The effect of health promotion devices feedback on employee behavior

The effect of health promotion devices feedback on

employee behavior

Master Thesis, MSc Supply Chain Management

University of Groningen, Faculty of Economics and Business

June 24, 2019

Stephan Christophi

Student number: s2518619

Email:

s.christofi@student.rug.nl

Supervisor

Prof. dr. ir. J.C Wortmann

Co-assessor

dr. C. Xiao

Acknowledgment: I would like to thank PhD candidate A. M. Bonvanie-Lenferink for providing timely and accurate feedback whenever needed. I would like to further mention that this study is embedded in her ongoing work.

The effect of health promotion devices feedback on

employee behavior

Abstract

Table of Contents

Control over frequency ... 20

Conceptual model ... 22

Methodology: ... 23

Sample and data collection: ... 23

Measures ... 25

Results ... 29

Multiple regression analysis for feedback on health ... 30

Independent sample t-test for control on health and reflexivity ... 31

One-way ANOVA for frequency on health and reflexivity ... 31

Two-way ANOVA for control and frequency on reflexivity ... 31

Regression Analysis for reflexivity on health ... 35

Exploratory analysis ... 35

Analysis of T3 questionnaire ... 38

Discussion ... 42

Limitations and future research ... 45

Introduction:

In the modern world there are plenty of desk-based jobs, all requiring prolonged sitting times and poor physical activity, both leading to unfavorable health conditions (Pedersen, Cooley, & Mainsbridge, 2014). Over the past years, extended sitting time has become an

occupational habit and has shown to affect an employee’s health, increasing exposure to chronic diseases such as diabetes and obesity (Ryde, Brown, Gilson, & Brown, 2015). To mitigate such sitting behavior and increase workplace health, actions are required to be taken (Pedersen et al., 2014), actions that can be taken both at organizational and individual level.

Actions can be taken in the form of health promotion programs, keeping employees

productive and healthy. Such programs have shown to decrease absenteeism and health care costs which in return decreased employee-related costs (S.G., 2001). Although at a company level, this is already a benefit, this research wants to address the advantages of health promotion programs in providing the feeling of positive health among employees. This study aims to show the effect of a health promotion program, which provides controlled feedback to employees regarding their sedentary time, on an employee’s health. Employees are expected to reflect upon the feedback received and correct their sitting behavior, which in return will lead to a feeling of positive health. The degree at which people reflect on and adapt their behavior, known as reflexivity, is a pivotal factor on how effective these actions are but yet little is known on how to improve it (Schippers, Den Hartog, Koopman, & Van Knippenberg, 2008). One element which is taken in consideration to improve reflexivity within this study, as previously mentioned, is controlled feedback received. This control gives people the ability to regulate the amount of feedback received improving the learning effect of feedback.

Prior work introduced the notion that when a person has to change a health-related habit, (s)he must be inspired by a hint in order for him to employ in an aware decision about stopping an existing behavior. Such a hint can be given in the form of feedback, and when this feedback is linked to unfavorable effects, like chronical diseases, the chances of following a new behavior may grow (Aarts & Dijksterhuis, 2000). The ability though, to reflect upon a behavior and improve it at an individual level, was not exploited in the depth needed (Ellis, Carette, Anseel, & Lievens, 2014). Self-management thus, has an important role on whether employees accept and act upon feedback received concerning their sitting behavior (Lorig & Holman, 2003). The role of technology within self-management was also previously

addressed in the form of the quantified self (QS), where individuals can track physical and behavioral information about themselves (Swan, 2013). However, the active role of an employee in determining the frequency of feedback received has not yet been addressed. Moreover, feedback includes a repeating cycle with multiple steps, with each cycle

influencing following cycles (Manuel & Smither, 2002), showing the learning effect when receiving feedback.

This study is embedded in the work of PhD candidate Anne Bonvanie-Lefferink, who started the project and is still working on it. Within her study, an experiment using a health

promotion program was done, data was collected, and surveys were filled in by participants. Surveys were filled in both before and after the health promotion program took place, and the data collected from the surveys was used to answer the research question of this project.

Here we want to further investigate the effect that feedback has on an employee’s sitting behavior and whether an active role of an employee, controlling the frequency of feedback received, might influence the results. Additionally, addressing the importance of reflexivity at an individual level and how this can be influenced by the feedback received will also be a new insight within this study.

Literature Review:

Health:

Health as defined in 1948 by the World Health Organization (WHO) is “a state of complete physical, mental and social well-being and not merely the absence of disease or infirmity”. Chronic diseases have almost become a standard and by following the WHO definition of health, people with such diseases would be considered decisively ill. This underrates the human ability to self-manage and operate in such a way that will give the feeling of satisfaction, despite having a chronic disease (Huber, 2014). The need for complete health would, in this way, mean that most people are unhealthy for most of the time. Changing the definition of health into a more dynamic one is a demanding and complicated target, dealing with different stakeholders.

A recent study introduced a new definition of health, which was named “positive health” (Huber, 2014). This new definition, as explained by Huber (2014) could arguably be preferred due to the unrealistic character of the word “complete” which is neither measurable nor operational. In this work, health is described as “the ability to adapt, self-manage and cope with the challenges of life”. Within this new concept, the acknowledgement of a person’s strength in being responsible for his own health is highly embraced (Huber, 2014). Further, Huber divided positive health in six dimensions, body functions, mental wellbeing, life meaning, quality of life, participation and daily functioning which can all be influenced by maintaining healthy behaviors.

altered to increase health. It was previously mentioned that poor health can be caused from inadequate physical activity and that an interference of such behavior can reduce early unemployment due to health-related issues (van den Berg, Schuring, Avendano, Mackenbach, & Burdorf, 2010).

A healthy work environment was described by the WHO (2009) as one where there is continuous improvement in terms of health and well-being for all employees by considering for example personal health resources in the workplace. The WHO (2009) further mentioned ways to boost personal health within a workplace, such as for example providing healthy food options, promoting walking and being flexible on work breaks to encourage physical activity. Employees encounter a number of health risks at work, resulting among others in cancer, musculoskeletal, neurological, respiratory and mental disorders (WHO, 2009). Further, in 2009 it was estimated that 37% of back pain is associated with worksite risk factors which leads to work absence and thus economic loss (WHO, 2009).

An employee’s health and well-being is of integral interest for hundreds of millions of working people worldwide (WHO, 2010). This is supported by an abundance of data exhibiting that in the long term, firms that encourage and support employees’ health are part of the most prosperous and productive firms and additionally experience higher rates in employee retention (Grossmeier et al., 2016).

Chronic diseases, inadequate mental and physical health have been found to be factors influencing employees towards early retirement (Börsch-Supan, Brugiavini, & Croda, 2009; Harkonmäki, Rahkonen, Martikainen, Silventoinen, & Lahelma, 2006). On the other hand, employees with good mental health and job satisfaction found to be more likely to extend their retirement age (Browne et al., 2019; Virtanen et al., 2014). Further, low self-perceived health was also found to have a strong correlation with early retirement among European employees (van den Berg et al., 2010).

Although chronic diseases can lead to early retirement, treatment and self-management of such diseases have also advanced with the years (Gignac et al., 2019). Developing research displayed that employees are more frequently able to maintain employment for longer, even when having a chronic disease (Sokka et al., 2010). The Health Council in the Netherlands even encouraged the government to pay great attention on ways to assist employees in delaying retirement referencing sustainable employability in the beginning of people’s careers as an encouraging factor (van der Mark-Reeuwijk et al., 2019), showing the importance of such interventions for employee health. With both notions of health and health care evolving, moving towards the idea of preventing health problems from occurring rather than concentrating on curing a disease, the basic assumption is to enable an individual to monitor himself and self-manage his health with available tools (Swan, 2012).

Self-management

condition and to enforce the behavioral, emotional and cognitive reactions needed to retain an acceptable quality of life. In current years self-management has become a familiar term for healthy behaviors and behavioral interventions and intents to aid individuals to take an active role in maintaining a certain degree of wellness.

A previously conducted study examined self-management among employees in the sense of employee job crafting and how employees, without supervision, controlled their behavior in order to operate and act in preferable ways (Zeijen, Peeters, & Hakanen, 2018). Zeijen et al. (2019) examined the effect of self-management on the relationship between work engagement and job crafting behaviors and discovered that self-management strategies had a mediating effect.

The importance of self-management strategies on employee career outcomes was also researched (Murphy & Ensher, 2001). Results demonstrated that employees that used self-management strategies and self-set career goals expressed greater perceived success in their careers compared to those that did not (Murphy & Ensher, 2001).

The above studies indicated the effectiveness of self-management among employees. The present study wants to further investigate how self-management, with the help of technology, functions among desk-job employees in managing their own sitting behavior and ultimately their health.

Quantified self (QS)

A term emerging in big data science, were individuals use technology and sensors for self-tracking and collection of physical and behavioral information is the term “quantified self” (QS) (Swan, 2013). The term was created with the development of the Quantified Self movement in 2007 by Gary Wolf and Kevin Kelly (Kelly & Wolf, 2007; Lupton, 2013). Since then, the concept and the correlated movement became very attractive and matured swiftly (Lupton, 2013). This development emphasizes the increasing desire of many to learn about their behaviors by self-observing personalized data. Data collection and quantification aided by QS creates feedback loops for behavioral alteration and promotes sense capabilities that ordinary senses would not be able to build (Swan, 2013).

Wearable devices and mobile applications now, aid users to track parameters associated to well-being and health, such as sleep quality, diet, heart rate, mood and activity levels (Stiglbauer, Weber, & Batinic, 2019). Providers of such devices declare and assure that tracking aids in feeling better, sleeping better and moving more (Nokia, 2018). Several reports back this declaration, but trustworthy evidence is limited and benchmark evaluations are incomplete (Hermsen, Frost, Renes, & Kerkhof, 2016).

Considering the importance that health-related QS technologies have acquired in our society, attention must be given to solid empirical evidence on the suggested advantages of such devices on health-related parameters (Stiglbauer et al., 2019). The development of QS technologies brought attention to gamification strategies, with introductory evidence demonstrating that such principles could increase enjoyment and thus engagement (Sardi, Idri, & Fernández-Alemán, 2017). This was backed up by a more recent study where it was found that gamification designs seemed capable to prompt social experiences which led to emotional attachment which in return influenced the acceptance of a tracking-self device (Hassan, Dias, & Hamari, 2019).

greater engagement with the tracking and setting clear goals led to increased physical health and positive emotions reported by the participants.

Literature on health and psychology claims that people are happier when healthy, but the significance of the quantified self within a workplace and its impact on health and wellbeing has not been elaborately analyzed within academic literature (Moore & Robinson, 2016). In a systematic review it was mentioned that for a successful utilization of wearable QS technology in a work environment, extensive teamwork is necessary between technology designers, researchers and firms (Khakurel, Melkas, & Porras, 2018). Some of the possible advantages of wearable QS devices for employees involve employee improvement in physical and mental health and a rise in operational efficiency and safety, which in return would lead to employer benefits (Khakurel et al., 2018).

A previously conducted study provided empirical contribution on elements supporting the acceptance of wearable QS devices among university employees and students (Khakurel, Immonen, Porras, & Knutas, 2019). Khakurel et al., found that employees are expected to shape a positive approach towards using QS devices and eventually become more motivated to use them if they learn that the device will stimulate them to be physically more active or by adding attractive engagement features to the device (e.g. gamification, data). Privacy and wearability seemed to be a concern among employees with straightforward effect on the motivation to use QS devices (Khakurel et al., 2019). Finally, Khakurel et al. (2019) conclude that QS devices would function well among employees as long as there were no privacy issues and the design was appropriate.

Sitting habits

Desk-job employees generally exceed the suggested maximum time working while keeping a sitting position (Goossens, Netten, & Van Der Doelen, 2012). Further, the increased number of desk-based employees also increased the number of people facing risks of developing a chronic disease due to extended sitting times and a lack of physical activity (Church et al., 2011). The immobile nature of sitting leads to muscular tension which in return can lead to fatigue, and with inadequate recovery, can raise long-term health problems (Hamberg-van Reenen et al., 2008). With the increase of such risks, an increase in health promotion programs within organizations has also been seen, with 81% of worksites providing such programs in 1990 and almost 90% in 2000 (Leeks et al., 2010). To avoid such health issues, office workers must improve their sitting behavior (Straker, Abbott, Heiden, Mathiassen, & Toomingas, 2013). A working-day of an individual having a desk-based job consists of lengthy sitting periods, which eventually becomes a habit, and to alter such a habit, individuals must be motivated to intentionally interfere (Aarts & Dijksterhuis, 2000). To build awareness of the risks of such sitting habits and motivate a change in employee behavior, employers are required to focus on health promotion programs within their organization. Such programs will aid in explaining the health risks developed by such a habit and more importantly, propose ways to alter such behavior (Pedersen et al., 2014).

The need for means that objectively measure sitting behavior within an office environment was previously mentioned, as well as the need for a measuring tool able to administer sitting patterns during an entire working day (Thorp et al., 2012; Zemp, Taylor, & Lorenzetti, 2016). A previously conducted study took these aspects into account by using a “smart” office chair which could objectively monitor the sitting behavior of desk-job employees and provide feedback to the user when a certain sitting period was reached (Roossien et al., 2017). Roossien et al., intended to explore how the feedback received created employee awareness on their sitting behavior and how this affected their sitting habits, but their results found no connection between feedback and altering ones sitting behavior. Another study tested the effect of a standing computer workstation to reduce sedentary time, showing that standing could be advantageous for some users, but while standing, perceived discomfort increased with time (Lin, Barbir, & Dennerlein, 2017). Lin et al. (2017) thus proposed that an adjustable sit-stand workstation could possibly help employees, giving them the ability to alter between sitting and standing. Considering that healthy sitting behavior is not clearly defined (Healy et al., 2013), instructions should include mixture of posture and duration leading to a shift between sitting and standing while working at a desk (Lin et al., 2017).

Enhanced understanding of the underlying framework of sitting behavior can propose and establish a plan of action to discontinue prolonged sitting time. Being aware, for instance, of the uninterrupted sitting periods of an employee, can establish ways directed towards that specific individual to change his behavior (Ryde et al., 2015). Such awareness can be generated by providing feedback to the employee on his sitting behavior. An indication of the benefits an employee can obtain by discontinuing prolonged sitting periods and increased daily energy usage was already mentioned in previous studies where it was suggested that periodical movement during working hours enhances an employee’s overall health (Pedersen et al., 2014).

H0A: Receiving Health Self-Management Applications (HSMA)-generated feedback on health behavior improves the health of employees

Reflexivity

The human ability to reflect on his/her behavior even though very beneficial, oftentimes is not exploited (Ellis et al., 2014). A term used to explain this ability is reflexivity and although seen as an iterative method divided in three components, reflection, planning and action (West, 1996), most research considers it as a unidimensional construct (Schippers, Hartog, & Koopman, 2007).

Reflexivity is in recent literature usually understood as a group process but can be operationalized at an individual level too (Schippers et al., 2007). At a team level, reflexivity is seen as the extent to which teams jointly reflect upon and modify their working procedures (West, 1996) and a combination of both feedback and reflexivity revealed a better overall team performance (Konradt, Schippers, Garbers, & Steenfatt, 2015). While reflexivity involves how something can be improved, feedback attempts to find information on how large the distance is between the current performance and the achieved goal (Schippers et al., 2007).

Further, Anseel et al., (2009), argued that when fused with feedback, reflexivity is more powerful, particularly when performance must be adapted. This interaction though, is still not explained in a satisfactory level, due to the lack of studies combining both feedback and reflexivity to test change in behavior and performance at individual level. One study exhibited that the combination of feedback with reflexivity, at team level, would improve task knowledge, transformation and finally performance (Konradt et al., 2015). Along with this finding, it was further suggested that feedback received preceding reflexivity was the most powerful in enhancing performance (Konradt et al., 2015), showing the mediating effect of reflexivity, at team level.

Kondradt et al. (2015) did give some first insights, showing a positive relation between feedback and reflexivity at team level, but the optimal reflexivity level has not been addressed nor found (Schippers, Edmondson, & West, 2014). Furthermore, it was discussed, due to the importance of reflexivity, that it is critical to grasp which factors motivate an individual to be reflective and establish a theory about the incentives of reflexivity (Schippers et al., 2008). Further, it was suggested that team reflexivity relies on what is discussed regarding previous activities and that team feedback can aid in making teams conscious of current information shortfalls (Johnson, Hollenbeck, DeRue, Barnes, & Jundt, 2013), showing the key role that feedback has on reflexivity.

A previous research discovered that future team performance can be enhanced by teams reflecting on previous activities (De Dreu, 2007). Following the same line of thought and by considering how self-management, in terms of health, involves one’s competencies to monitor his condition and implement behavioral reactions (Barlow et al., 2002), it can be assumed that self-management has an important role for reflexivity. Team reflexivity engages teams to continually renegotiate, which helps teams identify differences between the desired and actual situation which in return leads in process adaptation in order to attain a desired result (Schmutz, Kolbe, & Eppich, 2018). As with self-managing a healthy behavior, team reflexivity depicts that a team needs to take an active role to continuously observe and adapt in order to alter a process and maintain or even improve performance.

A previous study indicated that accepting feedback is a crucial mediator in the process of working with feedback received (Christensen-Salem, Kinicki, Zhang, & Walumbwa, 2018). In order for an individual to reflect, plan and take action he/she must first accept the feedback received. Similar with self-management and QS, feedback was detected to have a vital role for reflexivity too. Feedback was found to enhance adaptation and finally improve team performance (Konradt et al., 2015). It was also suggested that feedback stimulates teams in terms of reflection, adaptation and self-correction (DeShon, Kozlowski, Schmidt, Milner, & Wiechmann, 2004). This provides reason to consider reflexivity as a mediator between the feedback received and the positive health of desk job employees. To investigate the effect of reflexivity on employee health the following hypothesis has been constructed:

H0B: Individual reflexivity mediates the effect of feedback on the health of employees in such a way that the impact of feedback on health is more positive for employees with high reflexivity

Feedback

The deliberate distribution of information regarding a behavior in order to promote a behavior adjustment is frequently referred to as feedback (Van Velsor, Leslie, & Fleenor, 1997). Feedback has also previously been explained as “being at the heart of autoregulation” and a vital factor in advocating adaptation and change (DiClemente, Marinilli, Singh, & Bellino, 2001). It is a remarkable and dominant tool to influence learning and is a valuable principle in improving a person’s performance (Chhokar & Wallin, 1984) and has been characterized as a requirement for both motivation and learning within performance-oriented systems (Ilgen, Fisher, & Taylor, 1979).

further clarified that feedback can be used to guide an individual’s behavior, explaining the strong significance of feedback in changing health behaviors (DiClemente et al., 2001). Individual learning and performance can be highly influenced by receiving feedback in different theoretical contexts (Lam, DeRue, Karam, & Hollenbeck, 2011), one of which goal-setting theory (Locke & Latham, 1990). Feedback can contribute by giving information on how good an individual is performing in relation to the goal set. Additionally it can show what the effect of such change has on a person’s behavior (Lam et al., 2011). The character of feedback and the manner that feedback is given can alter its effect (Hattie & Timperley, 2007) which exhibits the need to investigate how feedback acts and how it can be used in the most adequate way to improve health.

Researchers previously mentioned personalized feedback in a study conducted to test the effect of such feedback in altering one’s level of physical activity. It was found that personalized feedback increased awareness concerning the lack of physical activity but, in the short term, participants did not change their behavior. This depicted the role and potential of personalized feedback as a tool in adjusting a behavior, but at the same time it revealed that feedback alone is inefficient (Godino et al., 2013). A linked field of exploration involves the power of technology (e.g. apps) in aiding personalized feedback and ultimately altering one’s behavior (Cook, 2019). Modern technology can produce such feedback without human clarification, which in return will provide low-effort instructions that inspire smaller, but more frequent changes in behavior (Rabbi, Pfammatter, Zhang, Spring, & Choudhury, 2015). Frequently interrupting prolonged sedentary times with a small walk can be such an instruction.

Frequency of feedback

A previous study hypothesized that high feedback frequency will positively influence performance and learning until a certain point (Lam et al., 2011). Lam et al. (2011) proposed that at the point when frequency hits too high levels, the information provided becomes staggering and individuals can not react to it and handle it. The crushing amount of information leads to the individual spending more time on processing the information given rather than on the task performance itself (Lam et al., 2011). Lam et al. (2011) further displayed that feedback frequency reveals a curvilinear, inverted-U relationship with task performance and effort to achieve the task (Figure 1), offering a greater role on an individual’s control over the amount of feedback he/she receives. According to Lam et al., we also hypothesized the following: H1: The relationship between feedback frequency and reflexivity displays an inverted-U relationship, in such a way that frequency of feedback strengthens the positive effect of feedback on reflexivity up to a certain point; beyond this point, the positive effects of frequency of feedback on reflexivity will experience a decline.

Control over frequency

Research done in recent years showed that control over when feedback is received has further enhanced the positive learning effects of feedback, although the reasons for benefitting from such control remain rather ambiguous. One reason posed was that an individual takes an active role in his learning process, developing a more thorough processing of the related information and increases motivation. It was further proposed that self-control conditions fit the exact wants of the learner (Chiviacowsky & Wulf, 2002) and realize the participants psychological need for autonomy (Hartman, 2007). This was backed up by another research, where the self-control and yoked group received feedback at the same frequency (Chiviacowsky, 2014). Chiviacowsky (2014) discussed that participants with no control performed worse than participants with control, declaring that an attentive look must be given to participant feedback autonomy.

Another experiment, done in the context of motor learning, further explored the effect of self-control on the reception of feedback (Hansen, Pfeiffer, & Patterson, 2011). Yoking a traditional group of participants with no control to a group with self-control on the reception of feedback demonstrated interesting results. Generally, the traditional group did not learn as effectively as the self-control group. Hansen et al. (2011) assumed that the process of making a decision on the amount of feedback received is a tool explaining the remarkable learning improvements of this situation. Further, they proposed that the imbalance between the possible learning moment and the administered feedback provided to the yoked group led to a decrease of feedback retention.

A recent study though, explained that the benefits of controlling the feedback received had smaller effects than what formerly was believed (Barros, Yantha, Carter, Hussien, & Ste-Marie, 2019). Barros et al. (2019) further suggest that informational factors better assist the advantages of controlled feedback rather than motivational factors.

The results mentioned above state that having control over the feedback received has positive effects, but the link between feedback frequency and control has not yet been mentioned. The effect of controlling the frequency of received feedback has not yet been studied but some first indications towards the importance of control were presented. One form of control was previously recommended by an experimental research, saying that people should be granted the option to turn off feedback warnings when needed (Van Dantzig, Geleijnse, & Van Halteren, 2013). In their study, Van Dantzig et al. (2013) also supported the notion that autonomy can be very useful by mentioning that feedback received at ill-timed moments is regarded as irritating and undermines autonomy.

It is thus expected that being autonomous and having the ability to control the frequency of feedback received will increase motivation and provide participants with an active role in the learning process. Further, when looking at figure 1, designed by Lam et al. (2011), participants will be able to actively restrict the frequency of feedback received, keeping it at a level with optimum results for each person individually. This in return, is expected to prevent feedback from becoming overwhelming and even irritating. Consequently, we hypothesize the following:

Conceptual model

Figure 2 below depicts the theoretical framework, presenting the relationships between the variables. Control over the frequency of feedback and frequency of feedback received are the two independent variables (IV’s) which are also the variables manipulated to test the change in reflexivity and health. Employee health is the dependent variable (DV). Individual

reflexivity, as mentioned above, is hypothesized to mediate the effect of feedback on an employee’s health.

Methodology:

To examine the hypotheses previously mentioned in the literature review a survey study was conducted. This research was both exploratory and confirmatory (theory testing). The sample used was subject to an experiment, providing state-of-the-art knowledge on the effect of feedback on a desk-job employee’s sitting behavior, showing the exploratory nature of the study. After the experiment was completed, the hypotheses were tested, illustrating whether theory is further confirmed (confirmatory) by the facts of the experiment (Karlsson, 2016).

Sample and data collection:

The experiment was conducted by Anne Bonvanie-Lenferink before the start of this project and took a period of six weeks. For the experiment a sample of 44 employees was used, all having desk-jobs at the Faculty of Economics and Business of the University of Groningen. All participants had to be healthy and work at least four days a week in order for the results to be generalizable for all desk-job employees. Different types of sensors were installed at each employee’s workspace, such as sensors in their keyboards to test their typing speed and others in their chairs to examine their sitting habits. Based on the output data provided by the sensors, participants would get a feedback alert on their phone advising them to have a short break from sitting and take a short walk.

The 44 participants were split into two groups for the experiment, giving a between-subjects design to the survey (Charness, Gneezy, & Kuhn, 2012). The first group, the self-control group, had control over the frequency of feedback received, having the ability to choose on the frequency of the alerts they would receive. The second group, the yoked group, had no control over the frequency of feedback. Each participant within the yoked group was coupled to one participant of the control group and would get the same number of alerts as the participant they were linked to.

developed by Anne and had to be filled, one before (T1) and one after the experiment (T2) took place, depicting that the survey had a within-subject design as well (Charness et al., 2012). The choice of which participants should be yoked together was done with the help of the first questionnaire. Within this questionnaire questions were provided on how

participants get to work, the distance they travel to get to work, how often they work and how physically active they are on a daily basis. This helped in making groups of two

participants with relatively same lifestyles. Additionally, each participant had to fill in a third, briefer questionnaire (T3) after each week at which the experiment took place. The data collected from these questionnaires would aid in giving information on the level of reflexivity of each participant in the course of the six-week experiment. All questions associated with reflexivity were previously developed and used in a previous study (Schippers et al., 2007), demonstrating the validity of these questions. To measure health, questions were used which were taken from a previous study conducted by the Institute of Positive Health

(MijnPositiveGezondheid). Within that study six dimensions of health were measured, but our study took only three into consideration, namely daily functioning, quality of life and mental wellbeing. See Appendix A for an overview of all questions used.

The three constructs needed for analysis where health as a dependent variable (DV), reflexivity as a moderator and feedback as the independent variable (IV). Further, feedback was divided into two parts, providing two IV’s. One was the frequency of feedback received by each participant, with frequency having three levels, high, medium and low. Low got 1-3 SMS’s per day, medium 4-8 and high 9-12. The second IV element was whether a participant had control or not over the frequency of feedback received. It must be clearly indicated that feedback was not present in the T1 questionnaire. The reason for this was because T1 was filled in before the participants received any kind of feedback.

Measures

The first step done before the analysis could start was the recoding of the questions that were asked in a negative way. The questions recoded within each questionnaire can be seen in table 1. It should be mentioned here that after the initial recoding, reverse coding took place for all questions Q2, Q3 and Q4 of T2. The reason for this was because initially these questions had a reverse Likert scale in comparison with T1, with 1 representing “totally agree” and 5 representing “totally disagree”.

Table 1 Recoded items

Questionnaire Questions recoded

T1 Q33.3 and Q35.1

T2 Q2.3 and Q4.1

T3 Q5.1

For the analysis to start, the main variables had to be constructed. As mentioned in the literature review, health and reflexivity were split in six and three smaller elements

respectively. Reflexivity in this study was also split three parts, namely reflection, planning and action (West, 1996). Health on the other hand took only three of the six dimensions proposed by Huber (2014), namely daily functioning, quality of life and mental wellbeing. Which questions are associated to which of the six elements of health and reflexivity can be found in Appendix C.

To test internal consistency, a reliability analysis was conducted, giving a Cronbach’s Alpha value for each of the six elements used. All elements except action for both T1 and T2 had a Cronbach’s Alpha higher than a = 0.7. For the action dimension Q35.1 and Q4.1 were removed from T1 and T2 respectively giving a Cronbach’s Alpha higher than a = 0.7. These high Alpha values implied that all items within each construct are correlated and measure the same construct consistently.

Table 2

Since all items showed high reliability, a new variable was created for each element by taking the mean value of all items within each construct. The reason for calculating the mean for each element of health and reflexivity separately was done in order to avoid giving more weight to one element over the others. After calculating the mean value for all variables shown in table 2, a mean value was calculated for the final constructs of health (DV) and reflexivity (mediator). Following, the mean value of both health and reflexivity in

questionnaire T1 was subtracted from the mean value of health and reflexivity in

questionnaire T2. This was done in order to show the change in health and reflexivity once the participants received feedback. The new variables used for analysis were named health change (DV) and reflexivity change (mediator). Next, participants with control over the frequency of feedback got a value of 1 for the variable control and participants without control got a value of 0. Further, in order to group participants in terms of frequency,

different values were given to the different levels of frequency. Low frequency got a value of 1, medium frequency a value of 2 and high frequency a value of 3. This was done so that participants from both groups (no control and self-control) would be also categorized according to the different levels of frequency. For participants in the self-control group that switched in frequency level during the experiment the average frequency was calculated and used in the analysis. The participant in the no-control group got the same level of frequency as the self-control participant they were yoked to. The new variables created are shown in table 3 together with some descriptive statistics for each variable.

Table 3

Descriptive Statistics

Variable N Minimum Maximum Mean Std. Deviation

Results

The goal of this study was to test whether feedback given to employees concerning their sitting behavior had an effect on their reflexivity and ultimately on their health. Two properties of feedback were taken into account, the frequency of feedback given to each participant and whether or not a participant had control over the frequency of feedback received.

The first step in the analysis tested the effect of feedback frequency on reflexivity and

whether this effect was non-linear as assumed by H1. If the effect was indeed non-linear, the variable of frequency would have to be handled as such in the analysis to follow. In order to do so a new variable was created named frequency squared which was used in a curvilinear regression. Before the regression was conducted, only the cases of participants without control were selected, as H2 hypothesized that frequency of feedback received for self-control participants would have a linear effect on reflexivity and health. The results (table 4) were not statistically significant and thus did not provide enough evidence to back up the assumption that there is a non-linear relationship between frequency and reflexivity (p = 0.238) and thus all variables used for further analysis were seen as linear.

Table 4

Curvilinear regression analysis frequency on reflexivity Dependent Variable: Reflexivity change

Multiple regression analysis for feedback on health

Once the non-linear relationship was not proven, a regression analysis was conducted to test the effect of both independent variables, control and frequency on our dependent variable, health. The results show low correlations between the variables (Appendix D, table D1) and significance levels higher than the wanted p = 0.05 (Table 5, 6). These results suggest that the effect of control over the frequency of feedback received and the level of frequency of feedback received on health are not statistically significant. This could be enough to stop the analysis once there is no relationship seen between IV and DV. Even so, for the purpose of this study further analysis was conducted.

Table 5 Coefficients

Dependent Variable: Health change

Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) .224 .389 .577 .569 Control -.011 .280 -.008 -.039 .969 Frequency -.103 .172 -.115 -.600 .554 Table 6

Multiple regression analysis control and frequency on health Dependent Variable: Health change

Model

Sum of

Squares df Mean Square F Sig.

1 Regression .211 2 .106 .181 .836b

Residual 15.781 27 .584

Independent sample t-test for control on health and reflexivity

An independent sample t-test was conducted to examine whether participants with control over the frequency of feedback had differences regarding reflexivity when compared to participants without control. The analysis shows a higher mean for the yoked group (µ = 0.1262) when compared to the self-control group (µ = 0.0347). Looking at the significance level, this higher mean is not statistically significant (p = 0.757) (Appendix D, table D2, D3). The same analysis was done to test the difference regarding health between the control and yoked group. The yoked group shows a higher mean (µ = 0.0248) compared to the control group (µ = 0.0139), but again the difference is not statistically significant (p = 0.966) (Appendix D, table D4 & D5).

One-way ANOVA for frequency on health and reflexivity

Then, to examine the differences between the various levels of frequency in terms of

reflexivity a one-way ANOVA test was conducted. Although there is a difference on the mean values, the difference is not statistically significant (p = 0.720) (Appendix D, table D6 & D7). The same test took place to analyze the difference in terms of health between the three levels of frequency. Again, the mean values are dissimilar, but these differences are not statistically significant (p = 0.748) (Appendix D, table D8 & D9).

Two-way ANOVA for control and frequency on reflexivity

A two-way ANOVA test for performed in order to test the effect of both independent variables “control over the frequency of feedback received” and “level of frequency of feedback received” on the mediator, reflexivity change. The test conducted a 2 (control: yes vs. no) by 3 (low vs. medium vs. high) ANOVA test on reflexivity change. Evidently, the effect of control over the frequency of feedback received on the mediator reflexivity is not

Table 7

Tests of Between-Subjects Effects

Dependent Variable: Reflexivity change

Source

Type III Sum

of Squares df Mean Square F Sig.

Partial Eta Squared Corrected Model 1.093a _{5 .219 .312 .901 .061} Intercept .423 1 .423 .604 .445 .025 control .172 1 .172 .246 .624 .010 frequency .696 2 .348 .497 .614 .040 control * frequency .533 2 .266 .380 .688 .031 Error 16.815 24 .701 Total 18.088 30 Corrected Total 17.908 29

a. R Squared = .061 (Adjusted R Squared = -.135)

Looking at the means of reflexivity change between participants with control and no control it seems that participants with control have a lower mean than those without. This shows that participants with no control over the frequency of feedback received have a bigger change in terms of reflexivity compared to the self-control participants, and even though the difference is quite small, this was not expected (Table 8). Further, the results show that a medium level of frequency of feedback received has the best results when control is not taken into account, with a mean value of 0.303 (table 9). This can be a first indication towards H1, hypothesizing that a medium frequency would be most beneficial for most participants. If looking at figure 1, medium frequency would be assumed to be the top of the inverted-U relationship between frequency and reflexivity.

Table 8

Estimates “control”

Dependent Variable: Reflexivity change

Control Mean Std. Error

95% Confidence Interval Lower Bound Upper Bound

no control .202 .233 -.280 .684

Table 9

Estimates "frequency"

Dependent Variable: Reflexivity change

Frequency Mean Std. Error

95% Confidence Interval Lower Bound Upper Bound

"low" -.076 .253 -.599 .447

"medium" .303 .289 -.293 .899

"high" .143 .281 -.437 .722

Going further into the two-way ANOVA results, table 10 shows all 6 combinations of the 2 by 3 matrix and the respective means in reflexivity change. These results display that the greater mean change for participants with control is at high level of frequency, showing that H2 could be supported but still this is only an indication towards this hypothesis as the results are not statistically significant. Further, the strongest positive effect of feedback on participants without control is seen when the frequency level is medium (mean value = 0.581). This can be taken as a further indication towards H1 as explained before, but again with low statistical significance.

Table 10

Estimated marginal means Control * Frequency Dependent Variable: Reflexivity change

Control Frequency Mean Std. Error

95% Confidence Interval Lower Bound Upper Bound

Figure 3 shows the mean values depicted in table 10 in a clearer way, providing comparison between the self-control and no-control group in terms of mean reflexivity change for the different levels of frequency.

As can be seen here, with control, participants have greater levels of reflexivity as the frequency of feedback increases, giving indications towards H2, hypothesizing that when an individual has control over the amount of feedback received, a higher frequency of feedback strengthens the positive effect of feedback. Whereas for participants without control, reflexivity increases going from low to medium frequency, but once frequency reaches high levels, a decline in reflexivity is observed, as hypothesized by H1.

Regression Analysis for reflexivity on health

After testing the effect of control and frequency on reflexivity change (mediator), a test was needed to analyze whether the mediator had an effect on health change (DV). This was done with a linear regression analysis using health change as the DV and reflexivity change as the IV. Table 11 shows that the effect of reflexivity on health is not statistically significant, p = 0.133 which is greater than the wanted 0.05 p-value needed. The 13.3% chance that the relationship between reflexivity and health is random is rather small, indicating that there can be a relationship between the two. However, given the threshold of 5%, no conclusive assumptions can be made.

Table 11

Regression Analysis Reflexivity and Health Dependent Variable: Health change

Model

Sum of

Squares df Mean Square F Sig.

1 Regression 1.262 1 1.262 2.398 .133b

Residual 14.731 28 .526

Total 15.993 29

Exploratory analysis

as a moderator, while gender was excluded because after a correlation analysis it was seen that gender did not correlate with any of the other variables

Multiple Regression: Reflexivity on health with BMI as moderator

A multiple regression was conducted to test whether BMI had a moderating effect on the relationship between reflexivity and health. The results (p = 0.034) revealed a significant effect (table 12). BMI alone was not statistically significant, p = 0.161 (table 13). The interaction though between reflexivity and BMI was statistically significant (p = 0.028) showing the strong moderating effect (B = 0.095, t = 2.329) of BMI on the relationship between reflexivity and health. The results showed that participants with a high BMI generally scored higher in the relation between reflexivity and health when compared to people with a normal BMI (Appendix D, figures D1, D2 & D3). It seems that participants with a normal BMI, when reflecting upon the feedback received, do not perceive that their actions have an effect on improving their health. On the other hand, participants with higher BMI, once acting upon the feedback they receive regarding their sitting behavior, get the feeling that their actions have a relatively strong impact on their health.

Table 12 ANOVA

Dependent Variable: Health change

Model

Sum of

Squares df Mean Square F Sig.

1 Regression 4.342 3 1.447 3.385 .034b

Residual 10.689 25 .428

Table 13 Coefficients

Dependent Variable: Health change

Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 1.120 .736 1.522 .141 Reflexivity change -2.309 1.103 -2.517 -2.094 .047 BMI -.042 .029 -.245 -1.445 .161 BMI_reflexivity .095 .041 2.800 2.329 .028

Next, a regression analysis was conducted to test whether BMI had a moderating effect on both the relationships of reflexivity with control and reflexivity with frequency. Both analyses revealed that the results were not statistically significant and further that there was no moderating effect of BMI on the relationship between control and reflexivity (table 14) and neither on the relationship between frequency and BMI (table 15).

Table 14 ANOVA

Dependent Variable: Reflexivity change

Model

Sum of

Squares df Mean Square F Sig.

1 Regression .793 3 .264 .387 .763b

Residual 17.064 25 .683

Total 17.857 28

Table 15 ANOVA

Dependent Variable: Reflexivity change

Model

Sum of

Squares df Mean Square F Sig.

1 Regression .568 3 .189 .274 .844b

Residual 17.289 25 .692

Analysis of T3 questionnaire

It was previously mentioned that at the end of each week of the experiment, a third

questionnaire had to be filled in by all participants which measured reflexivity. The reason for doing so was in order to understand the change in reflexivity of each participant during the first 5 weeks of the experiment. This was checked by a two-way repeated measures ANOVA. Week 6 was excluded from the analysis because the no-control group did not fill in the T3 questionnaire in that week and thus could not be compared with the self-control group. The reason for providing these questionnaires weekly was mainly to test the effect feedback had on reflexivity during the experiment and if there was a change, negative or positive during the course of the experiment. It could be that for example, reflexivity would increase during the first two weeks and then decline again during the last weeks. These changes would not have been seen if we only took T1 and T2 in consideration as only the start and end

measurements would have been taken into account.

After conducting a reliability analysis to test for internal consistency and further to examine the relationship between the items used to measure reflexivity within questionnaire T3, it was found that questions Q3, Q4 and Q5 were used to measure reflexivity. Reflexivity was again split in three elements for the same reasons as explained in the T1 and T2 analysis. After reverse coding question 5.1 a reliability was tested and a PCA was conducted. The questions used to calculate each element and their equivalent Cronbach’s Alpha value is shown in table 16.

Table 16

Internal consistency and Factor Analysis Variable Cronbach’s Alpha

The reflexivity mean for each participant for each week was then calculated in the same way as for questionnaires T1 and T2. First, the mean value for reflection, planning and action was calculated separately and following, these 3 values were used to find the overall mean value of reflexivity for each participant for each week.

Figure 4: Profile plot for high frequency

Figure 5 depicts a profile plot of all participants without control on the frequency of feedback received for all different levels of frequency. It can be seen here, even though the results are not statistically significant, that low frequency shows overall better results in terms of

Figure 5: Profile plot for “no control” participants with different levels of frequency

As mentioned in the literature review both health and reflexivity are split in three smaller dimensions. Health dimensions are daily functioning, quality of life and mental wellbeing and reflexivity dimensions are reflection, planning and action. A correlation analysis was

Discussion

The behavioral experiment conducted failed to support the hypotheses constructed at a statistically significant level. Although not significant, the results provided some indications on the possible effect of feedback on altering the sitting behavior of desk-job employees and finally on improving their health. Further, some differences between the self-control and no-control group were seen for the different levels of frequencies at which feedback was received.

H0A, hypothesizing that feedback received by Health Self-Management Applications would improve health could not be supported, with slightly more participants (56,67%) reporting a positive effect on perceived health after receiving feedback (Appendix B, table B5). Further the effect of feedback frequency and control on health have shown to have very low statistical significance (table 6), suggesting this study cannot support the assumption that feedback positively influences employee health. Next, 70% of the participants reported a positive reflexivity change once receiving feedback (Appendix B, table B6), depicting that participants did alter their sitting behavior but not at the level expected. Reflexivity did not show to have a mediating effect on the relationship between feedback and health as hypothesized by H0B (table 11). Participants, after receiving feedback, did not act upon the alerts in an adequate manner in order to change their sitting behavior and obtain a feeling of positive health.

Looking further into the impact of the frequency of feedback received and the control over the frequency of feedback received on the level of participant reflexivity, there were no statistically significant results to propose that there is an effect (table 7).

There are some indications found though (table 10, figure 3), towards H1 and H2, but the difference in mean values are very small, indicating that the results should be interpreted with great caution. By looking at figure 3, the no-control group reported greater reflexivity change when the frequency of feedback received was at a medium level. If looking at figure 1 and H1, medium frequency could be seen as the top point of the inverted-U shape

was not supported in a sufficient level by the analysis. For this reason, the analysis continued by looking at the effect of frequency on reflexivity as a linear relationship, instead of a non-linear relationship as was expected. As for the control-group, it seems that as frequency of feedback increases, reflexivity increases too, as hypothesized with H2. Again, both indications suffer from low statistical significance and very small margins of difference between the different levels of frequency, making it hard to draw conclusions towards accepting H1 and H2.

Due to low statistical significance of the analysis conducted and not being able to sufficiently support the hypotheses formulated in the beginning of this study, the role of some control variables was tested. Gender seemed to have no moderating effect in neither of the

relationships discussed. BMI on the other hand, revealed a strong moderating effect on the relationship between reflexivity and health. It was shown that when individuals have a higher BMI and follow the feedback received regarding their sitting behavior, they perceive that their actions have higher impact on their sitting behavior and on their health compared to people with a normal BMI. This can be a contribution to studies to follow, but it must be acknowledged that within this research the number of participants with a high BMI was quite small (N=13, table B7).

After not being able to sufficiently support the hypotheses formulated with the help of the data collected by this behavioral experiment, the following step was to look into the

literature for possible reasons explaining people do not respond to feedback regarding their sitting behavior as expected.

A previous study revealed that the recognition of the seriousness of disrupting long sitting periods is low, proposing that promoting awareness is of high importance (Van Dantzig et al., 2013) in order to successfully change such a behavior. This can be a first indication in

Along with the many advantages of using self-tracking devices to deliver feedback for habit change, some disadvantages also seem to arise. Feedback in the form of text messages is much simpler to ignore than information given by another person (Ledger & McCaffrey, 2014). Additionally, Ledger and McCaffrey (2014) mention that solutions given by digital technology are effortlessly abandoned and this can be backed up by the fact that more than half of the people who possessed a fitness tracker no longer used it. Literature on sustained use of interventions on behavior change found that there is an acute fall in willingness to self-monitor yourself after 10-14 days (Burke et al., 2008) and a linear drop in using wearable technology (e.g Fitbit). During our experiment, many participants stated that the feedback they received was not accurate, which could justify a participant’s choice to neglect the notifications received and lead to results which do not support our hypotheses.

Four previously conducted studies, even though indicating limited positive effects within their descriptive results, experienced problems in the form of inadequate statistical power leading to null results (Cowan, Bowers, Beale, & Pinder, 2013; Pereira, Quintal, Nunes, & Bergés, 2012; Quintal, Pereira, & Nunes, 2012; Rodgers & Bartram, 2011). The same problem occurred within this study, with indications being shown towards the different hypotheses proposed, but with low statistical significance. This could arguably be explained by the relatively small sample size (Hermsen et al., 2016) used for this study. The low statistical power of the results should thus be handled with utmost attention, because there is an increased chance that the effect sizes will be exaggerated.

received at inconvenient moments. This in return would lead to a reduced effect of feedback on altering their sitting behavior and ultimately improving their health.

Finally, it might be argued that reflexivity is a personal characteristic trait and thus difficult to change within a 6-week experiment and thus could lead to unreliable results. However, it was previously found that once people received feedback, they showed a higher level of reflection than people who did not receive any kind of feedback (Konradt et al., 2015). This could indicate that there should be a change in reflexivity within the 6-week experiment, with participants filling in a questionnaire measuring reflexivity both before and after receiving feedback. It was also found that reflexivity is an evolution process referring to activities that people accomplish between performance events (Schippers, West, & Dawson, 2015). Within our experiment, reflexivity indicated how reflexive a person assumed he was before receiving any kind of feedback and how this developed after each week at which the experiment took place. By doing this, the change in terms of reflexivity for each participant was measured.

Limitations and future research

The low statistical significance of the results revealed that the effect of feedback on reflexivity and health was not strong enough in order to come to valid and strong conclusions. Some indications are shown though, which could lead to some interesting avenues for research to follow. Results suggested that participants with control on the amount of feedback received, even though slightly and not statistically significant, overall performed better than participants with no control. This could further be investigated with a bigger sample and with a more accurate feedback platform.

As mentioned in the discussion, people seem to ignore feedback provided by digital

modality could earn more attention, something out of the scope of this study. A previous study explained that SMS text messages, even though believed to be a practical and effective way to deliver feedback (Hall, Cole-Lewis, & Bernhardt, 2015), are very hard to deliver at the exact moment needed. Once there is a delay between the moment that a behavior occurs and the delivery moment of the text message, performance, in the form of behavioral change can be negatively influenced (Bittner & Zondervan, 2015). This suggests that research to follow, should take feedback timing into consideration when setting up an experiment similar to the one performed in this study.

The nature of the work done by desk-job employees at the University of Groningen usually require high cognitive skills, skill variety and information processing. Lam et al. (2011) proposed that this can be a reason for which participants perceive feedback to have a

negative effect on their work performance and thus do not react upon the feedback received in an adequate level. Future research can take the nature of the job done among desk-job employees as an indicator when studying the effect of feedback on altering sitting an individual’s sitting behavior.

Conclusions

References

Aarts, H., & Dijksterhuis, A. (2000). Habits as knowledge structures. American Psychological Association. https://doi.org/10.1037/0022-3514.78.1.53

Anseel, F., Lievens, F., & Schollaert, E. (2009). Reflection as a strategy to enhance task performance after feedback. Organizational Behavior and Human Decision Processes, 110(1), 23–35. https://doi.org/10.1016/j.obhdp.2009.05.003

Ashford, S. J., & Cummings, L. L. (1983). Feedback as an individual resource: Personal strategies of creating information. Organizational Behavior and Human Performance, 32(3), 370–398.

Barlow, J., Wright, C., Sheasby, J., Turner, A., & Hainsworth, J. (2002). Self-management approaches for people with chronic conditions: a review. Patient Education and Counseling, 48(2), 177–187.

Barros, J. A. C., Yantha, Z. D., Carter, M. J., Hussien, J., & Ste-Marie, D. M. (2019). Examining the impact of error estimation on the effects of self-controlled feedback. Human Movement Science, 63(December), 182–198.

https://doi.org/10.1016/j.humov.2018.12.002

Bittner, J. V, & Zondervan, R. (2015). Motivating and achievement-eliciting pop-ups in online environments: A user experience perspective. Computers in Human Behavior, 50, 449– 455.

Börsch-Supan, A., Brugiavini, A., & Croda, E. (2009). The role of institutions and health in European patterns of work and retirement. Journal of European Social Policy, 19(4), 341–358.

Browne, P., Carr, E., Fleischmann, M., Xue, B., & Stansfeld, S. A. (2019). The relationship between workplace psychosocial environment and retirement intentions and actual retirement: a systematic review. European Journal of Ageing, 16(1), 73–82.

Burke, L. E., Sereika, S. M., Music, E., Warziski, M., Styn, M. A., & Stone, A. (2008). Using instrumented paper diaries to document self-monitoring patterns in weight loss. Contemporary Clinical Trials, 29(2), 182–193.

Caljouw, Simone R, de Vries, R., & Withagen, R. (2017). RAAAF’s office landscape The End of Sitting: Energy expenditure and temporary comfort when working in non-sitting postures. PloS One, 12(11), e0187529.

Charness, G., Gneezy, U., & Kuhn, M. A. (2012). Experimental methods: Between-subject and within-subject design. Journal of Economic Behavior & Organization, 81(1), 1–8.

Chhokar, J. S., & Wallin, J. A. (1984). A field study of the effect of feedback frequency on performance. Journal of Applied Psychology, 69(3), 524–530.

https://doi.org/10.1037/0021-9010.69.3.524

Chiviacowsky, S. (2014). Self-controlled practice: Autonomy protects perceptions of

competence and enhances motor learning. Psychology of Sport and Exercise, 15(5), 505– 510.

Chiviacowsky, S., & Wulf, G. (2002). Self–control feedback: does it enhance learning because performance get feedback when need it? Res Q Exerc Sport, 4(73), 408–4015.

Christensen-Salem, A., Kinicki, A., Zhang, Z., & Walumbwa, F. O. (2018). Responses to

Feedback: The Role of Acceptance, Affect, and Creative Behavior. Journal of Leadership and Organizational Studies, 25(4), 416–429.

Church, T. S., Thomas, D. M., Tudor-Locke, C., Katzmarzyk, P. T., Earnest, C. P., Rodarte, R. Q., … Bouchard, C. (2011). Trends over 5 decades in US occupation-related physical activity and their associations with obesity. PloS One, 6(5), e19657.

Clark, N. M., Becker, M. H., Janz, N. K., Lorig, K., Rakowski, W., & Anderson, L. (1991). Self-management of chronic disease by older adults: a review and questions for research. Journal of Aging and Health, 3(1), 3–27.

Conner, M., & Norman, P. (1996). Predicting health behaviour. Buckingham. open university Press.

Conner, M., & Norman, P. (2002). Health behaviors. Health Psychology, 8, 1–37. Cook, W. (2019). The effect of personalised weight feedback on weight loss and health

behaviours: Evidence from a regression discontinuity design. Health Economics (United Kingdom), 28(1), 161–172. https://doi.org/10.1002/hec.3829

Cowan, B. R., Bowers, C. P., Beale, R., & Pinder, C. (2013). The stroppy kettle: an intervention to break energy consumption habits. In CHI’13 extended abstracts on human factors in computing systems (pp. 1485–1490). ACM.

effectiveness: a motivated information processing perspective. Journal of Applied Psychology, 92(3), 628.

De Wind, A., Geuskens, G. A., Ybema, J. F., Blatter, B. M., Burdorf, A., Bongers, P. M., & Van der Beek, A. J. (2014). Health, job characteristics, skills, and social and financial factors in relation to early retirement-results from a longitudinal study in the Netherlands.

Scandinavian Journal of Work, Environment & Health, 186–194.

DeShon, R. P., Kozlowski, S. W. J., Schmidt, A. M., Milner, K. R., & Wiechmann, D. (2004). A multiple-goal, multilevel model of feedback effects on the regulation of individual and team performance. Journal of Applied Psychology, 89(6), 1035.

DiClemente, C. C., Marinilli, A. S., Singh, M., & Bellino, L. E. (2001). The role of feedback in the process of health behavior change. American Journal of Health Behavior, 25(3), 217– 227. https://doi.org/10.5993/AJHB.25.3.8

Ellis, S., Carette, B., Anseel, F., & Lievens, F. (2014). Systematic Reflection. Current Directions in Psychological Science, 23(1), 67–72. https://doi.org/10.1177/0963721413504106 Gignac, M. A. M., Smith, P. M., Ibrahim, S., Kristman, V., Beaton, D. E., & Mustard, C. A.

(2019). Retirement Expectations of Older Workers with Arthritis and Diabetes Compared with Those of Workers with No Chronic Diseases. Canadian Journal on Aging/La Revue Canadienne Du Vieillissement, 1–19.

Godino, J. G., Watkinson, C., Corder, K., Marteau, T. M., Sutton, S., Sharp, S. J., … van Sluijs, E. M. F. (2013). Impact of Personalised Feedback about Physical Activity on Change in Objectively Measured Physical Activity (the FAB Study): A Randomised Controlled Trial. PLoS ONE, 8(9), 1–9. https://doi.org/10.1371/journal.pone.0075398

Goossens, R. H. M., Netten, M. P., & Van Der Doelen, B. (2012). An office chair to influence the sitting behavior of office workers. Work, 41(SUPPL.1), 2086–2088.

https://doi.org/10.3233/WOR-2012-0435-2086

Grossmeier, J., Fabius, R., Flynn, J. P., Noeldner, S. P., Fabius, D., Goetzel, R. Z., & Anderson, D. R. (2016). Linking workplace health promotion best practices and organizational financial performance: tracking market performance of companies with highest scores on the HERO scorecard. Journal of Occupational and Environmental Medicine, 58(1), 16– 23.