Violence is a major public health problem,1 particularly important in the work environment. Homicide is the third-leading cause of occupational fatality and the second-leading cause of occupational fatality for women.2 Although much is known about work-related homicides, research on nonfatal violence and relevant risk factors is limited. Nearly 2 million acts of nonfatal work-related violence occur annually in the United States alone.3
Hospital and health care workers are at high risk for violence, particularly nonfatal violence.4,5 Violence against nurses specifically is a major occupational health problem.6–9 On the basis of the 1992 Minnesota Workers’ Compensation files, nurses accounted for more than 7% of the total work-related assault cases leading to more than 3 days of lost time; women's assault rate was twice that of men.10 Biologic,11 psychologic,12 and sociocultural13 theories have been generated to explain causes of violence. Understanding the factors that place persons at risk for violence is critical to development of effective interventions. The current study, following up on a smaller case–control study,14 was designed to identify environmental and other exposures associated with the risk of work-related violence. Such factors may provide a basis for interventions to reduce the risk of work-related violence.
Licensing is required for both registered nurses (RNs) and licensed practical nurses (LPNs) who practice in Minnesota. The target population was defined as licensed RNs and LPNs who had worked in Minnesota during the 12 months before the date they completed the survey. With approval by the University of Minnesota Institutional Review Board, we randomly sampled 6,300 nurses from the population (n = 79,128) of currently active RNs (n = 57,388) and LPNs (n = 21,740) who were licensed in the state of Minnesota as of 1 October 1998. Besides name, license type, and address, the state database included birth date, sex, and year of first licensure.
Selection of Cases and Control Subjects
We initially mailed a questionnaire to the entire sample of 6300 nurses to determine employment status and the incidence and consequences of work-related violence.15 On the basis of the responses, we identified 475 cases (those who reported at least 1 event of physical violence during the previous 12 months) and 1425 control subjects. Control subjects were selected randomly from all months during the study period in which the nurses indicated having worked but before any reported physical assaults to those nurses. This sampling method ensured that the distribution of sampled calendar months represented the distribution of months worked.
Physical assault was defined as being hit, slapped, kicked, pushed, choked, grabbed, sexually assaulted, or otherwise subjected to physical contact intended to injure or harm. Violence was work-related if it occurred in the work environment or during any activities associated with the job (including travel). This is consistent with the definition used by the U.S. National Institute for Occupational Safety and Health (NIOSH).16
We assessed exposures based on previous research on violence and evidence from other areas of the injury epidemiology literature.17 General exposures included work experience (years worked as a licensed nurse; years worked in department), average patient contact hours per shift, average number of nurses and number of overall staff located in the immediate work environment on the shift worked most often, primary facility and department/unit/area worked, the main patient population, and primary professional activity. Factors pertinent to environmental design included accessibility of exits and physical barriers preventing view of others in the work environment and level of lighting. Environmental protection factors (assault deterrents in the immediate work environment) included video monitor, metal detector, security alarm/panic button, controlled access, security personnel, or escort/body guard. Personal protection factors included cellular telephone and personal alarm.
For both the initial survey and the nested case–control study, we sent up to 4 follow-up mailings. These mailings included a cover letter providing information for participant consent, together with the pertinent survey, and a postage-paid return envelope.
The initial survey collected the following data: (1) months in which the nurses worked in a nursing position in the previous 12 months; (2) demographic information; and (3) information on physical and nonphysical work-related violence events during the study period. Overall, 79% responded (an estimated 78%, adjusting for the estimated eligible fraction among nonrespondents for age, gender, license-type, and location).18,19 The response rates for RNs and LPNs, respectively, were 81% (79%, adjusted) and 75% (73%, adjusted).15
A conceptual model based on a priori hypotheses served as the foundation for a causal model20 that in turn guided survey design and analysis.21,22 The survey questionnaire ascertained exposures for the month before and during the incident itself for cases; if multiple events were reported, cases were surveyed about the month before the earliest event. For controls the questionnaire ascertained exposures for random months, selected as described above. Questionnaires specific to the respective month were sent to all participants.
We obtained responses to the full case-control questionnaire from 324 cases (68%) and 946 control subjects (66%). However, we focused primarily on patient- or client-initiated work-related assaults (310 cases, 96% of all physical assaults). For each exposure of interest, we selected confounders for multiple logistic regression using the principles in Maldonado and Greenland,23 and based on directed acyclic graphs.20–22 These methods identify parsimonious models and exclude covariates that should not be entered into the regression because they could introduce bias.
To account for variability from sampling and also from uncertainty about adjustment weights and eligibility fractions, we calculated bootstrap confidence intervals (CIs)24 for all odds ratios. Potential response bias was controlled by inversely weighting observed responses by probabilities of response,25 estimated as a function of the following characteristics available from the licensing database: age; sex; license type; and type of home address (metropolitan versus nonmetropolitan). To adjust the weighting for unknown eligibility among nonrespondents, we estimated the probability of eligibility from these same factors.18 The entire weighting procedure was recalculated on each bootstrap iteration. Validation procedures, reported elsewhere, were conducted for self-reported physical assault injury occurrences and various workplace exposures.20 We conducted sensitivity analyses to determine the potential effect of an unmeasured confounder on the multivariate odds ratios.20,26
On the basis of the initial survey, 96% of nurses were women. On average, participants were 46 years of age (±SD, 10.1); 75% were RNs, and the rest were LPNs. The estimated physical assault rate was 13.2 per 100 persons per year (95% CI = 12.2–14.3). The assault rate was lower for RNs (12.0; 10.9–13.3) than for LPNs (16.4; 14.2–18.7).15
Characteristics of cases and controls are shown in Table 1. Cases and control subjects were similar by sex and age. Cases were less likely to have bachelor's degrees or higher and more likely to be working primarily in nursing homes or long-term care facilities and with geriatric patients.
Table 2 provides risk estimates at 3 levels of analyses: unadjusted; partially adjusted for a minimal set of confounders20,21; and the corresponding fully adjusted multivariate analysis, weighted for nonresponse and unknown eligibility.
Nurses at greatest risk of assault were those working in nursing homes or long-term care facilities (2.6; 1.9–3.6) and emergency (4.2; 1.3–12.8) and psychiatric (2.0; 1.1–3.7) departments. Risk increased for each additional hour of shift duration (1.05; 0.99–1.11).
Of all the environmental factors, the amount of lighting was most strongly associated with risk. The odds of assault were doubled when lighting was less bright than daylight (2.2; 1.6–2.8). Other elements of environmental protection (such as video monitors and security personnel) had little apparent effect. Risk was substantially reduced among nurses who provided their own cellular telephones or portable alarms (0.30; 0.15–0.71). However, cellular telephones provided by the employer provided no apparent protection (1.0; 0.70–1.5).
We found increased risks of work-related physical assault among nurses who worked in nursing home or long-term care facilities and also among those working in psychiatric and emergency departments. Other studies27,28 have identified similar risks using designs different from the present study. We also found increased risk of assault in environments that were not fully illuminated. A previous case–control study of occupational homicide has identified reduced risks with bright exterior lighting;29 however, the importance of interior lighting had apparently not been considered. Although every hour of patient contact increased risk at least 5%, both nursing and total staffing might moderate this risk. Further research may confirm this finding. The lower risk among nurses carrying their own cellular telephone or personal portable alarm is apparently not due tothe availability of the telephone itself, since those provided by employers conferred no protection.
Our information on both the exposures and the outcome was based on self-report, which is a potential weakness. We attempted to minimize this bias by limiting the recall of violent events to the previous 12 months30 and the recall of exposures to a 1-month period within the preceding year,14 as has been done in previous studies. To further minimize information bias, nurses were contacted again by mail to clarify ambiguous or missing information.20 We also conducted validation substudies of environmental exposures and health care treatment.20 Potential response bias was controlled for by Horvitz and Thompson reweighting25 using weights adjusted for the probability of being eligible among nonrespondents.18 Sensitivity analyses conducted on key exposures of interest26 suggest that the results are not due to unmeasured confounding.20
In summary, we estimated the incidence of violence in licensed nursing professionals, a large occupational population, and identified relevant risk and protective factors. These results can guide further investigation of relevant factors, and perhaps lead to effective methods for reducing the substantial risk of physical assault in health care settings.
1.Rosenberg M, Fenley M. Violence in America: A Public Health Approach
, New York: Oxford University Press; 1991;14–50.
2.U.S. Department of Labor. Bureau of Labor Statistics. National Census of Fatal Occupational Injuries in 2001, USDL 02-541, 2002. Available at: http://stats.bls.gov/iif/oshcfoi1.htm
. Accessed April 5, 2005.
3.Duhart DT. Violence in the workplace, 1993–1999, National Crime Victimization Survey, Bureau of Justice Statistics Special Report, U.S. Department of Justice, Office of Justice Programs; 2001.
4.Peek-Asa C, Howard J, Vargas L, et al. Incidence of nonfatal workplace assault injuries determined from employer's reports in California. J Occup Environ Med
5.National Institute for Occupational Safety and Health (NIOSH). Violence: Occupational Hazards in Hospitals.
Washington, DC: U.S. Department of Health and Human Services, Public Health Service, Centers for Disease Control and Prevention, DHHS (NIOSH); 2002; Publication No. 2002-101.
6.Carter R. High risk of violence against nurses. Nurs Manag (Harrow)
7.Arnetz JE, Arnetz BB, Soderman E. Violence toward health care workers: prevalence and incidence at a large regional hospital in Sweden. AAOHN J
8.Baxter E, Hafner RJ, Holme G. Assaults by patients: the experience and attitudes of psychiatric hospital nurses. Aust NZ J Psych
9.Mahoney BS. The extent, nature, and response to victimization of emergency nurses in Pennsylvania. J Emerg Nurs
10.LaMar W, Gerberich SG, Lohman W, et al. Work-related physical assault. J Occup Environ Med
11.Niehoff D. The Biology of Violence: How Understanding the Brain, Behavior, and Environment Can Break the Vicious Circle of Aggression
. New York: Free Press; 1999.
12.Dollard J, Doob L, Mowrer O, et al. Frustration and Aggression
, New Haven, Connecticut: Yale University Press; 1939, in Blue HC, Griffith EE. Sociocultural and therapeutic perspectives on violence. Psych Clin N Am
13.Bandura A. Aggression: Social Learning Analysis
. Englewood Cliffs, NJ: Prentice-Hall; 1973.
14.Lee SS, Gerberich SG, Waller LA, et al. A case-control study of work-related assault injuries among nurses. Epidemiology
15.Gerberich SG, Church TR, McGovern PM, et al. An epidemiological study of the magnitude and consequences of work-related violence: The Minnesota nurses’ study. Occup Environ Med
16.National Institute for Occupational Safety and Health (NIOSH). Current Intelligence Bulletin 57, Violence in the Workplace: Risk Factors and Prevention Strategies
, Washington, DC: U.S. Department of Health and Human Services, Public Health Service, Centers for Disease Control and Prevention, DHHS (NIOSH) Publication; 1996. No. 96-100.
17.Robertson LS. Injury Epidemiology: Research and Control Strategies
, 2nd ed. New York: Oxford University Press; 1998:167–194.
18.Mongin SJ. Adjustment for nonresponse in the Minnesota Nurses Study. Division of Environmental and Occupational Health, University of Minnesota, 2001. Available at: http://enhs.umn.edu/NewFiles/resreports.html
. Accessed April 5, 2005.
19.Church TR, Yeazel MW, Jones RM, et al. A randomized trial of direct mailing of fecal occult blood tests to increase colorectal cancer screening. J Natl Cancer Inst
20.Minnesota Nurses’ Study, Center for Violence Prevention and Control, Division of Environmental Health Sciences, School of Public Health, University of Minnesota, Minneapolis, Minnesota Web site. Available at: http://www1.umn.edu/cvpc/research.html
. Accessed April 5, 2005.
21.Greenland S, Pearl J, Robins J. Causal diagrams for epidemiologic research. Epidemiology
22.Hernan MA, Hernandez-Diaz S, Werler MM, et al. Causal knowledge as a prerequisite for confounding evaluation: An application to birth defects epidemiology, Am J Epidemiol
23.Maldonado G, Greenland S. Estimating causal effects, Int J Epidemiol
24.Efron B, Tibshirani J. An Introduction to the bootstrap. In: Cox DR, Hinckley DV, Reid N, et al., eds. Monographs on Statistics and Applied Probability
. New York: Chapman and Hall, 1993.
25.Horvitz DG, Thompson DJ. A generalization of sampling without replacement from a finite universe. Am Stat Assoc J
26.Rothman KJ, Greenland S. Modern Epidemiology
Philadelphia, PA: Lippincott-Raven Publishers; 1998.
27.Yassi A, Tate R, Cooper J, et al. Causes of staff abuse in health care facilities. AAOHN J
28.Barlow CB, Rizzo AG. Violence against surgical residents. West J Med
29.Loomis D, Marshal SW, Wolf SH, et al. Effectiveness of safety measures recommended for prevention of workplace homicide. JAMA
Copyright © 2005 Wolters Kluwer Health, Inc. All rights reserved.
30.Gabel CL, Gerberich SG. Risk factors for injury among veterinarians. Epidemiology