Workplace Violence and Occupational Stress in Healthcare Workers: A Chicken-and-Egg Situation—Results of a 6-Year Follow-up Study Nicola Magnavita, MD Aggregate Professor, Department of Public Health, Universita` Cattolica del Sacro Cuore, Rome, Italy
Key words Cause-effect, psychosocial factors, violence, workplace, occupational exposure, job strain, mental health Correspondence Dr. Nicola Magnavita, Universita` Cattolica del Sacro Cuore-Occupational Medicine Unit, Largo Gemelli 8, 00168 Rome, Italy. E-mail: [email protected] Accepted: March 9, 2014 doi: 10.1111/jnu.12088
Abstract Objectives: Violence at work (WV) is an important occupational hazard for health care workers (HCWs). A number of surveys addressing the causes and effects of WV have shown that it is associated with work-related stress. However, it is not clear what direction this relationship takes, that is, whether job strain facilitates aggression against HCWs or WV is the cause of job strain. Methods: From 2003 to 2009, HCWs from a public health care unit were asked to self-assess their level of work-related stress and to report aggression that occurred in the 12-month period preceding their routine medical examination. In 2009, physical and mental health and job satisfaction were also assessed. A total of 698 out of 723 HCWs (96.5%) completed the study. Findings: Job strain and lack of social support were predictors of the occurrence of nonphysical aggression during the ensuing year. HCWs who experienced WV reported high strain and low support at work in the following year. The experience of nonphysical violence and a prolonged state of strain and social isolation were significant predictors of psychological problems and bad health at follow-up. Conclusions and Implications: The relationship between work-related distress and WV is bidirectional. The monitoring of workers through questionnaires distributed before their periodic examination is a simple and effective way of studying WV and monitoring distress. Clinical Relevance: The findings of the present study may facilitate the subsequent design of participatory intervention for the prevention of violence in healthcare facilities. This should always be accompanied by measures designed to reduce strain and improve social support.
The occurrence of workplace violence (WV) against health care workers (HCWs) is a frequent and pervasive phenomenon (Warren, 2011). Although patients and visitors are mainly responsible for aggression, studies suggest that much of the nonphysical violence faced by HCWs is from colleagues and superiors (Magnavita & Heponiemi, 2011). An ever-increasing number of studies on WV have already established that all professional categories may become victims, even if nurses and physicians experience higher assault rates than other groups (Magnavita & Heponiemi, 2012). Some jobs, such as
Journal of Nursing Scholarship, 2014; 00:0, 1–11. C 2014 Sigma Theta Tau International
those connected with psychiatric or emergency services and first aid, are at greater risk for violence (Magnavita & Heponiemi, 2012). However, all healthcare professions are subject to the threat of violence, as demonstrated by studies on radiologists (Magnavita & Fileni, 2012; Magnavita, Fileni, Pescarini, & Magnavita, 2012), one of the safer, but not entirely risk-free, specialties. In the literature, the variety of methods used, the lack of systematic studies, and the tendency for HCWs not to report all events (Ferns, 2006) make it difficult to assess the exact frequency of the phenomenon. Annual rates of physical 1
Workplace Violence and Stress
aggression against HCWs extend from a low of 3.1% (Kamchuchat, Chongsuvivatwong, Oncheunjit, Yip, & Sangthong, 2008), to a prevalent 11% to 25% (Estryn´ et al., 2009; Gerberich et al., Behar et al., 2008; Gascon 2004; Roche, Diers, Duffield, & Catling-Paul, 2010), or even higher (35%–80%) levels (Chen, Hwu, Kung, Chiu, & Wang, 2008; Franz, Zeh, Schablon, Kuhnert, & Nienhaus, 2010; Zampieron, Galeazzo, Turra, & Buja, 2010). Nonphysical aggression rates are even more difficult to calculate; evaluations ranged from 38% to 90% ´ et al., 2009; Gerberich in a 12-month period (Gascon et al., 2004; Kamchuchat et al., 2008; Roche et al., 2010; Zampieron et al., 2010). All of these studies clearly demonstrate that violence against HCWs is a serious problem. In order to control this phenomenon, it is essential to gain an understanding of the underlying mechanisms. In 2007, the Italian Ministry of Health issued a Recommendation focusing on violence in healthcare facilities (Ministry of Health, 2007). A first-stage objective was to assess the risk for WV by collecting workers’ reports and conducting ad hoc surveys among HCWs. Unfortunately, since no funding was made available for these surveys, the recommendation was generally disregarded on account of a severe shortage of resources. At present, although there is a widely felt need to understand why violence breaks out and what can be done to prevent it, Italian healthcare institutions still lack policies, strategies, and administrative or behavioral provisions to counteract WV. WV has a significant impact not only on workers’ health, but also on productivity. Physical and nonphysical violence among HCWs is associated with a reduction in job satisfaction, increased occupational strain, and poor ´ et al., 2013; Magnavita & patient care outcomes (Gascon Fileni, 2012; Magnavita & Heponiemi, 2011, 2012; Roche et al, 2010), and an association has been reported between distress and WV (Gates, Fitzwater, & Succop, 2003; St. Pierre & Holmes, 2008), although the direction of this association needs to be clarified. It is unclear if the condition of job strain is created by WV or if it is caused by other factors (such as insufficient staff, high workload), and this condition then increases the incidence of WV. Since the direction of the relationship cannot be documented in cross-sectional studies, researchers base their interpretation mainly on their own beliefs. In their pioneering study on 6,300 Minnesota nurses, Gerberich et al. (2004) emphasized the impact of violence on workers’ health, with consequent distress and psychological and behavioral changes, whereas Zampieron et al. (2010) claimed that nurses are at the highest risk for aggression when they are overtired, distressed, and dissatisfied with their work, and asserted that the prevention of WV should be based on improved work organization. 2
Magnavita
The choice of one of these two pathogenetic hypotheses is of prime importance for the design of preventive measures. If the primum movens is aggression, we must increase the resilience of workers and arrange counseling intervention. If, on the contrary, it is mainly the state of distress of workers that causes violence, we should direct preventive intervention primarily toward eliminating occupational stress factors and lowering levels of staff distress. In order to clarify this issue, we conducted this research in a public healthcare unit in which we had already carried out a cross-sectional study. The latter had found a yearly rate of physical aggression of roughly 9%, and of nonphysical aggression of 25% to 30%, and had reported that nonphysical violence was associated with high job strain, low support, low perceived organizational justice, and high psychological distress (Magnavita & Heponiemi, 2012). In the present study we used a longitudinal method, because it was more suitable for evaluating the direction of the relationship between WV and job strain. The first hypothesis we examined was that job strain promoted acts of violence. A worker suffering from a state of distress might adopt behavior or work procedures that exposed him to acts of violence. We therefore expected perceived job strain to be significantly associated with acts of violence occurring in the following year. The second hypothesis we considered was that violence increased job strain. According to this hypothesis, a significant correlation would be observed between the occurrence of violence and stress perceived in the year following these violent acts. We also expected to find that the state of health of workers would be adversely affected by acts of violence and stress. A second aim of our study was to ascertain whether a regular collection of data by the Department of Health Surveillance could contribute to WV risk assessment, thus avoiding the need for specific and costly ad hoc surveys.
Method Participants The local Roma F (RMF) health unit is a public health authority that includes two hospitals and numerous local services in the province of Rome. During their routine medical examinations, HCWs from the RMF unit were invited to answer a questionnaire about their own occupational risks and state of health. In 2003, the questionnaire included a baseline assessment of work stress, anxiety, and depression. In 2005, 2007, and 2009, workers were invited to reassess their self-perceived level of work stress and were asked to describe their experience Journal of Nursing Scholarship, 2014; 00:0, 1–11. C 2014 Sigma Theta Tau International
Workplace Violence and Stress
Magnavita
of WV with reference to the previous 12 months. In 2009, physical and mental health and job satisfaction were also assessed. At the beginning of the study, all workers who had been exposed to any occupational risk (e.g., chemical, biological, physical, or organizational and psychosocial hazards) for at least 1 year (i.e. 1,413 out of the 2,407 employees working in the RMF unit) were invited to participate in the survey. Since only two of them refused, at baseline the cohort comprised 1,411 HCWs. During the observation period, workers who changed health unit or job, and workers who had not undergone routine medical examination at least every 2 years, were excluded from the study. In 2009, the eligible population comprised 723 persons. Complete responses were obtained from 698 workers, or 96.5% of HCWs called for routine examination. In Italy, all workers exposed to occupational risks must undergo regular medical examination. For many years, the employees who participated in this study had been accustomed to filling out questionnaires while waiting for their medical examination, so they were aware that the results would remain confidential and would be used in their interest. Although participation was not obligatory, most of the employees chose to complete the questionnaire. The main reason given for non-participation or for incomplete answers was lack of time. This research project was approved by the Ethics Committee of the Universita` Cattolica del Sacro Cuore of Rome.
Questionnaires Occupational stress was measured using the Demand/Control/Support (DCS) Questionnaire (Karasek, 1979) in its Italian version (Magnavita, 2007b). The questionnaire contained five items for the demand scale, six for the control scale, and six for the support scale. A Likert 4-point scale graded from 1 = never to 4 = often was used for demand and control scales, while the third scale (support) was graded from 1 = strong disagreement to 4 = strong agreement. The perceived job strain was calculated in two ways: (a) a continuous variable (demand/control [D/C]), obtained by dividing demand by control (weighted by item numbers); and (b) a categorical variable, obtained from the various combinations of high and low levels of demands and decision latitude, resulting in four types of work situations: (1) high strain, (2) low strain, (3) active, and (4) passive jobs. Validation of the translation ensured that the Italian version maintained the characteristics of the original and that the internal reliability (Cronbach’s α) of each subscale was satisfactory (in this cohort, values were 0.705–0.801 for demand; 0.661–0.680 for control; and 0.841–0.858 for support). Journal of Nursing Scholarship, 2014; 00:0, 1–11. C 2014 Sigma Theta Tau International
Violent incidents were registered using the Violent Incident Form (Arnetz, 1998). Physical aggression was defined as forceful, hostile, or aggressive behavior that may or may not cause harm. Nonphysical aggression was defined as any frustrating or unfriendly act (words, attitudes, actions) that creates a hostile work environment. The questionnaire showed good reliability (SpearmanBrown coefficient = 0.91; Magnavita & Heponiemi, 2011). Anxiety and depression were assessed using the Goldberg Questionnaire (Goldberg, Bridges, Duncan-Jones, & Grayson, 1988) in its Italian version (Magnavita, 2007a). This short questionnaire, which was developed to be used by nonpsychiatrists, contains two scales of nine binary items that assign 1 point for each positive response. The resulting scores range from 0 to 9 for each variable. In this study, the reliability coefficient (Cronbach’s α) was 0.809 for the anxiety scale and 0.765 for the depression scale. Mental health was assessed by the 12-item General Health Questionnaire (GHQ-12; Goldberg, 1978) in its Italian version (Piccinelli, Bisoffi, Bon, Cunico, & Tansella, 1993). Using the Likert scoring method from 1 to 4, we formed a “psychological problems” variable, ranging from 12 to 48. In this study, the reliability coefficient (Cronbach’s α) of GHQ-12 was 0.871. Physical health was self-assessed by responding to a single question: “How would you rate your overall health?” Answers were scaled from 1 (severe problems) to 5 (very good). The satisfaction derived from work was measured by the final question in the Italian version (Magnavita et al., 2007) of the Job Satisfaction Scale (JSS) developed by Warr, Cook, and Wall (1979 p. 146): “Taking into consideration every aspect of your present job, how do you feel about your job as a whole?” The scaled answers ranged from “very dissatisfied” (1 point) to “extremely satisfied” (7 points).
Statistical Analysis Descriptive statistics were used to characterize the study population. Repeated-measures analysis of variance (ANOVA) was used to compare stress levels recorded in different years. The Bonferroni post-hoc test was then used to determine which means differed. Logistic regression analysis was used to study the distribution of violent episodes across gender, jobs, and departments. Odds ratios (OR) and 95% confidence intervals (CI) were calculated. In a second step, logistic regression analysis was used again to ascertain the association between stress measured in each of the three temporal moments and violence experienced in the following year. Initially each variable was included individually in a univariate logistic 3
Workplace Violence and Stress
regression model, and the crude odds ratio was calculated; this value was then corrected by inserting the confounders (age, gender, profession, and type of healthcare activities). In a third step, the effect of violence on work-related stress in the following year was calculated using logistic regression. In this case, high job strain and low social support were respectively set as the dependent variable in separate equations, while violent events (physical assault, threats, harassment) reported in the 12 months preceding the measurement of the stress-related variables were considered to be independent variables. Finally, three separate backward stepwise linear regression models were fitted to identify predictors of psychological problems (GHQ score), physical health, and job satisfaction. In each model, baseline levels of anxiety and depression, together with age, sex, frequency of aggression, and the average values of job strain and support, were entered as predictors. Statistics were performed using the IBM/SPSS, Version 20.0 statistical package (IBM Corporation, Armonk, NY, USA).
Results A total of 251 male healthcare workers (36%) and 447 female workers (64%) completed the longitudinal study. The occupational characteristics of the population and the psychosocial values measured at the beginning and end of the investigation are shown in Table 1 (available with online version of the article).
Changes in Stress Levels Repeated-measures ANOVA revealed an overall significant difference in the means of all stress-related variables (p < .000). However, Bonferroni comparisons demonstrated that each variable followed a particular change pattern. Demand oscillated with an increase in 2005, a decrease in 2007, and a further increase in 2009, thus reaching levels not significantly different from those of 2003. Control increased in 2005, underwent no significant changes between 2005 and 2007, but showed another significant increase in 2009, thus rising to levels significantly higher than those of 2003. Job strain underwent a slight but significant decrease in 2005 and a more evident reduction in 2007, but remained stable in 2009. A gradual decline in social support was observed from 2003 to 2009. At the end of our observations of the cohort, control significantly increased, while job strain and support were found to be lower than baseline values. In each of the two-yearly examinations, workers reported physical and nonphysical aggression at the workplace. The frequency of physical aggression during the 4
Magnavita
12-month period preceding each routine medical examination ranged from 8.5% to 11.6%, while nonphysical aggression was reported by 22.5% to 28.5% of HCWs (see Table 1, available with online version of the article).
Workplace Violence Across the entire study period, a total of 172 workers (24.6% of the cohort) reported being the victim of at least one act of physical aggression in the 12-month period preceding the survey; 24 of them (3.4%) reported physical assault in two separate years and 5 (0.7%) in all the years investigated. The perpetrators of this violence were almost exclusively patients or visitors, three fourths of whom were male. Nonphysical aggression was more frequent than the physical aggression. A total of 367 HCWs (52.6%) reported having been exposed to at least one act of nonphysical aggression in the preceding year, and 135 (19.4%) reported being the victims of nonphysical aggression in more than one of the years studied. Patients, their relatives, and friends were responsible for more than half of these episodes; however, most of the remaining cases of violence were perpetrated by colleagues or superiors. The occurrence of WV was severely under-reported; only in exceptional cases was a report made to the Accident and Emergency Department, the Health Department, or the police. Both physical and nonphysical aggression was equally prevalent in male and female workers (Table 2, available with online version of the article). Nurses and doctors were the professional groups most exposed to aggression. A comparison with all other workers indicated that HCWs had a sevenfold increased risk for physical violence, while the risk for nonphysical violence was more than double (see Table 2, available with online version of the article). Results showed that workers in psychiatric and emergency services are at greatest risk for physical aggression, since about half of all violence of this kind is concentrated in these healthcare areas. If the risk for an office employee is taken to be 1, for workers in mental health services, the risk for undergoing physical aggression is 45 times higher (OR 45.9, 95% CI 15.9–132.2), and is 13 times higher (OR 13.0, 95% CI 5.3–31.9) for undergoing nonphysical violence. Staff in accident and emergency departments were at a significantly higher risk for violence (OR 6.9, 95% CI 2.5–19.0 due to physical aggression; OR 4.0, 95% CI 2.1–7.7 for nonphysical aggression). Workers in surgical and medical wards and in outpatient wards also reported a higher risk for physical violence than office workers, while the incidence of episodes of nonphysical violence was the same in these departments and offices (see Table 2, available with online version of the article). Journal of Nursing Scholarship, 2014; 00:0, 1–11. C 2014 Sigma Theta Tau International
Workplace Violence and Stress
Magnavita
Table 1. Descriptive Characteristics of the Observed Cohort 2003 N. of workers (% response rate) 1411 Male workers N (%) 594 (42.1) Female workers N (%) 817 (57.9) Age, mean ± s.d. (years)2 40.5±9.6 Length of work, mean ± s.d. (years)2 11.7±9.8 Physicians, N (%) 270 (19.1) Nurses, N (%) 728 (51.6) Technicians1 , N (%) 211 (15.0) Clerks, N (%) 202 (14.3) Psychiatry 83 (5.9) A&E 112 (7.9) Inpatient wards 418 (29.6) Outpatient 339 (24.0) Laboratory and services 246 (17.4) Office 213 (15.1) Anxiety, range 0–9, mean±s.d.2 3.1± 2.7 Depression,range 0–9, mean±s.d.2 1.4± 1.9 Demand, range 5–20, mean + s.d. 13.7 ± 2.7 Control, range 6–24, mean + s.d. 16.4 + 3.1 Support, range 6–24, mean + s.d. 20.5 + 2.9 Job strain, mean + s.d. 1.04+0.35 Mental health, GHQ12 (range 12–48), mean + s.d. Physical health (range 1–5), mean + s.d. Job satisfaction (range 1–7), mean + s.d. Cases of violence Physical aggression (N,%) Non-physical aggression (N,%)
2005
2007
2009
p
698 (96.5) 251 (36.0) 447 (64.0) 38.9±8.7 11.6 ± 8.8 125 (17.9) 419 (60.0) 70 (10.0) 84 (12.0) 67 (9.6) 88 (12.6) 226 (32.4) 151 (21.6) 83 (11.9) 83 (11.9)
14.0 ± 2.7 16.4 + 3.0 20.1 + 2.9 1.04+0.32
13.6 ± 2.8 16.8 + 3.0 19.9 + 3.1 1.01+0.31
2004 59 (8.5) 164 (23.5)
2006 66 (9.5) 201 (28.8)
13.8 ± 2.9 16.9 + 3.0 19.8 + 3.2 1.01+0.29 23.5±5.6 3.2±1.1 4.4+1.4 2008 81 (11.6) 157 (22.5)
0.000 0.000 0.000 0.000
(1 ): this category includes: laboratory technicians, radiology technicians, physiotherapists, psychologists, ancillary personnel, and blue collar workers (2 ): measured at baseline
Effect of Distress on Subsequent Violence Logistic regression analysis showed that stress-related variables were not good predictors of the occurrence of physical attacks against HCWs. Social support measured in 2007 significantly reduced the risk for aggression in the following year (Table 3, available with online version of the article). Stress variables proved to be better predictors of nonphysical violence. Nonphysical aggression was more frequent in workers with high-demand jobs, low decisional latitude, and a high D/C ratio in the previous year. A significant increase in the ORs of aggression in the following year could be seen for stress variables measured in 2003 and in 2007. Social support significantly reduced the ORs for nonphysical violence in all the observation period (see Table 3, available with online version of the article).
in subsequent years. A significant association was found between physical aggression and low social support. Nonphysical aggression was significantly associated with high job strain and low support in the following year (Table 4, available with online version of the article). WV and job strain significantly affected mental and physical health (Table 5, available with online version of the article). Psychological problems at the end of the observation period were significantly associated with anxiety and depression at the baseline, female gender, perceived job strain and low social support, and nonphysical aggression. Physical health was negatively associated with age, depression, female gender, job strain, and nonphysical aggression, and positively associated with social support. Job satisfaction was negatively associated with job strain and positively associated with support.
Effect of Violence on Distress and Health
Discussion
Workers who had experienced violence in the previous year were at greater risk for work-related stress
We found that job strain and lack of social support were predictors of the occurrence of nonphysical aggression
Journal of Nursing Scholarship, 2014; 00:0, 1–11. C 2014 Sigma Theta Tau International
5
6
1.15 (0.67–1.99) 1 6.3 (2.1–19.2)∗∗∗ 3.6 (1.3–10.4)∗ 1 16.9 (4.8–59.2)∗∗∗ 1.6 (0.4–6.9) 2.3 (0.7–8.0) 0.7 (0.2–3.3) 1.0 (0.2–5.1) 1
0.76 (0.44–1.31) 1 7.9 (2.3–27.7)∗∗∗ 6.2 (1.9–20.3)∗∗ 1 34.9 (4.5–268.4)∗∗∗ 11.7 (1.5–92.9)∗ 7.5 (0.99–57.1) 6.4 (0.8–50.8) 4.2 (0.5–38.0) 1
2006 1.12 (0.69–1.80) 1 7.3 (1.6–33.7)∗ 14.7 (3.6–60.9)∗∗∗ 1 66.5 (8.7–506.1)∗∗∗ 15.5 (2.0–120.9)∗∗ 8.4 (1.1–63.5)∗ 7.1 (0.9–55.4) 3.1 (0.3–30.2) 1
2008 0.91 (0.63–1.30) 1 6.8 (3.1–14.7)∗∗∗ 6.9 (3.4–14.0)∗∗∗ 1 45.9 (15.9–132.2)∗∗∗ 6.9 (2.5–19.0)∗∗∗ 4.9 (1.9–12.7)∗∗∗ 3.2 (1.2–8.8)∗ 2.1 (0.7–6.5) 1
Total1 1.41 (0.98–2.01) 1 1.2 (0.7–2.1) 1.6 (1.01–2.6)∗ 1 2.5 (1.2–5.0)∗ 2.0 (0.99–3.9) 0.8 (0.4–1.5) 0.9 (0.4–1.7) 1.1 (0.5–2.2) 1
2004 0.71 (0.50–1.00) 1 2.4 (1.4–4.1)∗∗ 2.0 (1.3–3.2)∗∗ 1 4.7 (2.3–9.5)∗∗∗ 1.6 (0.8–3.1) 1.4 (0.8–2.6) 1.1 (0.6–2.1) 0.5 (0.2–1.2) 1
2006 0.71 (0.48–1.04) 1 4.2 (2.1–8.6)∗∗∗ 4.3 (2.3–8.1)∗∗∗ 1 14.8 (6.1–35.6)∗∗∗ 6.5 (2.8–15.1)∗∗∗ 1.6 (0.7–3.5) 2.0 (0.9–4.7) 1.7 (0.7–4.4) 1
2008
Non-physical aggression
0.74 (0.54–1.01) 1 2.5 (1.5–4.0)∗∗∗ 2.9 (2.0–4.3)∗∗∗ 1 13.0 (5.3–31.9)∗∗∗ 4.0 (2.1–7.7)∗∗∗ 1.3 (0.8–2.1) 1.4 (0.8–2.4) 1.0 (0.5–1.9) 1
Total2
(∗ ) p
Key words Cause-effect, psychosocial factors, violence, workplace, occupational exposure, job strain, mental health Correspondence Dr. Nicola Magnavita, Universita` Cattolica del Sacro Cuore-Occupational Medicine Unit, Largo Gemelli 8, 00168 Rome, Italy. E-mail: [email protected] Accepted: March 9, 2014 doi: 10.1111/jnu.12088
Abstract Objectives: Violence at work (WV) is an important occupational hazard for health care workers (HCWs). A number of surveys addressing the causes and effects of WV have shown that it is associated with work-related stress. However, it is not clear what direction this relationship takes, that is, whether job strain facilitates aggression against HCWs or WV is the cause of job strain. Methods: From 2003 to 2009, HCWs from a public health care unit were asked to self-assess their level of work-related stress and to report aggression that occurred in the 12-month period preceding their routine medical examination. In 2009, physical and mental health and job satisfaction were also assessed. A total of 698 out of 723 HCWs (96.5%) completed the study. Findings: Job strain and lack of social support were predictors of the occurrence of nonphysical aggression during the ensuing year. HCWs who experienced WV reported high strain and low support at work in the following year. The experience of nonphysical violence and a prolonged state of strain and social isolation were significant predictors of psychological problems and bad health at follow-up. Conclusions and Implications: The relationship between work-related distress and WV is bidirectional. The monitoring of workers through questionnaires distributed before their periodic examination is a simple and effective way of studying WV and monitoring distress. Clinical Relevance: The findings of the present study may facilitate the subsequent design of participatory intervention for the prevention of violence in healthcare facilities. This should always be accompanied by measures designed to reduce strain and improve social support.
The occurrence of workplace violence (WV) against health care workers (HCWs) is a frequent and pervasive phenomenon (Warren, 2011). Although patients and visitors are mainly responsible for aggression, studies suggest that much of the nonphysical violence faced by HCWs is from colleagues and superiors (Magnavita & Heponiemi, 2011). An ever-increasing number of studies on WV have already established that all professional categories may become victims, even if nurses and physicians experience higher assault rates than other groups (Magnavita & Heponiemi, 2012). Some jobs, such as
Journal of Nursing Scholarship, 2014; 00:0, 1–11. C 2014 Sigma Theta Tau International
those connected with psychiatric or emergency services and first aid, are at greater risk for violence (Magnavita & Heponiemi, 2012). However, all healthcare professions are subject to the threat of violence, as demonstrated by studies on radiologists (Magnavita & Fileni, 2012; Magnavita, Fileni, Pescarini, & Magnavita, 2012), one of the safer, but not entirely risk-free, specialties. In the literature, the variety of methods used, the lack of systematic studies, and the tendency for HCWs not to report all events (Ferns, 2006) make it difficult to assess the exact frequency of the phenomenon. Annual rates of physical 1
Workplace Violence and Stress
aggression against HCWs extend from a low of 3.1% (Kamchuchat, Chongsuvivatwong, Oncheunjit, Yip, & Sangthong, 2008), to a prevalent 11% to 25% (Estryn´ et al., 2009; Gerberich et al., Behar et al., 2008; Gascon 2004; Roche, Diers, Duffield, & Catling-Paul, 2010), or even higher (35%–80%) levels (Chen, Hwu, Kung, Chiu, & Wang, 2008; Franz, Zeh, Schablon, Kuhnert, & Nienhaus, 2010; Zampieron, Galeazzo, Turra, & Buja, 2010). Nonphysical aggression rates are even more difficult to calculate; evaluations ranged from 38% to 90% ´ et al., 2009; Gerberich in a 12-month period (Gascon et al., 2004; Kamchuchat et al., 2008; Roche et al., 2010; Zampieron et al., 2010). All of these studies clearly demonstrate that violence against HCWs is a serious problem. In order to control this phenomenon, it is essential to gain an understanding of the underlying mechanisms. In 2007, the Italian Ministry of Health issued a Recommendation focusing on violence in healthcare facilities (Ministry of Health, 2007). A first-stage objective was to assess the risk for WV by collecting workers’ reports and conducting ad hoc surveys among HCWs. Unfortunately, since no funding was made available for these surveys, the recommendation was generally disregarded on account of a severe shortage of resources. At present, although there is a widely felt need to understand why violence breaks out and what can be done to prevent it, Italian healthcare institutions still lack policies, strategies, and administrative or behavioral provisions to counteract WV. WV has a significant impact not only on workers’ health, but also on productivity. Physical and nonphysical violence among HCWs is associated with a reduction in job satisfaction, increased occupational strain, and poor ´ et al., 2013; Magnavita & patient care outcomes (Gascon Fileni, 2012; Magnavita & Heponiemi, 2011, 2012; Roche et al, 2010), and an association has been reported between distress and WV (Gates, Fitzwater, & Succop, 2003; St. Pierre & Holmes, 2008), although the direction of this association needs to be clarified. It is unclear if the condition of job strain is created by WV or if it is caused by other factors (such as insufficient staff, high workload), and this condition then increases the incidence of WV. Since the direction of the relationship cannot be documented in cross-sectional studies, researchers base their interpretation mainly on their own beliefs. In their pioneering study on 6,300 Minnesota nurses, Gerberich et al. (2004) emphasized the impact of violence on workers’ health, with consequent distress and psychological and behavioral changes, whereas Zampieron et al. (2010) claimed that nurses are at the highest risk for aggression when they are overtired, distressed, and dissatisfied with their work, and asserted that the prevention of WV should be based on improved work organization. 2
Magnavita
The choice of one of these two pathogenetic hypotheses is of prime importance for the design of preventive measures. If the primum movens is aggression, we must increase the resilience of workers and arrange counseling intervention. If, on the contrary, it is mainly the state of distress of workers that causes violence, we should direct preventive intervention primarily toward eliminating occupational stress factors and lowering levels of staff distress. In order to clarify this issue, we conducted this research in a public healthcare unit in which we had already carried out a cross-sectional study. The latter had found a yearly rate of physical aggression of roughly 9%, and of nonphysical aggression of 25% to 30%, and had reported that nonphysical violence was associated with high job strain, low support, low perceived organizational justice, and high psychological distress (Magnavita & Heponiemi, 2012). In the present study we used a longitudinal method, because it was more suitable for evaluating the direction of the relationship between WV and job strain. The first hypothesis we examined was that job strain promoted acts of violence. A worker suffering from a state of distress might adopt behavior or work procedures that exposed him to acts of violence. We therefore expected perceived job strain to be significantly associated with acts of violence occurring in the following year. The second hypothesis we considered was that violence increased job strain. According to this hypothesis, a significant correlation would be observed between the occurrence of violence and stress perceived in the year following these violent acts. We also expected to find that the state of health of workers would be adversely affected by acts of violence and stress. A second aim of our study was to ascertain whether a regular collection of data by the Department of Health Surveillance could contribute to WV risk assessment, thus avoiding the need for specific and costly ad hoc surveys.
Method Participants The local Roma F (RMF) health unit is a public health authority that includes two hospitals and numerous local services in the province of Rome. During their routine medical examinations, HCWs from the RMF unit were invited to answer a questionnaire about their own occupational risks and state of health. In 2003, the questionnaire included a baseline assessment of work stress, anxiety, and depression. In 2005, 2007, and 2009, workers were invited to reassess their self-perceived level of work stress and were asked to describe their experience Journal of Nursing Scholarship, 2014; 00:0, 1–11. C 2014 Sigma Theta Tau International
Workplace Violence and Stress
Magnavita
of WV with reference to the previous 12 months. In 2009, physical and mental health and job satisfaction were also assessed. At the beginning of the study, all workers who had been exposed to any occupational risk (e.g., chemical, biological, physical, or organizational and psychosocial hazards) for at least 1 year (i.e. 1,413 out of the 2,407 employees working in the RMF unit) were invited to participate in the survey. Since only two of them refused, at baseline the cohort comprised 1,411 HCWs. During the observation period, workers who changed health unit or job, and workers who had not undergone routine medical examination at least every 2 years, were excluded from the study. In 2009, the eligible population comprised 723 persons. Complete responses were obtained from 698 workers, or 96.5% of HCWs called for routine examination. In Italy, all workers exposed to occupational risks must undergo regular medical examination. For many years, the employees who participated in this study had been accustomed to filling out questionnaires while waiting for their medical examination, so they were aware that the results would remain confidential and would be used in their interest. Although participation was not obligatory, most of the employees chose to complete the questionnaire. The main reason given for non-participation or for incomplete answers was lack of time. This research project was approved by the Ethics Committee of the Universita` Cattolica del Sacro Cuore of Rome.
Questionnaires Occupational stress was measured using the Demand/Control/Support (DCS) Questionnaire (Karasek, 1979) in its Italian version (Magnavita, 2007b). The questionnaire contained five items for the demand scale, six for the control scale, and six for the support scale. A Likert 4-point scale graded from 1 = never to 4 = often was used for demand and control scales, while the third scale (support) was graded from 1 = strong disagreement to 4 = strong agreement. The perceived job strain was calculated in two ways: (a) a continuous variable (demand/control [D/C]), obtained by dividing demand by control (weighted by item numbers); and (b) a categorical variable, obtained from the various combinations of high and low levels of demands and decision latitude, resulting in four types of work situations: (1) high strain, (2) low strain, (3) active, and (4) passive jobs. Validation of the translation ensured that the Italian version maintained the characteristics of the original and that the internal reliability (Cronbach’s α) of each subscale was satisfactory (in this cohort, values were 0.705–0.801 for demand; 0.661–0.680 for control; and 0.841–0.858 for support). Journal of Nursing Scholarship, 2014; 00:0, 1–11. C 2014 Sigma Theta Tau International
Violent incidents were registered using the Violent Incident Form (Arnetz, 1998). Physical aggression was defined as forceful, hostile, or aggressive behavior that may or may not cause harm. Nonphysical aggression was defined as any frustrating or unfriendly act (words, attitudes, actions) that creates a hostile work environment. The questionnaire showed good reliability (SpearmanBrown coefficient = 0.91; Magnavita & Heponiemi, 2011). Anxiety and depression were assessed using the Goldberg Questionnaire (Goldberg, Bridges, Duncan-Jones, & Grayson, 1988) in its Italian version (Magnavita, 2007a). This short questionnaire, which was developed to be used by nonpsychiatrists, contains two scales of nine binary items that assign 1 point for each positive response. The resulting scores range from 0 to 9 for each variable. In this study, the reliability coefficient (Cronbach’s α) was 0.809 for the anxiety scale and 0.765 for the depression scale. Mental health was assessed by the 12-item General Health Questionnaire (GHQ-12; Goldberg, 1978) in its Italian version (Piccinelli, Bisoffi, Bon, Cunico, & Tansella, 1993). Using the Likert scoring method from 1 to 4, we formed a “psychological problems” variable, ranging from 12 to 48. In this study, the reliability coefficient (Cronbach’s α) of GHQ-12 was 0.871. Physical health was self-assessed by responding to a single question: “How would you rate your overall health?” Answers were scaled from 1 (severe problems) to 5 (very good). The satisfaction derived from work was measured by the final question in the Italian version (Magnavita et al., 2007) of the Job Satisfaction Scale (JSS) developed by Warr, Cook, and Wall (1979 p. 146): “Taking into consideration every aspect of your present job, how do you feel about your job as a whole?” The scaled answers ranged from “very dissatisfied” (1 point) to “extremely satisfied” (7 points).
Statistical Analysis Descriptive statistics were used to characterize the study population. Repeated-measures analysis of variance (ANOVA) was used to compare stress levels recorded in different years. The Bonferroni post-hoc test was then used to determine which means differed. Logistic regression analysis was used to study the distribution of violent episodes across gender, jobs, and departments. Odds ratios (OR) and 95% confidence intervals (CI) were calculated. In a second step, logistic regression analysis was used again to ascertain the association between stress measured in each of the three temporal moments and violence experienced in the following year. Initially each variable was included individually in a univariate logistic 3
Workplace Violence and Stress
regression model, and the crude odds ratio was calculated; this value was then corrected by inserting the confounders (age, gender, profession, and type of healthcare activities). In a third step, the effect of violence on work-related stress in the following year was calculated using logistic regression. In this case, high job strain and low social support were respectively set as the dependent variable in separate equations, while violent events (physical assault, threats, harassment) reported in the 12 months preceding the measurement of the stress-related variables were considered to be independent variables. Finally, three separate backward stepwise linear regression models were fitted to identify predictors of psychological problems (GHQ score), physical health, and job satisfaction. In each model, baseline levels of anxiety and depression, together with age, sex, frequency of aggression, and the average values of job strain and support, were entered as predictors. Statistics were performed using the IBM/SPSS, Version 20.0 statistical package (IBM Corporation, Armonk, NY, USA).
Results A total of 251 male healthcare workers (36%) and 447 female workers (64%) completed the longitudinal study. The occupational characteristics of the population and the psychosocial values measured at the beginning and end of the investigation are shown in Table 1 (available with online version of the article).
Changes in Stress Levels Repeated-measures ANOVA revealed an overall significant difference in the means of all stress-related variables (p < .000). However, Bonferroni comparisons demonstrated that each variable followed a particular change pattern. Demand oscillated with an increase in 2005, a decrease in 2007, and a further increase in 2009, thus reaching levels not significantly different from those of 2003. Control increased in 2005, underwent no significant changes between 2005 and 2007, but showed another significant increase in 2009, thus rising to levels significantly higher than those of 2003. Job strain underwent a slight but significant decrease in 2005 and a more evident reduction in 2007, but remained stable in 2009. A gradual decline in social support was observed from 2003 to 2009. At the end of our observations of the cohort, control significantly increased, while job strain and support were found to be lower than baseline values. In each of the two-yearly examinations, workers reported physical and nonphysical aggression at the workplace. The frequency of physical aggression during the 4
Magnavita
12-month period preceding each routine medical examination ranged from 8.5% to 11.6%, while nonphysical aggression was reported by 22.5% to 28.5% of HCWs (see Table 1, available with online version of the article).
Workplace Violence Across the entire study period, a total of 172 workers (24.6% of the cohort) reported being the victim of at least one act of physical aggression in the 12-month period preceding the survey; 24 of them (3.4%) reported physical assault in two separate years and 5 (0.7%) in all the years investigated. The perpetrators of this violence were almost exclusively patients or visitors, three fourths of whom were male. Nonphysical aggression was more frequent than the physical aggression. A total of 367 HCWs (52.6%) reported having been exposed to at least one act of nonphysical aggression in the preceding year, and 135 (19.4%) reported being the victims of nonphysical aggression in more than one of the years studied. Patients, their relatives, and friends were responsible for more than half of these episodes; however, most of the remaining cases of violence were perpetrated by colleagues or superiors. The occurrence of WV was severely under-reported; only in exceptional cases was a report made to the Accident and Emergency Department, the Health Department, or the police. Both physical and nonphysical aggression was equally prevalent in male and female workers (Table 2, available with online version of the article). Nurses and doctors were the professional groups most exposed to aggression. A comparison with all other workers indicated that HCWs had a sevenfold increased risk for physical violence, while the risk for nonphysical violence was more than double (see Table 2, available with online version of the article). Results showed that workers in psychiatric and emergency services are at greatest risk for physical aggression, since about half of all violence of this kind is concentrated in these healthcare areas. If the risk for an office employee is taken to be 1, for workers in mental health services, the risk for undergoing physical aggression is 45 times higher (OR 45.9, 95% CI 15.9–132.2), and is 13 times higher (OR 13.0, 95% CI 5.3–31.9) for undergoing nonphysical violence. Staff in accident and emergency departments were at a significantly higher risk for violence (OR 6.9, 95% CI 2.5–19.0 due to physical aggression; OR 4.0, 95% CI 2.1–7.7 for nonphysical aggression). Workers in surgical and medical wards and in outpatient wards also reported a higher risk for physical violence than office workers, while the incidence of episodes of nonphysical violence was the same in these departments and offices (see Table 2, available with online version of the article). Journal of Nursing Scholarship, 2014; 00:0, 1–11. C 2014 Sigma Theta Tau International
Workplace Violence and Stress
Magnavita
Table 1. Descriptive Characteristics of the Observed Cohort 2003 N. of workers (% response rate) 1411 Male workers N (%) 594 (42.1) Female workers N (%) 817 (57.9) Age, mean ± s.d. (years)2 40.5±9.6 Length of work, mean ± s.d. (years)2 11.7±9.8 Physicians, N (%) 270 (19.1) Nurses, N (%) 728 (51.6) Technicians1 , N (%) 211 (15.0) Clerks, N (%) 202 (14.3) Psychiatry 83 (5.9) A&E 112 (7.9) Inpatient wards 418 (29.6) Outpatient 339 (24.0) Laboratory and services 246 (17.4) Office 213 (15.1) Anxiety, range 0–9, mean±s.d.2 3.1± 2.7 Depression,range 0–9, mean±s.d.2 1.4± 1.9 Demand, range 5–20, mean + s.d. 13.7 ± 2.7 Control, range 6–24, mean + s.d. 16.4 + 3.1 Support, range 6–24, mean + s.d. 20.5 + 2.9 Job strain, mean + s.d. 1.04+0.35 Mental health, GHQ12 (range 12–48), mean + s.d. Physical health (range 1–5), mean + s.d. Job satisfaction (range 1–7), mean + s.d. Cases of violence Physical aggression (N,%) Non-physical aggression (N,%)
2005
2007
2009
p
698 (96.5) 251 (36.0) 447 (64.0) 38.9±8.7 11.6 ± 8.8 125 (17.9) 419 (60.0) 70 (10.0) 84 (12.0) 67 (9.6) 88 (12.6) 226 (32.4) 151 (21.6) 83 (11.9) 83 (11.9)
14.0 ± 2.7 16.4 + 3.0 20.1 + 2.9 1.04+0.32
13.6 ± 2.8 16.8 + 3.0 19.9 + 3.1 1.01+0.31
2004 59 (8.5) 164 (23.5)
2006 66 (9.5) 201 (28.8)
13.8 ± 2.9 16.9 + 3.0 19.8 + 3.2 1.01+0.29 23.5±5.6 3.2±1.1 4.4+1.4 2008 81 (11.6) 157 (22.5)
0.000 0.000 0.000 0.000
(1 ): this category includes: laboratory technicians, radiology technicians, physiotherapists, psychologists, ancillary personnel, and blue collar workers (2 ): measured at baseline
Effect of Distress on Subsequent Violence Logistic regression analysis showed that stress-related variables were not good predictors of the occurrence of physical attacks against HCWs. Social support measured in 2007 significantly reduced the risk for aggression in the following year (Table 3, available with online version of the article). Stress variables proved to be better predictors of nonphysical violence. Nonphysical aggression was more frequent in workers with high-demand jobs, low decisional latitude, and a high D/C ratio in the previous year. A significant increase in the ORs of aggression in the following year could be seen for stress variables measured in 2003 and in 2007. Social support significantly reduced the ORs for nonphysical violence in all the observation period (see Table 3, available with online version of the article).
in subsequent years. A significant association was found between physical aggression and low social support. Nonphysical aggression was significantly associated with high job strain and low support in the following year (Table 4, available with online version of the article). WV and job strain significantly affected mental and physical health (Table 5, available with online version of the article). Psychological problems at the end of the observation period were significantly associated with anxiety and depression at the baseline, female gender, perceived job strain and low social support, and nonphysical aggression. Physical health was negatively associated with age, depression, female gender, job strain, and nonphysical aggression, and positively associated with social support. Job satisfaction was negatively associated with job strain and positively associated with support.
Effect of Violence on Distress and Health
Discussion
Workers who had experienced violence in the previous year were at greater risk for work-related stress
We found that job strain and lack of social support were predictors of the occurrence of nonphysical aggression
Journal of Nursing Scholarship, 2014; 00:0, 1–11. C 2014 Sigma Theta Tau International
5
6
1.15 (0.67–1.99) 1 6.3 (2.1–19.2)∗∗∗ 3.6 (1.3–10.4)∗ 1 16.9 (4.8–59.2)∗∗∗ 1.6 (0.4–6.9) 2.3 (0.7–8.0) 0.7 (0.2–3.3) 1.0 (0.2–5.1) 1
0.76 (0.44–1.31) 1 7.9 (2.3–27.7)∗∗∗ 6.2 (1.9–20.3)∗∗ 1 34.9 (4.5–268.4)∗∗∗ 11.7 (1.5–92.9)∗ 7.5 (0.99–57.1) 6.4 (0.8–50.8) 4.2 (0.5–38.0) 1
2006 1.12 (0.69–1.80) 1 7.3 (1.6–33.7)∗ 14.7 (3.6–60.9)∗∗∗ 1 66.5 (8.7–506.1)∗∗∗ 15.5 (2.0–120.9)∗∗ 8.4 (1.1–63.5)∗ 7.1 (0.9–55.4) 3.1 (0.3–30.2) 1
2008 0.91 (0.63–1.30) 1 6.8 (3.1–14.7)∗∗∗ 6.9 (3.4–14.0)∗∗∗ 1 45.9 (15.9–132.2)∗∗∗ 6.9 (2.5–19.0)∗∗∗ 4.9 (1.9–12.7)∗∗∗ 3.2 (1.2–8.8)∗ 2.1 (0.7–6.5) 1
Total1 1.41 (0.98–2.01) 1 1.2 (0.7–2.1) 1.6 (1.01–2.6)∗ 1 2.5 (1.2–5.0)∗ 2.0 (0.99–3.9) 0.8 (0.4–1.5) 0.9 (0.4–1.7) 1.1 (0.5–2.2) 1
2004 0.71 (0.50–1.00) 1 2.4 (1.4–4.1)∗∗ 2.0 (1.3–3.2)∗∗ 1 4.7 (2.3–9.5)∗∗∗ 1.6 (0.8–3.1) 1.4 (0.8–2.6) 1.1 (0.6–2.1) 0.5 (0.2–1.2) 1
2006 0.71 (0.48–1.04) 1 4.2 (2.1–8.6)∗∗∗ 4.3 (2.3–8.1)∗∗∗ 1 14.8 (6.1–35.6)∗∗∗ 6.5 (2.8–15.1)∗∗∗ 1.6 (0.7–3.5) 2.0 (0.9–4.7) 1.7 (0.7–4.4) 1
2008
Non-physical aggression
0.74 (0.54–1.01) 1 2.5 (1.5–4.0)∗∗∗ 2.9 (2.0–4.3)∗∗∗ 1 13.0 (5.3–31.9)∗∗∗ 4.0 (2.1–7.7)∗∗∗ 1.3 (0.8–2.1) 1.4 (0.8–2.4) 1.0 (0.5–1.9) 1
Total2
(∗ ) p
