Work-related threats and violence in human service sectors: The importance of the psycho-social work environment examined in a multilevel prospective study

1.1 Psycho-social work environment and work related violence

Viitasara & Menckel (2002) have presented a model for analysis of work-related violence and threats consisting of three levels: the individual, the situational and the structural level [17]. In their model, work-related threats and violence can be regarded as an end point in long-term processes with origins in the underlying situational and structural factors. Even though work-related violence occurs in a specific situation, the structural factors are present in time and space in every violent situation. The present study builds upon their model and focuses on the model’s structural level; namely the importance of the psycho-social work environment as a risk factor for work-related threats and violence. Employees in the health care sector apply themselves as instruments to achieve a supporting and non-aggressive milieu and good care outcomes, which entails that they are affected by their environment. Thus, it is important that the work environment is functional, supportive and effective.

Preventive strategies can operate at the individual, situational and the structural level. The literature on work-related violence and threats has identified several risk factors such as gender, age, work-experience, working time and occupations [10, 12] at the individual level. Also, specific situations are found to be risky. For instance, employees who are in close contact with patients, setting limits for patients’ behaviour, and frequently doing physical patient handling are at higher risk of work-related violence than other employees in the same setting [18]. Psychological theories of aggression assert that the occurrence of aggression can seldom be reduced to one single factor, but is more likely to be influenced by several factors simultaneously. Work-related violence and threats stem from the interplay of a wide range of personal, situational and social factors and must be understood within the framework of the general social context in which they occur [19]. Thus, work-related violence is a complex phenomenon, and there is no single explanation for why some employees are exposed to work-related violence and others not [20].

The staff’s performance in the form of burn-out or engagement are related to several structural factors (for instance demands, role conflicts, social support) in the work environment [21]. Thus, theoretical explanations as well as empirical data support the integration of the psycho-social work environment into research on work-related threats and violence. The theoretical explanation for linking work-related violence with psycho-social work environment is that a poor psycho-social work environment increases the risk of work-related strain and burn-out, which in turn decreases job performance and the motivation to perform well at work [22]. Poor performance (i.e., low quality of care, long waiting times, treatment errors and mistakes) is likely to result in a higher risk of exposure to workplace violence, as has been shown in a number of studies [23–26], presumably by generating feelings of frustration and aggression in patients [27, 28].

At the structural level, empirical data also supports the integration of the psycho-social work environment with research on work-related threats and violence. Studies have found a positive relation between victims of work-related violence and elements of the psycho-social work environment, such as low levels of influence on their own work, quantitative demands, high emotional demands, poor utilisation of their resources, role conflicts, and low supervisor and co-worker support [29–36]. However, the results are difficult to apply as recommendations, as the studies have several major limitations. First, mono sectors approaches make it difficult to generalize the results, because the psycho-social work environment varies according to type of sector [37–40]. It can be discussed whether the associations between psycho-social work environment and work-related violence found at for example psychiatric wards or in elder care, may be generalized to other sectors [29, 33]. Thus, sector differences appear to be an important factor to study [41]. Second, the cross sectional nature of the studies fails to exclude that for instance high level of emotional demands or low levels of influence may be the consequence and not risk factor of violence at work. So far, only one follow-up study has examined the prospective association between psycho-social work environment and work-related violence and found associations between low quality of team work, high levels of time pressure and increased risk of work-related violence [42]. Finally, the studies do not take into account that employees, and thus their psycho-social work environment, are clustered within workplaces [43]. In sum, these limitations highlight the need for research that uses a follow-up design, compares different sectors and uses multi-level analysis.

1.2 Work-related threats

Threats and violence are mutually associated, as threats of violence often present as an antecedent for work-related violence [19] and the underlying mechanism between work environment and violence and work environment and threats may be identical. For instance, Glomb (2003) found that prior to an aggressive incident, individuals were more likely to have engaged in less severe violent behaviors [44]. Also, Dupre & Barling (2006) found that psychological aggression toward supervisors was positively associated with physical acts of aggression directed toward supervisors [45]. Furthermore, the health consequences of threats of violence may be as serious as those of direct exposure to violence. For instance, a study found that violence and threats at work had a significant impact on the health of the victims one year after exposure [7]. As threats are more frequently than violence [5], important information about prevention of work-related violence can be achieved by identifying associations between work environment and threats.

Therefore, it is important to identify associations between psycho-social work environment and both work-related threats and violence.

1.3 The present study

The purpose of the present study is to investigate the associations between the psychosocial work environment and work-related threats and violence. The study uses a one-year follow-up design to investigate the associations in four human service sectors characterized by a high risk of violence and threats.

The specific purposes of the study are:

To identify associations between psycho-social work environment and violence in four high risk sectors.

2.1 Procedure and participants

The study population consists of a cohort established in 2010 with one-year follow-up [5]. The study population was employed in the following settings: Different types of public psychiatric wards (open wards, acute wards, secure wards), the elder care sector, at special schools (schools for pupils with diagnosis’ such ADHD, Autism) and in prison and probation services. The sectors were chosen because previous research in Denmark has found that these sectors have high frequencies of work-related violence and threats [12].

Participants working in psychiatric wards, elder care sector and at special schools received and filled out paper-and-pencil questionnaires during a planned meeting at the worksite. Invited participants were employees in jobs with client contact and who had no more than three weeks of absence at the time of survey distribution. As the purpose of the study was to analyze the associations between work environment and violence at work, the last criterion was to ensure that employees were sufficiently aware of the present working environment. Two of the research team members attended the meetings and the completed questionnaires were returned to the researcher at the end of the meeting. Employees, who did not participate in the meeting, received a questionnaire and a prepaid/stamped envelope to be returned directly to the researchers.

A web-based questionnaire was used for participants in the Prison and Probation Services – a decision made by top management. It was stated in the cover letter of the questionnaire, that participation in the study was voluntary and that the data would be treated confidentially.

The participants were recruited using convenience sampling. The workplaces were all public workplaces and they were contacted in formal ways by meetings with the top managerial level in the municipality (elder sector) or county (psychiatric wards). Hereafter local leaders were assembled and invited to participate. Special Schools are organized somewhat differently, therefore, each school was directly approached. PPS was contacted formally by meetings with the top managerial level. Three municipalities and one county were not able to participate. With regard to the PPS all the staff was included in the project and all invited special schools participated.

The baseline sample consisted of 5477 persons: Psychiatry (N = 930/response rate: 85%), Special Schools (N = 758/response rate: 91%), Eldercare (N = 966/response rate: 80%), and Prison and Probation services (N = 2823/response rate: 60%). Eligible for follow-up were employees still at the same worksite, still in jobs with client contact and with no more than three weeks’ absence at the time of survey distribution. A one-year follow-up was chosen because research has shown that a too long or short time span between baseline and follow up increases the risk of recall errors [46]. Furthermore, in relation to psycho-social work environment and mental health a one-year follow-up is recommended [47]. Altogether, 3584 participated in the follow-up study with an overall follow-up response rate of 65%. However, for the current study, we excluded the administrative staff (e.g. non-uniformed personnel) from the Prison and Probation service, because they represented many different occupations including consultants, cooks, teachers, priests, cleaners and nurses. We considered their psycho-social work environment to be too heterogenic to pool the different occupations together even though they all more or less have contact with prisoners. Hence, the final follow-up study sample consists of 3011 persons: Psychiatry (N = 698/response rate: 75%), Special Schools (N = 535/response rate: 71%), Eldercare (N = 610/response rate: 63%), and Prison and Probation services (N = 1168/response rate: 61%). To work in these occupation, a formal education is required and therefore most participant were educated (e.g either nurses, teachers, prison guards etc). Only 2.9% didn’t have an education and thus, we didn’t assess it as necessary to adjust for education.

According to Danish law, approval by the Ethics Committee written informed consent is not required in questionnaire and register-based projects.

2.2 Instruments

To measure the psycho-social work environment, we used the second version of the Copenhagen Psychosocial Questionnaire (COPSOQ) [48], which is a standardized instrument designed to monitor the psycho-social work environment. COPSOQ have acceptable intra-class correlation reliability for all scales, has shown good criteria validity and are translated into several languages. The internal consistency of the COPSOQ is comparable to other influential questionnaires such as the Job Content Scale [49] and the Effort-Reward Imbalance Questionnaire [50].

The questionnaire included most of the dimensions of seven influential psychosocial theories reviewed by Kompier in his paper on models of psychosocial factors at work [51]. The COPSOQ-II includes 50 items covering 5 broad dimensions; work demands, work organisation and job content, interpersonal relations and leadership and finally values at work. Each dimension consists of 2–5 items with a total of 18 scales. The items are measured by means of a five point Likert scale ranging from ‘always’ (100) to ‘never’ (0).

Scale scores range from 0 to 100 where high values of the scales indicate higher values of the dimension except high demands, high emotional demands, role conflicts and work-family conflict where high values of the scales indicate lower values of the dimensions. The 18 scales are labelled: quantitative work demands (e.g high number of demands at work; time pressure), emotional demands, influence over own work situation, possibilities for development, commitment to the work-place, meaning of work, predictability, rewards at work, role clarity, role conflict, work-family conflict, social community at work, social support from colleagues, social support from nearest supervisor, quality of leadership, mutual trust between employees, trust regarding management, organizational justice and finally work-family conflicts. The wording of the items and scales are available elsewhere [48].

In the current study Cronbach’s alpha (α) in nearly all scales was above 0.75, which indicates a high degree of internal consistency of the scales [52]. Mutual trust between employees was an exception with α= 0.61. The scales were dichotomised into a group of the 25% most exposed to a poor working environment and a group of the 75% least exposed. The dichotomisation was based on the whole study population (N = 3011) and not in each sector separately. In this way the definition of high demands, many role conflicts, low mutual trust between employees etc. is identical across sectors, thus making the psycho-social work-environment comparable across sectors. This means that the relative proportion of participants exposed for instance to high demands, was different depending on sector.

2.3 Outcome measures: Work-related violence and work-related threats

Based on the definition of Wynne et al. (1997) [53], we applied a broad definition of work-related violence including threats of violence as well as physical violence. We used a checklist consisting of 11 different types of violent incidents and 7 different types of threats of violence used in previous research in Sweden [54]. Types of physically violent behaviors included being spat on, hit, hit with an object, scratched/pinched, shoved, held firmly, punched with a fist, kicked, bitten, having a hard object thrown at you, and clients using a weapon or a weapon-like object. Threatening behaviors included being threatened with beatings, written threats, threatened in a scolding manner, threatened in an insulting manner, threatened over the phone, threatened involving objects, and threatened indirectly (i.e. threats towards family), but with no physical contact. Respondents were asked to indicate how often during the past year they had experienced each of these different types of threats or violence at the workplace. For both work-related violence and threats the frequency of occurrence was measured with a five point Likert-like scale ranging from never (0) to almost daily (4). The items were computed into two sum-scales labelled ‘threats of violence’ and ‘physical violence’ respectively. The two scales were dichotomised into the 25% most exposed versus the 75% least exposed. The dichotomisation was done in each sector separately.

2.4 Other measures-neuroticism

Neuroticism may be both a substantive and a biasing factor in self-reporting of strain and stress [55]. As far as we know, no previous studies examining the associations between work environment and work-related violence and threats have adjusted for the potential biasing effect of neuroticism. Therefore, we adjusted for the possible confounding effect of neuroticism in the analyses.

Respondents completed the Eysenck Personality Inventory – EPS Adult [56]. The questionnaire contains 6 items. Response categories for all questions were “yes” or “no”. The scale was dichotomised into the 25% scoring highest on the scale versus the 75% scoring lowest on the scale.

2.5 Demographic questionnaire

Background information about gender, work experience, working time and time of working hours in direct contact with patients, clients, pupils or prisoners was retrieved from the baseline survey.

Work experience: The participants were asked how many years they had been working at the current workplace. The answer should be written in years and months and the answers were divided in two groups representing more versus less than two years of working at the present workplace.

Working time: The participants were asked at what time of the day they usually were at work. The response categories were: Fixed daytime (06–18), only in the afternoon (15–24), only night times (22–06), non-fixed working time without working at nights, non-fixed working time including working at nights and other working times. The answers were divided in two groups: daytime work or non-daytime work.

Working time/hours in direct contact with patient, client, pupil or prisoners: The participants were asked about what percentage of the working-day they considered themselves to be in direct contact with patients, clients, pupils or prisoners. The response categories were <25%, 25–49%, 50–75% and >75% and were coded in two groups: more or less than 50% of working hours in direct contact with patients, clients, pupils or prisoners.

2.6 Statistical analyzes

The data did not meet the statistical assumptions of normality and homoscedasticity required for linear regressions [52]. Therefore, we applied multilevel logistic regressions and calculated adjusted odds ratio (OR) and 95% confidence intervals (CI) for the association between psycho-social work-environment measured in 2010 and work-related violence and work-related threats measured in 2011. Statistical analyses were performed using SPSS (Version 20), except from the multilevel logistic regression, which was performed in Stata 13.

2.7 Potential confounders

We adjusted for the possible confounding effect of neuroticism in the analyses. Furthermore, we adjusted for structural/organizational factors that might affect both the risk of work-related violence and threats and the character of the psycho-social work-environment. These factors included work-experience (more or less than two years working at the actual workplace), working time (day time or non-daytime) and working hours in direct contact with patient, client, pupil or prisoners (more or less than 50% of working hours). Furthermore, we adjusted for gender and baseline threats and violence. Two exceptions must be mentioned: First, in the elder sector, 96.9% of the participants were women. Therefore, we excluded men from the analyses. Second, in the special schools, 97.9% of the employees worked day time only. Therefore, only employees working daytime contributed in the analyses and no adjustment for working hours was carried out in this occupation.

As it can be seen from Table 3, several work-environment factors were associated with work-related threats. Especially high quantitative demands, many work-family conflicts, low organization justice, low role clarity, low rewards at work, high emotional demands, many role conflicts, low predictability and low level of influence over own work-situation were associated with work-related threats. The adjustment for potential confounders in model 1 did not change the size of the associations to any important degree. However, adjusting for baseline exposure to work-related threats (model 2) reduced the size of the associations.

As it can be seen from Table 4, several work-environment factors were associated with work-related violence. Especially many role conflicts, high emotional demands, low predictability, many work-family conflicts, low role clarity, low quality of leadership and low support from nearest supervisor were associated with work-related violence. As in the analysis of work-related threats above, adjustment for potential confounder did not change the estimates except from adjustment for baseline violence.

As can be seen in Table 5, in each single sector only few associations were statistically significant due to smaller strata compared to the total sample. Some differences between the four sectors were found. First, in the Prison and Probation Services the associations between the psycho-social work environment and work-related threats were in general stronger compared to the three others sectors. Second, in the special schools and at the psychiatric wards several risk estimates were below 1 and thus the direction of the association was in some cases opposite compared to the two others sectors. For instance, low support from nearest supervisor was a protective factor for work-related threats at psychiatric wards, but a risk factor in two other sectors. Only high quantitative demands, low rewards at work, many work-family conflicts and low organizational justice were associated with work-related threats in all four sectors.

As in the sector specific analysis with threats, the sector specific analysis with violence lacked power in each stratum. As can be seen in Table 6, some differences between the four sectors were found. First, in the Prison and Probation Services the associations between the psycho-social work environment and work related violence were in general stronger compared to the three others sectors. Second, in the special schools and at the psychiatric wards several risk estimates were below 1 and thus the direction of the association was in some cases opposite compared to the elder sector and the Prison and Probation Service. For instance, low mutual trust between employees was not a risk factor for work-related threats at special schools and at psychiatric wards, but was a risk factor in the two other sectors. Only low role clarity, many role conflicts, many work-family conflicts and low trust regarding management were positively associated with work-related violence across sectors.

4.1 Strengths

The present study has important strengths. First, the response rates are high and above the mean usually seen in organizational surveys [63]. Second, previous research has primarily been based on a single sector or cross-sectional designs. By examining participants from four sectors at the same time, using the same measurements and a follow-up design, this study contributes to the existing literature with more reliable results. Furthermore, the study also identifies certain factors in the psycho-social work environment (high quantitative demands, high emotional demands and many role conflicts) that prospectively and across four sectors are associated with work-related threats and violence. Third, instead of a single item to measure work-related threats and violence, we used a 19-items check-list questionnaire. A single items approach (e.g.‘have you been physically assaulted at work during the past year?’) in some extent leaves it up to the respondents to define what constitutes work-related violence. Thus the reported frequencies of work-related threats and violence may be affected by personal experience, interpretation and the culture of the workplace [64]. Fourth, we controlled statistically for several structural and one personal risk factor which may strengthen the results. Finally, the analyses were carried out using multilevel analysis and thus to some degree took the social context at each workplace into account.

4.2 Limitations

However, the results of the present study should be considered in the light of some potential limitations. First, the participants from special schools, psychiatry and elder care are not necessarily representative as data from these three sectors were based on a convenience sample primarily located in two regions in Denmark. Even though the general sample size is large, the participating workplaces from psychiatry, eldercare and special schools were recruited in a non-random sampling method. Furthermore, some workplaces refused to participate and we cannot rule out the risk of selection bias. If workplaces with poor working environment and/or high levels of exposure to threats and violence in general refused to participate, we may have underestimated the associations, thus reducing the external validity of the findings [65]. In addition, the method of data collection was different across the four sectors. Employees at the psychiatrics wards, elder sectors and the special schools filled out paper-and-pencil questionnaires during a meeting, whereas participants in the Prison and Probation Services were given a web-based questionnaire to fill out, when there was time for it during the workday and a computer was available. This may have affected the response rate (which was lower in Prison and Probation Service compared to the three other sectors) and may again introduce selection bias. If only employees with interest in the topic responded (i.e. those who were most exposed to work-related threats and violence), this may have inflated the associations. However, an additional analysis showed no difference in gender, work experience and education between responders and non-responders at baseline in the Prison and Probation Service (data not shown). Furthermore, only healthy employees participated in the current study. Those on sick leave, including sick-leave due to work-related violence and threats, did not participate. This may have lead to underestimation of the associations between psycho-social work environment and work-related threats and violence. Furthermore, analysing each sector separately affected the sample size for each analysis, and this decreased precision of the estimates. Another limitation relates to the data coming entirely from self-report and hence the possibility that mono method bias or unmeasured third variables are present [66]. A further limitation is the use of dichotomized measures, i.e. threats and violence, which creates a loss of information and perhaps a simplification of the relationships in question. However, dichotomizing also has some advantages as measures such as the odds ratio, often provide more realistic and meaningful assessments of the strength of relationships [67]. The final limitation concerns recall bias. Even though the time span for being exposed to work-related violence and threats was one year as recommended by researchers [47], the answers concerning the frequency of exposure in a questionnaire may be influenced by recall errors [46]. However, it is unlikely that the employees have forgotten important episodes compared to more frequent and everyday like episodes.

We adjusted for baseline violence and threats. However, it can be discussed whether the reported baseline is a reliable baseline, or just an arbitrary point in time [68]. The frequencies of work-related violence and threats at baseline may for example be affected by number and type of clients, pupils, patient and prisoners in the different workplaces and sectors. Therefore, adjusting for baseline violence and threats may inadvertently weaken, rather than strengthen the results by reducing the power of the explanatory variables.

4.3 Future research

More research is needed on how psycho-social work environment is related to work-related violence/threats. First of all, and most important, larger studies with many participants are needed to increase the statistical power. Next, more multi sector studies are needed to confirm or reject that the associations between psycho-social work environment and work-related violence have sector specific characteristics. Additionally, future research may supplement exposure data with the use of diaries as a repeated measure of work-related violence and threats to minimize recall errors. Finally, our results should be replicated in follow-up studies using other and several time lags to identify the dynamics between psycho-social work environment and work-related violence and threats. Longitudinal studies with several measurement points over an extended period of time might add to our knowledge of dynamics between psycho-social work environment and work-related threats and violence