, 0.02) for children who startled easily sometimes or often and by − 0.16 (95% CI − 0.32, 0.01) for children who seemed in a daze sometimes or often during COVID-19. Interestingly, when we repeated the analysis using September 5th as a cutoff date (as around 20% of the responses were received before this date) the associations provided greater contrast. Thus, psychosocial functioning scores were 7.71 CI (95% CI 0.71, 14.71) higher for respondents after September 5th and statistically significant at 5% and total PedsQL scores were 6.19 (95% CI − 0.94, 13.32). We also found statistically significant decrements in the proportion of reports related to angry outburst − 0.31 (95% CI − 0.48, − 0.15) and being in a daze − 0.26 (95% CI − 0.46, − 0.07). However, these analyses did not adjust for clinical diagnosis, and thus, for heterogeneity across the compared groups.

To control for heterogeneity across clinical profiles we randomly matched observations before and after September 25th, upon clinical diagnosis.³ Given the sample size, we were only able to match PMD patients. Table 8 reports results from comparing early vs late COVID-19 HRQoL and problematic behaviors using a sample of 65 matched pairs with PMD. The evidence shows that total PedsQL scores were 8.14 higher (95% CI − 1.33, 17.62) for PMD respondents after September 25, 2020. Similarly, emotional functioning scores were 11.61 (95% CI 0.86, 22.37) higher but statistically significant at 5%, for PMD participants after September 25, 2020 and physical functioning score were 11.57 higher (95% CI − 0.52, 23.97) after the cutoff date. Furthermore, we also found statistically significant decrements in the proportion of reports that a child startled easily − 0.27 (95% CI − 0.49, − 0.05) among PMD respondents September 25, 2020.

The major strength of this study is the timely collection of rich and novel data during COVID-19 which ascertained exposure-outcome relationships consistently across genetic disorders that strongly predispose children to syndromic ASD. Furthermore, we used different measures, HRQoL and observed behaviors, to assess the relationship between COVID-19 induced environments and children’s outcomes. Thus, we used validated and reliable HRQoL measures which are holistic constructs (Barile et al., 2013; Moriarty et al., 2003), that are highly correlated with widely used health metrics including morbidity, mortality, and healthcare and provide data required for comprehensive health economics, and policy evaluations for Drugs et al. (2006), Chim et al. (2010), and Excellence (2013). Furthermore, a high number of subjects with rare diseases is also a strength as the combined diagnosis matched sample of 230 individuals of HRQoL data during the COVID-19 pandemic.

We acknowledge the limitations of our findings. First, the present study used a cross-sectional design as the data on family and children’s environments were only collected during the pandemic. Also, the sample of the study might not be representative for all patients diagnosed with syndromic autism due to heterogeneity in clinical and behavioral profiles. Thus, replication of this study in other patients diagnosed with syndromic autism such as fragile X syndrome, MECP2 duplication syndrome would be a valuable contribution to the literature. The study’s limitations also include proxy (parent/caregiver) perspectives regarding a child’s HRQoL. This implies that reported estimates might be impacted by bias due to proxy parental reporting of children’s HRQoL as the evidence shows that caregivers tend to underestimate the HRQoL of their children compared with a child’s self-report (Wolke, 2016; Verrips et al., 2001). COVID-19 questions had not been fully validated when the data was collected which is also a limitation. Furthermore, due to the nature of data collection we were not able to confirm the genetic diagnosis in each participant. Despite the data from our study sample deriving from socioeconomically diverse populations of patients the sample is not diverse in terms of race/ethnicity and family make-up. Thus, reported results might not be generalizable to more diverse samples. Finally, the empirical methods used may not fully control for selection effects caused by non-random participation in our data collection. Thus, the reported empirical relationships are not necessarily causal.

The findings of this study have a number of implications for HRQoL and the living environments for children with syndromic autism, as well as, for autistic children more generally. The results of our study imply that caregiver access to mental health treatments must continue despite challenges imposed by an outbreak. This is particularly urgent given the recent evidence which shows that caregivers’ stress and anxiety levels might be directly linked to HRQoL outcomes in children (Spinelli et al., 2020; Ueda et al., 2021). Thus, it is imperative to design novel targeted interventions such as telehealth and integrated artificial intelligence systems aimed at ensuring that access to health services continues during potential future outbreaks.

Our study shows that time spent outdoors and time watching TV had opposite effects on HRQoL and on problematic behaviors in ASD children. Consequently, interventions incentivizing time outdoors and supporting play time outdoors might meaningfully improve HRQoL in children with syndromic autism. Understanding specific outdoor activities and their contribution to HRQoL among children with syndromic autism might be another fruitful area of investigation. Further studies conducted from multiple perspectives are needed to understand the mechanisms of HRQoL outcomes in children with syndromic autism. Our study demonstrates that HRQoL and behavioral outcomes reported earlier during the pandemic were generally worse compared to those reported later. While our models are not able to fully explain those results, these findings might suggest evidence of adaptive response due to psychological coping mechanisms which should be explored in future research.

Overall, our study implies that COVID-19-type measures aimed at stopping the transmissibility of infections will likely result in negative lifestyle changes associated with decrements in HRQoL and increase the likelihood of problematic behaviors in children with syndromic ASD. However, certain restrictions might increase activities which are likely to have a positive impact on children’s HRQoL and behaviors such as time spent outdoors. Strategies aimed at increasing time outdoors or time away from digital devices might protect ASD children from harmful consequences of restrictions to prevent the spread of respiratory infections.