Master s Thesis Human Resource Studies Tilburg University, Faculty of Social and Behavioural Sciences, 24 May 2017

Size: px
Start display at page:

Download "Master s Thesis Human Resource Studies Tilburg University, Faculty of Social and Behavioural Sciences, 24 May 2017"

Transcription

1 JOB STRESS, AND THE MODERATING ROLE OF PERCEIVED ORGANIZATIONAL SUPPORT 1 The mediating role of work overload in the relationship between team absenteeism and job stress, and the moderating role of perceived organizational support Master s Thesis Human Resource Studies Tilburg University, Faculty of Social and Behavioural Sciences, 24 May 2017 Name: Nina Frances Gibson ANR: Theme: Job stress in healthcare Project Period: September 2016 April 2017 Name of the supervisor: dr. R.S.M. de Reuver Second reader: dr. F.C. van de Voorde

2 JOB STRESS, AND THE MODERATING ROLE OF PERCEIVED ORGANIZATIONAL SUPPORT 2 Abstract Job stress is a growing problem in healthcare, leading to employee health issues, and lower quality of patient care. Based on the JD-R model, this study expected a positive relation between the job demand work overload and job stress. Moreover, a relation between team absenteeism and work overload was expected by creating a situation of understaffing. Work overload was expected to mediate the relation between team absenteeism and job stress. Moreover, based on the buffer hypothesis of the JD-R model, perceived supervisor support (PSS) was hypothesized to alleviate employees job stress. A multi-level design was chosen to quantitatively test the hypothesized relationships. Eight individual- and team-level variables were included as control variables. The final sample, which was a stratified sample based on team membership, consisted of 502 employees belonging to 82 teams, corresponding to a response rate of 24.98%. The findings indicate that individual work overload leads to higher levels of individual job stress. However, team absenteeism did not affect individual work overload, which implies that also the mediation of the team absenteeism job stress relationship by work overload was rejected. Moreover, the moderation of the relation between work overload and job stress by the PSS (measured on level 2) was insignificant. Finally, several limitations, suggestions for future research and practical implications are offered. Future researchers are recommended to further investigate the role of job satisfaction and team stability, which had significant correlations, and effects on team absenteeism, as well as on work overload and job stress. Keywords. Job stress, perceived supervisor support, team absenteeism, work overload, workload, JD-R model, multi-level research, multi-level analysis.

3 JOB STRESS, AND THE MODERATING ROLE OF PERCEIVED ORGANIZATIONAL SUPPORT 3 Introduction Job stress has received increasing attention throughout jobs and industries due to its negative health and wellbeing related outcomes (Cooper, Dewe, & O'Driscoll, 2001). Researchers have found evidence for the negative influence of job stress on physiological and mental health leading to e.g. depressions, anxiety and heart disease, when it is experienced over a longer period of time (Cooper et. al., 2001). In addition, it might influence unhealthy habits such as smoking or excessive alcohol consumption, which in turn can lead to decreased health (Johnson et al., 2005). Compared to other sectors, employees in healthcare are scoring particularly high on job stress (Bakker, Demerouti, Taris, Schaufeli, & Schreurs, 2003). A reason for this might be the recent financial pressure on hospitals, which has led to providing more complex care with less personnel (Weissman et al., 2007). As a consequence of the latter, individual levels of work load have increased, contributing to higher levels of job stress (Schaufeli, Bakker, & van Rhenen, 2009). Drawing on the widely recognized job demandsresources (JD-R) model, when job demands such as work overload exceed a certain limit, they can positively influence job stress and in turn cause health problems (e.g. burnout) (Bakker & Demerouti, 2007; Schaufeli et al., 2009). This process is also referred to as the health impairment or energy depletion process (Bakker & Demerouti, 2007). In the past, several researchers have confirmed that work overload impacts job stress positively (e.g. Schaufeli et al., 2009). While previous research has focused mainly on individual level antecedents of job stress, there is still a lack of research on how team processes can affect job stress (Camden & Ludwig, 2013). The present research attempts to fill this gap by examining team absenteeism as an antecedent of job stress. In hospitals, team absenteeism is a long recognized problem as it leads to poor patient care, and is accompanied by high administrative costs and overtime pay out (Camden & Ludwig, 2013). In the context of this research, it is expected that team absenteeism can produce a situation of understaffing within a team, which leads to work overload of individual team members (Cherry, Ashcraft, & Owen, 2007). The situation of understaffing is created when absent team members are not replaced and the team has to operate with less members, leading to work overload for the remaining team members. Besides, a situation of understaffing is formed when other team members cover absent colleagues, because these team members have to spend more personal time at work and, as a consequence, they may miss shifts themselves in order to compensate the extra time they were scheduled (Camden & Ludwig, 2013). The latter leads to again more team absenteeism. Because of the increased work overload caused by team absenteeism, team members might experience more pressure, which

4 JOB STRESS, AND THE MODERATING ROLE OF PERCEIVED ORGANIZATIONAL SUPPORT 4 causes increased job stress in turn (Bakker & Demerouti, 2007). Therefore, this paper argues that team absenteeism leads to more job stress through an increased work overload. In other words, work overload is expected to mediate the positive relationship between team absenteeism and job stress. When investigating how to reduce or prevent job stress, job resources as presented by the JD-R model offer possible solutions as they can help individuals to cope with high job demands (Bakker & Demerouti, 2007). Job resources are those social, physical, psychological or organizational features of the job that have the potential to either help individual employees achieving their job goals, reduce the negative effects of job demands or motivate personal growth, learning and development (Bakker & Demerouti, 2007). An important job resource in the context of job stress is perceived supervisor support (PSS), which can be defined as employees perceptions of how well their supervisor supports them and cares about their wellbeing (Rhoades & Eisenberger, 2002). When employees perceive to be highly supported, they are reciprocating by showing higher commitment, loyalty and effort (Rhoades & Eisenberger, 2002). Moreover, employees that experience a high PSS may feel better equipped to cope with higher job demands such as work overload (George, Reed, Ballard, Colin, & Fielding, 1993; Robblee, 1998). The choice of the job resource PSS is based on previous research that suggests that particularly supervisor support helps employees in healthcare in coping with job stress (e.g. Constable & Russell, 1986). However, those studies led to different results and, therefore, there is a need to investigate this effect further (Healy & McKay, 2000). If this study finds support for PSS alleviating the effect that work overload has on the level of employees job stress, this might help organizations to design and stimulate behaviours on the work floor that can lower stress levels, thereby increasing the wellbeing of employees. The aim of this research is twofold. First, it is the aim to find out more about the concept of job stress and how it is influenced by team absenteeism and the resulting work overload. According to the JD-R Model, job demands can elicit a health impairment process, leading to increased job stress and, in turn, to negative employee outcomes such as decreased performance and absenteeism (Bakker & Demerouti, 2007). Several researchers (e.g. Hakanen, Bakker, & Schaufeli, 2006) have found clear empirical evidence to support the health impairment process. However, instead of testing whether absenteeism is an outcome of job stress, which has been shown by several researchers before (e.g. Schaufeli et al., 2009), this research has a focus on team absenteeism as a potential antecedent of job stress. Thereby, the researcher takes into account dynamic processes in teams and, thereby, tries to fill a gap in research (Camden & Ludwig, 2013). As a second aim of this research, PSS is investigated as a possible moderator

5 JOB STRESS, AND THE MODERATING ROLE OF PERCEIVED ORGANIZATIONAL SUPPORT 5 in the relationship between work overload and job stress. Thereby, this study contributes to existing evidence on the JD-R model and the buffer hypothesis of job resources. While in the past research on job stress was mainly cross-sectional, this research includes data from two points in time, which can be referred to as a longitudinal design and, thereby, is a first step towards drawing conclusions on causalities between variables (Zapf, Dormann, & Frese, 1996). Based on the aforementioned, the following research question is formulated: To what extent is the relation between team absenteeism and job stress mediated by work overload and can PSS buffer the relationship between work overload and job stress? The next section will elucidate the expected relations by using the JD-R model. Moreover, the hypothesized relationships will be illustrated in Figure 1 at the end of the section. Theoretical framework The relation between work overload and job stress This section zooms in on the relation between work overload and job stress by using the JD-R model as theoretical underpinning. Job stress is defined as an individual's awareness or feeling of personal dysfunction as a result of perceived conditions or happenings in the work setting (Parker & DeCotiis, 1983, p.161). This definition implies that job stress requires a certain level of awareness of the individual experiencing it and that this feeling or awareness must be negative and unwanted. Thereby, job stress is clearly distinguished from the positive motivational excitement caused by challenging, but attainable goals (Parker & DeCotiis, 1983). Moreover, Parker and DeCotiis (1983) distinguished two dimensions of job stress, i.e. time stress and anxiety. While time stress refers to the feeling of being under time pressure at work, anxiety refers to feelings of discomfort related to the job such as one feeling guilty when taking time off from the job (Parker & DeCotiis, 1983). Work overload is defined as the perceived magnitude of work-role demands, and the feeling that there are too many things to do and not enough time to do them (Parasuraman Purohit, Godshalk, & Beutell, 1996, p. 280). It describes a situation in which an individual s workload becomes so high, that coping is merely possible by investing an extremely high amount of energy (Schulz, Kirschbaum, Prüßner, & Hellhammer, 1998). The JD-R model is a general model that can be used for a wide range of industries and professions and is well known in occupational stress research (Bakker & Demerouti, 2007). Drawing on the JD-R model, working conditions can be analysed with regards to their impact on employee outcomes such as performance and health (Bakker & Demerouti, 2007). Working conditions are divided into two broad categories: job demands and job resources. Job demands

6 JOB STRESS, AND THE MODERATING ROLE OF PERCEIVED ORGANIZATIONAL SUPPORT 6 are physical, psychological, social or organizational features of the job (e.g. work overload, work-family conflict, and demanding interactions with clients) that require a certain amount of effort, which is related to physiological and/or psychological costs (Bakker & Demerouti, 2007). Although job demands are not automatically negative aspects of a job, they can transform into job stressors when meeting them involves high effort from which the employee has not recuperated (Meijman & Mulder, 1998). Work overload can be classified as a job demand (e.g. Bakker & Demerouti, 2007). According to the JD-R model, a high level of job demands (e.g. work overload) elicits a health impairment process which can cause job stress (Bakker & Demerouti, 2007). Previous research confirms that work overload is an important antecedent of job stress (e.g. Maslach, Schaufeli, & Leiter, 2001). Therefore the following hypothesis is formulated: H1: Workload is positively associated with job stress. The relation between team absenteeism and work overload This section focuses on the relationship between team absenteeism and individual work overload. Teams can be defined as people working together to achieve something beyond the capabilities of individuals working alone (Marks, Mathieu, & Zaccaro, 2001, p.356). In healthcare, employees from different disciplines are working together to provide effective and safe patient care by relying on teamwork (Weller, Boyd, & Cumin, 2014). Team absenteeism occurs when one or several team members miss a scheduled shift (Huczynski & Fitzpatrick, 1989). In this study, team absenteeism refers to the relative amount of absent working hours of a whole team as compared to the total working hours of that team during a certain period of time. Absent durations for a certain period are measured as a ratio of absent working hours divided by the amount of total working hours as defined in the team members contracts (Schaufeli et al., 2009). Due to the dependencies between team members, the actions of one team member are affecting the actions of other team members (Camden & Ludwig, 2013). For example, when a team member is absent, other team members have to sacrifice personal time to cover the shift. The latter may cause more absenteeism in turn, as these team members want to compensate for their additional time spent at work (Camden & Ludwig, 2013). However, absent team members are often not replaced due to the extra costs this bears on the hospital. As a consequence, the team has to operate with less personnel, which leads to understaffing and causes work overload (Unruh, Joseph, & Strickland, 2007). In other words, when replacement is not possible, the team still has to fulfil the same amount of tasks and tasks can mostly not be delayed, especially not when the team s work is directly related to primary patient care (Unruh et al., 2007).

7 JOB STRESS, AND THE MODERATING ROLE OF PERCEIVED ORGANIZATIONAL SUPPORT 7 Previous research has confirmed that team absenteeism is a long recognized problem in healthcare, which causes understaffed units, staffing instability, and chaotic working environments, which ultimately results in lower quality patient care (Cherry et al., 2007; Unruh et al., 2007). Moreover, it harms the work environment and the morale of team members and is positively related to job stress (Unruh et al., 2007). Consequently, it is likely that the more team absenteeism occurs, the more severe the situation of understaffing is, and the more work overload will be experienced by team members. Therefore, it can be expected that team absenteeism has a positive relation with individual work overload, and the following hypothesis is formulated: H2: Team absenteeism positively affects the level of individual work overload. The mediation of the relation between team absenteeism and job stress by work overload As aforementioned, based on the JD-R model, work overload represents a job demand which, when exceeding a certain level, is related to higher levels of job stress (Bakker & Demerouti, 2007). Moreover, team absenteeism is thought to increase work overload by creating a situation of understaffing. Earlier research has shown that understaffing results in a higher patient-to-nurse ratio (i.e. one nurse has to handle more patients) and that the latter leads to increased levels of burnout (Aiken, Clarke, Sloane, Sochalski, & Silber, 2002). Therefore, it is expected that team absenteeism is leading to higher levels of job stress by causing increased levels of work overload. In other words, the relation between team absenteeism and increased job stress is expected to be (partially) explained by high work overload. This leads to the following hypothesis: H3: The relation between team absenteeism and job stress is partially mediated by work overload. The moderation of the relation between work overload and job stress by perceived supervisor support (PSS) This section focuses on perceived supervisor support (PSS) as a possible moderator of the relationship between work overload and job stress. According to the JD-R model, job demands and job resources are interrelated, which means that they can influence the effect of one another (Bakker & Demerouti, 2007). Job resources can influence the health impairment process by reducing the positive effect that a specific job demand such as work overload has on the level of job stress an employee experiences (Schaufeli et al., 2009). Therefore, job resources can help employees coping with job stress that is caused by work overload (Bakker & Demerouti, 2007).

8 JOB STRESS, AND THE MODERATING ROLE OF PERCEIVED ORGANIZATIONAL SUPPORT 8 Perceived supervisor support (PSS), which is the degree to which employees perceive that the supervisor values their contribution and wellbeing, is an important job resource in the context of job stress because employees that score high on PSS are found to be more satisfied with their jobs and show less stress signs such as burnout and fatigue (Rhoades & Eisenberger, 2002). Moreover, PSS is related to favourable organizational outcomes such as increased performance and reduced turnover (Rhoades & Eisenberger, 2002). PSS theory is grounded in Perceived Organizational Support (POS) Theory which is in turn based on Social Exchange Theory, stating that employees personify their organization and will reciprocate perceived support from the organization with higher commitment, loyalty and performance (Rhoades & Eisenberger, 2002). Employees infer from actions taken by their supervisors whether the organization cares about their well-being (Rhoades & Eisenberger, 2002). Especially in a hospital setting, it has been found that employees are more attached to their teams or departments, as compared to the organization as a whole (Veld & Van De Voorde, 2014). Therefore, it is expected that in the context of this research, perceived supervisor support is a more important job resource as compared to perceived organizational support, and the main focus will be on PSS. Based on the JD-R model, PSS is expected to reduce the positive effect that job demands such as work overload have on job stress (Bakker & Demerouti, 2007). Consequently, the employee is likely to be better able to cope with job demands when perceiving high supervisor support. In line with the latter, previous research suggests that PSS may have a buffering effect on the relation between job demands and job stress, because of the accessibility of material aid and emotional support in times of high job demands (George et al., 1993; Robblee, 1998). According to Cropanzano, Howes, Grandey and Toth (1997), there is a possibility that PSS has a direct effect on job stress. Moreover, Viswesvaran, Sanchez and Fisher (1999) argue that it seems plausible that job stress levels can be decreased by PSS in both high and low exposure to work overload. However, in line with the JD-R model, this study expects that a job resource such as PSS can help especially in high job demand situations because in low job demand situations, PSS is less salient as compared to a high demand situation (Bakker & Demerouti, 2007). To illustrate the latter, employees are more likely to seek their supervisors support when experiencing work overload. Therefore, job resources (e.g. PSS) can lessen job stress in situations of high job demands (e.g. work overload) because they help the employee in coping with it (Väänänen et al., 2003; Bakker & Demerouti, 2007). In situations of lower levels of work overload, supervisor support can still be useful for the employee, but to a lesser extent. Bakker, Demerouti and Euwema, (2005) deliver additional support for this assumption,

9 JOB STRESS, AND THE MODERATING ROLE OF PERCEIVED ORGANIZATIONAL SUPPORT 9 i.e. work overload results in lower levels of burnout when employees receive social support or have a good relationship with their supervisor, as compared to a situation of low social support. Therefore, based on the JD-R model and its empirical evidence, the following hypothesis is formulated: H4: The positive relation between work overload and job stress is buffered by PSS. Figure 1. Conceptual model Methods Design The present study is a quantitative empirical study with a longitudinal design. The researcher made use of company data from 2016, combined with survey data from Absenteeism percentages on team-level from 2016 were retrieved from the organization s archival. Moreover, the variables work overload, PSS and job stress were measured on an individual level by means of a survey in Procedure Data collection was performed in 2017 at a large hospital in the Netherlands. A stratified sampling design was used in order to have a sample with an appropriate representation from every team, thereby making sure that the sample is representative for the whole population. Moreover, random probability sampling was used to randomly select employees within the teams for the final sample. First, employees were sorted in ascending order per team and uneven

10 JOB STRESS, AND THE MODERATING ROLE OF PERCEIVED ORGANIZATIONAL SUPPORT 10 numbers in the list were selected for the sample. 50 percent of each team was selected, with a minimum of 10 and a maximum of 25 employees. Teams consisting of less than ten employees were drawn into the sample as a whole. Moreover, employees who had a dual employment were only chosen once for the sample based on the team for which they have worked most during the previous three months. Using the described procedure resulted in a sample consisting of a representative group in terms of size from every team, i.e. a stratified sample. Moreover, by choosing uneven numbers from the list, members are randomly drawn, also referred to as random probability sampling. By means of the Qualtrics online survey software, an including a cover letter explaining the aim of the study and a link to the survey was sent to the sampled employees. Participation was voluntarily. Sample Out of the 4826 employees working at the hospital, 2010 were included in the sample. Finally, 795 employees filled in the survey, which equals a response rate of 39.55%. List-wise deletion of missing values led to an effective number of 646 employees (32.14%). Further, teams where the absenteeism rates were unknown were excluded from the final sample. This resulted into a sample consisting of 502 employees (response rate = 24.98%), belonging to 82 teams. On average, the number of respondents within a team was 6.12, whereby the largest team contained 28 respondents and the smallest team 2 respondents. Of those 502 employees, the large majority of the respondents was female (84.5%). Moreover, the average age was years with a standard deviation (SD) of With regards to the educational background, most employees had a higher (43.4%) or lower (44.2%) vocational educational background (NL: HBO, MBO). The average working hours as written in the contract were hours (SD = 6.86). Average organization tenure was 17.5 years (SD = 11.15), and average team tenure was 11.7 years (SD = 9.17). Regarding type of team, 89.90% of the teams in the final sample were care teams (being in charge of primary patient care) and the remaining 10.10% were knowledge teams (supporting the whole organization in terms of e.g. finance, HR and IT related matters). The latter percentages are representative for the hospital, which contains around 81% care teams and 19% knowledge teams. Measures Job stress. Job stress was measured using a shortened version of the thirteen-item scale from Parker and DeCotiis (1983). 6 items were chosen based on their relevance in the organization s context. Three items represented the time stress dimension of job stress, while the other three items represented the anxiety dimension of job stress. A 5-point Likert scale was used and the response categories were: (1) totally disagree, (2) disagree, (3) neutral, (4) agree,

11 JOB STRESS, AND THE MODERATING ROLE OF PERCEIVED ORGANIZATIONAL SUPPORT 11 and (5) totally agree. Items were translated into Dutch. Examples of the included items are: Working here leaves me little time for other activities, and I feel guilty when I take time off from job (Parker & DeCotiis, 1983, p. 169). Factor analysis was found to be appropriate (KMO > 0.6, Bartlett s test of sphericity p <.001). Principal component analysis with Varimax rotation revealed one component with loadings above.30, explaining 52.70% of the total variance (Appendix A). Moreover, the scree plot confirmed the presence of one component. Therefore, construct validity was considered as good. Concerning reliability, Cronbach s alpha (α) was.816, which is considered as good (Evers, Van Vliet-Mulder, & Groot, 2000). Besides, all items contributed positively to the scale as all values for Cronbach s α if item deleted were lower than Cronbach s α of the construct when the item was included. Perceived supervisor support (PSS). PSS was measured using an eight-item scale from Rhoades, Eisenberger and Armeli (2001). While the original scale measured perceived organizational support (POS), this research adapted the wording, asking for perceived supervisor support instead. Moreover, while the original scale was measured on a 7-point Likert scale, ranging from (1) strongly disagree to (7) strongly agree, the present study used a 5-point Likert scale with the response categories: (1) totally disagree, (2) disagree, (3) neutral, (4) agree, and (5) totally agree. Items were translated into Dutch. Examples of the included items are: My supervisor really cares about my well-being and My supervisor cares about my opinions (Rhoades et al., 2001, p.828). Factor analysis was found to be appropriate (KMO > 0.6, Bartlett s test of sphericity p <.001). Principal component analysis with Varimax rotation revealed one component with loadings above.30, explaining 62.66% of the total variance (Appendix A). Moreover, the scree plot confirmed the presence of one component. Therefore, construct validity was considered as good. Concerning reliability, Cronbach s alpha (α) was.913, which is considered as good (Evers et al., 2000). The Cronbach s α if item deleted of the 8 th item was larger than Cronbach s α, namely.924. It is a contra-indicative item that was recoded before the analysis: If given the opportunity, my supervisor would take advantage of me (Rhoades et al., 2001, p.828). The difference of.11 was, however, considered as negligible and the item was not deleted from the scale. Work overload. Work overload was measured using a six-item scale of the Questionnaire on the Assessment and Experience of Work 2.0 (QEEW 2.0) (Van Veldhoven, Prins, Van Der Laken, Dijkstra, 2015). The QEEW 2.0 scale is a reduced scale of the original QEEW scale for work overload by Van Veldhoven and Meijman (1994), which contained 11 items. Responses were measured on a 4-point Likert scale with the following response categories: (1) never, (2) sometimes, (3) often, and (4) always. Examples items are: Do you

12 JOB STRESS, AND THE MODERATING ROLE OF PERCEIVED ORGANIZATIONAL SUPPORT 12 have to work very fast? and Do you have too much work to do? (Van Veldhoven & Meijman, 1994). Factor analysis was found to be appropriate (KMO > 0.6, Bartlett s test of sphericity p <.001). Principal component analysis with Varimax rotation revealed one component with loadings above.30, explaining 58.14% of the total variance (Appendix A). Moreover, the scree plot confirmed the presence of one component. Therefore, construct validity was considered as good. Concerning reliability, Cronbach s alpha (α) was.852, which is considered as good (Evers et al., 2000). Moreover, all items contributed positively to the scale as all values for Cronbach s α if item deleted were lower than Cronbach s α of the construct when the item was included. Team absenteeism. Team absenteeism was measured using archival data from the organization. The percentages indicate absence duration, i.e. the total number of absent hours relative to the total amount of working hours that employees were supposed to work based on their contracts. Thereby, the number of absence spells (frequencies) are not considered (Schaufeli et al., 2009). Control variables. In order to test the hypotheses, the following eight variables have been included as control variables because of their potential influence on job stress, work overload and/or team absenteeism. On the individual level (level 1), the variables age, education, weekly working hours, job satisfaction, and team tenure were included, while on the team level (level 2), the variables team stability, team type and team size were included. Age. According to Mayes, Barton and Ganster (1991), older employees are better able to cope with high job demands and therefore experience less job stress. Education. Employees holding jobs that require a higher education were found to experience higher levels of job stress (Bonanno, Galea, Bucciarelli, & Vlahov, 2007). The variable was recoded so that 1 = Secondary education, 2 = lower vocational education (MBO), 3 = higher vocational education (HBO), and 4 = University. The category for elementary level of education (elementary school) was excluded since there were no respondents in that category. Moreover, the category for other forms of education was excluded as it did not contribute to a meaningful interpretation of the value for education as an ordinal variable. Weekly working hours. Weekly working hours as defined in the contract are related to high levels of job stress and ill health outcomes and are, therefore, included in the analysis (Dembe, Erickson, Delbos, & Banks, 2005). Job satisfaction. Several studies have found that job satisfaction is negatively related to job stress and positively to employee turnover (e.g. Coomber & Barriball, 2005).

13 JOB STRESS, AND THE MODERATING ROLE OF PERCEIVED ORGANIZATIONAL SUPPORT 13 Team tenure. Team tenure is the time span in which an individual employee has worked in a certain team (Schippers, Den Hartog, Koopman, & Wienk, 2003). Team tenure has been shown to be related to professional experience which in turn leads to lower levels of job stress because more experienced individuals have developed better coping strategies (Klassen & Chiu, 2010). Team stability. Stable team compositions leads to greater team cohesion and higher levels of coordination (Gruenfeld, Mannix, Williams, & Neale, 1996). Therefore, it can be expected that team members in stable teams can better deal with work overload and as a result experience less job stress than teams which are less stable. Team stability was measured by means of three items concerning the state of the team, ranging on a 5-point Likert scale. An example question was: This team is very stable; I expect to work together with these team members for a long time. Factor analysis was found to be appropriate (KMO > 0.6, Bartlett s test of sphericity p <.001). Principal component analysis with Varimax rotation revealed one component with loadings above.30, explaining 66.86% of the total variance (Appendix A). Moreover, the scree plot confirmed the presence of one component. Therefore, construct validity was considered as good. Concerning reliability, Cronbach s alpha (α) was.746, which is considered as good (Evers et al., 2000). Moreover, all items contributed positively to the scale as all values for Cronbach s α if item deleted were lower than Cronbach s α of the construct when the item was included. Team type. Team type is relevant in this study as there are teams working in the primary process in the hospital (i.e. care teams) and teams that are supporting the organization in terms of human resources, communications, facilities etc. (i.e. knowledge teams ). Team absenteeism in care teams is expected to affect work overload and job stress more severely than in knowledge teams, as care teams have to take care of patients and, therefore, work cannot easily be rescheduled. In contrast, in knowledge teams, it might be easier to reschedule work to adapt to absent employees. Team type was included as a dummy, with 1 = care team and 0= knowledge team. Team size. Research suggests that team size can affect work overload of team members, such as when teams are larger, there is an increased workload being caused by greater needs of communication and coordination (Cassera, Zheng, Martinec, Dunst, & Swanström, 2009). Statistical analysis The analysis was conducted with the statistical software SPSS 22 for Windows. First, descriptive statistics including the means and the standard deviations, and Pearson s correlations, were calculated. Second, the model was tested by hierarchical regression analyses

14 JOB STRESS, AND THE MODERATING ROLE OF PERCEIVED ORGANIZATIONAL SUPPORT 14 using the MIXED procedure in SPSS for multi-level models. By using multi-level analysis, the variance that is caused by being part of a specific team is taken into account (Field, 2013). The independent variable, team absenteeism, was measured on level 2 (team level). All other main variables, i.e. work overload, job stress and PSS, were measured on level 1 (individual level). This corresponds to a so-called design (Bauer, Preacher, & Gil, 2006). To ensure that the design is appropriate for the data, the intra-class correlations (ICC) were calculated. The ICC1 score represents the percentage of variance in the outcome variable that is accounted for by the clustering (i.e. team membership) (Field, 2013). The ICC2 represents the reliability of the team mean scores (Heck, Thomas, & Tabata, 2010). To calculate the ICC s, a one-way ANOVA test was run. The ICC scores for PSS, work overload and job stress were calculated to verify whether it was appropriate to measure them at the individual level. Based on those outcomes, it will be decided whether the chosen design is most appropriate. In general, because models are being compared, maximum likelihood (ML) was used to estimate the parameters in the multi-level analysis (Field, 2013). The -2 Log-Likelihood and the chi-square test were used to check whether the fit of the model improved significantly. Moreover, the independent variables at level 1 including the control variables were grand mean centred, which means that for a certain variable, the mean score of all scores is subtracted from every individual score (Field, 2013). In multilevel models, centring variables helps to fight multicollinearity between variables and makes the model more stable (Field, 2013). To test the hypotheses, first, a null model with a random intercept was run (M0) with job stress as dependent variable. According to Hayes (2013), it is good practice to estimate intercepts as random in multilevel analysis. Second, the control variables were entered as predictors (M1). The independent variable team absenteeism was added in Model 2 (M2). In Model 3 (M3), the independent variable work overload was entered to test Hypothesis 1, the direct effect of work overload on job stress. To test hypothesis 2, the direct effect of team absenteeism on work overload, first a model with only the control variables as predictors and work overload as dependent variable was run (M0). In Model 1 (M2), the independent variable team absenteeism was entered. To test the mediation in hypothesis 3, the guidelines of MacKinnon, Fairchild and Fritz (2007) were used. Those guidelines require a significant effect between the independent variable (team absenteeism) and the mediating variable (work overload) and a significant relation between the mediating and the dependent variable (job stress). These required relations were tested in hypothesis 1 and 2. If those two relations have turned out to be significant, an additional bootstrap confidence interval tests whether the mediation is significant. Bootstrapping is

15 JOB STRESS, AND THE MODERATING ROLE OF PERCEIVED ORGANIZATIONAL SUPPORT 15 preferred to other methods such as the commonly used Sobel test, because in contrast to the Sobel test, it does not require the assumption of a normal sampling distribution of the indirect effect (Hayes, 2009). Finally, to test hypothesis 4, the moderation by PSS of the relation between work overload and job stress, the first model that was set up to test hypothesis 1 was extended by adding two variables: PSS (M4) and the interaction term PSS*work overload (M5). In this step, the random effect of work overload was included. In addition, in case of a significant interaction effect, simple slope analyses were performed in order to verify whether the slopes were significantly different from zero. Results Descriptive statistics The results of the means, standard deviations and Person s r correlations are presented in Table 1. As the theoretical model hypothesized, PSS is significantly and negatively related to job stress (r = -.31, p <.01) and work overload (r = -.18, p <. 01). Moreover, work overload and job stress are significantly and positively associated as it was assumed (r =.46, p<.05). However, in contrast to the expectations, team absenteeism seems unrelated to work overload and job stress. Concerning the control variables, as expected, team stability is significantly and negatively related to job stress (r = -.18, p <.01) and working hours are significantly and positively related to job stress (r =.10, p <.05). Regarding level of education, there is no significant correlation with job stress. However, level of education is negatively related to team absenteeism (r = -.20, p <.05), indicating that teams with higher absenteeism rates are associated with lower educated employees. Furthermore, as expected, job satisfaction is negatively related to job stress (r = -.42, p <.01); further, it is also negatively related to work overload (r = -.24, p <.01) and team absenteeism (r = -.11, p <.01) and positively to PSS (r =.34, p <.01) and team stability (r =.30, p <.01). Against the expectations, age is not significantly related to job stress. Moreover, for team type, no significant correlations with either job stress or work overload were detected based on Pearson s r. However, team type is negatively correlated with team absenteeism (r = -.19, p <.01), indicating that care teams have lower absenteeism rates as compared to knowledge teams. Finally, team size is positively related to team absenteeism (r =.09, p <.01), however, no relation with work overload has been detected. Finally, while team tenure is not significantly related to job stress, it is positively and significantly related to team stability (r =.17, p <.01).

16 JOB STRESS, AND THE MODERATING ROLE OF PER CEIVED ORGANIZATIONAL SUPPORT 16 Table 1 Means, standard deviations and Pearson s r bivariate correlations (n=502) Mean/% SD Team absenteeism (1) PSS (2) Work overload (3) ** 1 Job stress (4) **.46* 1 Team stability (5) **.20** -.15** -.18** 1 Job satisfaction (6) *.34** -.24** -.42**.30** 1 Level of Education (7) ** Age (8) ** 1 Working hours (9) * -.11* ** -.19** 1 Team type (10) % -.19**.09* ** 1 Size team (11) * ** * Team tenure (12) ** **.51** -.18** Note. * = significant at p<.05, ** = significant at p<.01; n = total number of cases; SD = standard deviation; Mean = sample mean; PSS = Perceived supervisor support; 1 Dummy variable with knowledge team as reference category.

17 JOB STRESS, AND THE MODERATING ROLE OF PERCEIVED ORGANIZATIONAL SUPPORT 17 Concerning the ICC s (Table 2), it can be concluded that job stress and work overload have rather low ICC1s (below 20%). Hence, there is a relatively high amount of individuallevel differences that are not explained by team membership and including the variables at the individual level is considered appropriate. However, PSS has an ICC1 value above 20% (ICC1 = 28.4%). Moreover, the mean score reliability (ICC2) is higher than.70 (ICC2 =.704), which can be considered as a threshold (lower limit) for aggregating data on team level (Klein & Kozlowski, 2000). The high ICC2 score indicates that a considerable amount of variability in PSS is explained by team membership and, therefore, measuring the variable at level 2 by taking the group means seems more appropriate. Because of these outcomes, PSS was aggregated to team level (level 2), which is a change to the initial conceptual model. Table 2 ANOVA outcomes and ICC1, ICC2 per variable MSW MSB F sig ICC1 ICC2 PSS (28.4%).704 Job stress (2.9%).153 Work overload (16.7%).542 Note. Formula: ICC1 = (MSB MSW) / (MSB + ((k-1) * MSW)); ICC2 = ((MSB MSW)/MSB) (Bartko, 1976; Bliese, 2000). k (average number of individuals in each group) = 6.12; n (number of groups) = 82 The results of the multi-level analyses can be found in Table 3 and 4. Hypothesis 1 states that work overload has an effect on job stress. In model 3 (M3) of table 3, it can be seen that while taking into account control variables and team absenteeism, work overload had a significant effect on job stress (B =.48, p <.001; table 3, M3). Moreover, the model s fit improved significantly when adding workload (χ 2 = 85.45, df = 1, p <.001; table 3, M3). Therefore, hypothesis 1 was accepted. Regarding hypothesis 2, which states that team absenteeism positively influences the level of work overload, no significant results could be found (B = -.00, p >.05; table 4, M1). Moreover, the change in -2 Log-Likelihood was not significant (χ 2 =.02; df = 1, p >.05; table 4, M1). Therefore, the fit of the model did not improve significantly when team absenteeism was added. Hypothesis 2 was rejected. Consequently, hypothesis 3, which proposed that work overload mediates the effect between team absenteeism and job stress, had to be rejected as well, as the guidelines from

18 JOB STRESS, AND THE MODERATING ROLE OF PERCEIVED ORGANIZATIONAL SUPPORT 18 MacKinnon et al. (2007) were not met. The guidelines for mediation require that the relations between the independent variable (i.e. team absenteeism) and the mediator (i.e. work overload), as well the relation between the mediator (i.e. work overload) and the dependent variable (i.e. job stress) are both significant (MacKinnon et al., 2007). Because team absenteeism had no significant effect on work overload, the requirements were not met. Therefore, hypothesis 3 was rejected. Finally, the results of hypothesis 4, which hypothesized a moderation of the relation between work overload and job stress by PSS, can be found in table 3 in model 5 (M5). The results show that there is no significant moderation effect by PSS (B = -.00, p >.05; table 3, M5). Moreover, the change in -2 Log-Likelihood was not significant when the interaction term was added to the model, which means that the model did not improve in terms of fit (χ 2 =.01; df = 1, p >.05; Table 3, M5). In addition, no significant direct effect of PSS on job stress was detected (B = -.07, p >.05; table 3, M5). Therefore, hypothesis 4 was rejected.

19 JOB STRESS, AND THE MODERATING ROLE OF PERCEIVED ORGANIZATIONAL SUPPORT 19 Table 3 Multi-level analysis predicting job stress (H1, H4) M0 B(SE) M1 B(SE) M2 B(SE) M3 B(SE) M4 B(SE) M5 B(SE) Fixed part Individual level Intercept 2.14 (.03)*** 2.08 (.17)*** 2.07 (.19)*** 2.03 (.17)*** 2.26 (.25)*** 2.26 (.25)*** Age -.00 (.00) -.00 (.00) -.00 (.00) -.00 (.00) -.00 (.00) Working hours.01 (.00).01 (.00).01 (.00).01 (.00).01 (.00) Education.03 (.04).03 (.04).02 (.04).02 (.04).02 (.04) Job satisfaction -.38 (.04)*** -.38 (.04)*** -.30 (.04)*** -.30 (.04)*** -.30 (.04)*** Team tenure.00 (.00).00 (.00) -.00 (.00) -.00 (.00) -.00 (.00) Work overload.48 (.05)***.46 (.05)***.53 (.06)*** Team level Team stability.01 (.05).01 (.04).04 (.04).04 (.04).04 (.04) Team type.00 (.07).01 (.07) -.03 (.06) -.01 (.06) -.01 (.06) Team size -.00 (.00) -.00 (.00) -.00 (.00) -.00 (.00) -.00 (.00) Team absenteeism.00 (.01) -.00 (.01).00 (.01).00 (.01) PSS -.07 (.05) -.07 (.06) Work -.00 (.10) overloadxpss 1 Random part τ 2 (intercept.00 (.01).00 (.00).00 (.00).00 (.01).00 (.00).00 (.00) variance) σ 2 (residual).36 (.03)***.30 (.02)***.30 (.02)***.25 (.02)***.25 (.02)***.25 (.02)*** Work overload.00 (.00) Deviance -2 LL Parameters Change in LL/(df) (8)*** (1) (1)*** (1) (2) Note. *p<.05, **p<.01, ***p<.001; job stress is the dependent variable; PSS = Perceived Supervisor Support, B = beta coefficients (unstandardized parameter estimates); SE = standard error; -2LL = -2 Log-Likelihood; df = degrees of freedom; 1 Interaction term calculated as PSS (level2)*work overload (level 1).

20 JOB STRESS, AND THE MODERATING ROLE OF PERCEIVED ORGANIZATIONAL SUPPORT 20 Table 4 Multi-level analysis predicting workload (H2) Model M0 B(SE) M1 B(SE) Fixed part Individual level Intercept 2.27 (.17)*** 2.28 (.19)*** Age.00 (.00).00 (.00) Working hours.00 (.00).00 (.00) Education -.01 (.04) -.00 (.04) Job satisfaction -.13 (.03)*** -.14 (.03)*** Team tenure.00 (.00).00 (.00) Team level Team stability -.06 (.05) -.06 (.05) Team type.11 (.07).11 (.07) Team size.00 (.00).00 (.00) Team absenteeism -.00 (.01) Random part τ 2 (intercept variance).02 (.01)*.02 (.01)* σ 2 (residual).20 (.01)***.20 (.01)*** Deviance -2 LL Parameters Change in -2LL/(df).03 (1) Note. Significance levels: *p<.05, **p<.01, ***p<.001; workload is the dependent variable; B = beta coefficients (unstandardized parameter estimates); SE = standard error; -2LL = -2 Log-Likelihood; df = degrees of freedom. Conclusion and discussion This study aimed to develop a better understanding of how team absenteeism in healthcare institutions can affect work overload and job stress of individual employees and, in addition, to get insights into perceived supervisor support (PSS) as a possible stress buffer. A multi-level design was chosen for testing whether individual job stress can be explained by team absenteeism, and whether this relation is mediated by individual work overload. Besides, PSS was investigated as a moderator. The design of the research was quantitative and longitudinal, using archival company data on team absenteeism from 2016, as well as employee

21 JOB STRESS, AND THE MODERATING ROLE OF PERCEIVED ORGANIZATIONAL SUPPORT 21 self-reported data from 2017 on job stress, PSS, work overload, and the control variables team stability and job satisfaction. Out of 2010 employees that were drawn based on stratified sampling, 502 employees belonging to 82 teams, were included in the final sample, which resulted into a response rate of 24.98%. The study s results can be interpreted in several different ways. Out of the four hypotheses, only the first one (H1) could be confirmed, all others had to be rejected. First, as hypothesized, work overload had a significant effect on job stress (H1), which is in line with the health impairment process of the JD-R model (Bakker & Demerouti, 2007). Therefore, employees who experience work overload, display higher levels of job stress. Second, there was no significant effect from team absenteeism on work overload (H2). Therefore, the possibility of mediation between team absenteeism and job stress via work overload (H3) had to be excluded based on the mediation requirements by MacKinnon et al. (2007), i.e. we fail to confirm the hypothesis that team absenteeism leads to more job stress via work overload. There are various reasons for this effect not to be significant. The results of this study reveal the importance of job satisfaction, as it had a significant negative effect on the level of job stress that is reported by the employees, and a negative correlation with team absenteeism. The latter leads to the question whether job satisfaction plays a mediating role between team absenteeism and job stress. Previous research by Weisman, Alexander and Chase (1980) suggests that understaffing, a situation that can be caused by team absenteeism, is associated with job dissatisfaction. Moreover, Aiken et al. (2002), have shown that a higher patient-to-nurse ratio, which can be caused by understaffing and team absenteeism, leads to job dissatisfaction. Drawing on previous research, it might be that team absenteeism causes hindered teamwork and decreased employee morale, which are both leading to employee dissatisfaction (Cherry et al., 2007). Moreover, job satisfaction has been found to be negatively correlated with job stress by several researchers (Flanagan & Flanagan, 2002). Based on the latter, it is possible that high team absenteeism leads to job dissatisfaction, which in turn leads to higher levels of job stress, implying a mediation effect. Furthermore, this study showed that team stability was positively related to job satisfaction and PSS, and negatively related to work overload and job stress. As previous research suggests, team stability leads to higher team cohesion, which in turn causes higher job satisfaction (DiMeglio et al., 2005). Based on the latter, future researchers should further investigate the role of job satisfaction and team dynamics such as team stability and team cohesion, as those factors might play an important role in relation to team absenteeism and job stress.