Challenging Behaviors across COVID-19 in Young Children with Rare Neurogenetic Conditions: A Seven-Year, Cross-Syndrome Analysis

Abstract

Challenging behaviors are common in rare neurogenetic conditions and significantly impact family life. Families were especially vulnerable during COVID-19, yet little is known about behavior trajectories considering pre-pandemic risks. The Purdue Early Phenotype Study has followed children with and without neurogenetic conditions since 2017, including during the pandemic. Analyses examined patterns and predictors of challenging behaviors among 228 children (Williams=52; Angelman=64; Prader-Willi=29; fragile X=23; controls=60), ages 1 month–10 years (57% male, predominantly White), focusing on how pre-pandemic features predicted later trajectories. Across classes, challenging behaviors increased. Latent class analysis showed that higher-risk family profiles predicted steeper increases in internalizing behaviors. Findings suggest COVID-19 layered complex additional effects onto dynamic early behavioral trajectories in families of children with rare neurogenetic conditions.

Lay Summary

Families of children with rare neurogenetic conditions commonly experience challenges with their child’s behavior which can be quite stressful to manage. This study followed 228 children with and without neurogenetic conditions from 2017 on, including during the COVID-19 pandemic. Here, our goal was to understand how characteristics of the child and family predicted behavioral changes during the pandemic. Overall, challenging behaviors increased during the pandemic but increased more substantially for families with less support and more high-risk features. These findings show that the pandemic layered on additional challenges for families of children with neurogenetic conditions and highlight the importance of supporting families during major public health disruptions.

The COVID-19 pandemic induced drastic shifts in day-to-day life for families (Biroli et al., 2021). These shifts were even more pronounced in several notable subgroups, including families of children with rare neurogenetic conditions (NGCs) and other disabilities, who were particularly vulnerable during the pandemic due to both direct health consequences (Macaluso et al., 2024) and reduced access to medical and educational supports (Kowanda et al., 2021). For children with NGCs and their families, challenging behaviors are of notable concern (Neo & Tonnsen, 2019, Weber et al., 2019; Willgoss et al., 2021) and can be difficult to manage given limited access to local experts well-versed in the needs of specific patient subgroups, even prior to the pandemic (Vohra et al., 2014). While it is likely that a disruption such as a global pandemic might intensify these challenges, patients with rare NGCs and their families also entered the pandemic with unique experiences – including pre-existing isolation, access barriers, and complex mental health and medical needs – that also shaped how the pandemic impacted their day-to-day lives. Examining how the pandemic intersected with these experiences is important to understanding and addressing the persistent needs of families impacted by neurogenetic conditions, particularly given the differential impact of COVID-19 on families of children with disabilities (Kowanda et al., 2021; Martin et al., 2023). More broadly, the experiences of these families can also serve as a unique model for understanding how genetic, behavioral, and environmental experiences intersected with the COVID-19 pandemic, informing broader developmental models of risk and resilience in response to major public health events.

The present study examines impacts of the COVID-19 pandemic on challenging behaviors of children with NGCs, here defined as genetic conditions typically associated with intellectual and developmental disability. Challenging behaviors are a common co-occurring feature of many NGCs and, depending on the condition, might include internalizing, externalizing, or self-injurious behaviors. Challenging behaviors are known to impact day-to-day family life (Currie et al., 2024) and also be impacted by the child’s environment (Hessl et al., 2001). However, causes of behavioral elevations tend to vary. Genetic variability associated with each NGC likely plays a direct role. Additionally, behavioral challenges can emerge secondary to adaptive behavior deficits and functional needs, such as gaining attention from a caregiver or avoiding a non-preferred activity or situation (Schmidt et al., 2016). Importantly, behavioral challenges emerge early in development across many NGCs (Neo & Tonnsen, 2019) and represent a top treatment priority for many NGC patient communities (Weber et al., 2019; Willgoss et al., 2021), signaling the significant impact these behaviors have on the day-to-day experiences of families.

Traditional diathesis stress models – which frame the onset of psychopathology as an interaction between preceding factors and acute challenges – would generally predict that children who experienced high rates of challenging behaviors prior to the pandemic would experience even greater risk for pandemic-related concerns. Given higher rates of medical and behavioral challenges among children with NGCs, many NGC families relied on medical and educational supports that were largely disrupted by COVID-19. Based on data that have emerged since COVID-19, these assumptions appear to be generally supported as many families experienced critical disruptions in support services (Currie et al., 2023; Martin et al., 2023; Nicholas et al., 2022). In families of children with autism, challenging behaviors generally worsened during the pandemic (Colizzi et al., 2020), and behaviors worsened disproportionately for families reporting pre-pandemic risk factors, including those experiencing low income and/or food insecurity (Panjwani et al., 2021). In other words, it is well-established that children have been impacted by the pandemic, including through disruptions in their day-to-day supports.

Similar elevated risk might be expected for caregivers managing high parenting stress and caregiving burden prior to COVID-19, as was the case in many NGC families. Indeed, the negative impact of COVID-19 lockdowns on caregivers is well-established, with reports that stress increased (Lam et al., 2024), health-related quality of life was poor (Fuerboeter et al., 2021; Rihm et al., 2022), and pre-pandemic challenges were exacerbated (Martin et al., 2023). The negative experiences of caregivers motivated the U.S. Surgeon General to release a full report on “Parents Under Pressure,” highlighting parent well-being as a critical public health priority (Murthy, 2024). Unfortunately, even prior to COVID-19, caregivers often reported social isolation and lacked the appropriate supports and resources to support their own health and well-being (Boettcher et al., 2021). The fragmented infrastructure supporting both caregivers and their children with rare conditions was taxed even further during the pandemic, as stay at home orders, access to in-person services, and infection risks further complicated the existing landscape of care (Kowanda et al., 2021; Martin et al., 2023). Pandemic-related service interruptions placed additional demands on caregivers, requiring them to assume greater responsibility in managing their children’s educational, vocational, and functional needs (Martin et al., 2023). Indeed, data collected across the pandemic largely confirmed expectations that NGC families experienced serious risk factors for both child and caregiver well-being.

However, the experiences of the pandemic also varied substantially across families depending on their unique characteristics and experiences, some with potentially protective effects. For example, in the context of child challenging behaviors, it is possible that lockdowns created a more predictable and manageable environment for some children with disabilities, potentially easing environmental triggers that contribute to challenging behaviors. Similarly, caregivers entered the pandemic with prior lived experience managing unexpected change, isolation, and access barriers. For example, families of children who had undergone hematopoietic stem cell transplant and associated prolonged isolation felt uniquely prepared to manage COVID-related isolation by drawing on prior experiences and learned coping skills (Chardon et al., 2021). Indeed, the experiences of families tend to encompass both risk and resilience; however, little is known about how such experiences unfolded during the pandemic and subsequently, how their child’s behaviors were impacted.

The present study leverages a cross-syndrome framework to examine the differential pandemic-related changes in challenging behaviors across families with and without NGCs. Methodologically, comparing profiles across syndromes – rather than examining NGCs in isolation, or relative to “typically developing” controls – permits a more nuanced understanding of variability within NGCs for several reasons. First, this approach allows researchers to identify factors that might impact many children with disabilities due to similar patterns emerging across several genetic subgroups, distinct from patterns that only emerge for specific genetic diagnoses. These comparisons can inform both the genetic etiology of challenging behaviors, as well as whether broad or syndrome-specific translational approaches are needed. Second, examining challenging behaviors in these groups can provide a framework for characterizing genetic etiology in tandem with an extreme environmental stressor (the COVID-19 pandemic), identifying patterns of risk and resilience that can inform broader public health approaches to disaster relief and emergency support. Together, these comparisons also facilitate translational efforts to design specific interventions and supports by differentiating experiences that impact many NGC subgroups (signaling broad translational relevance) from those impacting smaller groups uniquely (signaling local relevance and need for NGC-specific studies).

Here, we focus on four rare NGCs with established clinical phenotypes associated with intellectual and developmental disability. Children diagnosed with Angelman syndrome (AS) tend to have significant cognitive impairment and limited language skills. Early behaviors are often bound by limited motor skills; the majority eventually walk but often rely on assistive support (Horsler & Oliver, 2006). Children with fragile X syndrome (FXS) are often hyporesponsive as infants but hyperresponsive in toddlerhood. Anxiety and symptoms typically associated with autism are prevalent, though the phenotype varies by sex due to x-linked inactivation (Hunter et al., 2014). In Prader-Willi syndrome, children tend to demonstrate an uptick in challenging behaviors that coincide with the presentation of hyperphagia in early childhood, yet are often described as ‘typically’ behaving in infancy and toddlerhood; in later childhood, aggression and obsessive-compulsive behaviors often develop as well (Godler et al., 2022; Smith et al., 2003). The most salient challenge for children diagnosed with Williams syndrome (WS) tends to be social anxiety, and individuals tend to exhibit lower rates of aggression and fewer symptoms associated with autism. Children also tend to have stronger expressive than receptive language (Strømme et al., 2002). Groups differ significantly in their developmental skills, ranging from severe care needs in Angelman, moderate needs in Prader-Willi, and mild to moderate needs in fragile X and Williams syndrome; this variability facilitates understanding of how challenging behaviors might emerge secondary to other developmental functions.

Importantly to the present study, these groups also differ in frequency and severity of challenging behaviors. In general, across these groups, early infant temperamental profiles do not cleanly map onto later behavioral phenotypes, with complex emergence of features and patterns across development. When comparing community priorities for treatment across groups, internalizing problems such as anxiety are reported as a shared, common concern (Tsai et al., 2018; Weber et al., 2019; Willgoss et al., 2021). Although the manifestation of anxiety tends to vary across groups, often presenting as specific phobias in Williams syndrome (Pitts et al., 2016) or social avoidance in fragile X (Lozano et al., 2022), Prader-Willi communities also report concerns related to externalizing features such as opposition and aggression, which sometimes occur in conjunction with hyperphagia, a debilitating persistent hunger that presents in many people with Prader-Willi that can motivate extreme food seeking behaviors (Tsai et al., 2018). Thus, while some behavioral challenges are shared across groups, the nature and severity of features vary substantially.

The present study examined the long-term trajectories of young children with and without neurogenetic conditions whose caregivers reported on challenging behaviors before and after the COVID-19 pandemic. Our research questions are: (1) How did pre-pandemic family features distinguish challenging behavior trajectories among children with NGCs? We hypothesized that less access to support services and poorer caregiver mental health would contribute to greater child challenging behaviors and steeper increases in these behaviors over time, particularly in response to the COVID-19 pandemic due to loss of resources and educational support. (2) How did challenging behaviors vary across NGCs, controlling for these pre-pandemic features? We expected challenging behavior patterns to differ across syndromes, consistent with established behavioral phenotypes. However, these phenotypes tend to be based on early and late childhood features, which we expected to emerge differently across groups. For example, we expected challenging behaviors would generally increase over time in each syndrome group, but for fragile X and Williams syndrome groups to increase in internalizing symptoms, and Prader-Willi groups to increase in both internalizing and externalizing symptoms. We expected generally lower levels of challenging behaviors in Williams syndrome and Angelman syndrome, in part reflecting severe motor delays in Angelman syndrome.

Methods

Participants

Data were drawn from the Purdue Early Phenotype Study (PEPS; Neo et al., 2021), where participants represent a variety of low-incidence syndrome communities with elevated risk for challenging behaviors, including Angelman syndrome, fragile X syndrome, Prader-Willi syndrome, Williams syndrome, and a comparison group of low-risk controls (LRC). Participants were the child’s primary caregiver and served as proxy reporters for children ages 1 month to 10 years old. Children were required to be under three years old upon their first assessment for purposes of the broader PEPS study, with the youngest child enrolled at one month old. From the first assessment on, children are followed indefinitely, with the oldest child assessed in this sample being ten years old. Caregivers were predominantly white, non-Hispanic, and highly educated. Additional demographic information for both caregiver and child are presented in Table 1. Families in the United States were recruited via online methods, including virtual parent support groups, social media, and syndrome-specific listservs. Any participants who did not provide pre-pandemic data were excluded from the present study.

Table 1.

Demographic characteristics by syndrome group.

Characteristic	WS N = 521	AS N = 641	PWS N = 291	FXS N = 231	LRC N = 601	Overall N = 2281	p-value2
Child Age	3.73 (1.32)	3.56 (1.14)	3.51 (1.10)	3.70 (1.27)	3.91 (1.26)	3.70 (1.22)	0.431
Child Sex							0.060
Male	28 (54%)	32 (50%)	13 (45%)	19 (83%)	38 (63%)	130 (57%)
Female	24 (46%)	32 (50%)	16 (55%)	4 (17%)	22 (37%)	98 (43%)
Child Race							0.233
White	43 (83%)	55 (86%)	24 (83%)	19 (83%)	57 (95%)	198 (87%)
BIPOC	9 (17%)	9 (14%)	5 (17%)	4 (17%)	3 (5.0%)	30 (13%)
Caregiver Education							0.106
2 year degree or less	7 (14%)	15 (24%)	4 (15%)	5 (23%)	4 (6.7%)	35 (16%)
4+ year degree	43 (86%)	47 (76%)	23 (85%)	17 (77%)	56 (93%)	186 (84%)
Income	105 (81)	116 (80)	103 (50)	105 (43)	115 (63)	111 (69)	0.700
Depression	2.92 (3.40)	2.32 (2.53)	2.55 (2.72)	4.38 (3.65)	2.68 (2.79)	2.77 (2.98)	0.153
Anxiety	2.12 (2.77)	1.84 (2.84)	2.00 (2.30)	3.33 (2.35)	1.63 (1.97)	2.01 (2.52)	0.048
Stress	5.2 (3.2)	5.5 (3.6)	5.4 (3.9)	7.8 (3.8)	6.2 (4.2)	5.8 (3.8)	0.089
Parenting Stress	48 (26)	44 (20)	30 (17)	47 (20)	74 (17)	51 (25)	<0.001
Early Intervention Hours	2.9 (6.7)	5.9 (9.9)	3.1 (4.9)	3.8 (9.3)	0.5 (3.2)	3.2 (7.4)	<0.001
Developmental Quotient	69 (13)	52 (9)	73 (9)	68 (17)	95 (13)	71 (20)	<0.001
CBCL Internalizing t-score	52.84 (9.73)	53.44 (8.92)	46.1 (8.28)	55.32 (9.5)	44.25 (9.19)	48.79 (10.19)	<0.001
CBCL Externalizing t-score	53.89 (12.63)	50.94 (10.16)	45.41 (8.72)	52.80 (11.78)	45.75 (8.56)	48.59 (10.83)	<0.001
CBCL Total t-score	55.31 (11.19)	55.02 (9.46)	46.83 (8.18)	55.09 (10.79)	45.18 (8.81)	50.01 (10.77)	<0.001

¹Mean (SD); n (%).

²Kruskal-Wallis rank sum test; Pearson’s Chi-squared test; Fisher’s exact test. The Benjamini-Hochberg (BH) false discovery rate (FDR) correction (Benjamini & Hochberg, 1995) was applied to correct for multiple comparisons.

Procedure

The PEPS study began collecting data in 2017, with the goal of providing a convenient mechanism for families across the United States to document the development of their child with a rare condition. This project largely relied on a flexible convenience sample approach, whereby any family interested in contributing data could complete surveys. Families were prompted to contribute surveys every 6 months (if under 3 years) or every 12 months (if older than 3 years). They were not formally asked to complete a specific number of assessments and were permitted to skip assessments if inconvenient to their family, with the understanding that they could also return to the project at any time. Families were recruited through social media and word of mouth, and through secondary enrollment via other funded projects; for example, Angelman and fragile X families were often co-enrolled in a telehealth-based study of early development that enrolled families from 2017–2023. In contrast to many natural history studies, the PEPS study uniquely leveraged a cross-syndrome approach, with the same questionnaires given to children across a variety of rare conditions. The PEPS study also did not have a specific geographic or clinic affiliation, allowing for broader geographic diversity of participants.

Upon initial enrollment and then at 6 or 12-month intervals, caregivers were emailed a link to complete a set of developmentally appropriate online questionnaires. At each timepoint, caregivers were also asked to complete a brief phone interview that included questions regarding the child’s skills on a day-to-day-basis. Caregivers were compensated $15 for each assessment and $10 for each phone interview via Amazon gift cards.

Measures

Demographic Variables.

Caregivers reported their child’s date of birth, gender, and hours per week of early intervention services. Caregivers also reported their own education level and annual household income.

Child Challenging Behaviors.

The Child Behavior Checklist for Ages 1½-5 (CBCL; Achenbach & Rescorla, 2000) is a 100-item caregiver-report measure of challenging behaviors in children as part of the Achenbach System of Empirically Based Assessment. The CBCL measures total problems, internalizing problems, and externalizing problems across 6 subscales. The CBCL had strong internal consistency across total, internalizing, and externalizing subscales in an overlapping sample of children with rare neurogenetic conditions (α=.82-.94; Neo et al., 2021). We have previously reported cross-sectional challenging behavior profiles in a smaller subset of PEPS participants (Neo & Tonnsen, 2019). Here, we focus on total, internalizing, and externalizing raw scores in an expanded, longitudinal sample. Of note, although some participants in the sample aged past 5 years, items on the older version of the CBCL were not appropriate for the developmental ages of the children in the sample. Given our focus on changes in specific behaviors across development, we elected to continue administering the CBCL 1.5–5 “out of range,” focusing on changes in raw scores rather than standardized score conversions that are age- or sex-dependent. Given the conversion of raw to T-scores on the CBCL does not vary by age (e.g. the same raw score corresponds to the same T-score, regardless of child age), this approach allowed us to maintain maximum variability in scores while examining within-person changes in symptoms over time, aligning with our analytic plan. Since this project was started over a decade ago, some progress has been made in developing questionnaires that are chronologically and developmentally appropriate for rare conditions; however, measures for cross-syndrome cohorts of very young children are still lacking.

Child Adaptive Behaviors.

The Vineland Adaptive Behavior Scale – Third Edition Interview Form (VABS-3; Sparrow et al., 2016) is a comprehensive semi-structured caregiver interview of the adaptive behavior of individuals from ages 0–90+. The VABS-3 produces an Adaptive Behavior Composite score made up of communication, daily living skills, socialization, and motor skills subscales, each made up of individual subdomains. The subscales and ABC have been shown to have strong internal consistency in a normative sample (α=.94-.99; Pepperdine & McCrimmon, 2018). Here, we computed a developmental quotient for each child by dividing their average age equivalents across domain by their observed age at assessment. This age equivalent preserved variability in the VABS-3 scores while accounting for the child’s age, enabling adaptive skills to be included in latent class analyses as a single variable.

Caregiver Mental Health.

The Depression Anxiety and Stress Scales–21 items (DASS-21; Lovibond & Lovibond, 1995) is a self-report measure of the severity of symptoms of depression, anxiety, and psychological distress over the past week. Subscales of depression, anxiety, and stress are measured by 7 items each, scored on a 4-point Likert scale from 0 (did not apply to me at all) to 3 (applied to me very much, or most of the time). The DASS-21 total score and subscales have strong internal consistency in a sample of caregivers of children with autism (α=.83-.94; Lai et al., 2015). We examined the raw score from each subscale in analyses.

Parenting Stress.

The Parenting Stress Index–Short Form (PSI-SF; Abidin, 1995) is a 36-item self-report measure used to assess perceived stress experienced by parents. Subscales include parental distress, parent-child dysfunctional interaction, and difficult child. Participants respond on a 5-point Likert scale from 1 (strongly agree) to 5 (strongly disagree). PSI-SF has been shown to have strong internal consistency in a sample of caregivers of children with autism (α=.93; Lai et al., 2015). We examined the PSI-SF total raw score in analyses.

Analysis Plan

All analyses were performed using R Statistical Software (version 4.4.1 or greater). A significance level of 5% (α = 0.05) was used for all analyses. Participant characteristics were summarized using descriptive statistics, as appropriate. All continuous variables were assessed for normality. No adjustments were needed based on results of these preliminary analyses. Within subscales, missing data was addressed using mean imputation within CBCL, DASS-21, and PSI-SF subscales. Remaining missingness resulted in observation deletion for the specific analysis. Mixed effects models centered age at the sample mean across all observations (3.3 years) and probed both linear and quadratic age effects; both linear and quadratic effects (Age²) and relevant interactions were retained.

Predicting COVID Challenging Behaviors from Latent Class.

Our first research question was the degree to which changes in challenging behavior changed across the COVID-19 pandemic varied according to families’ pre-pandemic characteristics. To address this question, we conducted a latent class analysis – based on each family’s most recent assessment prior to the pandemic – that included demographic variables (child sex, child race, caregiver race, income, education), caregiver clinical features (depression, anxiety, stress, parenting stress), and child clinical features (child developmental quotient, early intervention hours). The low risk control group was constrained to a singular class to focus comparisons on NGC group experiences. The goal of this analysis was to define key profiles of family experiences prior to COVD-19.

The latent class analyses (LCA) used cross-sectional data using the observation prior and closest to the COVID-19 lockdown date of March 11th, 2020. Across the sample, children had between 1 and 7 outcome observations (Median=2, Mean=2.6). An observation prior to the pandemic onset date was always used unless a child had no pre-COVID observations, in which case the next available observation was used. LCA models with 2 to 5 factors were fit using Mplus version 8 (Muthén & Muthén, 2017). Child age, child sex, child race, and caregiver education were specified as categorical variables, and genetic diagnosis group was specified as a nominal variable. The remaining variables were specified as continuous LCA indicators. The Bayesian Information Criterion (BIC) was used to select the best fitting model. If BIC values were within 1% of each other, the solution with the smaller number of classes was retained in preference for parsimony. Following LCA models, control group was re-integrated into the dataset.

We next examined the association between latent class membership and child challenging behaviors, including in response to the COVID-19 pandemic. Here, we used longitudinal linear mixed effects models that account for nested data structure to predict challenging behavior trajectories (total, internalizing, and externalizing) across latent class groups. Predictors included age (Age, Age²), the main effect of COVID-19 lockdowns (Post versus Pre), latent class (1, 2, 3, LRC), and the interaction between these variables (COVID × Class, Age × Class, Age² × Class, Age² × COVID); with comparisons referenced to the LRC group. Longitudinal data used the generalized additive model for location, shape, and scale (GAMLSS; Stasinopoulos et al., 2017) implemented in the PersonAlytics (Tueller et al., 2023, 2024) R package (Team R, 2012). This approach implements generalized linear mixed effects models with over 70 distribution options. If multiple distributions were good candidates for the outcome distribution, distribution was selected using BIC.

Predicting COVID Challenging Behaviors from Syndrome Group.

To examine whether syndrome groups varied in challenging behavior and COVID impacts, we repeated mixed effects models, now predicting the same challenging behavior trajectories (total, internalizing, and externalizing) from age (Age, Age²), the main effect of COVID-19 lockdowns (Post versus Pre), syndrome group (AS, FXS, PWS, WS), and the interaction between these variables (COVID × Syndrome, Age × Syndrome, Age² × Syndrome, Age² × COVID), with comparisons referenced to the LRC group. To allow for comparisons with the latent class models, we controlled for pre-pandemic predictors, including demographic variables (child sex, child race, income, education), caregiver clinical features (depression, anxiety, stress, parenting stress), and child clinical features (child developmental quotient, early intervention hours).

Results

In total, 228 participants contributed to the model, with 687 challenging behavior ratings across groups (AS=176, FXS=44, PWS=117, WS=151, LRC=199). Descriptive data are presented in Table 1, reflecting the assessment for each participant closest to but preceding pandemic onset. Based on initial distributions, categorical variables were collapsed to ease model interpretation, including by reducing race/ethnicity and education to two groups (white and non-Hispanic, non-white and/or non-Hispanic; 2 year degree or less versus 4 year degree or more).

Establishing Latent Classes.

Table 2 details final results of the LCA models. Although a 5-class model (4 NGCs classes) provided the best fit (BIC= 9217.8), this solution included a very small fifth class (n=11) and was only marginally better fit than the 4-class model (BIC=9219.8), which was therefore retained for parsimony over the poorer-fitting 3-class model (BIC=9386.4). Variable contributions by class are summarized in Table 2 and visually depicted in Figure 1; here, all variables were standardized to a 0–1 scale to ease interpretability.

Table 2.

Demographic and clinical features across latent classes.

Characteristic	LCR (Class 1) N = 621	HCR+LEI (Class 2) N = 811	HCR+HEI (Class 3) N = 251	Low Risk Controls (Class 4) N = 601	Overall N = 2281	p-value2
Williams	19 (31%)	27 (33%)	6 (24%)	0 (0%)	52 (23%)	<0.001
Angelman	29 (47%)	24 (30%)	11 (44%)	0 (0%)	64 (28%)	<0.001
Prader-Willi	13 (21%)	10 (12%)	6 (24%)	0 (0%)	29 (13%)	<0.001
Fragile X	1 (1.6%)	20 (25%)	2 (8.0%)	0 (0%)	23 (10%)	<0.001
Low Risk Controls	0 (0%)	0 (0%)	0 (0%)	60 (100%)	60 (26%)	<0.001
Child Age	3.39 (1.17)	3.78 (1.23)	3.68 (1.13)	3.91 (1.26)	3.70 (1.22)	0.128
Child Sex						0.600
Male	32 (52%)	45 (56%)	15 (60%)	38 (63%)	130 (57%)
Female	30 (48%)	36 (44%)	10 (40%)	22 (37%)	98 (43%)
Child Race
White	52 (84%)	71 (88%)	18 (72%)	57 (95%)	198 (87%)	0.033
BIPOC	10 (16%)	10 (12%)	7 (28%)	3 (5.0%)	30 (13%)	0.033
More Than One Race	0 (0%)	0 (0%)	0 (0%)	0 (0%)	0 (0%)
Caregiver Education						0.005
2 year degree or less	5 (8.8%)	21 (27%)	5 (20%)	4 (6.7%)	35 (16%)
4+ year degree	52 (91%)	58 (73%)	20 (80%)	56 (93%)	186 (84%)
Income	134 (73)	83 (40)	135 (111)	115 (63)	111 (69)	<0.001
Depression	0.66 (0.77)	4.18 (3.23)	3.83 (3.04)	2.68 (2.79)	2.77 (2.98)	<0.001
Anxiety	0.18 (0.39)	3.44 (2.83)	2.91 (2.64)	1.63 (1.97)	2.01 (2.52)	<0.001
Stress	3.0 (2.4)	7.5 (2.9)	6.8 (4.3)	6.2 (4.2)	5.8 (3.8)	<0.001
Parenting Stress	33 (17)	50 (23)	46 (20)	74 (17)	51 (25)	<0.001
Early Intervention Hours	3.6 (5.4)	0.6 (1.0)	17.5 (12.8)	0.5 (3.2)	3.2 (7.4)	<0.001
Adaptive Behaviors	63 (15)	62 (14)	64 (17)	95 (13)	71 (20)	<0.001

¹n (%); Mean (SD).

²Pearson’s Chi-squared test; Fisher’s exact test; Kruskal-Wallis rank sum test; The Benjamini-Hochberg (BH) false discovery rate (FDR) correction (Benjamini & Hochberg, 1995) was applied to correct for multiple comparisons.

Figure 1.

Latent class membership by pre-pandemic demographic and clinical predictors.

The three NGC classes were distinguished by a number of demographic, clinical, and contextual variables. Class 1 (n=62) was, on average, defined by high caregiver education, low caregiver mental health symptoms (depression, anxiety, stress), younger child age, present but low intervention uptake, and low parenting stress; we refer to this class as “Low Clinical Risk” (LCR). Class 2, the largest class (n=81) was defined by lower caregiver education and income, low child early intervention uptake, and higher caregiver depression, anxiety, stress, and parenting stress; we refer to this class as “High Clinical Risk and Low Early Intervention” (HCR+LEI). The final and smallest (n=25) NGC class, Class 3, was defined by greater racial and ethnic diversity, greater access to early intervention, and moderately elevated caregiver depression, anxiety, stress, and parenting stress; we refer to this class as “High Clinical Risk and High Early Intervention” (HCR+HEI)

Representation across NGCs varied by group. Nearly all fragile X families fell into the HCR+LEI group (n=20 of 23). The remaining groups were more evenly balanced between the LCR and HCR+LEI groups, with a minority also falling into the HCR+HEI group. Among the three NGC latent classes, child age, sex, and developmental quotient did not differ; the significant p-value in Table 2 reflects differences between the NGC classes and the low-risk control class (LRC). Relative to the NGC classes, the constrained low risk control class was characterized by lower racial and ethnic diversity and moderate maternal mental health symptoms (higher than LCR, but lower than the HCR classes); as expected, child developmental quotient and early intervention uptake was higher and lower, respectively, reflecting the inclusion criteria for this group.

Challenging Behaviors by Latent Class.

We next used mixed effects models to examine patterns of challenging behaviors by class, across COVID-19 pandemic. Unconditional models indicated a significant proportion of variability in challenging behaviors, supporting the nested mixed effects modeling approach. Model results for each outcome (internalizing, externalizing, total) are presented in Table 3.

Table 3.

Results of mixed effects models predicting challenging behavior from latent class of pre-COVID experiences.

Predictor	Internalizing	Externalizing	Total
Child Age	0.84 (p = 0.02)	0.13 (p = 0.74)	0.86 (p = 0.36)
Child Age2	−0.22 (p = 0.27)	−0.46 (p = 0.08)	−1.26 (p = 0.03)
COVID (ref: Pre-COVID)	2.03 (p = 0.06)	0.85 (p = 0.53)	2.59 (p = 0.40)
Class 1 (ref: LRC)	1.73 (p = 0.13)	1.01 (p = 0.50)	6.54 (p = 0.08)
Class 2 (ref: LRC)	5.96 (p = 0.00)	7.85 (p = 0.00)	23.84 (p = 0.00)
Class 3 (ref: LRC)	5.25 (p = 0.00)	6.85 (p = 0.00)	21.08 (p = 0.00)
COVID * Class 1	−2.92 (p = 0.03)	−0.35 (p = 0.83)	−2.82 (p = 0.46)
COVID * Class 2	−2.19 (p = 0.08)	−1.27 (p = 0.42)	−4.44 (p = 0.21)
COVID * Class 3	1.55 (p = 0.35)	0.14 (p = 0.95)	5.96 (p = 0.22)
Child Age * Class 1	0.01 (p = 0.99)	1.54 (p = 0.01)	2.56 (p = 0.08)
Child Age * Class 2	0.20 (p = 0.69)	2.32 (p = 0.00)	3.80 (p = 0.00)
Child Age * Class 3	0.00 (p = 0.99)	2.52 (p = 0.00)	3.68 (p = 0.03)
Child Age2 * Class 1	0.64 (p = 0.03)	−0.21 (p = 0.58)	0.80 (p = 0.34)
Child Age2 * Class 2	0.21 (p = 0.42)	−0.49 (p = 0.14)	−0.12 (p = 0.87)
Child Age2 * Class 3	−0.11 (p = 0.75)	−0.67 (p = 0.12)	−1.25 (p = 0.19)
Child Age * COVID	−1.28 (p = 0.02)	−0.30 (p = 0.59)	−2.54 (p = 0.09)
Child Age2 * COVID	0.11 (p = 0.70)	0.02 (p = 0.95)	0.59 (p = 0.44)

In the low risk control group (reference group), total challenging behaviors increased then decreased around the centered age of 3.3 years, as indicated by a significant negative quadratic term. Although age-related changes in externalizing behaviors were negligible, a one-year increase in child age was associated with a .84 point increase in internalizing scores. The COVID-19 pandemic was associated with marginally higher internalizing symptoms (p=.060) and slower linear changes in internalizing symptoms over time. However, COVID-19 was not associated with externalizing or total challenging behaviors.

Patterns in the syndromic classes varied relative to the low risk control reference group. For the Low Clinical Risk (LCR, Class 1) group, internalizing behaviors increased at a similar rate to low risk controls but demonstrated a positive quadratic effect, whereby their predicted trajectory started higher than controls at 1 year of age, decreased to be similar to controls by age 3.3 years, and increased to be 6 points higher than the low risk control group by age 6 years. This group also demonstrated steeper linear increases in externalizing behaviors over time but did not significantly differ from the low risk control group in total challenging behavior scores. Relative to patterns observed in the control reference group, for Class 1, the onset of COVID-19 was associated with lower levels and slower increases in internalizing behaviors over time; the additional comparisons did not differ from the control reference group.

Patterns in the High Clinical Risk groups (Class 2, Class 3) were similar to each other; these groups exhibited higher internalizing, externalizing, and total challenging behaviors relative to the control reference group by 5–8 points (21–24 points for total score), and both externalizing and total challenging behaviors and also increased at a steeper linear rate. Of note, Class 2 demonstrated marginal reductions in internalizing symptoms after COVID-19 onset, similar direction to Class 1 (p=.08). However, no significant quadratic age effects were detected for any outcome variable, and Classes 2 and 3 did not statistically differ from the reference group before and after COVID-19.

Challenging Behaviors by Syndrome Group.

We next repeated the mixed effects models to examine the degree to which syndromic groups exhibited different challenging behaviors across the pandemic, controlling for age and the demographic and clinical factors embedded in the previous latent class models.

Results of these models are depicted in Table 5. Internalizing, externalizing, and total scores tended to increase over time, although only the internalizing slope approached significance and no quadratic effects were detected. The COVID-19 pandemic was associated with marginally higher internalizing symptoms (p=.070) but was not associated with externalizing or total challenging behaviors.

Table 5.

Results of mixed effects models predicting challenging behavior from syndrome group and pre-COVID experiences

Predictor	Internalizing	Externalizing	Total
Child Age	0.67 (p = 0.08)	0.19 (p = 0.68)	0.56 (p = 0.60)
Child Age2	−0.35 (p = 0.12)	−0.20 (p = 0.49)	−1.02 (p = 0.12)
COVID (ref: Pre-COVID)	2.09 (p = 0.07)	0.71 (p = 0.63)	2.77 (p = 0.41)
WS (ref: LRC)	0.80 (p = 0.56)	4.76 (p = 0.01)	9.35 (p = 0.03)
AS (ref: LRC)	−0.46 (p = 0.79)	0.70 (p = 0.76)	1.59 (p = 0.77)
PWS (ref: LRC)	−2.84 (p = 0.05)	−1.98 (p = 0.33)	−6.37 (p = 0.18)
FXS (ref: LRC)	0.47 (p = 0.82)	1.25 (p = 0.66)	−0.08 (p = 0.99)
COVID * WS	−1.39 (p = 0.37)	−1.34 (p = 0.50)	−4.21 (p = 0.35)
COVID * AS	−2.30 (p = 0.09)	−1.65 (p = 0.35)	−3.70 (p = 0.36)
COVID * PWS	−0.32 (p = 0.85)	0.96 (p = 0.66)	1.34 (p = 0.79)
COVID * FXS	−0.47 (p = 0.85)	2.44 (p = 0.45)	7.72 (p = 0.29)
Child Age * WS	1.20 (p = 0.03)	2.43 (p = 0.00)	5.62 (p = 0.00)
Child Age * AS	−0.62 (p = 0.28)	1.89 (p = 0.01)	2.43 (p = 0.13)
Child Age * PWS	0.45 (p = 0.49)	1.51 (p = 0.07)	3.46 (p = 0.06)
Child Age * FXS	−2.16 (p = 0.07)	−0.12 (p = 0.93)	−3.43 (p = 0.31)
Child Age2 * WS	0.13 (p = 0.67)	−1.10 (p = 0.01)	−1.27 (p = 0.17)
Child Age2 * AS	0.76 (p = 0.02)	0.26 (p = 0.52)	1.62 (p = 0.08)
Child Age2 * PWS	0.68 (p = 0.06)	−0.34 (p = 0.45)	0.40 (p = 0.70)
Child Age2 * FXS	1.05 (p = 0.09)	−0.03 (p = 0.97)	1.92 (p = 0.28)
Child Age * COVID	−0.60 (p = 0.36)	−0.42 (p = 0.57)	−1.88 (p = 0.30)
Child Age2 * COVID	−0.29 (p = 0.37)	−0.08 (p = 0.83)	−0.06 (p = 0.95)

Note: Models control for demographic variables (child sex, child race, household income, caregiver education), caregiver clinical features (depression, anxiety, stress, parenting stress), and child clinical features (child developmental quotient, early intervention hours). Full table is presented in supplement as Supplemental Table 1.

Patterns in the syndrome groups varied relative to the reference group. Children with WS had higher externalizing behaviors relative to the low risk control reference group, increasing approximately 2.5 points per year faster; however, a significant negative quadratic effect indicated a plateau in externalizing behaviors over time. Total behaviors in WS were also higher and increased faster than the reference group. Children with AS had similar behaviors across subscales to the reference group, however, significant quadratic effects suggested reductions and then increases in internalizing behaviors from age 1 to age 6, and increasing externalizing behaviors at a rate of 1.89 points per year. Children with PWS had similar trajectories in externalizing and total behaviors compared to the reference group, and lower internalizing symptoms at age 3.3 (p=0.05) which did not differ with time. In children with FXS, both levels and trajectories of behaviors across subscales were similar to the reference group. Relative to controls and controlling for age, COVID was not associated with significant changes or trajectories in challenging behavior scores across syndrome groups.

Figure 2 depicts changes in total challenging behaviors by group, before and after the pandemic. Here, challenging behavior scores are averaged by COVID period (pre- and post-) and age group (binned by year), and plotted separately by syndromic group.

Figure 2.

Total challenging behaviors by age and syndrome groups, before and after the COVID-19 pandemic

Discussion

The present study examined predictors of challenging behaviors across the COVID-19 pandemic among children with and without neurogenetic conditions. Overall, latent class analysis identified distinct pre-pandemic family profiles based on demographic, caregiver mental health, child development, and service-related factors. Mixed effects models indicated that challenging behaviors increased for many children following COVID-19 lockdowns, with variation in trajectories across latent class membership and syndrome groups. Specifically, children in higher-risk latent classes (e.g., lower caregiver education, higher stress, low intervention uptake) showed higher levels and steeper increases in internalizing challenging behaviors compared to lower-risk classes. However, the pandemic’s impact on challenging behaviors in general was strongest for younger children and specific to internalizing behaviors. Mixed effects analyses across syndrome groups paralleled these findings, revealing elevated challenging behaviors across most syndrome groups, with the exception of fragile X. Together, these results suggest a complex layering of COVID-19 impacts over the dynamic early trajectories of challenging behaviors among rare NGC families. These findings highlight the combined influence of pre-pandemic family context and diagnostic group on behavioral responses to pandemic-related disruptions.

A key finding of this study was that challenging behaviors were already increasing prior to the pandemic in many syndrome groups, and across the study, behaviors continued to increase but at a somewhat slower rate of change. These findings were in line with our expectations that behavioral patterns would differ across syndromes, however, the ages at which they were surveyed during COVID-19 (primarily the preschool period) is a time of particular fluctuation in challenging behavior profiles among NGCs. The observed patterns (i.e., high externalizing behaviors in WS, limited differences in FXS compared to controls) differed from expectations based on later childhood features but were consistent with varied rates of change in challenging behaviors observed in many NGCs; for example, children with Prader-Willi and fragile X often experience periods of lower problem behaviors in early childhood, followed by upticks in challenging behaviors in later childhood (Neo et al., 2021; Zelenova et al., 2018). It is particularly important to note that the cohort of children in this study aged alongside the unfolding of the pandemic. Given PEPS is a first-of-its kind study, comparing these findings to past cohorts of young children with NGCs is also not possible, and dramatic changes in early therapeutics for NGCs complicate efforts to compare phenotypes across cohorts. For example, growth hormones are now nearly universally used to support development in young children with Prader-Willi but were not approved in the United States until 2000 (Deal et al., 2013). However, within the current cohort, comparisons of similarly aged PEPS participants before and after COVID suggest that the pandemic did amplify challenging behaviors among many children with NGCs, and future follow-up studies will be important in examining the long-term stability of these patterns, including among children who were born after the COVID-19 pandemic.

We also observed several meaningful predictors of challenging behavior trajectories. Within our latent class analyses, a substantial minority of families fell into the low clinical risk class (LCR; Class 1), which was characterized by high caregiver education, fewer caregiver mental health symptoms (depression, anxiety, stress), younger child age, present but low intervention uptake, and low parenting stress. We conceptualized this class as demonstrating lower clinical risks, and it is notable that a substantial proportion of NGC families aligned with this group. However, it is important to note that this class may not be representative of the broader NGC population, as the defining features of this class indicate a particularly affluent group. Of note, many families of children with Angelman syndrome also fell into this group, which may reflect a particular sampling bias in the Angelman community that should be explored in future work.

Importantly, the identified classes were not differentiated by child sex, age, or developmental skills, and were instead primarily differentiated by caregiver mental health and intervention uptake. In line with our expectations and traditional diathesis stress models, we observed two classes that constituted higher clinical risks within the NGC groups, both associated with high caregiver mental health symptoms. This was unsurprising, given the known high prevalence of caregivers with mental health challenges (Boettcher et al., 2021) and the negative impact of pre-pandemic risk factors on children’s outcomes in other populations (Panjwani et al., 2021). A pertinent defining factor between the two high risk classes was the presence or absence of high early intervention service uptake. One subclass had low early intervention uptake (HCR+LEI; Class 2), while the other had higher early intervention uptake and also more racial and ethnic diversity (HCR+HEI; Class 3). Importantly, class membership accounted for significant variability in challenging behavior trajectories, over and above the impact of COVID, where both high-risk groups (Classes 2 and 3) exhibited higher internalizing behaviors, with the HCR+LEI (Class 2) group also exhibiting higher total challenging behaviors. Framed by diathesis stress models, findings confirm that children and families in higher risk classes based on pre-pandemic features experienced significantly more challenging behaviors across domains. With parent mental health as a defining feature of the high clinical risk classes, these data support the importance of monitoring caregiver mental health in the context of children’s challenging behaviors.

Notably, our latent class analysis also highlighted that access to early intervention services may have functioned as a protective factor for families, as early intervention was associated with a smaller increase in challenging behaviors among families with similar caregiver mental health profiles. In the context of the pandemic where service delivery was largely impacted (Martin et al., 2023), regaining access to services for families at risk for greater challenging behaviors could serve as an important protective factor. While encouraging that the group with higher intervention uptake also tended to report more demographic diversity, the persistent high mental health risks that co-occurred in that group suggests persistent clinical needs despite high intervention access. This underscores the importance of addressing caregiver mental health needs, even among families accessing substantial early intervention support.

Beyond class membership, we also probed patterns unique to the specific syndromes, with pertinent findings in the fragile X syndrome groups. Nearly all fragile X families fell into the high-risk class characterized by high clinical risks and low early intervention access (HCR+LEI; Class 2). Children with fragile X also exhibited similar challenging behaviors to low-risk controls, as well as stability in challenging behaviors over time. These findings are interesting for several reasons. First, the high clinical risk of fragile X families is notable and raises questions about the intersecting role of broader fragile X associated conditions (Hagerman & Hagerman, 2002), which often co-occur in fragile X families due to the heritable nature of fragile X. For example, many women who give birth to children with fragile X are themselves “carriers” of the FMR1 premutation, a genetic atypicality that, in women, is typically associated with higher rates of mood disorders, executive function deficits, pragmatic challenges, and medical issues such as early menopause. Premutation carrier mothers of children with fragile X may also be supporting loved ones with fragile X and FMR1 mutations, including aging parents experiencing conditions such as FX-associated tremor and ataxia syndrome (FXTAS), a neurodegenerative condition most common in males (Hagerman & Hagerman, 2002). Given these challenges, it is unsurprising that many fragile X families fell into the “high risk” class, despite more normative levels of child challenging behaviors than other NGC groups. Here, too, findings may have been impacted by the known sex differences in fragile X, as females often present with fewer challenging behaviors than males due to X inactivation (Hardiman & McGill, 2018); thus, the finding of children with fragile X having similar levels of behaviors to controls may be driven by sex effects that are unique to this group. Due to our smaller sample size, we were unable to probe meaningful effects of child sex, however, these limitations could be addressed in future work. It is important to continue studying the unique experiences of fragile X families given the persistent high clinical risks reported by caregivers in our sample.

Limitations and Future Directions

Despite constituting the largest cross-syndrome assessment of early challenging behaviors in NGCs to date, this study includes several notable limitations. First, the study involved a mostly affluent and relatively homogenous sample, and small subgroup sizes (e.g. few females with fragile X, few non-white participants) which tempered our ability to probe the unique experiences of demographic subgroups. Future work in larger, more diverse cohorts may be positioned to explore effects of sex on challenging behaviors, particularly in children with fragile X. The large clustering of diverse families in the HCR+HEI (Class 3) group should also be investigated in future research, as these families appear to be accessing early intervention services but also experiencing caregiver mental health challenges. It is possible that more integrated and tailored supports that address the caregiver’s needs alongside the child’s needs may be better suited for these families, particularly if support programs are designed with the needs of diverse caregivers in mind.

Given the challenging behavior profiles of NGC families are known to shift substantially and non-linearly across early development, additional research is needed to examine how challenging behaviors differ among cohorts of pandemic-impacted children, relative to children whose early childhood periods were not impacted by COVID-19. Importantly, our goals were to leverage the diverse ages and enrollment dates of our cohort to test whether COVID-related changes existed and how they may have differentially impacted groups, even though behaviors increased over time, too. In other words, we demonstrated that in most syndrome groups, children who were 3 years old prior to COVID demonstrated lower scores than children who were 3 years old after COVID. Here, the ‘messy’ nature of our sample was beneficial, as children within the same syndrome groups and ages contributed data both before and after COVID; we did not have a single, same-aged cohort matriculating through the pandemic at the same ages. It will also be important to compare trajectories of children with NGC-associated challenging behaviors to cohorts of children without NGCs, whose trajectories are not impacted by known genetic etiologies. Without such comparisons, it is not possible to definitively isolate the impact of COVID-19, above and beyond the expected developmental shifts in challenging behavior features over time.

As we have previously reported, very few participants in the sample demonstrated clinically elevated behaviors across groups (Neo et al., 2021). It is possible that NGC groups may have also been less likely to display certain behaviors due to low motor skills (e.g. less likely to hit) or cognitive and language abilities (e.g. less likely to argue). Indeed, the lack of early childhood challenging behavior inventories for NGC populations is a major limitation in the field (Budimirovic et al., 2017; Thurm et al., 2020), particularly when this study was started in 2017. As such, data is contextualized via controls but limited in interpretation to within-group patterns (e.g. whether WS phenotypes differed before versus after COVID) rather than to determine whether groups were ‘typical’ relative to controls. Similarly, thorough cross-syndrome comparisons will require substantial subscale and item-level analyses that were not possible within the scope of the current project; such comparisons represent a fruitful direction for future work.

Conclusion

In sum, the present study examined patterns and predictors of challenging behaviors among children with and without rare NGCs across the COVID-19 pandemic. The distinct risk profiles identified via latent class analysis underscored that pre-pandemic family context and diagnostic group together shaped behavioral responses to pandemic-related disruptions. This cross-syndrome approach provided valuable insights into how the pandemic differentially affected children with varying genetic vulnerabilities, cognitive abilities, and behavioral profiles – which would not have been plausible to detect via the study of any one diagnostic group alone. Findings underscore the importance of monitoring potential risk factors for worsening challenging behaviors – such as parent mental health and access to early intervention services – to better support families as their children with NGCs grow and develop over time. Although the present study identified within-cohort associations as a first step to informing the impact of COVID-19 on challenging behavior trajectories in NGC groups, additional cohort studies and continued follow-up of this and other COVID-19 impacted cohorts is important to identifying the full scope of COVID-19 related impacts, and identifying the specific supports necessary to aid families most affected.

Table 4.

Challenging behavior average raw scores by latent class (Upper Panel) and syndrome group (Lower Panel).

Class	Internalizing	Externalizing	Total
Controls	5.45	9.46	24.17
Low Clinical Risk (C1)	7.44	9.98	30.78
High Clinical Risk + Low EI (C2)	11.24	16.05	45.80
High Clinical Risk + High EI (C3)	9.94	14.35	41.77
Syndrome Group	Internalizing	Externalizing	Total
Angelman	10.57	13.75	43.03
Fragile X	11.91	15.39	43.59
Prader-Willi	6.12	9.33	26.81
Williams	10.48	15.90	44.06
Controls	5.45	9.53	24.24