The DeVOS study started in June 2020 in the only 2 publicly funded, referral hospitals (Azienda Socio-Sanitaria Territoriale [ASST]) serving the provinces of Varese (ASST Sette Laghi, hospital 1) and Como (ASST Lariana, hospital 2). Located in the Lombardia Region, the 2 provinces include about 1.5 million inhabitants, corresponding to 15% of the regional population. Each ASST includes hub-and-spoke hospitals, hospitals dedicated to mothers and children and rehabilitations, and outpatient clinics. To address the feasibility of our protocol in structures at different underlying WPV risks, we included all the hospitals comprising the ASSTs. The study had 4 main objectives: (1) to quantify WPV before the implementation of the new protocol (2016-2020) and to document the extent by which the available information on WPV episodes was compliant with the guidelines [14] across different hospitals; (2) to design and implement a new protocol for WPV reporting and management (Figure S1 in Multimedia Appendix 1), comprising of a standardized and easy-to-access incident report procedure gathering first information on violence attributes, consequences and contributing factors; of a root cause analysis for more in-depth assessment of violence determinants and impacts; and of a web-based platform for comprehensive event management and data collection; (3) to estimate WPV prevalence in the entire HCW population and in different subpopulations and wards after the implementation of the new standard; and (4) to assess the role of work organizational factors by estimating the association between turn-over, downsizing, sickness leaves, night shift working (defined according to a published method [15]), and WPV occurrence. The specific aims of this paper are related to objectives 1 to 3, whereas the role of organizational factors, as well as the psychological impact of violence, will be addressed in dedicated works. In accordance with regional guidelines [14] and international standards [16], WPV was defined as any form of verbal abuse, threats, physical assaults (to persons or things) and sexual harassment occurring at the workplace and perpetrated by hospital patients or visitors or hospital employees.

During 2016-2020, incident reporting was based upon procedures and data collection forms that were neither standardized between hospitals nor specific for aggressions. The affected HCW notified the violence episode to the risk manager in 1 ASST and to the safety personnel in the other. The incident report data collection forms differed across hospitals and were also used for incidents other than WPV (eg, treatments adverse events, “near miss”), and several versions were adopted during the 5 years following changes in hospital’s organizations and managements. A root cause analysis was not implemented in either hospital.

Figure S1 in Multimedia Appendix 1 depicts the flow chart for the WPV reporting and management protocol of the DeVOS project. Assaulted HCWs were required to notify the WPV to the risk management office (hospital 1) or to the safety personnel (hospital 2) within 72 hours of occurrence by using a standardized, WPV-specific incident report data collection form, available on the hospitals’ intranet. The form was developed taking into account previous experience by the risk managers of the participating hospitals, regional guidelines [14], and existing literature on the topic [9,10,17,18], and it collected information on violence attributes (form, hospital ward, involved HCW, perpetrator, violence date and time, and environmental factors), consequences (physical, psychological, reactions, feelings during the aggression, and work injury), and contributing factors (sociocultural, structural, organizational, relational, and clinical). More details are in Table S1 in Multimedia Appendix 1. Most fields were not mutually exclusive, and the HCW could fulfill more than 1 choice. In the case of WPV involving multiple HCWs, the risk management office was required to check that each assaulted HCW filled in 1 incident report. HCWs signing the specific consent were further contacted by the psychologist for the root cause analysis, comprising of a structured interview including the Modified Overt Aggression Scale [19] and the Broset Violence Checklist [20]. Then, the HCW received to an email address of his or her choice a 1-time access link to the project web platform for a safe and easy completion of the questionnaires assessing the impact of the WPV event (presence of cognitive, emotional, and somatic symptoms [21]; the Maslach Burnout Inventory [22]; and the General Health Questionnaire-12 [23]) and psychosocial work conditions (the Italian version of the Health and Safety Executive’s Management Standards Indicator Tool [24]). These questionnaires were validated for use with Italian HCWs [21,24,25]. Furthermore, the psychologist contacted the HCW supervisor for a guided interview using a checklist on work content and context factors. During periodic training meetings with workforce coordinators, the risk management officers explained the new protocol and provided information on violence notification, field definitions on the incident report form to standardize data collection (including contributing factors), and the rationale and motivation for participation in the root cause analysis. The coordinators were then asked to instruct the HCW of their ward or unit. Additionally, in hospital 1, all newly hired HCWs were targeted by the same training. In both ASSTs, the new protocol became effective on November 1, 2021. One ASST (hospital 2) gradually introduced the protocol to the outpatient departments during the year 2022. Therefore, for that ASST, we included only the hub and the main spoke hospitals in the current analyses, corresponding to 71% of the workforce.

As part of the study, we developed a new web-based platform for centralized data collection, with customized access to data visualization or modification for risk management office personnel, the psychologist, and the data analyst. Due to ethical and confidentiality issues, we collected minimal sensitive information on the HCW and the aggressor. Furthermore, the easy visualization of questionnaire results by the psychologist allowed an immediate identification of high-risk workers for referral to supportive or mitigation resources.

The study received approval by the Ethical Committee of Insubria (IDs 82/2021 and 90/2021). Participants signed an informed consent on data confidentiality and protection and a separate informed consent to participate in the interview with the psychologist. The web-based platform for centralized data collection complied with the European Union legislation on data protection. The platform applied a complete anonymization of personal data, allowing the tracing of repeated events on the same worker. HCWs received no compensation for their participation in the study.

To be consistent with the literature mostly reporting WPV prevalence for a 12-month interval, for each year in the “before” study period (2016-2020) and by hospital, we estimated the 12-month WPV rate (per 100 workers) as the ratio between the number of reported WPV in the year and the number of HCWs in the payroll administrative records at the beginning of each year. The 95% CI for the yearly rate was estimated using the exact binomial distribution [26]. We estimated underreporting as the ratio between the observed and expected number of violent incidents. For the latter, we used the WPV rate estimated in 2015-2017 in a large public University hospital in Northern Italy by Viottini and colleagues [27] as 210 observed incidents each year over a population of 10,970 HCWs, along with the number of HCWs, again from payroll administrative records. The 95% CI for the ratio was estimated from the Poisson exact method [26]. To date, we completed the first 14 months since the implementation of the new reporting standard (November 1, 2021 to December 31, 2022). For the entire period, we reported on data completeness as missing data prevalence on the incident report data collection forms. To describe the violence attributes, we first assessed the co-occurrence of different types of violence in the same episode, by reporting percentages and mean number of violence types by episode. Then, we reported the distribution of major violence characteristics, consequences, and contributing factors, in the overall sample and by sex of the assaulted HCWs, to identify the presence of sex-related patterns. The association between violence characteristics and sex was formally tested through chi-square tests. The analyses of WPV underreporting and rate were restricted to a 12-month period (year 2022, n=166 WPV) adopting the same methods described above for the “before” period and using the number of HCWs in the payroll records at the beginning of the implementation of the new standard. The overall rate was then broken down as the sum of 2 mutually exclusive components: 1 being the rate of the first violent episode and the second being the rate of recurrent violent episodes on the same worker. Statistical analyses were performed using SAS software (version 9.4; SAS Institute Inc).

Between 2016 and 2020, a total of 400 WPV reports were notified in the 2 study hospitals. Data completeness was scanty: in 160 (40%) of the reports, the affected HCW was not identified (eg, replaced by a “generic” description of job title or by expressions such as “everyone present”), and when identifiable, in 70.8% (170/240) of reports, there was at least 1 missing information on age, sex, or job title. This prevented us from providing specific prevalence estimates by subgroups. Similarly, environmental details, consequences, and contributing factors were not available in 59% (236/400), 18% (72/400), and 50% (200/400) of reports. Yearly WPV rates were diverging between the 2 study hospitals, increasing from 0.52 (95% CI 0.34-0.77) per 100 HCWs to 1.0 (95% CI 0.75-1.31) in 1 hospital and declining from 1.92 (95% CI 1.42-2.54) to 0.57 (95% CI 0.32-0.93) in the other (Figure 1A). The yearly observed to expected ratio was significantly below 1 for most years, suggesting substantial (up to 73%) underreporting. In particular, the ratio from 2016 to 2020 changed from 0.27 to 0.52 (eg, reduction in underreporting) in the hospital with an increasing WPV rate and from 1.0 to 0.30 (eg, increase in underreporting) in the other.

During the first 14 months of implementation of the study standard, a total number of 205 WPV were reported, half (50%) of them on the same day of occurrence, and 75% within the first 72 hours. Data completeness was optimal: only in 2% (4/205) of reports the affected HCW was not identified, and when identifiable, only in another 2% (4/201) there was at least 1 missing information on demographics or job title information (Table S1 in Multimedia Appendix 1). The observed to expected ratio was 1.91 (95% CI 1.63-2.21) as compared to the “before” period in the same hospitals, which was consistent in the 2 hospitals (1.96 and 1.77). When using the reference rate from literature to estimate the expected number of WPVs, the observed to expected ratio was 1.14 (95% CI 0.98-1.32), being slightly lower in hospital 2 where the first notification was to safety personnel (0.83, 95% CI 0.60-1.13) rather than to the risk management office (1.28, 95% CI 1.07-1.52) as in hospital 1. The 12-month WPV rate was 2.08 (95% CI 1.79-2.42) per 100 HCWs (Table 1), being slightly lower in hospital 2 than in hospital 1 (1.52 vs 2.35; P=.02; Figure 1B). WPV rate (Table 1) declined with age, with the youngest group being at the highest risk (P<.001); it was higher in male than female HCWs (2.82 vs 1.83; P=.008); and it was more than doubled in nurses (3.49; 95% CI 2.93-4.15), as compared to nurse assistants (1.74) and physicians (1.08; P<.001). Finally, psychiatric wards (14.3) and emergency departments (13.3) were at higher WPV rates than the remaining wards; about one-fifth (0.41/2.08, 19.7%) of the rates are due to recurrences. Analyses by hospital (Table S2 in Multimedia Appendix 1) suggested higher rates in the youngest group in hospital 1 and in psychiatric wards in hospital 2, whereas associations by sex and job title were homogeneous across the hospitals.

Table 2 reports the co-occurrence of violence forms in the incident reports. Verbal abuse was the largest majority (186/205, 90.7%) Of these, verbal abuse was the only form of violence in 26 (14%) out of 186 incidents, whereas 134 (72%) also reported threats, 98 (52.7%) reported physical assaults, and 15 (8%) reported sexual harassment. The mean number of reported violence forms was 2.33. We report similar figures for the other violence forms. Table 3 reports the distribution of major violence characteristics and risk factors, by sex of the assaulted HCW. Male HCWs were more likely than female HCWs to be subject to physical assault (45/67, 67.2% vs 62/130, 47.7%; P=.01). Either alone or with a patient, visitors were more likely to assault female than male HCWs (40/130, 30.7% vs 8/67, 12%; P=.02). No sex-related differences were observed on violence time or location. Female HCWs were less likely than male HCWs to report a work injury to the insurance authority (4/130, 3.1% vs 15/67, 22.4%; P<.001) and were more likely than men to report psychological consequences (46/130, 35.4% vs 13/67, 19.4%; P=.02) and life-threatening feelings during the aggression (23/130, 17.7% vs 5/67, 7.5%; P=.05). Finally, 47.8% (32/67) of WPV in male HCWs was related to the clinical conditions of the perpetrator, which is greater than that in female HCWs (40/130, 30.8%; P=.02). In female HCWs, the single most frequently reported factor was the sociocultural one (eg, related to linguistic or cultural barriers or behaviors; 53/130, 40.8%). Of note, structural factors including safety (the lack of barriers, alarm systems, or escape or poor lightning) were among the least prevalent identified contributing factors, both in men and in women.