Background
There is compelling evidence of the positive effects of physical activity on health [
1,
2]. However, it has been reported that a large proportion of the population do not achieve recommended physical activity levels [
3,
4], even when active commuting (i.e. walking and cycling), an important contributor to overall physical activity levels among young and middle-aged adults [
5], is taken into account. There is also emerging evidence that time spent in sedentary activities is a risk factor for chronic conditions, such as diabetes and cardiovascular diseases [
6,
7] and premature mortality [
8,
9], irrespective of the overall physical activity levels. Sedentary activities are those with low energy expenditure, such as prolong sitting time at work or at home, sleeping, lying down, and watching television [
10].
Previous studies examining changes in physical activity over time have mainly focused on time trends in leisure time physical activities (LTPA) such as exercise, sports, and gardening [
11‐
20] and less attention have been given to time trends in active commuting and sedentary daily activities [
16,
17,
19,
21,
22]. US data suggest that participation in LTPA has declined or remained stable over time [
15]. In contrast, studies from Canada, Denmark, and Finland suggest that participation in LTPA have increased in more recent years [
17‐
20]. Few studies have examined changes in active commuting also showing variable results [
17,
19,
22]. A study found that participation in active commuting among adults in Finland declined from the 1970s to the early 2000s [
17]. In contrast, in the same period of time, active commuting increased among Canadian adults [
19]. A US study comparing data from 2001 to 2009 found that the overall prevalence of active commuting was low and that there were modest increases in walking while cycling levels were stable [
22]. The studies that have examined changes in sedentary daily activities have reported an increase in sedentary time spent at work or at home in recent years across all jurisdictions [
15,
17,
19,
23].
Variations in physical activity over time may be related to the effects of aging, to the different life experiences of generations of people born at different times (cohort effects), or these variations could also be the result of societal and environmental changes which affect the population as a whole (period effects). Only few studies have examined age, period, and cohort effects in LTPA [
16,
17,
24‐
26], active commuting [
17], and daily activity [
17,
26]. Given this gap in the literature, the goals of this paper are: 1) to examine age, period, and cohort effects in physical activity across three domains: leisure time, commuting, and sedentary behavior (sedentary time spent at work or at home) over 16 years in a representative sample of Canadians, 2) to examine whether age, period, and cohort effects are explained by changes in education, income, and body mass index (BMI), and 3) to examine whether changes in sedentary behavior over time influence the trajectories of participation in LTPA and active commuting.
Discussion
The major findings from this study are two-fold. First, there is a strong cohort effect in active LTPA, active commuting, and sedentary behavior, such that recent generations are more likely to report being physically active in leisure time and commuting, and at the same time to be more sedentary. Second, underlying these cohort effects is a substantial period effect to both increased participation in physical activity (active LTPA and active commuting) and increased sedentary behavior over time from 1994/95 to 2010/11. The higher participation in active LTPA and active commuting in more recent cohorts was related to this period effect (secular trend) of increasing physical activity over time. Another key finding is that, generally, those with sedentary behavior were less likely to engage in physical activities, particularly in active commuting.
Previous studies have examined changes over time in physical activities [
11‐
16,
23], but the examination of the contribution of age, period, and cohort to changes over time in physical activities across different domains has been more limited [
17,
20,
24‐
26]. The findings of higher participation in active LTPA are in line with some studies [
17,
20,
25] but not others [
24]. Two studies found that when compared at the same age more recent cohorts were more likely to exercise regularly or to be physically active in leisure time [
17,
25]. Whereas another study found that the volume and duration of leisure time physical activity was lower in more recent cohorts of Australians [
24]. Our finding of greater participation in active commuting among recent cohorts is in contrast to a study comparing active commuting by birth cohort in the Finish population [
17]. One possibility for the discrepancies between our findings for active LTPA and commuting with those in the literature is that our study used longitudinal panel data in which the same individuals were followed over time, while those studies are based on combined data from multiple cross-sectional surveys [
17,
24]. Also the Australian study [
24] examined average energy expenditure, whereas in our study we examined a binary variable identifying individuals who met recommended activity levels. The greater sedentary behavior in more recent cohorts is in line with the few studies that have compared sedentary behavior by birth cohort [
17,
26]. The Finish study used a similar measured as ours and found that sedentary behavior was greater in more recent cohorts [17]. In addition, a longitudinal study examining energy expenditure at work or at home among Chinese adults, found that more recent cohorts have lower levels of physical activity at work or at home [
26].
That sedentary behavior has increased over time is in keeping with previous research [
15,
17,
19,
23], while there is less agreement about trends over time in LTPA and active commuting. Likewise, the trend of increasing LTPA in more recent years in Canada is in line with studies from Canada [
19], Denmark [
20], and Finland [
17], and in contrast to two studies from Australia [
24] and Norway [
23]. Likewise, the higher participation in commuting activities is in accord with some studies [
19,
22], but not others [
17]. The increasing trend in participation in physical activities over time contrasts the increasing trend in obesity in the population. This paradoxical result may be explained by that leisure time and commuting-related activities only account for a small portion of the overall daily energy expenditure by adults, with physical activity at work accounting for most of the energy expenditure [
28]. It could also be that the levels of these types of activities are not enough to curb the obesity epidemic and overcome the effects of changes in eating patterns [
34,
35]. In addition to individual factors, the influence of physical and social environment on health behaviors is widely recognized [
36,
37]. Therefore, it is reasonable to hypothesize that changes in participation in physical activities is influenced by societal changes, such as social norms and/or policies with regards to physical activity over time. It is possible that health promotion efforts and increase awareness of the importance of exercise in the population over time has increased participation in physical activities. Unfortunately, we did not have data on this to test the hypothesis.
As has been found more generally, women were less likely to report participating in physical activities and more likely to be sedentary [
38‐
40]. We also found that those with higher SES (as shown by higher income and education) were more likely to report participating in active LTPA, but not in active commuting. As expected those who were obese were more likely to be sedentary and were also less likely to engage in active LTPA and commuting [
41,
42].
The inverse relationship between sedentary behavior and participation in physical activity concurs with a study combining cross-sectional data from 20 countries [
43]. However, a further study from Australia suggested the reverse: that sedentary activity was partially compensated by physical activity [
44]. The negative impact of modern changes in transportation, occupations and domestic activities on total energy expenditure along with increasing sedentary behavior has been noted [
15]. Recent findings also suggest that leisure time, in the context of sedentary lifestyles, is unlikely to be sufficient to prevent increasing population levels of overweight, obesity and chronic disease [
34]. In contrast, a recent meta-analysis found that high levels of physical activity attenuated the increased mortality risks associated with sedentary behavior [
45]. Furthermore, a prospective population-based study in the UK found that active commuting reduced the risks of cardiovascular diseases, cancer and all-cause mortality [
46]. Therefore, it is crucial to adopt a broader approach to understanding and influencing sedentary behavior in addition to increasing physical activity. One possibility is promoting walking or biking for transportation as a way to increase the overall physical activity levels in the population, and particularly for those in more sedentary occupations; and, at a more macro level, the importance of considering city planning, safety and the creation of appealing environments for walking and biking [
47,
48]. Perhaps another alternative is to examine the impact of breaking-up prolonged sitting time and how this might relate to increasing light and moderate-to-vigorous intensity physical activities. Studies suggest that there may be metabolic benefits to regularly breaking-up prolonged sitting time in addition to reducing overall sedentary time [
49,
50].This is particularly relevant at the workplace, where employers could develop and implement initiatives to improve and maintain the well-being of their workforce.
Limitations and strengths
An important limitation of this study is that our analyses are based on self-reported data. Studies have shown that self-reported measures are subjected to recall and measurement bias. We do not expect that the patterns seen over time to be affected by recall or measurement bias, as these biases are unlikely to vary over time. In addition, it has been suggested that self-reported physical activity measures may also be affected by social desirability bias [
51]. However, a systematic review of the literature comparing objective and self-reported measures for assessing physical activity indicated that studies show inconsistent findings: self-reported measures both over- and under-estimated accelerometer data [
52]. Combined data from the Canadian Health Measures Survey (CHMS) from 2007 to 2010 show that estimated minutes/day from the accelerometer and LTPA questionnaire were similar among those 12–59 years, while those 60 years or older over-reported their participation in LTPA [
53]. The study also showed overall weak correlations between self-reported and accelerometer LTPA. A comparison with American data showed discrepancies similar to those observed in CHMS [
54]. Furthermore, one study found that overweight or obese individuals overestimated energy expenditure from self-reported physical activity questionnaires [
55]. Another study comparing self-reported physical activity to accelerometer data found that the accuracy of self-reports was higher for men and for those with a lower BMI [
56]. We have somewhat accounted for the potential inconsistencies in self-reported physical activity by BMI and sex, as we have controlled for these variables in the models. Although cross-sectional studies have collected data with accelerometer [
52‐
54], we are not aware of longitudinal panel studies or studies examining age-period-cohort effects with synthetic cohorts (i.e. combining repeated cross-sectional surveys). Given the limited information on changes over time in more objective measures of physical activity, it is difficult to ascertain to what extent changes in self-reported measures reflect real changes in physical activity or if these changes are driven by reporting biases associated with awareness of the importance of physical activity for health. This is an area that merits further research.
In order to examine physical activity trajectories, participants were included if they provided information on at least three cycles of data collection and responded to the physical activity questions at baseline. We found differences related to age, income and BMI between participants included for analysis and those who were excluded. As such, the generalizability of our results may be limited by these differences. Another limitation of the study is the attrition due to drop-outs and mortality, particularly in the older cohort. We re-estimated the models by including indicators variables identifying those who died and those who dropped-out. We also re-estimated the models using the sample of participants with data in the nine cycles. Findings from these analyses did not change our conclusions.
The major strength of this study was the use of data from a large and representative survey of the Canadian population spanning 16 years, which allowed the examination of changes over time independently from the effects of aging and birth cohort. Another strength of the study is that the NPHS collected information on physical activities in three domains: leisure time, commuting, and daily activities, which allowed us to comprehensively examine the patterns of physical activities over the time period.