The association between workplace bullying and depressive symptoms: the role of the perpetrator

Concern has been raised about differences between employees labelling themselves as being bullied as compared with those who do not label themselves as being bullied (e.g. in terms of personality and mental health) [1, 33‐35]. Thus, the main analysis of the present study only included participants who labelled themselves as targets of bullying, whereas in an additional analysis we compared the depressive symptoms of the bullied and non-bullied respondents. To obtain a sufficiently large study population, we included questionnaire data from two cross-sectional surveys of the general working population in Denmark: we used self-reported questionnaire data from the Danish Working Environment Cohort Study in 2010 (DWECS 2010) and the Work and Health Study in 2012 (WH 2012). Although the two questionnaires were not identical, they were both appropriate for the purpose of the current study.

For the purpose of the present study, we excluded participants who did not label themselves as being bullied (n = 23,474) in both cohorts. Among the remaining participants (n = 3543), we aimed at including respondents who were potentially at risk of being bullied by any of the perpetrators presented in the surveys. Therefore, participants reporting not having a leader (n = 102) were excluded from the analysis, assuming that they were not being at risk of being bullied by a leader. Similarly, we also excluded participants who were self-employed (n = 66) or working as assisting spouses (n = 4), assuming that the former group did not have a leader, and the latter group had neither a leader nor colleagues in the formal sense. The data sets, however, did not include any information about respondents not having clients and subordinates. In addition, from the main analyses we excluded participants reporting multiple perpetrators in WH 2012 (n = 313), of whom 254 reported bullying by leaders, 286 by colleagues, 78 by clients and 44 by subordinates. Overall, we included data from 2478 bullied individuals (DWECS 2010: n = 958; WH 2012: n = 1520).

All statistical analyses were conducted using IBM SPSS Statistics 21. Frequency analysis was used to investigate the prevalence of the reported perpetrators. In a one-way ANOVA we tested the difference between the MDI scores among those bullied by leaders, colleagues, subordinates and clients, respectively. As equal variances in the MDI scores between the groups were not confirmed, Games-Howell pairwise multiple comparisons were conducted. Finally, an ANCOVA analysis was performed in order to test whether the variance of the MDI scores could be explained by the perpetrator when adjusting for age, gender, and occupational sector. A p-value < 0.05 was considered statistically significant.

The majority of the respondents belonged to the social and health care sector both in DWECS 2010 (n = 229, 23.9 %) and in WH 2012 (n = 331, 25.4 %), representing the biggest group of the study, whereas the least amount of respondents belonged to the graphical sector (n = 10, 1 % in DWECS 2010, n = 11, 0.8 % in WH 2012), representing the smallest group. The distribution of occupational sectors, age, gender and depressive symptoms (MDI) of the bullied respondents are presented in Table 1.

The results of DWECS 2010 showed that the most commonly reported perpetrators of workplace bullying were clients (41.5 %), whereas the least common perpetrators were subordinates (4.2 %). The prevalence of being bullied by leaders and colleagues were 23.8 % and 30.5 %, respectively. The mean MDI score was the lowest (mean: 8.9; SD: 8.2) among those being bullied by clients, and it was the highest (mean: 13.1; SD: 9.6) among those being bullied by subordinates (Table 2).

Table 3 reports the adjusted results concerning the mean MDI scores of the bullied respondents according to the perpetrators, and shows that the MDI score of those who were bullied by clients differs significantly from the other three groups. Compared to those being bullied by their clients, individuals who reported bullying from leaders (mean difference: 4.29; 95 % CI: 2.27-6.31), subordinates (mean difference: 4.41; 95 % CI: 0.33-8.49), and colleagues (mean difference: 3.55; 95 % CI: 1.64-5.46) had a significantly higher mean MDI score. No other statistically significant difference was found between the groups in this sample.

The results of WH 2012 showed that the most commonly reported perpetrators of workplace bullying were colleagues (60.3 %), whereas the least common perpetrators were subordinates (4.5 %). The prevalence of being bullied by leaders and clients were 26.3 % and 9.0 %, respectively. The mean MDI scores were the lowest (mean: 11.7; SD: 7.8) among those being bullied by their subordinates, and it was the highest (mean: 14.0; SD: 9.4) among those being bullied by their leaders (Table 2).

Respondents being bullied by leaders had a significantly higher mean MDI score than participants being bullied by colleagues (mean difference: 1.67; 95 % CI: 0.14-3.20), yet no other significant difference regarding the MDI scores was found among the four perpetrator groups. Nevertheless, we also found a considerably higher MDI score among those being bullied by a leader (mean difference: 2.39; 95 % CI: -0.25-5.03) compared with those being bullied by clients, although the difference was not statistically significant at conventional levels (Table 3).

In addition, we compared the mean MDI scores of the non-bullied and bullied groups. In DWECS 2010 the results showed that non-bullied respondents had significantly lower mean MDI scores compared to those who were bullied by leaders (mean difference: -6.38; 95 % CI: -7.62- -5.15) colleagues (mean difference: -5.66; 95 % CI: -6.77- -4.55), subordinates (mean difference: -6.75; 95 % CI: -9.68- -3.82), and clients (mean difference: -2.21; 95 % CI: -3.16- -1.25). We found the same results in WH 2012, where non-bullied respondents reported lower MDI scores than those bullied by leaders (mean difference: -6.56; 95 % CI: -7.65- -5.48), colleagues (mean difference: -4.80; 95 % CI: -5.53- -4.07), subordinates (mean difference: -4.50; 95 % CI: -7.06- -1.93, and clients (mean difference: -4.22; 95 % CI: -6.16- -2.28).

As an additional analysis, we compared the mean MDI scores of respondents being bullied by one perpetrator (one perpetrator: n = 1520) with those who were excluded from the main analyses due to reporting multiple perpetrators (two perpetrators: n = 280; three or more perpetrators: n = 33). While adjusting for age, gender and occupational sector, we found that those reporting two perpetrators had a significantly higher mean MDI score than those reporting only one perpetrator (mean difference: 2.52; 95 % CI: 0.96-4.08). The difference between those reporting one versus three or more perpetrators was of a similar size, although it was not statistically significant due to the small numbers (mean difference: 2.46; 95 % CI:-1.78-6.69). There was no difference between those reporting two and those who reported three or more perpetrators (mean difference: 0.06; 95 % CI: -4.37-4.50).

In DWECS 2010, the effect sizes of the significant relationships ranged from 3.55 to 4.41, whereas the effect size of the only significant association in WH 2012 was 1.67. These effect sizes can be quantified as up to 47 % of the standard deviation of the mean MDI score, although relative to the whole scale range (0-50) these differences are rather small. In order to make the pattern of DWECS 2010 and WH 2012 visually comparable, Fig. 1 illustrates the adjusted MDI scores of the bullied respondents depending on the self-reported perpetrators.

In DWECS 2010 we found that those being bullied by their clients had significantly lower scores on the MDI scale compared to those bullied by leaders and colleagues. Likewise, in WH 2012 we found an indication that those bullied by clients had less severe depressive symptoms compared to those bullied by leaders, yet due to the few respondents in the client category this association was not statistically significant at the conventional level. Furthermore, in WH 2012 those bullied by their colleagues had significantly less severe depressive symptoms than those who were bullied by their leaders.

Taken together, the overall pattern of the MDI scores of those being bullied by leaders, clients and colleagues is alike in the two surveys, whereas the influence of being exposed to bullying by subordinates on the MDI scores did not show a clear pattern. As one cohort showed significant differences between the MDI scores of those bullied by clients and leaders, and the other cohort only showed a non-significant tendency, it is hard to draw a universal conclusion.

One reason for the observed differences in occurrence of participants being bullied by clients could be the different distribution of occupational sectors. However, the two surveys were based on random samples of the general working population and very similar in composition with regards to the included occupational sectors. In both cohorts a similar (and the biggest) proportion of the respondents belonged to the “Social and Health” sector, which is typically including the largest amount of clients. Thus, the difference in occurrence of participants being bullied by clients is not likely to be due to the occupational composition of the two surveys.

Another reason for the different association between the various perpetrators and the MDI scores of the bullied respondents could be the slight difference in the definition of bullying between the two surveys. Although the dissimilarity may have influenced our results, this explanation is not highly likely as the overall prevalence of bullying was similar between the two studies.

In general, we made a considerable effort (e.g. investigating the distribution of different occupational sectors, sampling procedures, or the formulation of the definition of bullying) in order to illuminate the explanation for the remarkable differences in the two cohorts, yet none of them resulted in a convincing explanation.

We hypothesized that (1) those bullied by leaders reported more severe depressive symptoms compared to the other groups and (2) those bullied by clients reported the least severe depressive symptoms. Our hypotheses were only party confirmed by the results.

The results of DWECS 2010 support the findings of Clausen et al. [27], who proposed that bullying might be experienced less severely in case an employee is bullied by clients with impaired mental capacities e.g. in eldercare. However, in our study the response category “clients” also included customers, patients and students, and so could be interpreted more broadly, thus we cannot state that those who reported bullying from clients only provided service for patients with mental illnesses. Nevertheless, the result supports the notion that the less powerful the perpetrator is, the less severe the target’s depressive symptoms are [30]. Furthermore, in case of being bullied by clients, the targets may have the possibility to seek social support from their leader or colleagues in order to cope with bullying [37, 38].

Interestingly, the depressive symptoms of those who were bullied by their subordinates - as a group of less powerful perpetrators in terms of occupational hierarchy - were not any less severe than those being exposed to bullying by others. This finding goes somewhat against the role of the previously highlighted formal power in terms of health outcomes of bullying. Nevertheless, it may be explained by the Scandinavian organisational culture [1], in which the differences in power between individuals in various formal and informal positions are relatively small, and therefore being bullied e.g. by a leader or colleagues may have similar outcomes than being bullied by a subordinate. It is important to note, however, that due to the low prevalence of bullying by subordinates, we had low statistical power to test its difference from the other perpetrator groups.

In WH 2012, surprisingly, the depressive symptoms of those being bullied by clients were not statistically different from the other groups. Nevertheless, it still seems that being bullied by a leader is associated with worse psychological health in terms of depressive symptoms than being bullied by clients. In the same sample we found that those who were exposed to bullying by leaders reported significantly more severe depressive symptoms compared to those who were bullied by colleagues. These findings also support the argument concerning the organisational power between the perpetrator and the bullied respondents.

Finally, the current study was carried out among those respondents who reported workplace bullying: 9.7 % in DWECS 2010 and 11.9 % in WH 2012. This prevalence is somewhat higher than the prevalence in other Scandinavian countries, yet it is in line with findings showing an overall 11 % in studies using the self-labelling method with a definition [5]. Furthermore, according to the newest report available [39], the prevalence of workplace bullying in the general working population in Denmark was 11.6 % in 2014, thus our findings are in accordance with the prevalence of workplace bullying observed in other Danish studies.

The major strength of the present study is that it was based on two large, nationwide and representative samples. Furthermore, by analysing and comparing data from two cross-sectional surveys of the general working population, we challenged our own findings instead of accepting the results obtained from one survey, or merging them without thorough preliminary analysis. The comparability of the two data sets are supported in that the prevalence of workplace bullying was similar in the two surveys even though they used slightly different definitions of workplace bullying as well as response alternatives. A final strength is that we used a validated and widely used depression scale.

The study also has limitations. Due to the differences between the two surveys, we excluded respondents being exposed to bullying by multiple perpetrators in WH 2012 from the main analysis, although this analytical choice resulted in lower statistical power.

In addition, the findings of the present study may be even more pronounced in samples from other - non-Scandinavian - countries where the differences in workplace hierarchy are associated with bigger perceived power differences. Our samples represent the Danish general working population and therefore, the findings do not necessarily hold in countries where the ruling organisational culture is essentially different than in Scandinavia, which limits the generalizability of the results.

Furthermore, there could be several explanations for the observed associations between the perpetrator and the MDI score of the targets. First, due to the cross-sectional nature of the study, we cannot establish causal relations between being bullied by various perpetrators and the level of depressive symptoms of the targets. Although being exposed to bullying could explain the targets’ more severe depressive symptoms, it is also well-established in longitudinal studies that mental health problems at baseline are associated with an increased risk of subsequent exposure to bullying [22, 24, 40, 41]. In the present study, when comparing the depressive symptoms of the bullied and non-bullied respondents, we found in both cohorts that bullied respondents had higher scores on the depression scale than non-bullied respondents irrespective of the perpetrator. Furthermore, when comparing respondents who were bullied by only one perpetrator with those who were bullied by multiple perpetrators in WH 2012, the mean MDI scores of those being bullied by multiple perpetrators were higher than of those being bullied by only one perpetrator. These results indicate that non-bullied respondents have better mental health than bullied respondents, and that individuals with a poorer psychological health may be exposed to bullying by more people at their workplace. Nevertheless, based on the nature of cross-sectional study there are no grounds to decide whether the worse psychological health is the result or the cause of (the multiple) bullying experience.

Second, it is also plausible that there is bias in the reporting of who the perpetrators is due to other factors, for example reporting the leader as perpetrator due to a general dissatisfaction with the workplace or work climate.

Third, no information on exposures outside working life such as marital status or major life events, which may have confounded with the reported associations, was available in the cohorts. These factors are well-known risk factors for the development of depression, and if these factors are differentially distributed among the perpetrator groups, they may have biased the differences in depressive symptoms between the groups. However, the direction of the bias (under- or overestimation) is unknown.

Overall, alternative explanations for the differences between the groups could be reverse causality, differential misclassification, and unadjusted confounding.

Furthermore, due to the lack of response categories regarding the frequency of bullying in DWECS 2010, the data did not allow us to assess the health effects of workplace bullying based on the frequency of the exposure. Similarly, there was a difference in the way participants were asked about the perpetrators in the two surveys. In DWECS 2010, respondents could only choose one perpetrator, and therefore in case of experiencing bullying by multiple perpetrators, they had to indicate the perpetrator based on a hierarchy, of which we had no information. Consequently, we cannot be sure that in DWECS 2010 there was only one perceived perpetrator of the targets, which made it impossible to evaluate the depressive symptoms of those who perceived themselves as being exposed to multiple perpetrators. Overall, we cannot rule out that this dissimilarity may have influenced our results in a way that could not be further explored within the scope of the present study.

Finally, not everyone was at risk of being bullied by all perpetrators, although we aimed at including respondents being at risk to be bullied by anyone listed in the two surveys. However, in certain professions it may not be the case to work together with all the potential perpetrators and thus being at risk to be bullied by them. Nevertheless, our data did not let us to eliminate this issue beyond applying our inclusion criteria.