Keywords

1 Introduction

Sedentary Behavior is defined as any waking behavior characterized by an energy expenditure ≤1.5 metabolic equivalents (METs), while in a sitting, reclining or lying posture. A metabolic equivalent is deemed to be 3.5 ml O2/kg/min in adults without mobility impairment or chronic disease [8]. The various types of physical activities can be categorized into sedentary behavior (1.0–1.5 METs), light-intensity (1.6–2.9 METs), moderate-intensity (3–5.9 METs), and vigorous-intensity (≥6 METs) activities. Table 1 describes the metabolic equivalent of several physical activities in an office [9].

Table 1. Metabolic equivalents of physical activities in an office.
Full size table

Sedentary lifestyle has been reported to have severe negative effects on health, and this is now a global issue. A previous study of 17,013 Canadians suggested that people who spent the majority of their time sitting were 50% more likely to die in early ages than those who sit the least. The results of this study remained similar even after sex, age, smoking status and body mass index were controlled [10]. In Australia, more than half of the adult citizens are reported to be too inactive, and physical inactivity has contributed to the second major cause of cancer, following that of smoking [11].

As a matter of fact, office workers have already paid attention to their sedentary behaviors. A recent survey has indicated that prolonged periods of sitting has raised concerns among office workers, 88% of whom believed that time spent sitting is directly associated with their health status. Musculoskeletal complaints, followed by general health and obesity, were identified in the survey as their most significant considerations [12]. In particular, taking breaks is essential for office workers’ well-being. It is clear that break activities during working hours may maintain productivity over the entire working day [13]. Office workers may need to have a break from their work to restore energy and remain focused [14]. It is also probable that they self-interrupt or rest to manage their workload and productivity [15].

Peer effect, in which the behavior of an individual is affected by the behavior of their peers [16], have been evidenced to exist in the workplace in terms of wages, absenteeism, entrepreneurship decisions, and investment decision making in recent studies [4,5,6,7]. Nevertheless, little research has addressed whether peer effects have an impact on the sedentary behaviors of office workers. In this paper, peer effects on the resting behaviors of office workers, refer to how their resting behaviors are influenced by their peers, as including when to take a rest and what to do during the break periods are influenced by. This implies that resting decisions do not depend on one’s physiological needs in isolation, as individuals may take breaks when they realize that their peers are doing so. The purpose of this paper is to propose a model to quantify the peer effects on office workers’ break behaviors, and to explore how individual break behaviors can be affected by peers in the same workspace. Since sedentary behaviors are health hazards and sedentary office workers are keen to make a difference, the potential implications of this model will be discussed in the conclusion.

2 Methodology

We define an individual \( i \)’s peers as all the individuals excluding \( i \) in the same workspace \( w \). We then hypothesize a linear regression model to illustrate the existence of peer effects on individual resting behaviors:

$$ R_{i,w} = \varvec{ }\beta_{0} + \beta_{1} P_{i,w} + \beta_{2} H_{i} + \varepsilon_{io} $$

where \( R_{i,w} \) refers to the resting behaviors of the individual \( i \) in the workspace \( w \). \( P_{i,w} \) denotes the corresponding effects of resting behaviors of the peers. \( H_{i} \) is a vector of individual \( i \)’s physiological characteristics, \( \beta_{0} \) represents the constant term, and \( \varepsilon_{i} \) is the error term. \( \beta_{1} \) stands for the major coefficient of interest between peer effects and resting behaviors. As the duration of hourly rest periods can be estimated by oxygen uptake [23], it can be reasonably expected that physiological condition (such as age, gender, cardiovascular states) is a dependent factor of one’s resting behaviors. Therefore, the purpose of the following experiment is predominately to examine the relationship between \( R_{i,w} \) and \( P_{i,w} \).

To quantify the above items, we introduced a parameter, time interval \( m \) (unit: minutes), to quantify the peer effects among individuals. It designates the time difference between the starting time of the two nearest successive rest events of individual \( i \) and \( i \)’s peer \( j \), assuming that the earlier event has an impact on the later one. Figures 1 and 2 may explain it.

Fig. 1.
figure 1

An illustration of time interval m (j’s rest behavior has an influence on i)

Full size image
Fig. 2.
figure 2

An illustration of time interval m (i’s rest behavior has an influence on j)

Full size image

By using time interval \( m \) and the frequency of each time interval C(m), \( P_{i,w} \) can be calculated as the reciprocal of the weighted average of the time intervals (\( j \)’s rest behavior has an influence on \( i \)).

$$ P_{i,w} (t) = \frac{1}{D}\mathop \sum \limits_{1}^{D} \left( {\frac{{\mathop \sum \nolimits c(m_{j,i,w} )}}{{\mathop \sum \nolimits (m_{j,i,w} \cdot c(m_{j,i,w} ))}}} \right) $$

where \( t \), represents the controlled time threshold i.e. ignoring the time interval that is longer than \( t \); \( D \) refers to the number of observed days; \( m_{j,i,w} \) denotes the time interval that \( j \) influences \( i \) in the workspace \( w \), under the condition that the rest event of \( j \) happened earlier than the rest event of \( i \); and \( c(m_{j,i,w} ) \) designates the frequency of \( m_{j,i,w} \). The larger the results of \( P_{w,i} (t) \) is, the easier it is for \( i \) to be influenced by the peers. \( R_{i,w} \) can be calculated by dividing the average frequency of resting events by the average proportion of resting duration to the total working hours of individual \( i \):

$$ R_{i,w} = \frac{H}{D}\mathop \sum \limits_{1}^{D} \left( {\frac{{f_{i,w} }}{{t_{i,w} }}} \right) $$

where \( H \) represents the regulated working hours of the workspace \( w \), \( D \) refers to the number of observed days; \( f_{i,w} \) and \( t_{i,w} \) are the total amount of frequency and duration of \( i \) to take breaks in the workspace \( w \) respectively. The larger the result of \( R_{i,w} \), the more likely \( i \) is to take breaks.

Similarly, we define Individual Power (\( I_{i,w} \)) as the ability of individual \( i \) to influence the resting behaviors of \( i \)’s peers in the workspace \( w \) and \( R_{(i),w} \) as the resting behaviors of individual \( i \)’s peers in the workspace \( w \). \( I_{i,w} \) can be expressed as the reciprocal of the weighted average of the time intervals (\( i \)’s rest behavior has an influence on \( j \)):

$$ I_{i,w} (t) = \frac{1}{D}\mathop \sum \limits_{1}^{D} \left( {\frac{{\mathop \sum \nolimits c(m_{i,j,w} )}}{{\mathop \sum \nolimits ({\text{m}}_{i,j,w} \cdot c(m_{i,j,w} ))}}} \right) $$

where \( t \) represents the controlled time threshold (i.e. ignoring the time interval that is longer than \( t \)); \( D \) is the number of observed days; \( m_{i,j,w} \) denotes the time interval in which \( i \) influences the peer \( j \) in the workspace \( w \), under the condition that the rest event of \( i \) happened earlier than the that of \( j \); and \( c(m_{i,j,w} ) \) designates the frequency of \( m_{i,j,w} \). The larger the result of \( I_{i,w} (t) \), the more power \( i \) have to influence the peers.

\( R_{(i),w} \) can be expressed by dividing the average frequency of resting events by the average proportion of resting duration by the total working hours of all individuals excluding \( i \):

$$ R_{(i),w} = \frac{H}{D}\mathop \sum \limits_{1}^{D} \left( {\frac{{f_{(i),w} }}{{t_{(i),w} }}} \right) $$

where \( H \) represents the regulated working hours of the workspace \( w \), \( D \) is the number of observed days; \( f_{(i),w} \) and \( t_{(i),w} \) are the average frequency and average time for taking breaks in one day of all the individuals in the workspace \( w \), excluding \( i \). The larger the result of \( R_{(i),w} \), the more likely \( i \)’s peers are to take breaks.

By exploring the relationship between \( R_{i,w} \) and \( P_{i,w} \), we can identify the role of peer influence on individual resting behaviors. By contrast, the role can be double proved by testing the relationship between \( R_{(i),w} \) and \( I_{i,w} \), as individual \( i \) can be also treated as another individual \( j \)’s peer.

3 Research Experiment

Two-week field observations were conducted with 12 participants (F = 7, M = 5) of at three offices to detect whether individuals may take a break in response to others’ similar behavior. Office workers at a university who were healthy and had predominantly desk-based jobs were recruited by email. During the observation, researchers sat in the natural surroundings of the participants’ offices for two days, unobtrusively observing and recording their activities when they left their seats to take breaks and what they did during those periods. We also interviewed the participants at the end of each day to further explore their resting behaviors based on the observations. Several rules were set to identify and record resting behaviors.

  1. 1.

    The minimum unit used to record the time is one minute, the same as that of to record time interval. Therefore, the case may appear in which one independent rest event of one individual was recorded as happening at the same time that another rest event ended for the same individual, but they were recorded as two events rather than one whole event.

  2. 2.

    The recorded resting behaviors have been categorized into the following events, for which we also estimated the energy expenditure of each activity according to Table 1:

    1. (A)

      Looking for something/reaching for something (4.0 METs)

    2. (B)

      Talking with colleagues (1.8 METs)

    3. (C)

      Going to the toilet (3.0 METs)

    4. (D)

      Filling containers with water (3.0 METs)

    5. (E)

      Standing for a rest/stretching (1.8 METs)

    6. (F)

      Answering the telephone (standing/walking around) (1.8 METs)

  3. 3.

    A multitasking rest event was recorded as a single rest event. For example, refilling a teacup while talking with a colleague.

4 Results

Our finding indicates that one office worker’s resting behaviors may result in another one or more other office workers to show similar behavior. Such events have been witnessed frequently, confirming the existence of peer effects on the resting behaviors of sedentary office workers.

The statistical data provides strong support for the observational findings. We plotted a diagram of the frequency of each time interval (\( C(m) \)) against occurred time intervals (\( m \)) of all participants during the observations. Figure 3 clearly demonstrates that it was when the time interval approached 0 that the successive events were most likely to take place. The frequency tends to decline dramatically with the growing time interval. A case that was observed on numerous occasions supports the perspective, where one person left the seat to go to talk to another person, causing that person stood up and replied. Another example was when one individual went to refill his/her teacup and invited his/her colleague to do so as well, so both of them were prone to filling their cups with hot water. These two examples demonstrate how an office worker may have an almost immediate influence on the activities of their peers by verbal invitation or enquiry. In addition, a previous paper has suggested that there are probably many circumstances in which office workers invite each other to take breaks [17].

Fig. 3.
figure 3

Frequency of time interval occurs during the observations

Full size image

Table 2 demonstrates the linear regression estimates of the impact of peer effects on resting behaviors under certain time interval threshold controls. These controls were based on Fig. 3, where saltation occurs (at a time interval of approximately 5, 10, 15, 20, 25 and 30 min). The results in row (1) show that peer effect (\( P_{i,w} \)) is a significant indicator for predicting individual resting behaviors (\( R_{i,w} \)) when the time interval threshold is at approximately 20 min. Row (2) describes how the reasonability of the above conclusion can be confirmed by the relationship between the peers’ resting behaviors (\( R_{(i),w} \)) and individual power (\( I_{i,w} \)) under the same control. The conclusions are in conformity, since peers’ resting behaviors can also be predicted and indicated by the effects of individuals.

Table 2. Linear regression estimates of resting behaviors and peer effects
Full size table

5 Conclusion and Recommendation

In conclusion, it is evident that peers have a noticeable effect on individual break behaviors within a certain range. The effects are mutual and significant, indicating that, in turn, an individual is also able to affect the break behaviors of peers. Thus, certain design implications are derived regarding the application of peer effects on office workers to reduce their sedentary behaviors.

Social interactions are direct influential factors on rest frequency. Taking some offices as an example, it was clear that they had a culture of staff inviting each other to go for a drink break, while others demonstrated the practice in which that one person would bring drinks for the peers who remained sitting for a long time [17]. The promise of peer effects provides policy makers with suggestions that can be employed to increase the overall performance by realizing peer groups [18]. Combined with this research, managers may apply the above model to evaluate the resting behaviors of their employees to test whether they are sufficiently active. The model also makes it possible to recognize employees who positively and actively influence peers’ resting behaviors, and those whose behaviors are easily affected by peers, following which managers may attempt to group them together in the same workspace. Consequently, the overall activity of the workforce is expected to be augmented.

Social norms are major barriers to behavior changes among office workers [19]. Team leaders may properly utilize the peer effect on employees to improve their well-being, as senior managers can implement tremendous shifts in the organizational norms [20]. For instance, the leaders themselves could take the initiative to spend less time siting -and thereby encourage their employees to follow suit. This may lead to health improvements for both leaders and employees.

Moreover, social networks, such as online discussion forums and social networking sites, showed a promising consequence regarding a decrease in prolonged sitting [21, 22] and another strong factor is that communication technologies can be employed to encourage peer effects [16]. A previous survey indicated that users are extremely unsatisfied with the current reminders such as pop-up windows on computer screens and wearable devices with vibrating inactivity alerts [17], it is therefore possible for designers to apply social media into the design of systems or products to reduce sedentary behaviors in the workplace by increasing the peer effect. One potential suggestion could be an application that enables a group of office workers to discuss the subject and thereby motivate each other to reduce their levels of physical inactivity.