Time spent cycling, walking, running, standing and sedentary: a cross-sectional analysis of accelerometer-data from 1670 adults in the Copenhagen City Heart Study

To achieve reliable measurements of physical activity types and stationary behaviours, only individuals having ≥5 days of measurements with ≥16 h of recordings per 24-h day were included in the analyses, regardless of whether it was weekdays or weekend days. All days marked as ‘sick days’ in the diaries were excluded.

Participant characteristics and outcome variables were described using medians with the first and third quartiles (Q1-Q3) and frequencies with percentages (%) as appropriate. We used medians instead of means because some of the continuous variables had skewed distributions.

To identify potential sources of selection bias, we compared the characteristics of the individuals that 1) did not give consent to wear accelerometers with those that gave consent, and 2) did not fulfil the inclusion criteria (i.e., ≥5 days with ≥16 h/day of accelerometer data) with those fulfilling the criteria, by assessing 95% confidence intervals (CIs) of medians and proportions.

Differences between population groups (i.e., sex, age groups, levels of education and BMI) for time spent in physical activity types, stationary behaviours, bed, and number of steps taken per day were assessed using Kruskal-Wallis rank sum tests (i.e., a p-value <0.05 were considered to indicate a difference between groups). We used 95% CIs to assess which groups that were different. Similarly, differences in the number of individuals spending time above or below the pre-specified health-related thresholds between population groups were assessed using Pearson’s Chi-squared test with Yate’s continuity correction and 95% CIs for proportions. The CIs for medians and proportions were calculated using the normal approximation method and the Wilson’s score method, respectively [31].

The distribution of both cycling and running time were skewed with a high number of individual means being equal or close to zero. Hence, to better illustrate the distribution of cycling and running time among those performing these activities, we presented the median time with Q1 and Q3 among those with an individual mean >0 min and >10 s, respectively. These time thresholds were chosen to exclude individuals not cycling and to exclude running time estimates that most likely were the result of misclassification.

In the fifth examination, 4543 individuals chose to participate out of 9215 invited (participation rate: 49.3%). Of these, 2335 gave consent to wear accelerometers (participation rate: 51.4%). After processing the raw accelerometer data, data from 2019 individuals were available of which 1670 fulfilled the inclusion criteria (82.7% of 2019) (Fig. 1).

The characteristics of the study population are presented in Table 1. The median wear time of accelerometers was 23.8 h/day and the median number of valid days was 6 days. The final study population consisted of 57.4% women. The median age was 61.8 years (Q1-Q3: 48.6–72.6). Similar proportions had a vocational education, higher education or a university education (25.4, 25.2 and 27.3%, respectively). The median BMI was 25.2 kg/m². About 43% rated their general health as good. Finally, about 17% were smokers. Some variables had missing values; however, the proportion was for most variables <1%.

A higher proportion of participants not giving consent to wear accelerometers were aged ≥75 years, were classified as obese, had no education, had had a white-collar occupation or been housewife/house husband for the longest time since completion of education, were widow/widower, reported their fitness to be worse compared to their peers, and reported their general health to be less good or poor, compared to consenting participants. With regards to leisure time physical activity, a higher proportion of non-consenting individuals reported being sedentary, while a lower proportion reported “regular physical activity and exercise (moderate physical activity)” compared to consenting participants. For details see Table A2.1 in Additional file 2.

The participants that did not fulfil the inclusion criteria (i.e., ≥5 days with ≥16 h/day) were younger, and a higher proportion had a university education, were students, unmarried, non-smokers, and had a lower systolic blood pressure, while a lower proportion were previous smokers, compared to eligible participants. Furthermore, a higher proportion reported “mainly sedentary work”, while a lower proportion reported “sitting or standing, from time to time walking during work”. Finally, a higher proportion reported “more strenuous [leisure time] physical activity” compared to eligible participants. See Table A2.2 in Additional file 2 for details.

Accelerometer data from 316 participants were lost due to different reasons. The primary reasons were incorrect initialisation of the accelerometers (i.e., non-recordings), accelerometers lost in postal services, and <1 h of wear time (i.e., data was not processed if total wear time <1 h).

To our knowledge, this is the first study reporting accelerometer-based estimates about time spent cycling, walking, running, standing and sedentary in an adult general population. We found a long duration and high prevalence of cycling and walking among the study participants. However, many Copenhageners also spent a lot of time being sedentary.

A shorter duration of cycling, walking and running was found among older individuals, individuals with the lowest educational levels and individuals being overweight and obese. The longest duration of time spent sedentary was found among men, and individuals being older, overweight and obese, but no differences were seen between educational levels.

Considering the high proportion of older individuals in our study population (i.e., 42% ≥65 years), we believe the daily cycling time and the prevalence of cycling (i.e., 8.31 min/day among the 61% cycling and about 20% cycled ≥15 min/day) is relatively high, and reflects the strong cycling culture in Copenhagen, the “City of Cyclists” [12]. Cycling is well-documented to lower the risk of mortality [3, 4, 6, 17, 26], cardiovascular disease and type 2 diabetes [5, 7], and is associated with other health outcomes, such as lower BMI [33], and higher health-related quality of life among elderly [29]. The observed cycling is hence likely to have a considerable positive effect on the public health of residents of Copenhagen [14]. Looking beyond health, cycling has both economic [34] and environmental benefits [35].

To our knowledge, this is the first study reporting cycling time in a general population measured with accelerometers during a week. Consequently, comparison of our results with other studies is limited. However, the median cycling time of 8.31 min/day (58.17 min/week) among those cycling is slightly shorter compared to self-reported estimates found in other cohort studies, ranging from 25.7 min/day in Denmark [4] to 10.6 min/day in the Netherlands [36]. In contrast, the prevalence of cycling in our study (61%) is higher or similar compared to estimates based on self-reported data reported in other studies of Belgian, Danish and Dutch general populations, ranging from 43% [37] to 69% [6, 17, 37]. These differences may be explained by the poor agreement between self-reported and direct measurements of physical activity [18], and other factors such as differences in attributes of the built environment known to affect cycling levels [37].

We found that 88% of the study population walked fast (i.e., ≥100 steps/min) ≥30 min/day. For most individuals, a walking cadence of 100 steps/min corresponds to physical activity of moderate intensity [25]. Thus, almost 90% of our study population fulfil a large part of WHO’s physical activity recommendations of ≥150 min of moderate-intensity physical activity per week by walking only [38]. However, comparison of these numbers with recent global and regional estimates of insufficient physical activity [39] is challenging since our estimates include all walking during waken hours, and the WHO recommendations are based on the accumulation of MVPA in bouts of ≥10 min [38]. Future comparison may be easier, since the Physical Activity Guidelines for Americans now have dropped the bout-requirement [40]. Walking with moderate or higher intensity is known to lower the risk of premature mortality [26, 41] and has beneficial effects on cardiovascular disease risk factors [2]. Thus, our finding of a high walking time of at least moderate intensity highlights the potential of improving public health through the promotion of walking [42, 43].

Similar to cycling, there are few previous studies of the general population with comparable accelerometer-based measurements of walking. The median walking time (i.e., 82.6 min/day) of our study is longer than self-reported [41] and accelerometer-based (i.e., count-based) [44] walking estimates from general population studies from high- and upper-middle-income countries. Again, self-reported measurements have in general low agreement with direct measurement of physical activity [18]. However, the count-based estimates of walking time in the NHANES 2005–2006 are substantially shorter (i.e., sum of slow, moderate and brisk walking: 28 min/day) [44]. Even if the categories “purposeful steps” (i.e., 40–50 steps/min; 66.9 min/day) and “faster locomotion” (i.e., ≥120 steps/min; 1.5 min/day) is added, the estimated walking time is still considerably shorter [44]. Interestingly, the number of steps/day in the NHANES 2005–2006 (i.e., uncensored: 9685 steps/day) [44] is similar to our findings of 9288 steps/day. Some of the differences in walking time are hence likely explained by differences in the accelerometer position and processing of the data.

We found that 41% spent ≥10 h/day sedentary. Several studies indicate that a sedentary time of 10–11 h/day or more increase the risk of incident cardiovascular disease, type 2 diabetes and mortality [8‐10]. Thus, a relatively high proportion of the adult population of Copenhagen may be at increased risk of cardiometabolic disease and mortality. This may in particular concern those with concurrent low levels of fast walking (i.e., older individuals, those with lower educational levels, and overweight or obese individuals), since the detrimental effects from excessive sedentary behaviour are dependent on the level of MVPA [11, 32]. Hence, this risk should be seen in light of the previously discussed relatively high levels of cycling and walking fast, which may reduce the risks associated with excessive sedentary behaviour [11, 32].

Our median sedentary time of 579 min/day is similar [20, 44, 45] or shorter [46] than other accelerometer-based estimates from cohort studies of general populations from high-income countries. Importantly, these studies used count-based classification of sedentary time (e.g., <100 cpm), which may not be directly comparable to our results that are based on posture-detected estimation of sitting and lying. For example, lying, sitting, and standing are all likely to result in <100 cpm. Hence, the differentiation of standing from sitting and lying may partly explain why our estimates are slightly lower than those reported by Diaz et al. that defined sedentary behaviour as 0–49 cpm [46]. Furthermore, the prevalence of being sedentary ≥10 h/day (i.e., 41%) is higher compared to other accelerometer-based studies of the general adult population (e.g., 23–24%) [9, 20]. This can be a result of the relatively high proportion of elderly (i.e., who have a longer duration of sedentary behaviour) in our study population compared to other studies.

We found a shorter duration of cycling, walking and running among older individuals, individuals with the lowest educational levels and individuals with a higher BMI. Although reported in terms of overall physical activity, this is in agreement with previous studies [47‐49]. We also found that women spent more time in some physical activity types (i.e., longer duration of walking, walking fast, MVPA, and a higher number of steps/day), which is contrary to what has been reported in summaries of previous studies where male sex is associated with higher levels of physical activity [48, 49]. With regards to sedentary behaviour, we found a longer duration among men, individuals being older, and individuals being overweight and obese. These findings is in line with previous research [50, 51]. However, we did not find any differences between educational levels, which has been found in previous studies [50, 51].

Considering the participation rate in the fifth examination of the CCHS (49.4%) and the percentage wearing accelerometers (51.4%), the risk of selection bias should be acknowledged. Based on differences in self-reported leisure time physical activity between participants not giving and giving consent to wear accelerometers (i.e., those not giving consent reported more sedentary behaviour and less regular physical activity and exercise), it is possible that our group-level estimates of sedentary behaviour and more vigorous activity types (e.g., cycling, walking fast and running) are under- and overestimated, respectively.

The high validity of the Acti4 software in detecting physical activity types and body postures from thigh-worn accelerometers [22, 24] is a strength of our study. Detection of cycling by Acti4 is based on continuous pedalling. This means that interrupted pedalling for >15 s during a cycling trip (e.g., waiting at traffic lights or freewheeling) will be recorded as time spent sitting or standing depending on the position of the right thigh, and not cycling. This can explain some of the previously mentioned differences between self-reported cycling time and our cycling time estimates, since self-reported estimates most likely include the recalled total travel time (e.g., going “from A to B”).

We chose our inclusion criteria (i.e., ≥5 days with ≥16 h/day) to achieve estimates of physical activity and stationary behaviours with high reliability. This excluded 349 (17.3%) participants from the analyses that were slightly different, again leading to further selection of the study population. We acknowledge the trade-off between external validity of the results and reliability of the measurements. However, no differences in time spent in the activities were seen when we compared our findings with those based on a less conservative inclusion criteria (i.e., ≥1 day with ≥16 h/day); therefore, this should not have a significant impact on our findings.

Our results are based on the individual mean time spent in the physical behaviours across the measurement period. Given our inclusion criteria (≥5 days of measurements with ≥16 h of recordings per 24-h day), the measurement period would for most individuals include both weekdays and weekend days. We acknowledge that physical activity patterns may be different on weekdays and weekends, but believe that the investigation of this lies beyond the scope of this paper.

We do not have information about where the measured physical behaviours take place (i.e., geographical location). However, we believe that reaching risk groups can be achieved despite the lack of information about where they are physically active. With the information about how people are active, city planners could, for example, nudge the target group (and potentially all of us!) towards a more active lifestyle (e.g., active transportation by cycling or walking, climbing stairs instead of using escalators or elevators, etc.). Policy makers could support local grassroots initiatives (e.g., running communities) or sports clubs. Researchers and others could mobilise knowledge about easy ways to increase physical activity to the target groups (e.g., through interest groups, senior- or patient organisations). Finally, different social media channels may offer other possibilities to reach risk groups in society.

In the light of substantial evidence of health benefits from cycling and walking, the long duration and high prevalence of cycling and walking found in this study population may contribute to a substantial reduction in the risk of developing NCDs and mortality [2, 3, 5, 7, 16, 26, 41].

Our findings may reflect Copenhagen’s strategy and investments over the past two decades to increase active transportation [12]. We believe city planning has a great potential in creating active, healthy societies that facilitate physical activity as part of daily living, promotes health and prevents NCDs [1, 52]. This should be a high priority for policy makers globally. However, our results also show that many Copenhageners spend much time sedentary, and that individuals being older, those with a short education and individuals being overweight and obese are least active through cycling, walking and running. WHO's Global action plan on physical activity 2018–2030 has the vision of “more active people for a healthier world” (1). The data in the present study are highly relevant for stakeholders to tailor initiatives at different societal levels to promote physical activity among the least active residents of Copenhagen. As previously discussed, this requires both population-level and individually focused approaches entailing collaboration between different sectors, such as policymaking, public health, city planning, business and industry, education, health care, mass media, and others [1, 40, 42].

Finally, these data provide unique opportunities to gain new knowledge about the role of physical activity types and stationary behaviours in both the development and prevention of NCDs. For example, by linkage of these estimates with national register data and by testing associations with risk factors for NCDs.