Sedentary behaviour surveillance in Canada: trends, challenges and lessons learned

Background Historical changes in the nature of sedentary activities have been observed in other countries, but it is not clear if similar trends exist in Canada. It is also unclear how changes in the measurement of sedentary behaviour affects national estimates. Our objective is to document all sources and measures of sedentary behaviour from Canadian, nationally representative surveys, and report on selected estimates of time spent in sedentary activities. Lessons learned can benefit the wider international surveillance community. Methods We describe and document all data sources of sedentary behaviour at the national level in Canada, and report on selected prevalence data from repeated cross-sectional surveys. We summarize amounts of total device-assessed sedentary time and self-reported sedentary activities (e.g., passive travel, leisure television, computer, video games, screen, and reading) by age group over time. Results Nineteen national surveys were identified. Changes in questions and/or response categories precluded direct assessment of trends over time for some measures; however, certain trends were observed. Accelerometer-measured sedentary time, leisure reading (among those < 50 years) and television/video viewing in younger age groups have remained relatively stable (with a possible slight decline in television/video viewing). Time spent in passive travel and leisure computer and electronic device use appears to have increased. Television and video viewing appears to have increased in older adults while their leisure reading appears to have fallen. Conclusions Changes in measurement of sedentary behaviour can affect estimates and reduce comparability over time. Total leisure screen use appears to have increased over time, reflecting the ways in which Canadians spend their free time and technological advances. The main public health message is the need for continued efforts to reduce leisure screen use, especially among youth and older adults.


Background
Canadians spend the majority of their days engaged in sedentary behaviour [1]. Sedentary behaviour includes activities undertaken at a low energy expenditure (≤ 1.5 metabolic equivalents) while sitting, lying or reclining, such as watching television, using a computer or sitting in a vehicle [2]. Systematic review evidence has identified that sedentary behaviour is a risk factor for several chronic conditions (e.g., cardiovascular disease, diabetes, obesity, cancer, and depression) and mortality [3][4][5]. Large doses of physical activity (~60-75 min/day) have been shown to offset the increased risk associated with prolonged sedentary behaviour [6]; unfortunately, the majority of Canadians fall well below these levels [7]. As a result, public health messaging emphasizes the importance of reducing prolonged time spent sedentary in addition to getting adequate physical activity and sleep [8]. The Canadian 24-Hour Movement Guidelines for Children and Youth include a recommendation to reduce leisure screen time to 2 h or less per day and to limit prolonged sitting [9]; guidelines for adults are under development.
The field of sedentary behaviour research is growing at a rapid rate [10], but prior to this interest, Canada had a long-standing record of measuring this behaviour. The 1985 Canada Fitness Survey asked respondents to selfreport their daily sitting time; this data was subsequently used in one of the earliest papers to document the link between prolonged sitting and mortality risk [11]. Additionally, several national surveys have included questions related to screen-time; a key sedentary behaviour. The growing public health need to monitor and report on sedentary behaviour and related factors led to a modernization of the Public Health Agency of Canada's (PHAC's) physical activity surveillance system which is now directed by the Physical Activity, Sedentary Behaviour and Sleep (PASS) Indicators [12,13].
Within the PASS Indicators, the PHAC reports on the sedentary behaviour levels of Canadians using the most recent and comprehensive data available [7]. While changes to survey content have permitted updates to measures/indicators to keep up with the sedentary practices of Canadians (e.g. types of screens used) [14], this has created challenges in obtaining a consistent time series of data. It is important to understand these measures in their historical context and whether trends in the types and duration of sedentary behaviour can be detected. Lessons learned can be applied in international comparisons and for the interpretation of trends in other countries.
Research in other high-income countries suggests that leisure time spent sedentary has remained fairly stable with slight variations over time [15][16][17][18]. However, the nature of leisure sedentary activities is in flux with a greater uptake of screen-based technologies beyond simply watching television [15,16,[19][20][21]. Sedentary transport has also increased, albeit at a slower rate than screen use [15,16], while other sedentary activities (e.g., socialization and crafts) have declined [16]. It is not clear if similar trends exist in Canada. The objective of this research was to document all sources and measures of sedentary behaviour from Canadian, nationally representative surveys, and report on selected data from those which are ongoing and provide estimates of time spent in various sedentary activities. We also hoped by examining how measurement has changed over time we could provide insights regarding comparability within and across countries and lessons learned for future surveillance of sedentary behaviour.

Methods
We describe and document all sources and measures of sedentary behaviour at the national level in Canada, and report on selected prevalence data from common surveillance health surveys.

Measures of sedentary behaviour
Any question(s)/item(s) related to sedentary behaviour were eligible for inclusion. Questions could comprise any domain (e.g., occupational, leisure, transportation, household) or type of activity (e.g., screen time, reading, hobbies, car travel). Both device-assessed and self-report measures were included.

Data collection procedures
An environmental scan of all national survey sources in Canada was carried out to identify measures of sedentary behaviour and an inventory created. The TAxonomy for Self-reported Sedentary behaviour Tools (TASST) framework [22] was used to describe the characteristics of identified survey questions including collection years, age of respondents, method of administration, sedentary behaviour type (e.g., total sitting, leisure television), number of questions, recall period (e.g., past 7 days), temporal unit (e.g., single day, week), assessment period (e.g., weekday only, weekend day only, total week), and response options (categorical or continuous). Table 1 includes the questions and response content for all surveys and highlights changes between years.
The prevalence of sedentary activities across survey years by age group was examined within four surveys: Canadian Community Health Survey (CCHS); Canadian Health Measures Survey (CHMS); General Social Survey (GSS); and, the Health Behaviour in School-Aged Children (HBSC) study. These four surveys were chosen as they provide ongoing, national estimates of time spent in various sedentary activities.

Canadian Community Health Survey (CCHS)
The CCHS began in 2000 and is an ongoing, crosssectional survey conducted by Statistics Canada that collects health information from a representative sample of the Canadian private-dwelling population ages 12 and older. For some years, the module assessing sedentary activity was asked of all respondents and for other years it was optional content selected by a subset of health regions or provinces and is only representative of those regions (Additional file 1: Table S1).

Canadian Health Measures Survey (CHMS)
The CHMS began in 2007 and collects self-reported and measured health information from a representative      sample of the Canadian household-dwelling population aged 3 (6 to 79 years in Cycle 1 to 79 years). The CHMS collects data from an interview-administered questionnaire conducted in the respondent's home, as well as from a visit to a mobile examination centre (MEC) where physical measures are taken. The parent/guardian of children aged 3-11 years answer the household questionnaire on their behalf. During their visit to the MEC, ambulatory respondents of the CHMS were invited to wear an Actical accelerometer (Philips Respironics, Oregon, United States) over their right hip during waking hours for seven consecutive days. The Actical records time-stamped acceleration in all directions and provides an index of movement intensity based on a count-per-minute value; an intensity of < 100 counts-per-minute was used to identify sedentary time [23,24]. A complete description of accelerometer data reduction procedures is available elsewhere [23][24][25].

General Social Survey (GSS) -Time-Use Surveys
The GSS collects information on living conditions, social life, well-being of Canadians, and specific social policy issues. Time-Use surveys began in 1985 and are conducted at approximately five-to seven-year intervals and employ a retrospective 24-h time diary to collect information on respondents' participation in, and time spent on, a wide variety of day-to-day activities. Starting in 2010, it accounted for multitasking. This analysis includes data from the cycles of the GSS with a time-use survey. Within the GSS, 'youth' only includes those aged 15 to 17 years.

Health Behaviour in School-Aged Children Survey (HBSC)
The HBSC is a nationally representative survey that collects information on the health and well-being, social environments and health behaviours of school students [26][27][28]. The HBSC is a collaborative study with the World Health Organization Regional Office for Europe with 49 participating countries across Europe and North America and collects data every four years on 11-, 13and 15-year old boys and girls. The PHAC funds the Canadian component of the HBSC which is conducted by Queen's University and began in 1994.

Statistical analyses
All analyses were conducted using SAS Enterprise Guide v.5.1 (SAS, Inc., Cary, NC). Descriptive statistics summarized amounts of various total sedentary time and specific sedentary activities (e.g., passive travel, leisure television, screen, computer, video games, and reading), across surveys using means or proportions and 95% confidence intervals (CIs). Cases missing data per sedentary behaviour outcome were omitted from the respective analyses. Using unadjusted linear regression analyses  ; 1986, 1992, 1998, 2005, 2010, 2015), Health Behaviour in School-Aged Children study (HBSC;1990, 1994, 1998, 2002, 2010, 2014. Only surveys with the same questions and responses have been joined by solid lines crude trends in average daily accelerometer-measured sedentary time across cycles were examined within each age group. Between-cycle pairwise contrasts were conducted to examine differences between cycles with a Bonferroni adjustment. Within all surveys, if a categorical response option with a range was included, we used the midpoint of each category to generate a continuous measure of time spent in the various sedentary activities (e.g., 0-1 h = 0.5 h), for the uppermost category we used the starting amount (e.g., more than 20 h = 20 h). Time spent in various screen-based activities was summed to provide a measure of total leisure screen time per day.  ; 1990, 1994, 1998, 2002, 2006, 2010, 2014). Only surveys with the same questions and responses have been joined by solid lines Where possible, the number of children and youth meeting screen time recommendations (≤ 2 h/day of recreational screen time) from the Canadian 24 Hour Movement Guidelines for Children and Youth [29] were estimated using proportions and 95% CIs. Results are presented by age group with between group differences assessed using independent sample t-tests or analysis of variance for continuous outcomes or chi-square for categorical outcomes. In the figures presented, data points that are derived from the same questions and response options have been connected with a solid line.
To account for the complex survey design and nonresponse bias and to correctly estimate variance, all analyses of CCHS [30] and CHMS [31] data were weighted using the survey weights generated by Statistics Canada. For the CHMS data, denominator degrees of freedom were set at 11 for all cycles except for Cycle 2 which used 13. To account for survey design effects in the CCHS, CHMS and GSS, the bootstrap technique was used to estimate 95% CIs. In the first five cycles the HBSC generated an unweighted national sample. Analysis of data from 2010 onwards incorporates population weights and controls for clustering at the school level. Table 1 provides characteristics using the TASST framework for the 19 identified surveys. Canada's long-standing history of collecting data on sedentary behaviour has lacked consistency across and within surveys, and several only administered one-to-two times. Although most surveys incorporate(d) a repeated crosssectional design providing estimates over time in the Canadian population (or a sub-sample), changes in questions and/or response categories make it difficult to directly assess trends (see Table 1). Nonetheless, we present cross-sectional estimates by survey year for specific sedentary activities to provide a visual pattern of changes over time using data from the CCHS, CHMS, GSS, and HBSC (Figs. 1, 2, 3, 4, 5, 6, 7 and 8). Pairwise contrasts, however, identified no significant between cycle differences. Sedentary time is higher in older age groups, with children engaging in significantly lower amounts and older adults engaging in significantly higher amounts compared to other age groups. In Cycle 5, sedentary time ranged from 7.7 (95% CI: 7.5-7.9) hours/day in children (6-11 years) to 9.9 (95% CI: 9.7-10.1) hours/day in older adults.  Table 1 for changes in questions).   ; 1990, 1994, 1998, 2002, 2006, 2010, 2014 Computer use outside of school or work  Total leisure screen time Figure 5 provides a historical overview of leisure screen time by age group across survey years. Total leisure screen time was identified using a sum of time spent on different screen-based media depending on survey and year (see Table 1). A transition towards asking respondents to estimate their total leisure screen time rather than separate activities occurred in both the CCHS (starting in 2017) and CHMS (2016-2017 children). Despite the changes in survey questions, trends in self-reported daily leisure screen time appear to be increasing in all surveys.

Watching television or videos
Within each survey total screen time assessed via similar questions appears to have increased across all age groups. Among youth, daily leisure screen time from the HBSC has consistently been higher than that reported in the CCHS and CHMS. Most recent estimates of total  (HBSC;1990, 1994, 1998, 2002, 2010, 2014 ; 1986, 1992, 1998, 2005, 2010, 2015). Only surveys with the same questions and responses have been joined by solid lines Figure 6 provides an overview of the proportion of children and youth meeting the screen time recommendations (≤ 2 h/day) from the Canadian 24-Hour Movement Guidelines [9]. The change in methodology has created a challenge in discerning whether the proportion of children and youth meeting the screen time guidelines has changed with time. Prior to more recent years (which reflect substantial changes to questions), adherence to screen time recommendations appeared to be declining.    (1986,1992,1998,2005,2010,2015) Discussion This is the first comprehensive review of sedentary behaviour surveillance data in Canada. Surveys differ with respect to their measures (e.g., questionnaire, 24-h recall, activity monitors), sampling frame (e.g., national level [CHMS], sub-national [CCHS]), and sample age (e.g. HBSC: youth in grades 6-10, CCHS: ≥ 12 years, CHMS: 3-79 years and GSS: ≥ 15 years). While this paper is focused on Canadians, the findings and lessons learned regarding measurement and surveillance are relevant to other countries with similar measures and surveys. We hope that the findings will help to inform interpretation of data in the surveillance of sedentary behaviour.

Reading time
Assessing trends in type-specific sedentary activities (i.e., television/video, computer, video games, reading) is difficult given the changing nature of the content and this precluded our ability to report with certainty on changes in estimates over time. At times, different data sources told different stories, adding to the complexity of interpretation. For example, in Fig. 2 we see that among adults aged 50-64 and ≥ 65 years, television viewing increased in the CCHS and CHMS, but remained relatively stable in the GSS. Research in older adults using the TASST framework found that using a recall period of a typical day in the past week was more sensitive to change than a past day recall [32].
Similar to what was observed in the United States using data from the National Health and Nutrition Examination Survey (2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016); television time appears to have remained relatively stable (amongst most age groups), while computer and electronic device use has increased [19]. Also similar to what was observed in the Australian Time Use Surveys; total leisure screen time looks to have increased [15]. A larger proportion of leisure time is spent engaged with different types of screens (e.g., tablets, smartphones) for different purposes (e.g., online games, shopping, communication). A temporal analysis of GSS data found that the proportion of Canadians reporting the use of computers during free time (e.g. email, on-line social networking, searching for information) increased nearly five-fold from 5% in 1998 to 24% in 2010 [33]. This uptake in screen time has implications given its association with chronic disease and mortality [3].
Leisure reading remained relatively stable in those younger than 50 years and although it appears to have declined in those older, their levels still remain higher. In comparison, the Australian Time Use Survey data suggest small reductions [15]. Reading is the only sedentary activity shown to be associated with beneficial outcomes including academic achievement and maintenance of cognitive function [34,35].
Pairwise contrasts between cycles found no significant differences in total sedentary time. While there were significant negative linear trends in youth, young adults and adults aged 50-64 years, the only cycle to exhibit a lower estimate was the most recent cycle (2016-2017). Longer and subsequent follow-up would be needed to identify if a trend truly exists or if this is an artefact. We are unaware of any other national or international studies to have examined trends in device-measured sedentary time. However, others have examined total self-reported sitting; and have found that it has remained fairly stable [19,36].
Within the GSS, passive travel time appears to have increased slowly. Other surveillance studies have also found that passive travel increased, albeit at a slower rate than screen use [15,16]. The estimates produced reflect all passive travel and not solely commute time to work/ school. According to data from the 2016 Census, Canadians (≥ 15 years) commuters spent an average of 26 min travelling to their work. From 2011 to 2016, the number of Canadians with longer commutes (≥ 60 min) rose by 5% [37,38].
Historically, epidemiologic studies have largely relied on self-report measures of sedentary behaviour. Selfreport measures are prone to systematic errors due to a respondent's inability to accurately estimate their frequency and duration of time spent in behaviours [39]. Recall frame relates to the number of hours, days or weeks an individual recalls a behaviour in the past. While longer recall frames are used to provide a better estimate of 'usual' activity, shorter recall frames are used to improve reliability and validity of a questionnaire. Earlier cycles of the CCHS and CHMS asked participants to recall their sedentary activities during a usual week over the previous 3 months. Asking individuals to provide a summative estimate of weekly sedentary activities (e.g., television watching) is likely to increase information bias [39]. Recent modifications to reduce this recall frame to a typical day in the past 7 days have likely led to improvements in recall accuracy. It is difficult to establish the construct validity of self-report measures of domain-and type-specific sedentary activities as this would require them to be compared to an accurate measure of the behaviour (e.g., combination of GPS and accelerometer, body camera, direct observation). However, criterion or predictive validity has been assessed demonstrating that different doses of these type-specific sedentary activities (e.g., screen time) are associated with health outcomes [35,[40][41][42][43].
The 24-h activity recall data collected within the GSS Time Use Survey is likely to provide a more accurate estimation of time spent in various behaviours versus questionnaire. Given that 24-h is a finite amount of time, respondents must consider their entire day and allocate activities accordingly. Time-use diaries have been shown to be more valid than questionnaires at quantifying time spent sedentary [44,45]. Unfortunately, time-use surveys are resource intensive placing a greater burden upon respondents, as well as those who undertake analysis. Given that space on a population health survey is always limited, the capacity to undertake these recalls is unlikely [46]. Balancing the needs and interests of multiple data users (e.g., surveillance, policy makers and researchers) in sedentary behaviour surveillance is a challenge. In our previous work [46], we describe that sedentary behaviour questions should be "concise, valid/reliable, evidence-based, and developed using best practices". Questions should also be "adaptable and able to assess various modes of sedentary behaviour" [14]. The International Sedentary Assessment Tool (ISAT) is a sedentary behaviour module that addresses these needs by providing: an itemized list of questions in order of their established relationship with health; the capacity to use any item separately; and, examples of modes (e.g., Smartphone, tablet) in brackets to allow for updates as new technologies/modes emerge [14]. Additionally, it recommends collecting responses in continuous format to allow for guideline assessment in the future [14].
Only recently have surveillance systems included the use of device-measured (e.g., accelerometer) sedentary time to overcome some of the potential sources of bias in self-report measures [39]. The CHMS strengthens Canadian surveillance by monitoring total sedentary time and patterns of behaviour (e.g., breaks, bouts) using accelerometers. While accelerometers help to reduce response bias, they may misclassify time spent standing stationary as sedentary time. It is important that as the field moves to also include wrist-worn accelerometers (which have shown to improve compliance in wearing the monitor) [47] that surveillance systems consider the accuracy of these devices for detecting sedentary time in addition to physical activity and how this will affect trends. While this is a rich data source that can provide more accurate information on the sedentary time of Canadians, the CHMS is not able to provide sub-national estimates (e.g., provincial/territorial) and lacks the ability to distinguish between domains and types of sedentary activities. This domain-and type-specific information is important for designing interventions and identifying key targets for behavioural modification. The complimentary nature of the self-reported and devicemeasured data can offer up rich insights into the behaviours of Canadians and other countries.
Ensuring that methodologies used to assess sedentary behaviour are consistent enhances the ability of a surveillance system to examine the behaviour in relation to health outcomes and assess differences across demographics. Surveillance systems benefit from questions that are sensitive to change, provide reliable estimates and sufficient face (i.e. are the questions understood by respondents) and predictive validity (i.e. can identify a dose-response between behaviour and outcome). Predictive validity is one of the key components for establishing causality as outlined by the Bradford Hill criteria [48]. Self-reported sedentary behaviour is often more strongly associated with health outcomes than device measures [35,[49][50][51]. Recall of a specific sedentary activity, such as television, is likely easier than a summation of all time spent sitting throughout the day, and has been found to be more strongly associated with health outcomes than total sitting [6]. Evidence suggests that a single question for total sitting is less valid than a summative measure when compared to a construct standard (e.g., activPAL or accelerometer) [52][53][54][55]. Recently the CCHS and CHMS have moved toward a single question for leisure screen time. This helps to overcome issues of multi-tasking where an individual may inadvertently 'double-count' time spent on multiple devices such as watching television while simultaneously using a tablet to surf the Internet. It may explain why estimates of total leisure screen time in the newer cycles are lower than those preceding where a summation of television, computer and video game time was used. It may also be that asking individuals to recall all of their leisure screen time increases the likelihood of response and recall bias compared to asking about activities separately.

Strengths and limitations
Strengths of this study include the use of data from large, population-based survey samples and examination of trends by age group. The study examined trends in device-measured total sedentary time which had not previously been done in a national cohort and which helped to overcome biases associated with self-reported total sitting. Previous examinations of trends in sitting time in other international cohorts have relied on self-reported measures [19,33]. This study also examined trends for a number of self-reported sedentary activities including 'inactive' video games which have not yet been examined in a national cohort. Unfortunately, we were unable to report on trends in occupational sedentary time. To date, no known Canadian national survey has collected information on time spent sitting exclusively at school or work. In the years covered by this analysis, the technologies used for screen time have proliferated making older measurement methodologies less able to reflect current technology use. These changes largely prevented the assessment of trends using statistical comparisons given that measures have changed over time. Our results focus on Canadians aged ≥ 6 years. There is very little national data for sedentary behaviour among infants and preschool-aged children and even among those aged 6-11 years there are few sources of data for some indicators (e.g., television and video watching). This is a limitation of the current surveillance landscape both in Canada and abroad. We hope that the new Canadian Health Survey on Children and Youth [56] will help to fill in some of these gaps.

Conclusions
We hope this inventory of measures demonstrates opportunities and challenges of measuring an evolving concept like sedentary behaviour through surveys. Although it seems like having static questions would be helpful for monitoring trends over time, it is not appropriate to ignore changes in behaviours and preferences. International surveillance systems can apply the lessons learned within this paper when examining trends and comparing internally and to other countries. While total device-assessed sedentary time has remained relatively stable, time spent in self-reported leisure sedentary activities appears to have changed. Youth are the greatest leisure screen users and appear to be changing the types of screens with which they engage; swapping television viewing for greater time spent on computers and other electronic devices. While older adults continue to have the lowest amounts of leisure screen time, they appear to be decreasing their time spent reading and increasing their time spent watching television and using electronic devices, which based on the evidence is a significant public health concern. Total leisure screen use appears to have increased over time a reflection of the changing ways in which Canadians spend their free time and technological advances. Future research into the risks associated with various types of sedentary activities and prolonged sedentary time and the health of Canadians (and other nations) across all ages (especially younger children) is needed.
Additional file 1: Table S1. Sedentary activities module uptake by region in the Canadian Community Health Survey. Authors' contributions SAP conceptualized and designed the study, did analyses, drafted the initial manuscript, and reviewed and revised the manuscript. AM designed the study, did analyses, and reviewed and revised the Article. KCR, GPB, WT conceptualized and designed the study and reviewed and revised the manuscript. All authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.

Availability of data and materials
The datasets analysed during the current study are available through the Research Data Centres (RDC) Program at Statistics Canada (https://www. statcan.gc.ca/eng/rdc/index).
Ethics approval and consent to participate Approval for the conduct of the CCHS, CHMS, GSS, and HBSC (and Queen's University) was obtained from Health Canada's Research Ethics Board and written informed consent and assent were obtained from all participants.

Consent for publication
Not applicable.