Sedentary behaviour levels in adults with an intellectual disability: a systematic review and meta-analysis

1.1. Sedentary behaviour (SB)

Sedentary Behaviour (SB) and physical inactivity are frequently seen as one and the same, however they are very different and should be addressed separately. While recommendations for movement and PA levels in adults are long established for health benefits, corresponding recommended levels for SB, other than to reduce SB, are not (Bull et al., 2020).

In an effort to provide clarity, in the literature, SB has been defined as ‘any waking behaviour characterized by an energy expenditure of ≤1.5 METs while in a sitting, lying or reclining posture’ for example watching television or working on a computer (Tremblay et al., 2017, p. 9). Hence SB constitutes too much sitting or stationary activity as opposed to physical inactivity which is too little exercise or physical movement. A scoping review revealed that many publications have confused physical inactivity and sedentary behaviour. Hence a much broader definition of SB was refined for the purposes of this systematic review to also include physical inactivity and thus support the thorough identification of the prevalence of SB among this population and capture all relevant, seminal pieces. The definition of SB for the purposes of this systematic review is:

‘Low physical activity as identified by metabolic equivalent (MET) or step levels or as measured by the Rapid Assessment of Physical activity questionnaire (RAPA) or the International Physical Activity questionnaire (IPAQ) or sitting, lying or reclining for more than 3 hours per day’.

A metabolic equivalent (MET), known as the resting metabolic rate, is an objective measurement scale used to classify activity types and levels. A MET is the amount of oxygen (O₂) burned at rest and is the equivalent of 3.5ml O₂ per kg bodyweight per minute (Jette et al., 1990) or 1kilocalorie per kg of bodyweight per hour (Newton et al., 2013).

In the general population, SB has been linked to increased cardio-metabolic risks, increased obesity and mortality, as well as increased cancer risk (de Rezende et al., 2014; Patel et al., 2010; Same et al., 2016; Thorp et al., 2011). Emerging evidence is highlighting the importance of reducing SB for improving cardio-metabolic health. The same body of evidence is supporting the adoption of a holistic public health approach to improving activity levels as well as reducing SB (van der Ploeg & Hillsdon, 2017). High levels of SB, even if minimum exercise guidelines are met, show increased risk of heart disease, diabetes, and stroke (Patel et al., 2010).

However, the detrimental impact of SB can be reduced by interspersing periods of PA throughout the day (Healy et al., 2008). While breaking up time spent being sedentary has been shown to prevent disability in older adults (Sardinha et al., 2015), there is no similar information on adults with ID. This systematic review was conducted to explore the state of the science of sedentary behaviour in adults with an intellectual disability. It is critical that this information is identified so that they may be supported to age in a positive way. Overall, SB is poorly understood. The aim of this systematic review is to understand the prevalence of SB in adults with an intellectual disability.

2.1. Research question

PICO, which is used for quantitative studies was used to define the question as follows (Schardt et al., 2007):

The research question to be addressed is:

‘What are the sedentary behaviour levels of Adults with an Intellectual Disability?’.

2.2. Eligibility criteria

The criteria for study inclusion in the review are as follows:

The criteria for exclusion in the review are as follows:

2.3 Information sources

2.3.1. Databases. The following four databases were used to perform the search:

In addition, the following sources were explored for grey literature sources:

2.3.2 Search strategy. The search strategy was refined into two concepts following the application of PICO. Concept 1 is ‘Sedentary behaviour or inactivity’ and Concept 2 is ‘Intellectual Disability’. Each of the two concepts were searched using MESH terms and keywords and then combined using OR. Then the total results of each concept were combined using AND. A figure representing the search strategy is available in extended data (Lynch, 2021). This search was repeated for each of the four databases. The resulting article list was the complete combined database search results. This list was screened for inclusion.

Search string. An example of the search string used for the Medline database is shown in Table 1.

2.3.3. Screening process. All identified articles from each database that is searched, as well as all grey literature sources, were combined and duplicates removed. Endnote software was used to store all the identified articles. The articles were stored in folders which were named after the search process used. Using the inclusion criteria as detailed above, all articles were initially screened by title and then by abstract. The remaining full text articles were retrieved and read thoroughly. Those that did not meet the inclusion criteria were omitted.

2.4 Quality assessment and risk of bias

The remaining articles were quality assessed by two separate assessors using two validated quality assessment tools from the National Institute of Health (NIH) (National Institute Health, 2020), the first for observational cohort and cross-sectional studies and the second for randomised controlled trials (RCTs). A third person was available as an adjudicator for any discrepancies. The tools used are available as extended data (Lynch, 2021).

There are different types of study quality assessment tools for the different study types. For Controlled Intervention Studies and Observational Cohort and Cross-sectional studies, 14 criteria were used to evaluate the study quality, while for Case-Control studies 12 criteria were used. 11 criteria were used to determine the study quality of RCTs. This means that a maximum quality score of 11, 12 or14 could be achieved depending on the study type. This quality score was used to determine if the study should be included in the review. Quality scores were divided into 3 main categories: Good, Fair or Poor. See Table 2 for details.

2.4.1. Quality scoring. Scores were attributed to distinct parts of the study design for example type of study, design and blinding, where a ‘yes’ answer gives a score of ‘1’, a ‘no’ answer a score of ‘0’ and could potentially highlight an issue with the article.

3.1. Screening Process

The PRISMA search flowchart is shown in Figure 1.

Figure 1. PRISMA search flowchart.

An excel spreadsheet served as the data extraction tool to summarise the remaining articles. Article details were captured under 25 category headings. Exclusion criteria eliminated 20 articles. Two assessors [LL, EB] reviewed and quality assessed each of the final articles. There were no big discrepancies in results so a third adjudicator [MMcC] was not required.

3.2. Quality assessment and risk of bias

The final number of articles that went forward for a full quality assessment was 28. Using the NIH’s quality assessment tools for observational, cohort and cross-sectional studies and Randomised Control Studies (RCTs) to assess the internal validity of each article and any sources of potential bias, (National Institute Health, 2020). only articles rated in the fair to good range by the two assessors [LL and EB] were included. Appropriate quality scores for inclusion in this systematic review were achieved by 25 articles. These 25 articles are summarised in Table 3.

The reasons for study exclusion are shown in Figure 1.

3.3 Data extraction

An excel spreadsheet served as the data extraction tool which captured 11 different categories from each study. This was used to summarise all the shortlisted studies. The categories that were captured are shown in a table which is available as extended data (Lynch, 2021).

3.3.1. Data items. The PICO framework was used to define what data will be sought from variables as follows:

3.3.2. Outcomes and prioritisation. The outcome of this investigation into sedentary behaviour determined the sedentary behaviour levels of older adults with an intellectual disability.

Primary outcome

3.4. Data Synthesis

Article data was grouped according to the sedentary behaviour (SB) assessment category used in each article. Four methods for quantifying SB were identified in the 25 articles that passed the quality assessment. These four methods were:

The data was scrutinised to establish the breakdown of SB time, steps and screen time by residence, age, level of ID and gender but this was not always possible because studies often used different age ranges e.g. 18–49, <45, 50+, and few studies analysed results by residence type, gender or ID level. Thus, this type of analysis was not always possible.

RevMan Review Manager Version 5.4.1 (The Cochrane Collaboration, 2020) was used to synthesise results in a graphical format called a Forest Plot. A Forest Plot is a graphical representation of a meta-analysis where individual study’s results are represented by a box, and lines which denote the 95% confidence interval (CI). The influence a study has on the overall meta-analysis, the study weighting, is denoted by the size of the box. The amount of result variance between individual studies is represented by the heterogeneity value, I², of the forest plot (Israel & Richter, 2011). A higher I² value means a greater difference is observed between studies which is not due to chance and a meta-analysis may be inappropriate, as studies may not have similar populations. Values for I² of greater than 50% are considered to be indicative of moderate heterogeneity, 75% or greater is considered high, while values of 25% are low and hence similar (Higgins et al., 2003). A random effects model was used in the Forest plots to account for variance in studies such as varied settings, measurement devices, age or mixed levels of ID.

The mean difference and standard error for each study were used to determine a pooled mean prevalence for SB. In addition, a cumulative mean of means was calculated to determine the pooled prevalence of SB. Pair-wise comparisons were calculated where data was available using means and standard deviations. Scales were adjusted on the Forest plots so results may be seen clearly.

4.1. Measurement devices

A variety of measurement devices were used to assess sedentary behaviour. The prominent devices for measurement were accelerometers which were used in 14 studies. However, 4 used pedometers, 1 used a personal activity monitor, 1 used a survey as well as pedometers and accelerometers, 3 used a questionnaire or survey, 3 used IPAQ and accelerometers and 2 studies used self-report and accelerometers.

4.2. Steps per day

Steps as a measure of physical activity or SB were used in 11 studies, which involved 985 participants. The objective measurement of steps per day in these 11 studies was obtained using accelerometers and pedometers as shown in Table 6, which also shows the mean and range of steps per day. As can be seen in Table 6 a variety of devices were used.

An RCT by Melville and colleagues observed that at baseline the 102 Scottish participants, who had mild to severe level of ID and lived in different residential settings, were sedentary for 65.5% of the day (Melville et al., 2015). Being female, older age, more severe ID and having mobility impairments were significant predictors for low levels of PA (Hilgenkamp et al., 2012).

4.2.1 Steps per day and age. Some studies found that age could be a contributing factor to the number of steps per day taken. A US based cross-sectional study investigating the sedentary behaviour of two different age groups of adults with ID, younger adults (aged 18–49 years) and older adults aged 50+, showed the average steps per weekday decreased with age. However, the authors felt this difference could be attributed to the younger group having more wear time. More than 40% of adults with ID and more than 55% of older adults with ID had <5000 steps per day (Dixon-Ibarra et al., 2013). Similarly, in the Dutch based Healthy Aging and Intellectual Disability (HA-ID) study, 257 eligible older adults aged 50+ years of all levels of ID and residential settings wore a pedometer for 14 days. The average number of steps per day and the number in each age group that met the daily step recommendation was inversely proportional to age groups. In the 50–59 years group (n=146) 17.8% had greater than 10,000 steps per day and 41.1% had greater than 7,500 steps per day. In the 60–69 years (n=83) 18.1% >10,000 and 34.9% >7,500 steps per day. In the 70–79 years group (n=25), 8% > 10,000 and 16%>7,500. In the 80–89 years group (n=3) no one had greater than 7,500 steps per day. Overall, 39% of participants performed <5,000 steps per day (Hilgenkamp et al., 2012).

Conversely, Woods et al. (2018) which examined the behaviour of 19 participants aged 18 to 62 years with Prader-Willi Syndrome, found the 18–30 years and 40+ age group had similar steps but the 30–40 years had less steps. A study with 131 US-based ambulatory community living adults with ID showed that ID and age were strong factors in the numbers of steps per day taken (Peterson et al., 2008). Conversely a Spanish study with 84 adults who had varying levels of ID and attended an occupational day centre observed no difference in age-related SB (Oviedo et al., 2017). Hence the age and step count per day relationship is inconclusive. A summary of studies with age-related steps per day is shown in Table 5.

4.2.2 Steps per day and gender. Some studies found that gender was a contributing factor to less steps per day. This was investigated by four studies. A 2011 Scottish study with 62 community-based adults with mild to moderate ID deduced that women were significantly more likely to be sedentary (Finlayson et al., 2011). However, Johnson and colleagues in a study investigating physical activity levels of 37 community-based ambulatory adults with ID found the average daily step count accumulated over 14 days was comparable for both genders (Johnson et al., 2014). Similarly, a study with 19 participants with Prader-Willi Syndrome found the mean steps per day for males was analogous to females (Woods et al., 2018). In contrast, the Dutch HA-ID study, found that 21.8% of male participants and 11.3% of females had >=10,000 steps/day, while 42.9% men and 29% women had >=7500 steps/day (Hilgenkamp et al., 2012). Hence the effect of gender on steps per day is inconclusive. Table 6 shows the mean steps per day by gender.

The forest plot shown in Figure 2 shows the gender pairwise comparison. According to this plot females take more steps per day than males, which is contrary to some study results (Westrop et al., 2019). The mean difference seen is 1,089.2 steps per day at 95% CI [-69.72, 287.57]. However, a high heterogeneity of I² = 79% is observed indicating it may not be appropriate to pool article results due to study differences (Higgins et al., 2003). In addition, as the diamond shape touches the line of no effect the overall effect is not significant.

Figure 2. Pairwise comparison of steps per day by gender (divided by 10).

4.2.3 Steps per day and day of week. Several studies highlighted the influence of weekday versus weekend on the daily step count. The Dixon-Ibarra et al. (2013) study showed significantly less steps were observed from weekdays to weekends for all adults with ID. For weekends, adults with ID had an average of 4530 (SD±2337) steps per day and older adults with ID had 3504 (SD±2239). Finlayson and colleagues (2011) also found participants were more active on weekdays than weekends. Similarly, the average step levels in a Spanish study (Oviedo et al., 2017) were higher on weekdays with 6523 (SD±2807) steps per day compared to 5378 (SD±3686) steps per day at the weekend Equally, Peterson et al. (2008) found that weekday steps per day ranged from 1796 to 21,744 while weekend steps per day ranged from 1189 to 30,931. There appears to be an influence of weekend versus weekday on step levels.

4.2.4 Summary steps per day. To calculate a pooled mean of steps per day, a forest plot was produced using each of the 11 study’s individual mean and standard error. The results which give a pooled study mean of 6,715 steps per day, at 95% confidence interval (CI) [6,086, 7,344] are shown in Figure 3. The variability between studies is very high with I² =88% indicating high heterogeneity, which may indicate that it is inappropriate to combine studies due to the potential variability in studys (Higgins et al., 2003).

Figure 3. Study steps per day means with SE (Divided by 100).

4.3. A cumulative mean of means was calculated for all 11 studies. This pooled mean result was 6,555 steps per day. Screen time

In total three articles used television (TV) viewing as a means of evaluating SB. Two articles quantified SB by the amount of time spent looking at a screen, whether that was watching TV, videos, DVDs, using a gaming console or computer. The third article, Melville et al. (2018) in a cross-sectional study of 725 people with an ID, used a proxy-based measure of subjective screen times. This showed that 50.9% of participants spent four or more hours per day watching TV. This study showed that increased screen time was associated with higher levels of ID, being male, having mobility issues, obesity, hearing issues and epilepsy. The second study, with 1,618 participants, which was a mixed methods study using mail and an online survey to gather information indicated that 61.5% of participants watched three or more hours of TV per day and 40% watched four or more hours per day (Hsieh et al., 2017) which was a similar time observed in the Carlson study which had 17 participants (Carlson, 2016). Hsieh and colleagues (2017) also found that men with ID spent more time watching TV than women with ID. Furthermore, time spent watching TV was higher for those living on their own or in family homes than group homes. Those with mild/moderate ID spent more time watching TV than severe/profound. No difference in TV watching was observed by age groups. Figure 4 illustrates the forest plot for screen time. This plot has low heterogeneity with I²=0%, which indicates that the two studies may be compared. It shows the mean screen time per day is 3.42 hours at 95% CI [3.32, 3.53].

Figure 4. Screen time.

4.3. Assessing sedentary behaviour by time

SB, which is time spent sedentary, was assessed using time (in either hours or minutes per day) in 13 studies which included 713 participants. Objective measurements were obtained by accelerometers and/or pedometers and in one case a personal activity monitor. The minimum sedentary activity observed was 4hrs/day and the maximum 24hrs/day (McKeon et al., 2013). A study with 17 participants with ID, showed that higher ratings of self-reported health status predicted less SB and greater PA minutes in persons with ID (Fitz Gerald & Hahn, 2014). A larger sample in a Spanish study which compared the activity and SB of 66 active and non-active individuals with mild and moderate ID and 31 older adults with no ID, found there were large amounts of SB even if groups met the PA guidelines for health. Furthermore, the number of sedentary bouts was greater in the ID groups than non-ID groups (Oviedo et al., 2019). Sedentary time was accumulated in bouts of 1–30 minutes in duration in a US-based study with 52 participants with all ID levels (Ghosh, 2020). Harris and colleagues (2019) demonstrated that 143 participants had a median of 7 breaks per day (95% CI, 4-11), where the median duration of breaks observed was 43.2 minutes (95% CI, 27.2-73.7). An accelerometer and the Bouchard scale were used to quantify activity levels of 37 participants with mild to moderate ID in Australia. The nine-point Bouchard scale defined level 1 as lying down, sleeping or resting and level 2 as seated activity. Using level 1 and 2 as indicators of SB, participants were sedentary for an average of 83.7% of each day (Temple & Walkley, 2003). Another Spanish study found that 84 adults with varying levels of ID who attended an occupational day centre spent 79.4% of their waking hours sedentary (Oviedo et al., 2017). A study looking at activity levels of 90 adults with ID living in group homes identified that participants are extremely sedentary during weekdays, spending the largest percentage of time in SB (mean = 67.3%, SD±12.0%) (Chow et al., 2018). These studies and the mean sedentary time per day are shown in Table 7.

With consideration to differences observed in SB between weekends versus midweek, three studies did identify differences but not consistently. Table 8 shows the measured SB in three such studies. Furthermore, a secondary analysis of two pooled RCTs showed a significant difference in break duration between weekdays 79.8 (SD ±151.6) minutes and weekend days 62.6 (SD ±55.7) minutes (Harris et al., 2019).

4.3.1 Sedentary time and gender. A 2011 Scottish study with 62 community-based adults with mild to moderate ID presented the average SB time per day for women as 19.56 hours and men 17.62 hours. On weekdays this was 19.46 hours for women and 17.24 hours for men (Finlayson et al., 2011). Figure 5 shows the paired comparison of mean sedentary minutes per day by gender. The Spanish study which had results for active (Act) and inactive (InA) groups of participants show both groups included separately here. This forest plot shows that men have less sedentary minutes per day than women, the mean difference is -234.3 [95% CI, -48.12, 1.26]. The Fitzgerald study appears to have the biggest influence on the pooled result due to its higher weighting. However moderate heterogeneity is present between studies as demonstrated by I²=57%. Hence it may be inappropriate to combine results (Higgins et al., 2003).

Figure 5. Mean sedentary minutes by gender (divided by 10).

However, if the Finlayson study is excluded from the calculation (as it appears to be an outlier), the I² value reduces to zero. See Figure 5.1. The Forest plot still shows that men have more sedentary minutes than women and as the lower and upper points of the horizontal plane of the diamond (i.e. the [95% CI, -293.9, -1.36]) both lie to the left, it means the resulting difference is significant. The mean difference is -153.7. No heterogeneity is present so it is appropriate to combine study results.

Figure 5.1.Mean sedentary mins by gender (excluding Finlayson) (divided by 10).

4.3.2 Summary SB time. The percentage of waking time spent in SB seen in these studies varied from 72% of wear time (Bellicha et al., 2020) to 83.77% (Temple & Walkley, 2003). The total daily time observed in SB in these studies varied from 437minutes or 7.28 hours to 1206.3 minutes or 20.1 hours.

For analysis purposes, all times were converted to minutes. Figure 6 shows a forest plot of studies with mean sedentary time in minutes. This demonstrates high heterogeneity (I²=99%) and hence high levels of variability among study results which means it may be inappropriate to pool results (Higgins et al., 2003). However, the plot provides a good visual representation of the results. The mean sedentary minutes was 599.9 minutes per day at 95% CI [520.3, 679.5] or 9.99hours. A pooled mean result for all studies was calculated using a mean of means formula (as used in Section 4.2). The resulting pooled mean of total sedentary minutes per day for all 13 studies is 606.3minutes or 10.1hours per day.

Figure 6. Mean time sedentary with SE (minutes divided by 10).

4.4. Diverse methods

Two studies used alternate methods to identify SB. The first study with 58 participants with ID using an ActiHeart device investigated Physical activity level (PAL). PAL is the ratio of total energy expenditure and resting energy expenditure as described by United Nations Food and Agriculture Organisation (2001). PAL cut-off points for activity levels were <1.4 for sedentary. The mean total physical activity level measured in this study was 1.39 (SD+/-0.15) which is indicative of a sedentary lifestyle (Moss & Czyz, 2018). The second study which used a question on how much sitting time people did to identify SB, found that 47% of the 920 participants sat for ‘all, most or a lot of the day’ (Tyrer et al., 2019).

Reporting guidelines

Harvard dataverse. PRISMA checklist and flow chart for ‘Sedentary behaviour levels in adults with an intellectual disability: a systematic review and meta-analysis’. DOI: https://doi.org/10.7910/DVN/HYMA0J

Data are available under the terms of the Creative Commons Zero "No rights reserved" data waiver (CC0 1.0 Public domain dedication).