Risk Factors for Workplace Bullying: A Systematic Review

Objective: The goal of this study was to systematically review risk factors for workplace bullying. Methods: The search was carried out in two databases. Studies with estimates of risk factors for workplace bullying were included in the review. We assessed the quality of the selected studies using an adapted version of the Downs and Black checklist. Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) and Meta-analyses of Observational Studies in Epidemiology (MOOSE) guidelines were used for reporting papers. Results: Fifty-one papers were included, and 70.6% were from European countries. Women were reported to be at higher risk of being bullied in most studies (odds ratio (OR) from 1.17 to 2.77). Authoritarian and laissez-faire leadership styles were positively associated with bullying. Several psychosocial factors, such as stress (OR from 1.37 to 4.96), and occupational risks related to work organization, such as flexible work methods, role conflict, role ambiguity, monotonous or rotating tasks, high demands, pressure of work, and unclarity of duties were strongly associated with bullying. Discussion: The findings highlight the central role of organizational factors in bullying. Policies to prevent bullying must address the culture of organizations, facing the challenge of developing a new management and leadership framework.


Introduction
Workplace bullying is a still relatively recent issue in occupational health research, with most studies having occurred in the last 30 years. It is defined as situations where a person repeatedly and over a period of time is exposed to harassment, abuse, offenses, or social exclusion, placing the individual in an asymmetrical position where he/she is not able to defend himself/herself from unethical behaviour [1][2][3][4]. Its occurrence in workplaces is high, as demonstrated by a systematic review of 102 estimates, which identified an overall prevalence of 14.6% [1]. Also, bullying can be considered one of the greatest threats for workers' well-being [5,6], reinforcing the importance of better understanding its causes and mechanisms.
In the 1990s, Leymann claimed that four main factors were related to the occurrence of bullying in organizations: deficiencies in work design, deficiencies in leadership behaviour, socially exposed position of the victim, and low moral standards in the organization [2][3][4], describing the 'work environment hypothesis'. Since then, several studies have been conducted in order to elucidate its impact and causes. Recent systematic reviews and longitudinal studies of the consequences of workplace bullying demonstrate that bullying is associated with mental health problems-such as depression [7], anxiety [8,9], suicidal ideation [10,11]-sleep problems [12,13], neck and back pain [14], for bullying, analysing other concepts of violence, not describing statistical tests or showing only correlation coefficients, and being methodological articles or reviews. A total of 70.6% (n = 36) of the included studies were from Europe, 11.8% (n = 6) from North America, 9.8% (n = 5) from Australia and New Zealand, and 7.8% from other countries-Japan (n = 2), Mexico (n = 1), and Ghana (n = 1). Regarding methodological issues, most of the studies were cross-sectional, with data either from national surveys or specific groups of workers. Response rates varied from 12.5% to 95.0%, but only two studies reached a response rate higher than 80%. The approaches, methods, objectives, and quality of these studies were very heterogeneous. Papers scored between 6 and 15 points in the Downs and Black assessment tool [23]. A total of 15 articles (29.4%) scored less than 10 points and 18 (35.3%) scored above 13. Missed points in the Downs and Black index were mostly due to lack of losses description, p-value reporting, and confounding evaluation.
Most of the studies reported that women are more likely to be bullied than men [26][27][28][29][30][31][32][33][34][35][36][37][38]. Statistically significant odds ratios and prevalence ratios ranged from 1.17 to 2.77. Also, 11 studies showed no association between gender and bullying [39][40][41][42][43][44][45][46][47][48][49]. Only two studies suggested a higher risk among men [50,51]. The results on the association between age and workplace bullying were also inconsistent. Although eight studies found that workers younger than 44 years of age are more likely to be bullied [30,32,37,41,45,50,52,53], nine studies showed no association between age and workplace bullying [29,36,39,[42][43][44][47][48][49]. Only one study suggested a higher prevalence of bullying in older workers [54]. Only five studies presented relevant data on the association between ethnicity and bullying. Three studies suggested a higher prevalence of bullying between 'non-white', 'multiracial individuals', and 'Asian and black', compared to white workers. The risk was increased in a range Only five studies presented relevant data on the association between ethnicity and bullying. Three studies suggested a higher prevalence of bullying between 'non-white', 'multiracial individuals', and 'Asian and black', compared to white workers. The risk was increased in a range from 1.30 to 2.30 times [29,34,50]. One study showed no differences in the Negative Acts Questionnaire (NAQ) mean score when comparing white workers and black or ethnic minorities [51], while one study showed 28% less prevalence of bullying in non-white workers [52]. The NAQ is a tool with several questions about negative acts that suggest workplace bullying during the previous six months [4].
The association between family structure and bullying also varied between studies. Four studies reported no association between workplace bullying and marital status [37,43,49,55]. On the other hand, three studies reported that single, separated, divorced, and widowed workers were more likely to suffer from bullying at work [34,46,56], while one study found a higher risk of bullying among married workers (odds ratio (OR) = 3.06 (1.41-12.94)) [42]. Two studies also found a strong risk of being bullied among workers who have small children at home (OR from 1.92 to 2.87) [35,36].
Most of the studies reported no association between education and bullying. However, data from the 5th European Working Conditions Survey suggested a strong association between lower educational level and bullying (OR = 5.51 (1.79-16.95)) [36]. At the same time, two Turkish studies with forest engineers reported a higher prevalence of bullying among those with a doctoral degree [37,55]. Cross-sectional Single question: 'Have you at least once during the last 12 months felt exposed to offensive actions or harassment at your work place?' For example-your actions or comments were ignored, you are not taken seriously, were ridiculed or patronized (y/n). T-test and chi-squared Age Although common among men (22%), offensive actions or harassment were twice as common in women (45%; PR 2.0; 95% CI 1.6-2.4, controlling for age). The majority of female engine room crew members reported harassment or bullying, but they were few in total numbers (11/19; 58%).  Job performance and absenteeism were unrelated to workplace bullying. Those with worse physical health (beta = 0.15, p < 001) and higher strain (beta = 0.11, p < 0.05) at Time 1 experienced more bullying at Time 2. There was stronger support for the importance of organizational factors in workplace bullying. While positive organizational resources, such as ethical leadership and POS, were not related to workplace bullying, destructive leadership (beta = 0.22, p < 0.001) and more team conflict (beta = 0.20, p < 0.001) at Time 1 were associated with higher levels of bullying at the Time 2. Effective organizational strategies were protective (beta = −0.11, p < 0.01). Full model explained 37% of bullying variance.  Dussault [66] 2015 Canada Sample of 288 adults 153 were attending evening undergraduate classes in organizational behaviour management at a Canadian regional university, and 135 were employed within a multinational company in construction.
Cross-sectional NAQ-r Structural Equation Modelling. Given the non-normality of the data, robust maximum likelihood estimation was used.
Covariates not described.
Transformational leadership was negatively related to work-related bullying (β = -0.57), perceived Person-related bullying (β = -0.57), and perceived physically intimidating bullying (β = -0.45). Transactional leadership was also negatively related to work-related bullying (β = -0.38), perceived Person-related bullying (β = -0.30), and perceived physically intimidating bullying (β = -0.14). Laissez-faire leadership was positively related to work-related bullying (β = 0.51), perceived Person-related bullying (β = 0.53), and perceived physically intimidating bullying (β = 0.51).   Female forest engineers were more exposed to humiliation compared to male (p ≤ 0.05), People in the 34-44 age group were more exposed to 'relevant to person' (p ≤ 0.05) and 'task-related' behaviours (p ≤ 0.05). No significant relationship was found between the marital status and the levels of exposure to bullying. A significant relationship was found between education level and humiliation (p ≤ 0.05), indicating that forest engineers with a doctor's degree were more exposed to humiliation compared to those with a bachelor's or a master's degree. None At Time 1, no statistically significant differences between victims and non-victims were found with respect to gender, age, civil status, and years of service in the profession. Statistically significant differences were found between workplace bullying victims and non-victims in contract type (p < 0.05), in years with the organization (p = 0.004), and in the position (p = 0.006), indicating that a higher percentage of workplace bullying victims have a stable contract, more years of service, and a longer period in the position. At Time 2, no statistically significant differences between victims and non-victims were found by gender, age, marital status, contract, and years of service in the profession, in the organization or in the position.   Agervold [74] 2009 Denmark 898 participants from 12 different local government social security offices (local authority-educated social workers with equivalent competence and general office personnel). 1023 questionnaires were distributed. Response rate = 88%.

None
Demands of work, pressure of work, autocratic management style, unclarity of duties, and social work climate were strongly associated with bullying. 12 A total of 343 men (10.95%) and 583 women (12.78%) had experienced bullying weekly or more, and for 6 months or more (Leymann's definition).
Using both Leymann's definition and self-reported exposure, 275 men (8.78%) and 488 women (10.70%) had been bullied. For men, the point prevalence was significantly higher among services activities, and lower among managers and professionals. For women, no significant difference was found according to economic activities and occupations. For men, the point prevalence ranged from 3.69% in construction to 14.63% in other community, social and personal service activities, and from 3.27% in physical, mathematical, and engineering science professionals to 17.74% in protective services workers.

Cross-sectional NAQ T-test
None (used a matched control group) There were significant differences between victims and non-victims on four out of five personality dimensions. Victims tended to be more neurotic and less agreeable, conscientious and extraverted than non-victims. However, a cluster analysis showed that the victim sample can be divided into two personality groups. One cluster (64% of the victims sample), did not differ from non-victims. On the other hand, a small cluster of victims tended to be less extrovert, less agreeable, less conscientious, and less open to experience but more emotionally unstable than victims in the major cluster and in the control group. 9 46 Pranjic [48] 2006 Croatia 511 physicians from 1 hospital and 7 health centres in Tuzla, Brčko District and Banja Luka region. Response rate = 73% (total of 700 in the target population).
Cross-sectionalMobbing questionnaire (produced by researchers)Chi-squared None Explicitly type A personality (people with a chronic sense of time urgency, usually busy and very competitive, even in non-competitive situations) was the only factor associated with the bullying report. Age, gender, hours of work, and job title were not associated with the bullying report. 11 47 Bilgel [49] 2006 Turkey 877 full-time government employees in the three main public sectors: health, education, and security. 25 primary healthcare units and one public hospital, nine schools (two kindergartens, four primary schools, three high schools) and 13 police stations were randomly selected.  Cross-sectional Scale with twenty types of bullying behaviour were taken from the literature, based on Rayner and Hoel definitions (in the past 12 months) Chi-squared None Sex was not associated with workplace bullying. Younger workers (18-30 years old) were more likely to be bullied (prevalence = 51%) than the others (31-40 yo = 40%; 41-50 yo = 34%; >50 yo = 35%). Bullying was more frequent among full-time workers (full-time, prevalence = 47%; part-time = 30%). Unqualified residential care staff (48%) and nurses (44%) presented higher prevalence of exposure to bullying, compared to doctors (31%), ancillary staff (27%), administrative staff (37%), therapists (37%) and psychologists (36%). 12 51 Cole [30] 1997 The relationship between personality traits and being a victim of bullying was also evaluated by a few studies. Five studies described statistically significant associations between some specific personality traits and bullying. Neuroticism was identified as a risk factor for bullying in three studies, with odds ratios from 1.23 to 1.28 [26,40,76]. Most of the personality traits evaluated by these studies were not associated with bullying. In a cluster of participants of one study, people with Type A personality [48], less extroverted, less agreeable, less conscientious, less open to experience and more emotionally unstable were also more likely to be bullied [76]. Nevertheless, another study showed that personality characteristics explained only 2% of the variance (adjusted R 2 = 0.02) of bullying [58].
A wide range of important occupational factors regarding work organization, management issues, type of job, and earnings were also evaluated across the studies. The occurrence of bullying varied across professions, and it was not possible to identify a pattern in this aspect. The results were also controversial concerning the type of work (if permanent or temporary). Civil servants were more likely to be bullied than other workers [34,38,45]. Workers with lower income [63], paid hourly [34], and less satisfied with their payment [35] were also more likely to be bullied. Regarding the association between years of work in the organization and the occurrence of bullying, studies showed either a positive association [43] or no difference [49]. Shift work was strongly and positively associated with workplace bullying. Odds ratios (OR) of this association ranged from 1.74 (US workers) [34] to 2.68 (Spanish workers) [36]. The magnitude of the association was stronger in the public (OR = 2.46) than in the private sector (OR = 1.94) [35].
Organizational change, lack of procedural justice [53], and poor psychosocial safety climate were strongly and positively associated with bullying [71]. Leadership style was reported as an important risk factor by nine studies. Passive laissez-faire leadership, evaluated by three articles, increased up to 4.3 times the risk of workplace bullying [64,66,69]. Destructive [61], dictatorial [53], and autocratic leadership [74] were also related to a higher occurrence of bullying. On the other hand, supportive leadership style [57], consideration of individual by leaders [64], transformational and transactional leadership [66], authentic leadership [69], and fair leadership [73] reduced up to 70% the risk of bullying.
Work stress was one of the most important occupational factors reported in empirical studies, and was always strongly and positively related to bullying [33,35,36,49,62]. The odds ratios varied from 1.38 [49] to 4.96 [36]. Lack of social support [30,53,70,72], low social capital [59], and effort-reward imbalance [60] were also strongly associated with a higher risk of bullying.
Lastly, worse physical and mental health increased the risk of being bullied both in cross-sectional and longitudinal analyses. Sick leave and being on sickness treatment were also positively associated with bullying [38,57,61,65,67].

Discussion
A total of 49 out of 51 studies on risk factors for workplace bullying came from high-income countries, particularly Europe. Thus, either epidemiological studies on workplace bullying have not been developed in low-and middle-income countries, or their findings have not been published internationally. In a globalized scenario, it is not plausible that this phenomenon is not happening in other countries, mainly in countries where working conditions are poor. Only two studies [42,74] had response rates higher than 80%. As Nielsen (2010) advised, representative and convenience samples provide significantly different estimates of the prevalence of bullying [1], which could distort effect measures. On the other hand, most of the studies measured bullying with validated instruments, which tend to provide a 'more objective' measure. Although a few adaptations of Leymann's inventory were described, the NAQ (77) has been used and validated in several working populations [77,78], thus it is an option that improves the comparability of results. Some studies also measured bullying 'subjectively' (self-labelled approach), providing its definition and asking whether the worker was bullied or not in a single question. A combination of both objective and subjective measures can also be a satisfactory method to improve specificity [74,79].
The identified studies evaluated workers in several professions and activities, with a great variability of sociodemographic profiles and occupational characteristics, which makes it difficult to summarize and compare the results. While some studies focused on sociodemographic determinants and personality traits as antecedents of bullying, others focused on managerial and organizational factors. The 'work environment hypothesis' proposed by Leymann in the early 1990s targeted the work organization as the main cause of bullying, while more recent studies tested new hypotheses regarding the relationship between individual factors, such as personality and bullying [21]. This demonstrates the complexity of the phenomenon and indicates that simple explanations focusing on one or a few aspects of bullying are not enough to study this theme [80].
Regarding demographic factors, the role played by women in a globalized labour market, the work organization of institutions, and also the sexist culture in work environments could explain the consistency of results showing a higher risk of workplace bullying among women. Regarding age and marital status, results largely vary according to the type of job. Possibly in certain types of jobs with a demand for rapid response or physical effort, such as in blue-collar jobs, older workers could be more likely to be bullied. Also, in a job requiring a flexible schedule, people with children could be the main targets. Despite the fact that studies usually show a higher vulnerability of black people and ethnic minorities for several exposures and health outcomes-such as racial discrimination [81]-very few studies evaluated the association between ethnicity and bullying and results on this matter are controversial.
Findings from four papers [26,40,48,76] are consistent with a systematic review [21], indicating a positive association between bullying and neuroticism. However, other personality traits were not associated with bullying. Effect sizes of the association between neuroticism and bullying were small, so residual confounding should be considered. The explanatory power of personality as a predictor of bullying was also very small [58]. One of the premises of occupational health actions is that the work environment and the work organization should be adequate to individual characteristics, not the opposite. Thus, interventions to prevent bullying focused on individual aspects tend to be limited. Notwithstanding, studies investigating the effect of work organization on bullying should evaluate these factors as confounders or effect modifiers.
The effect of sociodemographic, personality and some occupational factors on bullying varied across studies, while a poor work organization and poor working processes always increased the risk of bullying. This risk varies across professions because it depends much more on the work environment than the profession itself or the years of work. For example, since workers' resignation is not common among civil servants, they can be easy targets of bullying, mainly when organizations are not prepared to support employees not well adapted to employers' demands. Workers with lower income, paid hourly, and less satisfied with their payment, as well as those in shift work, are usually exposed to more precarious working conditions, making these individuals more vulnerable to bullying.
In a context of 'flexible restructuring' of capitalism [82], the fundamental role of the capital and labour conflict on the existence of workplace bullying is unveiled by the human resources management (HRM) ideology [80,83]. Workplace bullying plays an important role in the intensification of work processes, being a tool to deepen mechanisms to control workers [84]. As a 'functional discipline', HRM lies at the core of organizational design and practice, shaping the way organizations operate [83,85]. The importance of HRM as a main determinant of bullying is evidenced by the positive association between low moral standard, lack of procedural justice, organizational change, and low psychosocial safety climate with bullying.
HRM ideology determines leadership styles in organizations and, agreeing with a systematic review about bullying in nurses [22], authoritarian and laissez-faire leadership patterns were strongly related to workplace bullying in this review [53,64,66,69,74] while 'democratic' leadership styles-such as supportive [57], authentic [69], transformational [66,69] and fair leadership [73]-protected the organization against bullying. Leaders are selected to put in practice the organizations' 'values' and 'missions' and an authoritarian leadership pattern is still highly valued in several companies [86]. A poor work organization, where workers are under a lot of pressure and/or ethical values are secondary, is more likely to request authoritarian leadership. In some organizations, once its existence can improve productivity and accomplishments, bullying is institutionalized and works as an essential part of leadership and managerial practice [3,87]. This can also be one of the reasons why interventions against bullying are often ineffective [18].
Leadership patterns are strongly related to the work organization and might determine role conflicts, role ambiguity, flexibility in work methods, and unclarity of duties. All these problems in work methods and in management increase the risks of bullying, results that agree with former reviews [19,20]. These hazards are related to the post-Fordist models; with the restructuring of production, workers need to be 'super-qualified', polyvalent, and able to perform several tasks and activities [82,88]. These organizational aspects, while they intensify working processes, generate workloads such as work pressure, monotonous tasks, rotating tasks, high demands and stress, which also increase the risk of workplace bullying. On these aspects, a bidirectional association is also plausible, as the existence of bullying in an organization degrades the work environment [80,89] increasing workloads and intensifying working processes.
The findings from this review reinforce Leymann's theoretical model, highlighting the central role of organizational factors on bullying determination. HRM, leadership patterns, and organizational factors are key distal determinants, having an impact on the physical and psychosocial workloads which determine bullying.

Conclusions
We have a large amount of valuable data concerning workplace bullying in high-income countries, particularly in Europe. However, a major effort is still necessary to encourage research on workplace bullying in low-and middle-income countries. Discrimination and harassment are more often described in non-dominant or disadvantaged groups [90], so a social context with poorer working conditions could lead to a higher risk of workplace bullying.
The main limitations of our review were the low response rates in most of the selected papers, as well as the variability of measures to assess risk factors and outcome, reducing the comparability of findings. Also, most of the studies were cross-sectional and were not able to estimate the effect of all occupational factors on bullying, precluding strong inferences regarding the direction of associations. Therefore, future studies with a longitudinal design and representative samples (or at least a better description of losses) are important to clarify associations subjected to reverse causation and improve the interpretation of the results. It is necessary to deepen the understanding of the role of organizational factors and emphasize the role of human resource management on bullying causation. The effect of work schedules, breaks, and extra hours on bullying also need to be studied. Validated instruments, such as the Psychosocial Safety Climate-12, which addresses aspects of the management in work environments, can be fundamental to evaluate distal contextual factors. Considering the scarcity of information about the association between race and sexual orientation with bullying, future studies should also investigate this subject.
Bullying should be understood as a completely unacceptable and unethical behaviour in workplaces. Policies to prevent bullying must address the culture of organizations and face the challenge of developing psychosocial safety at work environments. Interventions promoting a new management and leadership framework, increasing democratic values, and promoting employee participation in work decisions, should be implemented and evaluated in order to provide better parameters for practice in occupational health.