Violence against health care workers is a major issue, with reported staff victimization ranging from 50% to 95% (Aljohani et al., 2021; Richardson et al., 2018). Not surprisingly, emergency department (ED) personnel represent more than 50% of reported incidents of assault within health care settings (Holland et al., 2021; Roppolo et al., 2020). All members across the ED workforce are subject to risk of violence, including nurses, physicians, secretarial, and ancillary staff (Carver & Beard, 2021; Copeland & Henry, 2017). Accordingly, best practices to prevent workplace violence include staff training and education, identification of potential events, de-escalation measures, appropriate multidisciplinary interventions, and coordination across an inclusive ED workforce (Carver & Beard, 2021; Mitra et al., 2018; Roppolo et al., 2020).
However, systematic reviews and meta-analyses found that more than 97% of research studies on ED workplace violence prevention in the past 11 years focused on physician and nurses (Aljohani et al., 2021; Chakraborty et al., 2022). Many of these studies used the Management of Aggression and Violence Attitude Scale (MAVAS) as an assessment of staff perceptions of the causes of violence and methods of violence management to determine Workplace Violence training needs and training effectiveness (Cheung et al., 2018; Soares & de Vargas, 2013). Although these two ED work groups lead and implement most workplace violence prevention efforts, all ED work groups interact with patients and families and can participate in prevention efforts and violence mitigation. In addition, if instruments, such as the MAVAS, are used for assessment in a new population, they need to be externally validated beyond nurses and physicians to inform sustainable ED workplace violence prevention programs (Messerly & Marceaux, 2020; Walters et al., 2016).
The framework Duxbury (2002) used to develop the MAVAS instrument incorporated explanations for the causes of patient aggression and violence in health care from different perspectives. These are the internal, external, and situational models. Each model addresses constructs of focus, including patient attributes, environmental factors, and staff-patient interaction, respectively (Duxbury, 2002,2003).
KEY POINTS
* All ED staff members are at highest risk for workplace violence.
* The MAVAS instrument is a reliable and valid measure of ED staff attitudes regarding workplace violence.
* The MAVIS instrument may help identify best practices to reduce workplace violence.
OBJECTIVE
The aims of this study were to (1) validate the MAVAS in an inclusive ED work group sample and (2) in the validated sample(s) to explore their reported perceptions of safety, approaches to manage violence and aggression, and experiences with violence in the ED to inform a multidisciplinary workplace violence prevention program.
METHODS
Study Design
This is an investigator-initiated single-site noninterventional retrospective cross-sectional survey design assessing the psychometric properties of the MAVAS within a convenience sample representing ED workforce groups. The Good Samaritan University Hospital Institutional Review Board (IRB) acknowledged the study as exempt status (IRB#: 2022.07.21.05). In addition, this study conforms with the Strengthening the Reporting of Observational Studies in Epidemiology guidelines (von Elm et al., 2007) and the Streiner and Kottner Recommendations for Reporting the Results of Studies of Instrument and Scale Development and Testing (Streiner & Kottner, 2014).
Population and Setting
The study was conducted within a Northeastern U.S. suburban, community, Level I adult trauma center with more than 120,000 annual admissions. All ED staff members employed by the institution during May 1, 2022, to June 30, 2022, were eligible to participate in the study. Our recruitment strategy included signs posted in private staff lounges and break rooms describing the study and a quick-response (QR) code and hyperlink address to access the confidential survey. A convenience sampling methodology was used. There was no financial incentive to participate in the survey and the average time to complete the survey instrument was 5 min.
Data Collection and Instruments
Data were collected via electronic responses on the SurveyMonkey Health Insurance Portability and Accountability Act (HIPAA) compliant software platform. Data were downloaded from the SurveyMonkey platform as a comma separated value (.csv) file at the end of the survey period. Full completion of the survey was voluntary. As such, some items may have been left blank if respondents were unsure of how they wanted to answer, and possibly through human error of missing an item. Partial completed data were included where appropriate in descriptive analyses for demographic items. Incomplete surveys (n = 16) were not included in the item analysis, reliability analysis, and exploratory factor analysis.
Demographic Variables
Self-reported demographic variables of gender, highest educational level completed, role/work group, years in current role, years employed in any ED setting, years in current ED setting were collected. These data were used to provide descriptives of the population sampled for psychometric validation of the MAVAS instrument. Demographic data were further used to stratify participant responses by ED work groups.
Management of Aggression and Violence Attitude Scale
The original 27-item MAVAS was based on three constructs of the causes of aggression and violence (n = 14 items) and perceptions of approaches to manage violence and aggression (n = 13 items). The three constructs of the causes of aggression are internal/biological (n = 4), external/environmental (n = 4), and interactional/situational (n = 5). The internal/biological construct consists of internal or biological items that can cause or influence aggressive and violent behavior. The external/environmental construct consists of items reflecting physical or social factors that may affect aggressive and violent behavioral tendencies. Finally, the interactional/situational construct focuses on interpersonal relationships or interactions that can increase or decrease aggressive and violent behaviors (Duxbury, 2003).
The MAVAS is scored on a 5-point Likert scale (i.e., 1 = "strongly agree" to 5 = "strongly disagree"). Higher scores reflect lower levels of the respondents' agreements with the items. The MAVAS score has been used to estimate staff attitudes regarding Duxbury's explanatory constructs of causes of aggression and violence, and to also compare those attitudes with perceptions of various approaches to manage patient violence and aggression (Pulsford et al., 2013). The MAVAS has demonstrated satisfactory psychometric properties of validity with factor analysis supporting a four-factor model (with each loading at .8 and above) and interitem and test-retest reliability estimates (r = .80, r = .97) across nursing and physician populations (Duxbury, 2002,2003; Pulsford et al., 2013; Wong & Chien, 2017). In addition, the MAVAS was found reliable (r = .87) in an adult inpatient mental health patient population (n = 20) during the initial piloting of the instrument (Duxbury, 2002).
Incidence of Assault/Violence
Although institutional policy includes reporting incidence of assault, staff incidence of assault or experienced violence in the workplace is underreported across patient care settings (McGuire et al., 2021, 2023; Mento et al., 2020). To record accurate data of staff incidence of assault, six investigator-developed items were included in the survey. The items were: (1) Have you been verbally assaulted at work within the past 12 months?; (2) If yes, did you file a report of the incident?; (3) Who did you file the report with?; (4) Have you been physically assaulted at work within the past 12 months?; (5) If yes, did you file a report of the incident?; (6) Who did you file the report with? The items measuring incidence of assault/violence comprise an index of self-reported occurrences and validity estimation include discriminant and convergent validity (Streiner, 2003).
Perception of Safety at Work
To decrease participant burden, perception of safety at work was estimated on a one-item Visual Analogue Scale of 0 = not safe at all and 100 = completely safe (Heller et al., 2016). The responses to perception of safety at work were then used to estimate convergent and discriminant validity with responses of reported assault within the past 12 months.
Open-Ended Item
As the issue of workplace violence in the ED is important to all to maintain a safe and healthy work environment, one open-ended item, "Please provide any further comments here" was the last item in the survey. The item was to allow the participants to provide any additional feedback regarding workplace violence, safety in the ED, or the study.
Statistical Analysis and Power Calculation
This was a noninterventional quantitative study to assess the psychometric properties of the MAVAS within an inclusive ED staff sample. The analysis was completed using IBM Statistical Package for the Social Sciences (SPSS) Statistics for Windows, Version 28 (IBM Corp; Armonk, NY) and Stata Statistical Software: Release 17 (StataCorp LLC; College Station, TX).
The suggested minimal sample size for adequate power with an exploratory factor analysis is a minimum of five participants per item (Kyriazos, 2018). The MAVAS is a 27-item instrument; therefore, 134 participants would provide adequate power to answer the primary aim of this study. For inferential analyses a two-tailed testing approach with a chosen significance level of p < .05 was used. Independent sample T tests were used to compare groups with continuous outcomes. The chi-square test for Independence and Fisher's exact test were conducted to analyze categorical data if more than 20% of cells had expected counts less than five. In addition, adjusted residuals (ARs) were estimated to determine categories with significant differences between observed and expected counts. Values less than -2 or more than +2 were the cutoffs for determining significant associations for individual cells (Sharpe, 2015).
Psychometric Testing
Validity
Constructs are unobservable, or latent, phenomena that one cannot measure directly, such as attitudes. In measurement research, scales that are composed of statement or question items, provide an indirect approach to estimate properties of an unobservable construct through a series of accuracy (validity) and repeatability (reliability) testing within a given population (Cotter et al., 2018; McElligott et al., 2018). Therefore, to establish the construct validity of the MAVAS within an inclusive ED workforce exploratory factor analysis was employed before further analysis of reliability or patterns of responses. First, the factorability of the MAVAS data within this sample was examined using the Bartlett's test of sphericity (acceptable p < .05) and Kaiser-Meyer-Olkin (KMO) test (index value of KMO > 0.6). The eigenvalues of the factors in the MAVAS were examined within the scree plot (in SPSS), in which the shape of the curve was expected to change direction and become horizontal after four observations above one. All factors above the elbow, or the break in the plot, would contribute to the largest proportion of the total variance of the construct of attitude toward violence and management of aggression in the sample. Varimax rotation was chosen to extract the significant domains of the MAVAS due to the assumption of the noncorrelated responses across factors.
Additionality to estimate the validity of the Visual Analogue Scale measure of perception of safety at work, correlations, and independent sample t tests of participants' reported physical assault within the past 12 months and Visual Analogue Scale responses were conducted. As reported physical assault was measured as a dichotomous variable and the Visual Analogue Scale perception of safety as a continuous variable, the point biserial correlation coefficient was used for this correlational estimate.
Reliability
Reliability of the MAVAS was established with Cronbach's [alpha] internal consistency estimates. In addition, reliability of the MAVAS four subscales was estimated.
Item Analysis
Estimates of central tendency and dispersion were conducted for the 27 items in the MAVAS. Comparison of mean and standard deviation estimates were made for the total sample of the inclusive ED workforce across items and subscales. Furthermore, as the MAVAS has not yet been evaluated in an ED sample beyond physicians and nurses, two subsamples of all ED work groups were created and responses were compared across items and subscale responses. These subsamples were (1) physician and nurses in which the MAVAS has been previously psychometrically validated and (2) the remainder of the participant in other ED work groups, excluding physician and nurses.
RESULTS
Response Rate and Missing Data
Of the total number of ED employees who met inclusion criteria, 134 of 338 participated in the study resulting with a response rate of 40%. Within the "physician and nurse" work group, there was no missing data for the demographic responses for "ED position." Within demographic items, there were less than one cases of missing data, ranging from (n = 8, 6%) of data missing for educational level to (n = 3, 2%) for gender. For testing reliability and item analysis, cases were excluded listwise deletion based on all variables in the procedure (n = 6, 5%) (Table 1).
Demographics
Of the participants, the majority were bachelor's prepared (n = 61, 48.4%), women (n = 93, 71.0%), and were nurses (n = 55, 41.0%). Study participants reported an average of (M = 8.2, SD = 8.4) years worked in their current profession, (M = 6.1, SD = 6.5) years worked in any ED, and (M = 5.4, SD = 8.5) years worked in the current facility's ED (Table 1). Participants were further categorized into following two work groups: "physician and registered nurses" (n = 77, 57%) and "other" (n = 57, 43%). The "physician and registered nurses" group included physicians, registered nurses (RNs), nurse practitioners, and physician assistants. The "other" work group included technician assistants, nurse assistants, registrars, unit clerks, concierge, phlebotomists, security, and social workers.
Item Analysis
MAVAS Items
The full sample indicated similar means for each subscale, with differences in subscale means the highest between the two work groups for "external environment." The highest average item indicating strong disagreement with the item statement on the MAVAS instrument was "Views - Approach 3" (M = 3.81, SD = 1.3) (Table 2). The lowest averaged items on the MAVAS instrument were "Internal - Bio 3" (M = 1.53, SD = 0.8) and "Views - Approach 13" (M = 1.53, SD = 0.8) (Table 2).
There were similarities between the two work groups, with few notable differences (Table 2). Among both the "physicians and registered nurses" work groups, and the "other" work group, "Internal - Bio 4" (M = 3.8, SD = 1.2) was the highest average item. Furthermore, "Internal - Bio 3" (M = 1.6, SD = 0.7) was the lowest average for both work groups. Similar to the full sample and "physician and registered nurses" work groups, the work group "other" averaged "Views - Approach 3" (M = 3.8, SD = 1.4) highest. Item "Views - Approach 13" (M =1.5, SD = 0.8) had the lowest average among the "other" work group.
For the majority of responses to items on the MAVAS, RNs' averages were above the sample mean (n = 25, 93%). Of the 27 physicians, (n = 24, 88.9%) found average responses below the sample's mean for the MAVAS items. Items that physicians' averaged higher than the sample mean included "External - Environmental 2," where physicians' averaged (M = 2.3, SD = 1.0) over the sample (M = 2.1, SD = 1.0) physicians had the lowest mean (M = 2.1, SD = 1.1) for item "Views - Approach 11" compared with the highest mean for NAs (M = 3.1, SD = 1.3).
For item "Views - Approach 4," physicians had the lowest mean (M = 2.7, SD = 1.2) versus security with the highest mean (M = 3.8, SD = 1.1). Similarly, on "Views - Approach 5," physicians once again had the lowest mean among the groups (M = 2.8, SD = 1.3), with security having the highest mean (M = 4.0, SD = 1.3). Nurses were found to have the highest mean (M = 2.5, SD = 1.1) for item "Views - Approach 7" compared with registrars who were found to have the lowest mean (M = 1.3, SD = 0.5).
"Interactional - Situational 3" (M = 2.22, SD = 1.2) indicated similarity among the diverse ED positions. Registrars (M = 2.17, SD = 1.6), NAs (M = 1.95, SD = 1.1), technical assistants (M = 1.82, SD = 1.3), and security (M = 1.73, SD = 1.2) were below the sample average. Overall, registered nurses (M = 2.49, SD = 1.1) and physicians (M = 2.44, SD = 1.3) were above the sample mean.
Aggression and Violence Items
There were no significant differences in incidence of verbal or physical assault, or perception of safety by gender ([chi]2[130] = 0.08, p = .772; [chi]2[129] = 0.01, p = .971; t[126] = 1.2, p = .263). In addition, years of experience was not correlated with incidence of verbal assault, physical assault, or perception of safety rb(130) = .002, p = .999, rb(129) = -.032, p = .715, and r(128) = .122, p = .171, respectively. However, significant variation in incidence of physical assault, [chi]2(130) = 34.81, p = .001 was found between ED workforce groups. Security guards reported significantly more incidence of physical assault than would be expected (AR =+2.8).
Reliability
The MAVAS instrument was found to be reliable (Table 3) with this study's sample (n = 27; [alpha]= .872). The subscale with the lowest Cronbach's [alpha] was "Internal - Biological" (n = 4; [alpha]= .521). Discriminant and convergent validity of perception of safety Visual Analogue Scale was established with rb(130) = .475, p < .001 and t(130) = 6.1, p < .001, respectively.
Validity
The MAVAS instrument was validated with this study sample (Table 4). The KMO value was 0.803, and Bartlett's test of sphericity was significant (p < .001), suggesting that the correlation matrix was appropriate for exploratory factor analysis. The exploratory factor analysis of the MAVAS yielded a seven-factor model, explaining 63.03% of the total variance. The factors were interpreted following a varimax rotation. Factors were suppressed under 0.4 to minimize cross-loadings and increase the likelihood that the factors represent meaningful and distinct underlying constructs. Factor 1 identified eight items with loadings ranging from 0.422 to 0.777. Factors 2, 3, and 5 loaded four items with loadings ranging from 0.530 to 0.827, 0.543 to 0.788, and 0.508 to 0.728, respectively. Factors 4 and 7 had two items with loadings ranging from 0.530 to 0.827 and 0.416 to 0.752, respectively. Factor 6 loaded with three items ranging from 0.543 to 0.788. There were no negative loadings. Two items, Views - Approach 5 "Physical restraint is sometimes used more than necessary" and Internal-Biological 1 "It is difficult to prevent patients from becoming violent or aggressive." cross-loaded on Factors 2 and 3 and Factors 5 and 7, respectively (Figure 1).
Although different than the four constructs validated in previous studies with nurse and physician samples, the seven-factor structure found with this sample's exploratory factor analysis supported the original conceptual framework used by Duxbury for the 27-item scale construction. The underlying latent constructs identified with this sample were (1) environmental and interactional causes of aggression and violence; (2) communication and progressive approaches of management; (3) reactional approaches of management; (4) consistency or effectiveness of management approaches; (5) internal psychological locus of causes of aggression and violence; (6) pharmacological approaches of management; and (7) internal physiological locus of causes of aggression and violence.
DISCUSSION
The response rate was sufficient, and all ED work groups were represented supporting the internal validity of the findings. In addition, there were fewer missing data than anticipated. The lack of significant variation in item analyses between the physicians/nurses and other work groups supports the conclusion that the MAVAS is an appropriate tool to measure perceptions of causes of violence and attitudes toward managing violence of an inclusive work group.
The MAVAS' reliability findings indicate that the MAVAS tool is consistent in measuring the proposed constructs of perceptions of aggression and violence. However, while overall the sample was consistent in their responses to the MAVAS instrument, opportunities for addressing learning needs were identified in the items from the subscale of "Internal - Biological," which include items such as "Aggressive patients will calm down automatically if left alone." This was the subscale with the lowest internal consistency, suggesting that the four Internal - Biological items originally established in the development of the MAVAS instrument were capturing more than one factor. The items were related to behavioral components that many nonbehavioral ED staff members may not have training or education in. The results indicate there can be tailored educations to meet the training needs for the diverse work groups that staff the ED.
In addition, the original subscale "Views - Approach" indicated there were items high variability in responses. Particularly, Items 2 (M = 3.84, SD = 1.33), 3 (M = 3.32, SD = 1.40), and 5 (M = 3.3, SD = 1.34) which were statements about the appropriateness and frequency of using restraints. Item 2 supports the attitude that restraints were for patients' own safety, whereas Item 5 supports the attitude that restraints were used too often. These results align with "Views - Approach" being the subscale with the second lowest estimate of internal consistency for this sample when using the original MAVAS' subscale structure for estimation. These results support the notion that restraint use is a highly complex issue with serious implication on patient safety (both protective and potential harm), liability, medicolegal ramifications, as well as a perceived alignment with law enforcement that differentially affects different job classes. This suggests an opportunity for further exploration of training and education needs in workplace violence for different members of the ED workforce.
Similarities found in subscales "Interactional - Situational." There may be some perception that there is a deference to physicians on topics such as pathophysiology of agitation and violence. It is possible, too, that the ED has culture in place that facilitates similar perceptions of how aggression and violence were handled for items related to "Interactional - Situational." Certain topics may discourage disagreement with management if someone perceives it is out of their level of training.
Interestingly, the validity of the MAVAS within an inclusive ED workforce sample was different than the four constructs validated in previous studies with nurse and physician samples. The seven-factor structure found with this sample's exploratory factor analysis supported the original conceptual framework used by Duxbury for the 27-item scale construction. With these new factors identified, the MAVAS can be used for assessment and training within an inclusive ED workforce.
LIMITATIONS
Although the sample was representative of the population of interest, this was a single-site design limiting generalizability of findings. Potential selection bias is also a potential limitation as convenience sampling was used and self-selection into the study may impact internal validity. In addition, though the sample size was sufficient for the proposed exploratory analysis, a larger sample size will provide the opportunity to perform an exploratory and confirmatory analysis within this population. As such, our planned next steps include validating the MAVAS across our system EDs using an inclusive ED workforce sample.
CONCLUSIONS
The validity of the MAVAS was established for an inclusive ED workforce sample. However, further testing with a larger inclusive ED workforce will allow for confirming the hypothesized underlying structure with an exploratory factor analysis and confirmatory factor analysis. As identified in the results, some work groups require specific education on topics that may not have been originally intrinsically linked to their roles. Multiple studies have demonstrated the effectiveness of universal training of ED staff in workplace violence prevention and management. The evidence from this study provides a framework for using the MAVAS instrument as a measure of all ED work groups' perceptions of violence and aggression in the ED, and identify learner needs in prevention and management of aggression and violence.
Acknowledgment
The authors thank Theresa Tomkin, MS-APRN, ANP-C, the Nursing Director of the Emergency Department, Good Samaritan University Hospital. She mentored the nurse ED residents through the Quality Improvement project that resulted in an interdisciplinary research study that had her full support. In addition, the authors express their gratitude to Sarah A. Eckardt, MS, BA, data scientist and biostatistician, for her expertise and patience with reviewing the manuscript and provided thoughtful direction.
REFERENCES