The impact of heavy nurse workload and patient/family complaints on workplace violence: An application of human factors framework

Abstract Aim To examine the relationships between workload factors at different systems levels (unit level, job level and task level), patients/family complaints and nurse reports of patient violence towards them using a human factors framework. Design This is a secondary analysis of cross‐sectional data. Methods Data from 528 nurses working in medical–surgical settings in British Columbia, Canada, were analysed. At the unit‐level workload factors included patient‐RN ratios, patient acuity and dependency; at the job‐level perceptions of heavy workload, undone nursing tasks and compromised professional standards due to workload; and at the task‐level interruptions to workflow. Results Workload factors at multiple levels were directly related to workplace violence. Nurses' increased reports of compromised standards (job level) and interruptions (task level) were related to increased reports of physical and emotional violence, and higher patient acuity (unit level) was related to increased reports of emotional violence. Patient/family complaints mediated the relationship between almost all the workload factors and workplace violence.

their capacity to deliver care effectively, resulting in anxious and frustrated patients who become violent (Pich, Kable, & Hazelton, 2017;Shields & Wilkins, 2009). Because patients/families are the main perpetrators of violence towards nurses, this study investigated whether patient/family complaints about nursing care delivery can serve as an indicator of patient violence towards nurses. In particular, the purpose of this study was to examine the relationships between workload factors at different systems levels, patients/family complaints and nurse reports of patient violence towards them.

| BACKG ROU N D
In one national Canadian study, when nurses reported inadequate staffing and resources in their work environments, they also reported more frequent exposure to physical and emotional workplace violence perpetuated by patients, even after controlling for nurse and job characteristics (Shields & Wilkins, 2009). A longitudinal cohort study of more than 34,000 nurses across eight European countries found heavy workload, operationalized as time pressure, was associated with a higher likelihood of violence from patients and harassment by other organizational employees (Camerino, Estryn-Behar, Conway, van Der, & Hasselhorn, 2008). A cross-sectional survey study of about 2,500 Australian nurses examined the effect of several workload factors on their exposure to workplace violence-operationalized as physical violence, emotional violence and threat of violence perpetuated by patients, families/visitors and organizational employees (Roche, Diers, Duffield, & Catling-Paull, 2010). Unanticipated changes in patient acuity were associated with a higher likelihood of threat of violence; higher RN staffing levels were associated with a lower likelihood of threat of violence and physical violence; and more undone nursing tasks were associated with physical, emotional and threat of violence (Roche et al., 2010). Although this study focused on various sources of workplace violence, patients by far were the most common perpetuators of all three types of workplace violence. More recently, Pich et al. (2017) found that nurses who had experienced verbal and physical workplace violence from patients in the last 6 months were two times more likely to have experienced heavy workload and time management issues. In this study, nurses reported workload, time management issues and inadequate staffing as the most common causes of workplace violence perpetuated by patients. Similarly, a descriptive study of 174 Jordanian emergency nurses reported workload was the most common cause of workplace violence, two times more than caring for patients with dementia or Alzheimer's disease (Darawad, Al-Hussami, Saleh, Mustafa, & Odeh, 2015).
The research evidence suggests an association between nurses' heavy workloads and negative patient outcomes, including patient complaints and dissatisfaction. For example, some studies showed that heavy nurses' workloads, operationalized as nurse staffing levels and higher number of undone nursing tasks, were associated with more frequent patient/family complaints (Aiken et al., 2010;Thomas-Hawkins, Flynn, & Clarke, 2008). Others found a relationship between heavy nurse workload and higher patient dissatisfaction (Kutney-Lee et al., 2009). Recently, the Canadian Centre for Occupational Health and Safety (2019) identified clients' complaints of unfair treatment as an important warning sign of workplace violence. To our knowledge, no previous research has examined the association between patient/ family complaints and workplace violence. In addition, the small body of research that examined associations between nurses' workload, patient complaints and violence towards nurses focused predominantly on unit-level workload factors.

| Theoretical framework
Workload is a complex construct requiring considerations of multiple factors at multiple levels (Carayon & Gurses, 2008;Holden et al., 2011). Despite this complexity, a majority of workplace violence research has examined workload factors at one systems level (e.g. unit level) without using a systematic theoretical framework. To address this limitation, the conceptualization of workload is informed by a human factors framework in this study (Holden et al., 2011;MacPhee, Dahinten, & Havaei, 2017). According to this framework, an interaction between demands and resources produces workload at multiple levels including unit level, job level and task level. Holden et al. (2011) conceptualized unit-level workload as staffing considerations; job-level workload as the general and specific demands of the job including the amount of work and the level of concentration required to complete it in a given day; and task-level workload as the demands (e.g. need for multitasking) and resources (e.g. technology) for a specific nursing task such as medication administration. Holden et al. (2011) validated their framework among nurses. Heavy workload at the three levels was associated with higher job dissatisfaction and burnout, but only job-level and task-level workload factors were associated with more frequent medication errors (Holden et al., 2011). Overall, this research showed that nurse and patient outcomes may vary, depending on the level of workload (i.e. unit, job, task). MacPhee et al. (2017) used the Holden et al. human factors framework to conceptualize workload and examine its impact on patient and nurse outcomes. In this study, unit-level workload factors included nurse reported patient-RN ratios, patient acuity and patient dependency; job-level workload factors included nurses' perceptions of heavy workload, nursing tasks left undone and compromised professionals standards due to workload; and task-level workload factors included the frequency of interruptions to workflow. The authors found heavy workload at all three levels was associated with adverse patient outcomes, such as patient falls and urinary tract infections, and negative nurse outcomes, such as emotional exhaustion (MacPhee et al., 2017).

| Research question
In this study, the MacPhee et al. (2017)

| Design
This study was a secondary analysis of cross-sectional survey data from 528 nurses working in medical and/or surgical settings in British Columbia (BC), Canada.

| Sample
In the larger study, a proportionate stratified random sample of registered nurses (RNs) and licensed practical nurses (LPNs), based on geographic regions (health authority) and employment status (fulltime, part-time and causal), was invited to participate in the study in 2015 (blinded). A total of 1,876 acute care nurses in direct care, leadership and educator roles participated. For this secondary analysis, the inclusion criteria consisted of direct care nurses from medical, surgical or medical-surgical areas, resulting in a sample of 528 nurses.

| Data collection
In the larger study, data were collected using an electronic survey platform. The study survey consisted of a series of researched de-

Workload factors
Unit-level RN Staffing Levels. A ratio of patients per RN was obtained based on two questions that asked nurses to identify the number of patients and the number of direct care RNs on their unit over the last shift. These questions were previously validated in the international RN4CAST research Sermeus et al., 2011).
Unit-level patient acuity and dependency. Two items based on the American Association of Critical Care Nurses' Synergy Model were used to assess patient acuity and dependency (Curley, 2007). While patient acuity refers to the instability, complexity and unpredictability of patients' condition, patient dependency describes patient's ability to perform their own activities of daily living. These definitions were provided on the survey and nurses were asked to rate the average acuity (1 = not at all acute, 4 = very acute) and dependency (very independent, 4 = very dependent) of their patients during the last month. For this analysis, acuity and dependency scores were dichotomized into 0 = not at all or somewhat acute, 1 = moderately or very acute; and 0 = very or somewhat independent, 1 = somewhat or very dependent. These questions were previously validated by MacPhee et al. (2017).
Job-level heavy workload. Three workload items based on the Canadian National Survey on the Work and Health of Nurses were used to assess the general amount of work nurses have to complete (Statistics Canada, 2006). The items asked about the frequency by which nurses arrived early or stayed late, worked through breaks to complete work and felt they had to complete "too much work" Nurses were asked to select all the activities that were necessary but left undone during their most recent shift due to lack of time. Cronbach's alpha was 0.81 suggesting good internal consistency.

Patient/family complaints
A single validated question, based on RN4CAST research, was used to identify the frequency by which nurses received complaints from their patients and their families over the last year . Response options ranged from never (0) to every day (6).

Demographics
A series of questions were used to assess respondents' age, gender, professional designation (RN vs. LPN), highest nursing education (diploma vs. degree), years of nursing experience and employment status (full-time vs. part-time or casual).

Workplace violence
Emotional and physical abuse. Two workplace violence items were adapted from the 2005 National Survey of the Work and Health of Nurses (Statistics Canada, 2006) and the study by Hesketh and colleagues (Hesketh et al., 2003). The questions asked participants to identify the frequency by which they had experienced emotional and physical abuse from patients and/or families in their primary workplace over the past year (0 = never, 6 = everyday). Hesketh et al. (2003) established the convergent validity of the items by linking more frequent workplace violence exposure to higher job dissatisfaction of Canadian nurses.

| Analysis
Key methods of data analysis were descriptive statistics, bivariate correlations and hierarchical multiple regression using the Statistical Package for Social Sciences for Windows 25.0 (SPSS Inc.). In particular, the first and second research questions were examined using hierarchical multiple regression; nurse characteristics were entered into the first regression model followed by unit-level, joblevel and task-level workload factors in the second, third and fourth models, respectively, followed by patient/family complaints in the fifth model. To save power, only variables that showed significant bivariate correlations with one or both of the outcome variables were included in the regression model. To further save power, demographic variables that were not related to workplace violence in the regression model were dropped from the model. The third research question was examined using Baron and Kenny's recommendations (1986) and Preacher and Leonardelli's Sobel test (2010).
According to Baron and Kenny, mediation is dependent on three to four conditions: (a) a significant beta coefficient when the independent variable is regressed on the outcome variable; (b) a significant beta coefficient when the independent variable is regressed on the mediator; (c) a significant beta coefficient of the mediator regressed on the outcome variable controlling for the independent variable; and (d) non-significant beta coefficient of independent variable regressed on outcome variable controlling for the mediator. While all four conditions are required to establish full mediation, partial mediation requires only the first three conditions.

| Ethics
Participants were informed that survey completion and submission would indicate consent to be included in the study. Ethics approval was obtained from the university behavioural ethics review board (Approval number: H14-00789). RNs. More than half of the sample had a nursing degree and a fulltime position. professional standards a few times a month. At the task level, on average, nurses were interrupted a few times a week. Nurses were exposed to physical and emotional violence an average of once a month.     The positive beta associated with job-level compromised standards and task-level interruptions suggests that when nurses compromised their professional standards due to workload and were interrupted during their work more frequently, they reported more frequent exposure to physical and emotional violence. Similarly, the positive beta associated with patient/family complaints suggests that more frequent complaints from patients and/or their families were associated with more frequent exposure to physical and emotional violence. The negative betas associated with professional designation and employment status suggests that LPNs and nurses in a full-time position were at a higher risk of physical and emotional violence compared with their RN colleagues and those in part-time or casual positions.   (Holden et al., 2011). Although human factors research is investigating influences of one system level on another (e.g. additive effects), this is a new area of study (Karsh, Waterson, & Holden, 2014).

| D ISCUSS I ON
A second key finding was that patient/family complaints were directly related to increased reports of both types of workplace TA B L E 2 Bivariate correlations between study variables (N = 528) f (0 = very or somewhat independent, 1 = somewhat or very dependent). *p < .05. **p < .01. ***p < .001.
violence, but there was a stronger relationship with emotional violence than physical violence According to Bandura's (1991) (Gallagher & Mazor, 2015;Pichert, Hickson, & Moore, 2008), there is a dearth of evidence related to its impact on nurse safety.
Two mid-range theories, in particular, shed light on study findings: stress theory (Russ-Eft, 2001;Staal, 2004) and spiral of incivility theory (Andersson & Pearson, 1999;Sommovigo, Setti, Argentero, & O'Shea, 2019). According to stress theory, too much stress and/or chronic stress (distress) has detrimental effects on cognition, attention and memory, which in turn adversely influence performance (Russ-Eft, 2001;Staal, 2004). Heavy workload is a source of distress to nurses, hindering their ability to provide effective patient care. Nursing evidence has established a link between heavy workload, job distress and poor performance (Kokoroko & Sanda, 2019;Li et al., 2017). We surmise that poor quality performance (due to workload factors) is negatively perceived by patients/families who initially respond with complaints.
According to the spiral of incivility theory, there is a spiralling process in response to uncivil acts which typically starts with little misbehaviours that can escalate to more serious acts of aggression (Andersson & Pearson, 1999;Sommovigo et al., 2019). If patients/ families perceive lack of quality care (e.g. lack of response to call lights) as an uncivil act towards them, this theory suggests that Change in R 2 6.5%*** 3.4%*** 9.1%*** 1.5%** 2.7%*** R 2 6.5%*** 9.8%*** 18.9*** 20.4%** 23.1%*** This spiralling process may be triggered when patients' coping strategies are suboptimal due to external causes (e.g. illness, pain) that put them in vulnerable situations (Sommovigo et al., 2019). A systematic review of 53 studies of service providers across a range of industries and sectors used this theory to explain how poor employee performance spirals into client aggression (Sommovigo et al., 2019). Among customer service providers outside of healthcare, poor employee performance was a precursor to client dissatisfaction and incivility such as making gestures (e.g. eye rolling) to express their impatience (Sliter & Jones, 2016). Future research should explore this theory in the context of nurse-patient relationships.
Use of the human factors framework provides a more granular examination of workload factors at different levels that precipitate patient complaints and workplace violence. In this study, certain nurse workload factors, such as job-level compromised professional standards and task-level interruptions, were a greater source of distress to nurses than unit-level and other job-level workload factors. Rodney (2017) (Pereira, Mueller, & Elfering, 2015).
Patient/family complaints were the strongest predictor of both physical and emotional violence, more so than workload factors. This finding may be because nurse workload is only one source of patient/ family complaints. Other factors in the healthcare context such as unmet expectations of patients and families, ineffective communication, lack of resources, wait times and poor care coordination are also known to result in patient dissatisfaction and complaints (Gallagher & Mazor, 2015;Lee, Moriarty, Borgstrom, & Horwitz, 2010;Najafi, Fallahi-Khoshknab, Ahmadi, Dalvandi, & Rahgozar, 2018). Evidence suggests that patient complaints are often dismissed by healthcare providers as attributions of patient personalities (Gallagher & Mazor, 2015). Patient complaints may be a precursor to emotional and physical violence: They may be the 'canary in the coal mine.'

| Limitations
A key strength of this study was its systematic conceptualization of workload using a human factors framework. Additionally, to our

| CON CLUS ION
This study's findings support the importance of using a multisystems human factors framework to examine those workload factors at unit,  TA B L E 5 Sobel test results for mediation effect of complaints on the relationship between workload factors and physical and emotional violence (N = 528) job and task levels that are associated with workplace violence. The study findings have implications for nurses, managers and policymakers. Heavy workload may be one of the root causes of patient/family violence towards nurses. Employers and policymakers must implement system-level specific strategies that alleviate nurse workload factors at multiple levels. While most efforts in healthcare target nurse staffing levels as the only unit-level indicator of heavy workload, new efforts must target workload factors including and beyond nurse-patient ratios. New workload management approaches, for example, advocate for assignments that create a match between patients' care needs and nursing competencies and level of experience (Georgiou, Amenudzie, Ho, & O'Sullivan, 2018).
Our findings also suggest patients and their families respond adversely to nurses' inability to provide effective care due to heavy workloads. Thus, it is important that nurses and managers take patient/family complaints about nursing care seriously. Future research should examine the impact of tracking patient/family complaints on patient and nurse outcomes. According to Gallagher and Mazor (2015), "like any adverse event, patient complaints have an epidemiology that can yield important lessons for prevention ….In many situations, patients and family members may be the first to detect lapses in safety or quality" (p. 352-353). Accordingly, it is important that nurses and managers apply a root cause analysis approach with patient/family complaints as they do with other patient adverse events in health care.

CO N FLI C T O F I NTE R E S T
None to declare.

AUTH O R CO NTR I B UTI O N S
Havaei has made substantial contribution to this study's conception and design, analysis and interpretation of the data. MacPhee has made substantial contribution to the larger study design and acquisition of the data. Both authors have been involved in drafting the manuscript and revisiting it critically for important intellectual content.

PATI E NT CO N S E NT
Not required.