Correlates of physical activity in children and adolescents with intellectual disabilities: a systematic review

Background Children and adolescents with intellectual disabilities (ID) participate in low levels of physical activity. To inform the development of interventions, we need to better understand factors associated with physical activity. The aim of this study was therefore to systematically review correlates of physical activity in children and adolescents with ID. Methods The review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. Ovid MEDLINE, Ovid Embase, Web of Science, ERIC, CINAHL and PsycINFO were searched between 1 January 1990 and 29 February 2020 to identify English-language studies, which examined correlates of free-living physical activity in children and adolescents ( 0 – 19 years) with ID. Study quality was assessed. Correlates were analysed using a narrative synthesis and classi ﬁ ed using the socioecological model as intrapersonal, interpersonal, organisational or environmental. Results Fifteen studies published between 2010 and 2019 met the inclusion criteria and were included in the review. Forty-eight individual correlates were identi ﬁ ed. Studies were predominantly focused on intrapersonal-level correlates. Of those correlates investigated in more than one study ( n = 6 ), having better motor development was positively associated with physical activity. Inconsistent results were found for age and cardiorespiratory ﬁ tness. Sex, percentage body fat and body mass index were not correlated. No interpersonal-level, organisational-level or environmental-level correlates were included in more than one study. Conclusions To date, we have limited and inconclusive evidence about correlates of physical activity in children and adolescents with ID. Only when future studies unravel correlates and determinants, across all domains of the socioecological model, will the potential opportunities to improve health by increasing physical activity levels be achievable.


Introduction
Intellectual disabilities (ID) are defined by significant limitations in both intellectual functioning and adaptive behaviour that onset before the age of 18 years (American Association on Intellectual and Developmental Disabilities 2020). Children and adolescents with ID experience many health inequalities in comparison with their typically developing peers, including higher rates of mental ill health, obesity, type 2 diabetes and other health conditions, which may persist into adulthood (Allerton et al. 2011;Einfeld et al. 2011;Maiano 2011). The severity of these health inequalities results in children and adolescents with ID having a life expectancy of up to 20 years less than people without ID (O'Leary et al. 2017). Therefore, it is essential that these health inequalities are addressed and methods to improve the health of children and adolescents with ID are identified and developed.
Supporting positive lifestyle changes in children and adolescents with ID is one potential method to reduce these inequalities (Emerson & Baines 2011). Specifically, physical activity (PA) has been shown to have numerous physical and mental health benefits in children and adolescents (Biddle et al. 2019). In order to gain clinically meaningful benefits, it is recommended that children and adolescents participate in an average of 60 min of moderateto-vigorous PA (MVPA) per day (Chief Medical Officers 2019). However, few children and adolescents with ID meet these PA guidelines but instead participate in less activity than their typically developing peers and are generally inactive (Boddy et al. 2015;Einarsson et al. 2015). Einarsson et al. (2015) reported that in a sample of children (n = 91), none achieved the weekly PA guidelines, whereas 40% of an age-matched and sex-matched comparison group of typically developing children met the guidelines. Therefore, increasing PA could help reduce the health inequalities of children and adolescents with ID (Maiano 2011).
Increasing PA through interventions is often used to promote a healthy and active lifestyle and to improve the health of children and adolescents (Biddle et al. 2015). However, a systematic review and meta-analysis found that existing interventions in children and adolescents with ID have been ineffective at improving PA levels . This systematic review suggested this was due to ineffective design, as interventions included structured activity sessions that were not sustainable post-intervention and were developed based on existing interventions for typically developing children. Generalising interventions designed for typically developing children and adolescents to children and adolescents with ID could be ineffective because of the different needs of these groups. For example, children and adolescents with ID have less autonomy and greater parental overprotection than their typically developing peers (Martin & Choi 2009;Downs et al. 2013). Having ID also increases the barriers to PA experienced by children and adolescents, such as reduced physical and social skills, parental concerns relating to bullying, and limited access to facilities and inclusive clubs .
Therefore, to increase PA levels in children and adolescents with ID, new interventions need to be developed to fit into their lives and to address barriers that they face. In accordance with the behavioural epidemiological framework and the Medical Research Council guidelines for Developing and Evaluating Complex Interventions, which inform intervention development, the first stage is understanding the evidence base, including factors that are associated with the behaviour of interest, that is, PA, and identifying a theory (Sallis et al. 2000;Craig et al. 2008). Multilevel interventions, such as those based on a socioecological model (Sallis et al. 2008), are considered to be essential in effecting behaviour change for promoting PA among children and adolescents (Mehtälä et al. 2014;Simon et al. 2014). According to the socioecological model, there are multiple levels of influence on PA, namely, intrapersonal (e.g. sex and motivation), interpersonal (e.g. friends support and family support) and community/environmental (e.g. equipment and PA programmes in the community) or organisational factors/resources (e.g. gyms and parks). Although correlates of PA have been extensively synthesised in both typically developing children and adolescents and children and adolescents with physical disabilities (Sterdt et al. 2014;Li et al. 2016), no studies have synthesised research relating to factors associated with PA in children and adolescents with ID, with existing reviews in this field focused primarily on synthesising qualitative evidence (Shields et al. 2012;Sterman et al. 2016;. Therefore, to address the gaps in the literature, this study aims to systematically review the existing evidence relating to correlates of PA in children and adolescents with ID using a socioecological approach.

Methods
This systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines (Moher et al. 2009). This review was registered with the International Prospective Register of Systematic Reviews (PROSPERO, registration number: CRD42019128417).

Search strategy
We searched Ovid MEDLINE, Ovid Embase, Web of Science Core collection, ERIC, CINAHL and PsycINFO from 1 January 1990 to 29 February 2020 using comprehensive terms related to 'ID', 'children or adolescents or youth' and 'PA' (full search strategies are presented as the Supporting Information). Age was limited in database searches where applicable and to preschool child, children and adolescents. In addition, a manual bibliography and citation search of included studies was conducted using Google Scholar.

Eligibility criteria
To be included in the review, studies had to include children and adolescents (0-19 years) and free-living PA, defined as activity conducted in the participants own environment and not as part of an intervention, as the outcome variable. All reported dimensions of PA were eligible for inclusion, for example, mode, frequency, duration and intensity of PA (Strath et al. 2013). PA intensities are categorised as follows: light [LPA; 1.6-2.9 metabolic equivalents of task (METs)], moderate intensity (MPA; 3.0-5.9 METs), vigorous (VPA; ≥6.0 METs) and MVPA (≥3.0 METs; Strath et al. 2013).
Correlates had to be investigated using one the following types of analyses: correlations, regression, analyses of covariance, multivariate analyses of covariance or chi-squared. For studies that included children and adolescents without ID or adults, at least 50% of participants had be aged ≤19 years and have ID, if results were not reported separately. Studies were also required to be full text, peer reviewed and published in English. Observational, cross-sectional, longitudinal (retrospective and prospective) and casecontrol study designs were eligible. Intervention studies were eligible for inclusion if relevant baseline data were available. Studies with small samples (<20 participants) or case series designs were excluded as these papers are less representative. Studies including children and young people with autism spectrum disorder (ASD) were excluded, unless the population had ID in combination with ASD. Studies that were laboratory based or where the outcome measure was structured exercise (including physical education or exercise training) were excluded. If it was not clear whether a study meet these eligible criteria, the corresponding study author was conducted for clarification.

Screening
Once duplicates were removed, all records were imported into COVIDENCE software (www.covidence. org) for title and abstract and full-text screening. All titles/abstracts and full texts were double screened by two researchers (L. S. and A. M. M.) with inter-rater reliability (Cohen's kappa) of κ = 0.57 (moderate agreement) and κ = 0.79 (fair agreement), respectively.

Data extraction
Data extraction was conducted using a structured tool created in Excel that was piloted prior to data extraction. Extracted data included author, setting, target population, study design and aim, sample characteristics (sample size, age, sex, level and causes of ID), PA measurement and outcome, and correlates of PA. To ensure consistency across studies and to limit bias in reporting, only correlates that were analysed at P < 0.05 were included. Across studies, a range of univariate, bivariate and multivariate analyses were reported (and where multivariate analyses included adjustments for other factors, these are reported alongside the main results). Whenever possible findings reported were those from the fully adjusted model. Variables that were conceptually similar were combined into one variable. Variables that were very much alike or gave a representation of closely related ideas or topics were identified by the team and combined based on group consensus. Where variables were combined, it was made it clear in the corresponding results table, along with a description of what the original variables were. The variable body mass index (BMI) was combined with BMI percentile and defined as BMI. The variable Fundamental Movement Skills (and its components) was combined with motor development under the title 'Motor development'. Therefore, this includes any variable relating to a child's fine and gross motor development. Two researchers (L. S. and L. A. H.-M.) each extracted data from 100% of the included studies, and one other researcher (A. M. M.) independently extracted data from 20% of included studies, with discrepancies resolved through consensus discussion.

Quality assessment of included studies
Study quality was appraised using the Standard Quality Assessment Criteria for Evaluating Primary Research Papers from a Variety of Fields (Kmet et al. 2004). Each item in the checklist was provided a score (2yes; 1partial; and 0no), and any items not relevant were marked 'N/A' and excluded from calculations of the summary score. A summary score was calculated for each paper by summing the total score obtained across relevant items and dividing by the total possible score. Resulting quality ratings were calculated in percentage form and were categorised as weak (<55%), moderate (55-75%) or strong (>75%) quality (Eddens et al. 2018). Quality appraisal was carried out independently by two researchers (L. S. and L. A. H.-M.) and a consensus reached. Quality assessment was for descriptive purposes only.

Analysis
To increase the robustness of the results, correlates that were investigated in two or more studies are the primary outcome of this review. For transparency and to highlight potential areas for future research, a table with all correlates identified is included as Appendix A, with selected correlates reported as secondary outcomes in the results. Because of heterogeneity across studies, meta-analysis techniques were not appropriate for this review. Instead, findings of included studies were combined in a narrative synthesis. However, effect sizes of associations for correlates were reported where this information was available from studies or where data were available to hand calculate or transform into effect sizes (see Appendix B). Correlates of PA data were synthesised based on a socioecological model (Sallis et al. 2008). Data were presented, where available, according to intrapersonal factors, interpersonal factors, organisational and environmental factors. Correlates were coded to reflect a positive correlation (+), negative correlation (À) or no correlation (0).

Literature searches
The searches yielded 7142 results once duplicates were removed. The results of the full searches and the screening are represented in Figure 1, including reason for full-text exclusion. Fifteen studies met the inclusion criteria and were included in this review.   (Vanhelst et al. 2013) and the Philippines (Eguia et al. 2015).
Thirteen studies measured PA objectively, while the remaining two measured PA subjectively using questionnaires (Lin et al. 2010;Vanhelst et al. 2013).
Sample sizes that were included in final analysis (retained participants that adhered to the measurement protocol) ranged from 20 (Shields et al. 2017) to 381 (Vanhelst et al. 2013). The mean age of participants ranged from 9.4 to 18.2 years, and five studies included participants over the age of 19 years; however, these participants represented <50% of each sample and the studies met inclusion criteria (Izquierdo-Gomez et al. 2015a;Izquierdo-Gomez et al. 2015b;Sundahl et al. 2016;Sit et al. 2017;Sit et al. 2019). Participants were predominantly male, with only one study reporting a larger number of female participants (Sundahl et al. 2016). The percentage of females in the studies ranged from 15% (Eguia et al. 2015) to 52% (Sundahl et al. 2016).

Quality assessment
Thirteen studies were rated as strong quality, with two of moderate quality; full quality assessment scores have been presented in Table 1. Because of these high ratings, it is assumed that study quality will not impact the results, and therefore, the results have not been stratified based on quality assessment scores.

Correlates of physical activity
Forty-eight correlates of PA were included for analysis and corresponded with the following levels of the socioecological model: intrapersonal (26 correlates), interpersonal (17 correlates) and environmental (5 correlates; see Appendix A). Effect sizes of associations for all correlates were reported where this information was available in studies, or the necessary information was available in studies for this to be calculated (see full details in Appendix B).

Correlates (reported in two or more studies)
Only six of the 48 correlates were reported in two or more studies, and none of those six were interpersonal or environmental variables.

Intrapersonal
Only 6/26 (23.1%) of the intrapersonal correlates were reported in at least two studies in the same dimension of PA ( Table 2). The six included intrapersonal correlates were age, sex, BMI, total percentage body fat, motor development and cardiorespiratory fitness. Two studies reported that motor development was positively correlated with PA, specifically, the PA dimension of number of steps per day (Eguia et al. 2015;Wouters et al. 2019).
The relationship between cardiorespiratory fitness and PA was inconsistent. Of the two studies that assessed cardiorespiratory fitness, Matute-Llorente et al.

Intrapersonal
Of the 20/26 (76.9%) intrapersonal correlates that were not included in two or more studies (see Appendix A), few significant correlations were found with PA. These secondary outcomes were discussed in detail within the research team, and a consensus was reached based on the expertise and experience within the team of the key findings selected for further discussion in the text as the most relevant for further investigation. However, all correlates are presented in Appendix A for further reference. Some key findings are as follows: sleep habits were positively correlated with subjectively measured PA (Vanhelst et al. 2013); positive health was not significantly associated with either VPA or total PA (counts per day) (Izquierdo-Gomez et al. 2015b); and socio-economic status was negatively correlated with MPA, MVPA and total PA and was non-significant when analysed with VPA (Izquierdo-Gomez et al. 2015b).
One study found no significant differences between those with ID and ASD and those with ID without ASD (non-ASD) groups across all included intensities of PA; therefore, having ASD in combination with ID was not correlated with LPA, MPA, VPA and MVPA (Boddy et al. 2015). Another study (Downs et al. 2016), which examined PA bouts, found that ID and ASD were not correlated with MPA or VPA at any bout duration. ID and ASD were not correlated with bouts LPA at 30, 60 and 180 s; however, children with ID but no ASD accumulated significantly more continuous LPA bouts lasting 5, 10 and 15 s and then children and adolescents with ID and ASD (Downs et al. 2016).
Down syndrome was found to be not correlated for any dimension of PA in which it was analysed.
Specifically, Down syndrome is not correlated with MVPA, steps per day, counts per minute (Wouters et al. 2019) and subjectively measured regular PA (Lin et al. 2010).

Discussion
This is the first study to systematically review correlates of PA in children and adolescents with ID. It highlights that previous research has predominantly investigated factors associated with PA behaviours at an intrapersonal level. All correlates reported in two or more studies were at an intrapersonal level, and the majority showed no significant association with PA, which contrasts with previous research on PA correlates in typically developing children and adolescents. Therefore, this review demonstrates that children and adolescents with ID have different PA behaviours in comparison with their typically developing peers, which has significant implications for future research. For children and adolescents without ID, a systematic review of reviews of correlates of PA demonstrated that being male was the strongest factor associated with PA (Sterdt et al. 2014). However, of the studies included within this review, no significant associations between gender and PA were identified. This is a highly relevant finding as PA research in people without ID is often viewed from a gender-specific perspective, with gender norms around PA widely reported and gender-specific interventions developed (Metcalfe 2018;Owen et al. 2018). This review suggests that gender is not as relevant to PA in children and adolescents with ID in comparison with typically developing children and adolescents. However, a systematic review of gender differences in adults with ID found that men complete significantly more MVPA than females (Westrop et al. 2019). Therefore, it could be that gender differences are not present until adulthood in people with ID; therefore, focused interventions in childhood and adolescence could potentially prevent these gender differences in PA from developing.
In adolescents without ID, Sterdt et al. (2014) identified that the majority of systematic reviews found a negative relationship between age and PA. Within the present review, however, three studies reported no association between age and PA, and two studies reported a negative association. Although this evidence is inconclusive, it suggests that the influence of age on PA is less present in children and adolescents with ID. This could be attributed to the lower autonomy and greater reliance on parents that children and adolescents with ID have, in comparison with their typically developing peers who, in general, become more autonomous with age, especially as they reach adolescence (Shields et al. 2012;Sterman et al. 2016).
In relation to the modifiable variables identified, this study showed that weight status was not associated with PA of any intensity. This is in contrast to research in typically developing children and adults with ID where obesity is negatively associated with PA (Jiménez-Pavón et al. 2010;Miguel-Berges et al. 2018;Bodde et al. 2013). Overweight/obesity is especially relevant for children and adolescents with ID as this is a significant secondary health problem in this population, with the prevalence of overweight/obesity increasing with age (Maiano 2011). However, the relationship between weight and PA in children and adolescents with ID may be more complex as obesity is a symptom of numerous disorders relating to ID, for example, Down syndrome (Melville et al. 2005). Therefore, the relationship between weight and PA may be less relevant to children and adolescents with ID; that is, children and adolescents with ID who are overweight/obese are not less likely to participate in PA, unlike their typically developing peers. However, PA may have more of a preventative role in relation to overweight and obesity, especially for children and adolescents who are more at risk, for example, individuals with Down syndrome. Although, in children and young people with ID, increasing PA through intervention is effective in reducing weight (Maiano et al. 2014). Therefore, PA is a promising mechanism for weight management in the population.
The only positive correlations identified within this study related to motor development and cardiorespiratory fitness. Motor development relates to children and adolescents having the basic physical skills required to be active, and increasing motor development has been identified as an effective mechanism to increase PA, particularly in preschool children (Engel et al. 2018;Holfelder & Schott 2014).
Considering that children and adolescents with ID will, in general, develop motor skills at a slower rate than typically developing children, interventions focused on supporting children and adolescents with ID to develop these fundamental movement skills could be a promising area for further investigation.
Cardiorespiratory fitness is an outcome of PA (Caspersen et al. 1985). Therefore, the inconclusive evidence on the relationship between cardiorespiratory fitness and PA, and the lack of a stronger relationship found for higher intensity activity, could suggest that children and adolescents with ID are not active for a sufficient duration and/or intensity to increase their fitness. This is concerning as cardiorespiratory fitness is a key indicator of health, such as cardiovascular and metabolic health (Ekelund et al. 2007;Ruiz et al. 2016). However, as parameters of cardiorespiratory fitness, for example, peak heart rate and heart rate variability (de Carvalho et al. 2018), vary between children and adolescents with and without Down syndrome, these results may not be as applicable to children with Down syndrome.
Using the socioecological model within this study enabled behaviours to be classified and better highlight which factors could have potentially relevant benefits on health-related PA. Disability-focused models, such as the medical model of disability (Marks 1997), focus only on the individual and not the wider environment; therefore, the socioecological model enables gaps in the literature to be better identified. As highlighted in this review, the focus of previous research on interpersonal factors has resulted in a limited evidence based related to wider socioecological factors. However, interesting variables were identified at wider socioecological levels that require further investigation, such as parental support being positively associated with PA.
Therefore, future research should aim to understand the relationships between wider socioecological factors and PA in children and adolescents with ID. This is highly relevant as previous qualitative research has demonstrated that children and adolescents with ID face numerous organisational and environmental barriers to PA ). In addition, Stanish et al. (2016) reported that adolescents with ID want to be doing more activity than they currently are, which suggests that there are wider, non-intrapersonal barriers that are limiting PA opportunities. Therefore, this should be further investigated using robust quantitative methods.
Furthermore, previous research in children and adolescents with ID has aimed to increase PA by generalising interventions developed in the general population (McGarty et al. 2018). However, this study highlights that correlates of PA that have been more established in children and adolescents without ID, for example, gender and weight status, are not consistent in children and adolescents with ID. Therefore, this trend of generalising interventions from typically developing is not appropriate, and a greater focus should be put on understanding the PA behaviours of children and adolescents with ID. In line with the behavioural epidemiological framework, the next stages of research should focus on more longitudinal research and predictors of PA over time.
In addition, greater focus should be given to emerging areas of research, such as developing motor development to improve PA levels.

Strengths and limitations
This study used a robust and structured methodology to review correlates of PA that was in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines, with two researchers completing all aspects of screening and data extraction. This study also included all dimensions of PA, which enabled potential differences in correlates between intensity or type of PA to be highlighted, and the inclusion of variables reported in two or more studies as the primary outcome reduced bias. The use of the socioecological model provided a framework to categorise correlates and to highlight where previous research has been focused and what future research directions are required. In relation to the included studies, 13 out of 15 studies measured PA using objective measures, which increases the validity of results compared with subjective measures (McGarty et al. 2014).
Not without limitation, the results included within this review are all based on cross-sectional research, and no study provided power calculations to justify sample size, which may limit the generalisability and validity of the results. Moreover, there are differences in data collection and reporting between the 15 studies. Some of the variation in findings in the studies reviewed may have been related to a variation in the data collection methods; although several studies used validated methods of measuring PA, some others used self-report or parent/teacher questionnaires. A wide variation is evident in sample size, with studies ranging from 20 to 381. Six (40%) of the 15 studies focused solely on children and/or adolescents with Down syndrome. It is well known that people with Down syndrome have a different health profile compared with people with ID without Down syndrome (e.g. Esbensen et al. 2007). Such a wide variation in the data collection methodology, cohort sizes and populations included between the studies can lead to a variation in results; therefore, some of the findings should be interpreted with caution.

Conclusions
In summary, this systematic review provides the first synthesis of data relating to factors correlated with PA in children and adolescents with ID. This provides valuable data and highlights priority areas for future research. These results provide preliminary evidence that the correlates of PA in children and adolescents with ID are different from the correlates previously identified for typically developing children and adolescents. This has significant implications for future research as it highlights that the PA behaviours of children and adolescents with ID need to be viewed as distinct from their typically developing peers. Therefore, future research should aim to better understand wider socioecological factors related to PA specifically in children and adolescents with ID.

Conflict of Interest
No conflicts of interest have been declared.

Source of funding
No external funding was received for the research reported in the paper.