Association between healthcare professionals’ perceptions of hospital attitudes towards workplace violence and the prevalence of workplace violence: a nationwide cross-sectional survey in China

Abstract

Objective

To examine the association between healthcare professionals’ (HCPs) perceptions of their hospitals’ attitudes towards workplace violence (HATWPV) and the prevalence of workplace violence (WPV) in China.

Design, settings and participants

This nationwide cross-sectional survey was conducted among HCPs from September 2022 to January 2023. Participants were recruited using a multistage sampling strategy across three provinces (Zhejiang, Henan, Guizhou) representing eastern, central and western China. A total of 2976 valid questionnaires from 92 medical institutions were analysed.

Primary and secondary outcome

The primary outcomes were self-reported physical violence and verbal threats in the past year. Secondary outcomes were HCPs’ perceptions of HATWPV: concern about experiencing WPV (CAE-WPV), satisfaction with hospitals’ procedures and systems for addressing WPV (SHPS-aWPV) and perceived institutional support for reporting WPV incidents (SR-WPVIs).

Results

A total of 2370 (79.6%) HCPs expressed CAE-WPV, 1438 (48.3%) were dissatisfied with SHPS-aWPV and 981 (33.0%) reported working in institutions without SR-WPVIs. After adjusting for sociodemographic, occupational and stress-related factors, negative perceptions across all three HATWPV dimensions were significantly associated with higher risks of both physical violence and verbal threats. The association varied by hospital level and region. For example, dissatisfaction with procedures showed a stronger link with verbal threats in primary care institutions and with physical violence in the eastern region.

Conclusion

HCPs’ perceptions of HATWPV are associated with the prevalence of WPV. Medical institutions should proactively assess these perceptions, establish transparent and supportive reporting mechanisms and integrate violence prevention into organisational culture and evaluation systems to promote a safer workplace.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• This large-scale, multicentre cross-sectional investigation encompasses primary, secondary and tertiary healthcare institutions, various specialties and various age groups, with strong geographical and institutional representativeness.

• Workplace violence and hospital attitude perceptions were measured using self-reported questionnaires, which may introduce recall and social desirability biases.

• The cross-sectional design precludes causal inference.

• Although the sample distribution mirrors the national macro-trend of health workforce concentration, participants from the eastern region are slightly over-represented.

• Data were collected after the relaxation of COVID-19 restrictions; despite adjusting for pandemic-related covariates, residual confounding (eg, department type, patient volume) may remain.

Introduction

Workplace violence (WPV) directed towards healthcare professionals (HCPs) in clinical settings and encompasses physical violence, verbal threats and medical disputes.¹ WPV has become a widespread concern globally, particularly in Asian countries.^{2 3} In China, reported WPV prevalence reaches 62.4%,⁴ with doctors and nurses as the primary victims.⁵ A recent large-scale online cross-sectional survey among Chinese HCPs further corroborated the high prevalence of WPV (58.2%), identifying emotional abuse (49.7%) and threats (27.9%) as the most common forms, and highlighting male gender, work shift, employment in secondary hospitals and longer working hours as significant risk factors.⁶ According to the Indian Medical Association, nearly 75% of doctors have experienced intimidation, verbal threats or physical violence while fulfilling their professional responsibilities.³ In Pakistan, 24 studies have reported WPV prevalence rates of 25–100%.⁷ In Bangladesh, 80% of nurses reported experiencing WPV in the past year.⁸ The rising prevalence threatens HCPs’ personal safety,⁵ contributes to HCPs attrition and undermines HCPs sustainability.⁶ Medical institutions serve as the primary interface between service providers and recipients; therefore, their attitudes and responses to WPV may therefore be associated with its occurrence.⁹

Previous WPV research has primarily focused on demographic characteristics and the violence prevalence. Specifically, individuals with certain attributes, such as longer service years, senior professional titles, demanding work arrangements including frequent night shifts and heavy patient loads, are associated with elevated WPV vulnerability. Furthermore, HCPs working in high-stress environments like emergency or mental health departments are more susceptible to experiencing WPV.^{6 10} Patient-related factors, such as limited medical knowledge, unrealistic expectations regarding treatment outcomes, dissatisfaction with healthcare services and high costs, have also been identified as major contributors to WPV.¹¹ Additionally, social factors, including stigmatisation of HCPs in public discourse and the normalised distrust of the medical profession, have been shown to exacerbate WPV.^{12 13} However, evidence-based research examining medical institutions’ behaviours and attitudes towards WPV remains limited. Existing studies have focused on interventions such as metal detectors, which do not reduce violent incidents.¹⁴ Cai et al proposed that tailored training models could enhance HCPs’ WPV management capacity.¹⁵ Other studies suggest that institutional efforts often target identifying violent patients and specific groups for intervention.¹⁶ Despite these findings, large-scale studies investigating the association with hospital attitudes towards workplace violence (HATWPV) are unavailable.

Our previous study on WPV incidents in county-level hospitals in eastern China revealed a strong correlation between HCPs’ perceptions of HATWPV and WPV prevalence.¹⁷ However, whether this association holds across regions with different healthcare resource allocation and economic development levels requires further investigation. Existing research indicates that WPV in developing countries, such as India and Pakistan, is closely linked to health resource distribution challenges stemming from socioeconomic disparities.^{7 18 19} For example, the risk of physical assault against HCPs in southern Italy is 9.33 times that in the northern regions.²⁰ Previous studies suggest that uneven medical resources distribution may overload certain institutions (eg, prolonged waiting times, reduced patient-provider communication), which has been associated with lower patient satisfaction and heightened provider-patient tensions.^{2 21} In China, pronounced regional imbalance in medical resources distribution is well documented. The eastern region holds substantial advantages over the western region in basic medical infrastructure, human resources and overall resource allocation.²² Moreover, urban medical institutions are also better equipped with hospital beds and human resources than rural ones.²³ Therefore, this study aimed to survey medical institutions across various regions and levels in China to further investigate the association between HCPs’ perceptions of HATWPV and WPV prevalence, aiming to provide evidence-based strategies for effective WPV prevention.

Methods

Study design and data collection

This nationwide cross-sectional survey of HCPs followed the Strengthening the Reporting of Observational Studies in Epidemiology guideline, using structured questionnaires (see Brief Questionnaire) and self-report methods. The questionnaire, written in Chinese, was administered to Chinese HCPs via the online platform Questionnaire Star between September 2022 and January 2023. Detailed inclusion and exclusion criteria are provided below. Prior to the completion of the questionnaire, all respondents were required to review and provide consent through an electronic informed consent form.

Study participants

The survey targeted HCPs who had direct contact with patients in medical institutions at various levels across the eastern, central and western regions of China.

Inclusion criteria

HCPs aged between 18 and 65 years, employed at any level or type of medical institution, holding relevant professional qualifications and practice certificates, volunteered to participate in the study and provided written informed consent were included in the survey.

Exclusion criteria

HCPs who were on authorised or sick leaves for ≥2 weeks during the survey period and who were working as hospital administrators were excluded.

Patient and public involvement

Patients and members of the public were not involved in the design, conduct, reporting or dissemination of this research.

Sample size and sampling

We estimated the total sample size using the standard formula ${\mathrm{P}}_{\mathrm{o}}$, with Z_α/2 =1.96 (α=0.05, two-tailed). Based on preliminary findings and an estimated WPV prevalence of 62.4%,⁴ we set the bound on error (δ) at 0.013, yielding a required sample size of 2672 participants. Accounting for a potential non-response rate of 10–15% and the likelihood of incomplete or invalid questionnaires, we increased the final estimated sample size to approximately 3000.

Sampling was conducted in three steps:

Step 1, based on geographical location, economic development, medical resource allocation and population density, this study selected Zhejiang (eastern China), Henan (central China) and Guizhou (western China) as the research provinces, as their medical resources and economic levels provide regional representativeness (see online supplemental table S1).^24–27 The higher proportion of participants from the eastern region aligns with the national concentration of healthcare human resources in eastern China.

Step 2, within each province, healthcare institutions were selected through quota sampling by region and institution type. The target sample size for medical institutions was allocated based on their scale and level (see online supplemental materials 1). Specifically, in each province, 3–5 tertiary hospitals (aiming to recruit 70–90 participants each), 6–8 secondary hospitals (aiming for 30–50 participants each) and 18–20 primary care institutions (aiming for 10–30 participants each) were selected.

Step 3, participants were selected using either cluster sampling or stratified quota sampling by department/professional title, depending on the number of HCPs in each institution. Each medical institution invited one to three coordinators, who received training on research practices before initiation and were required to strictly maintain confidentiality and impartiality throughout the study. To enhance representativeness and reduce self-selection bias, institutional coordinators proactively invited HCPs meeting predefined quotas through internal meetings, notices and WeChat groups, rather than open voluntary registration. All participants were informed about electronic informed consent, data anonymity and confidentiality prior to data collection.

Given the data collection occurred during the COVID-19 policy adjustment period when a higher non-response rate was possible, we conservatively distributed 4320 questionnaires to secure at least 3000 valid responses. A total of 2996 were returned (69.4% response rate), of which 2976 were valid (68.9% effective response rate). To evaluate potential sampling bias, we compared the demographic characteristics of the final sample with national data from the China Health Statistics Yearbook 2022 (online supplemental table S2). The standardised mean differences for key variables, including gender, age distribution and hospital tier, were all below 0.15. This indicates a negligible to small difference and suggests that the sample is reasonably comparable to the national population of HCPs in these structural characteristics.

Measures

Outcome measures

Physical violence and verbal threats resulting from WPV were designated as the primary outcome variables and quantitatively assessed through self-reported measures. Physical violence was defined as direct physical assaults on HCPs, including beating, kicking, slapping, stabbing, pushing, biting and pinching.²⁸ Verbal threats were defined as explicit verbal statements expressing potentially violent intent that caused psychological fear or distress among HCPs. The participants were asked the following question: ‘In the past year, have you experienced physical violence or verbal threats from patients, their relatives or friends?’ Responses were categorised using a 3-point scale (1=experienced physical violence, 2=experienced verbal threats without physical violence and 3=did not experience WPV).

In accordance with previous literature, several sociodemographic and occupational variables were included in the analysis (table 1). Uncertainty stress was selected as an independent variable based on the stress-aggression theory. We measured HCPs’ perceptions of HATWPV using a 5-point Likert scale, including: concern about experiencing WPV (CAE-WPV), support for reporting WPV incidents (SR-WPVIs) and satisfaction with hospital procedures and systems for addressing WPV (SHPS-aWPV). We developed these items through a comprehensive literature review, adapted them from WHO questionnaires and had them reviewed by a five-expert panel for content validity. The scale demonstrated acceptable internal consistency, with a Cronbach’s α of 0.726.

Table 1. Sociodemographic characteristics and difference analysis of WPV experienced by HCPs in China.

Demographic variables	Total n (%)	WPV
		Physical violence n (%)	Verbal threat n (%)	No WPV n (%)
	2976	94 (3.2)	505 (17.0)	2377 (79.9)
Gender* χ2=44.29; p<0.001
Male	927 (31.1)	35 (3.8)	218 (23.5)†	674 (72.7)
Female	2049 (68.9)	59 (2.9)	287 (14)	1703 (83.1)
Age‡ H=40.5; p<0.001
Age‡ (years mean±SD)	35.7±9.2	37.64±8.8	37.59±8.7	35.21±9.3
Marital status* χ2=29.23; p<0.001
Married	2124 (71.4)	79 (3.7)†	401 (18.9)†	1644 (77.4)
Unmarried	781 (26.2)	14 (1.8)	94 (12.0)	673 (86.2)
Other	71 (2.4)	1 (1.4)	10 (14.1)	60 (84.5)
District* χ2=24.8; p<0.001
Western	714 (24)	26 (3.6)	100 (14.0)	588 (82.4)
Central	903 (30.3)	33 (3.7)	196 (21.7)†	674 (74.6)
Eastern	1359 (45.7)	35 (2.6)	209 (15.4)	1115 (82.0)
Level of medical institution* χ2=32.92; p<0.001
Primary	1157 (38.9)	31 (2.7)	151 (13.1)	975 (84.3)
Secondary	715 (24.0)	16 (2.2)	125 (17.5)	574 (80.3)
Tertiary	1104 (37.1)	47 (4.3)	229 (20.7)†	828 (75)
Profession* χ2=57.98; p<0.001
Clinician	1468 (49.9)	56 (3.8)	321 (21.6)†	1109 (74.6)
Nurse	806 (27.1)	27 (3.3)	111 (13.8)†	668 (82.9)
Other medical profession	684 (23)	11 (1.6)	73 (10.7)	600 (87.7)
Education level§ H=22.8; p<0.001
Associate’s degree or below	738 (24.8)	9 (1.2)	79 (10.7)	650 (88.1)
Bachelor’s degree	1723 (57.9)	64 (3.7)	304 (17.6)	1355 (78.6)
Master’s degree or above	515 (17.3)	21 (4.1)	122 (23.7)†	372 (72.2)
Professional title§ H=80.2; p<0.001
Junior	1328 (44.6)	23 (1.7)	145 (10.9)	1160 (87.3)
Intermediate	1084 (36.4)	44 (4.1)	197 (18.2)	843 (77.8)
Senior	564 (19)	27 (4.8)	163 (28.9)†	374 (66.3)
Uncertainty stress* χ2=64.8; p<0.001
No	2186 (73.5)	45 (2.1)	324 (14.8)	1817 (83.1)
Yes	790 (26.5)	49 (6.2)†	181 (22.9)†	560 (70.9)
Years of work experience‡ χ2=35.52; p<0.001
<10 years	1474 (49.5)	40 (2.7)	194 (13.2)	1240 (84.1)
10–19 years	859 (28.9)	26 (3.0)	178 (20.7)†	655 (76.3)
≥20 years	643 (21.6)	28 (4.4)	133 (20.7)†	482 (75)
Working hours§ H=76.4; p<0.001
<40 hours/week	577 (19.4)	9 (1.6)	54 (9.4)	514 (89.1)
40–49 hours/week	1098 (36.9)	22 (2.0)	159 (14.5)	917 (83.5)
50–59 hours/week	530 (17.8)	19 (3.6)	111 (20.9)	400 (75.5)
≥60 hours/week	771 (25.9)	44 (5.7)†	181 (23.5)†	546 (70.8)
Work intensity* χ2=95.37; p<0.001
Low-moderate	1590 (53.4)	23 (1.4)	194 (12.2)	1373 (86.4)
High	1386 (46.6)	71 (5.1)†	311 (22.4)†	1004 (72.4)
CAE-WPV* χ2=97.52; p<0.001
No	606 (20.4)	6 (1.0)	29 (4.8)	571 (94.2)
Yes	2370 (79.6)	88 (3.7)†	476 (20.1)†	1806 (76.2)
SHPS-aWPV* χ2=113.89; p<0.001
Satisfied	1538 (51.7)	24 (1.6)	170 (11.1)	1344 (87.4)
Dissatisfied	1438 (48.3)	70 (4.9)†	335 (23.3)†	1033 (71.8)
SR-WPVIs* χ2=86.16; p<0.001
Supportive	1995 (67.0)	39 (2.0)	270 (13.5)	1686 (84.5)
Opposed	981 (33.0)	55 (5.6)†	235 (24.0)†	691 (70.4)

For statistical analysis, responses on the original 5-point Likert scales were dichotomised as follows: CAE-WPV (1–2=not concerned, 3–5=concerned); SHPS-aWPV (1–2=satisfied, 3–5=dissatisfied); SR-WPVIs (1–2=supportive, 3–5=opposed).

^*R×C list test.

^†Adjusted standardised residuals ≥2.58.

^‡Kruskal-Wallis H test.

^§Non-parameter Kruskal-Wallis H test and R×C list test.

CAE-WPV, concern about experiencing WPV personally; HCPs, healthcare professionals; SHPS-aWPV, satisfaction with their hospitals’ procedures and systems for addressing WPV; SR-WPVIs, support HCPs reporting of WPV incidents; WPV, workplace violence.

The following HCPs’ perceptions of HATWPV items were included in the survey:

1. Work environment safety dimension: Participants were asked, ‘How concerned are you about experiencing WPV?’ Responses were rated on a 5-point Likert scale (1=not at all concerned, 2=not concerned, 3=somewhat concerned, 4=concerned and 5=extremely concerned).

1. Policies and systems: Participants were asked, ‘How satisfied are you with the procedures and systems for handling violent incidents in your medical institution, such as reporting to the police, safety protocols and post-incident support?’ Responses were rated on a 5-point Likert scale (1=very satisfied, 2=satisfied, 3=neutral, 4=dissatisfied and 5=very dissatisfied).

2. Institutional culture: Participants were asked, ‘To what extent does your medical institution support reporting WPV incidents to authorities?’ Responses were measured on a 5-point Likert scale (1=strongly supported, 2=supported, 3=neutral, 4=opposed and 5=strongly opposed).

Workload intensity was measured using a 5-point Likert scale and subsequently categorised into low-to-medium intensity and high intensity. This categorical conversion was conducted in accordance with standard statistical procedures to facilitate comparative analysis while maintaining conceptual clarity.

Uncertainty stress was assessed using the Uncertainty Stress Scale developed by Professor Yang Tingzhong from Zhejiang University²⁹ (see online supplemental materials 1). This instrument is designed to assess the psychological pressure experienced by individuals due to uncertainty related to events or situations. It has been widely validated for use in Chinese populations and demonstrates strong internal consistency, with a Cronbach’α coefficient of 0.81. A higher score indicated greater levels of uncertainty stress. Participants were classified as experiencing positive uncertainty stress if their overall score exceeded 12.

Statistical analysis

All statistical analyses were conducted using SPSS V.26.0 (IBM, Armonk, New York, USA) and comprised three sequential steps.

Step 1: Sociodemographic characteristics, work-related factors, uncertainty stress and self-perceived HATWPV were summarised based on the occurrence of WPV. Continuous variables were expressed as the means and SDs, whereas categorical variables were expressed as frequencies and percentages. Group differences were examined using the χ² test, Kruskal-Wallis H tests or Mann-Whitney U tests. When significant differences were identified among multiple groups, adjusted standardised residuals were used for post hoc comparisons.

Step 2: After confirming no significant differences among institutions using multilevel models and assessing the absence of potential multicollinearity among explanatory variables via variance inflation factors, multivariate logistic regression was employed to evaluate the association between HCPs’ perceptions of HATWPV and WPV prevalence. To account for potential confounding, the model was adjusted for COVID-19-related variables (uncertainty stress, workload intensity and working hours).

Step 3: Exploratory subgroup analyses were performed based on hospital level and geographical region to examine the impact of regional healthcare resource disparities. We included interaction terms for ‘HATWPV×hospital level’ and ‘HATWPV×geographic region’ separately in the primary multivariate regression models to visualise the effect estimates of HATWPV across different hospital levels and geographical regions. Additionally, multivariate logistic regression analyses were used to further investigate subgroup differences in the impact of HATWPV on WPV prevalence, based on sociodemographic characteristics, work-related factors and uncertainty stress variables associated with WPV prevalence.

Step 4: We calculate the attributable risk and population attributable fraction of key attitude variables to clarify the association of HCPs’ perceptions of HATWPV with WPV. For population attributable fraction calculation, each attitude variable was dichotomised (exposed vs reference group) and entered into multivariable logistic regression models adjusted for sociodemographic factors, occupational variables, uncertainty stress. Exposure groups were defined as: CAE-WPV=‘somewhat concerned/concerned/extremely concerned’; SR-WPVIs=‘neutral/opposed/strongly opposed’; SHPS-aWPV=‘neutral/dissatisfied/very dissatisfied’; all others served as reference. Adjusted ORs (AORs) were used to compute population attributable fraction.

No missing data were present for any analytical variable. All statistical computations were two-tailed, and a p value of <0.05 was considered significant.

Results

Sociodemographic disparities in WPV prevalence across study groups

The study included 92 medical institutions. Their characteristics and distribution are presented in online supplemental table S3). Table 1 summarises the sociodemographic characteristics of HCPs in China. A total of 2976 HCPs participated, mean age of 35.7±9.2 years. Of these, 2370 (79.6%) expressed concern about WPV occurrence. Dissatisfaction with institutional protocols for managing WPV was reported by 1438 (48.3%) respondents, while 981 (33%) indicated their medical institutions did not support the reporting of WPV incidents. A total of 94 (3.2%) respondents reported experiencing physical violence, whereas 505 (17.0%) reported experiencing verbal threats. Significant differences in the prevalence of physical violence and verbal threats were observed across various demographic and occupational categories, levels of uncertainty stress and HCPs’ perceptions of HATWPV classifications. Post hoc multiple comparisons revealed that respondents reporting uncertainty stress, working ≥60 hours/week, experiencing high workload intensity, expressing CAE-WPV, reporting dis-SHPS-aWPV and perceiving insufficient SR-WPVIs were more likely to experience physical violence and verbal threats.

Association between HCPs’ perceptions of HATWPV and the prevalence of WPV

As shown in table 2, after adjusting for sociodemographic factors, work-related factors and uncertainty stress, cross-sectional analysis indicated significant associations between dimensions of HCPs’ perceptions of HATWPV and the prevalence of WPV. Specifically, reported physical violence was associated with a ‘very dissatisfied’ status regarding SHPS-aWPV (AOR = 5.24). Reported verbal threats showed clear graded associations across different dimensions: for CAE-WPV, AOR increased progressively with higher levels of concern, 2.83 (somewhat concerned), 4.25 (concerned) and 5.78 (very concerned); for SR-WPVIs, AOR rose with more negative attitudes toward reporting, from 1.55 (neutral) to 2.20 (opposed); and for SHPS-aWPV, AOR also increased with greater dissatisfaction, 1.63 (neutral), 2.53 (dissatisfied) and 4.33 (very dissatisfied). Subgroup analyses across different demographic and occupational strata confirmed the robustness of these associations, although with reduced statistical power due to smaller sample sizes in subgroups (online supplemental table S4). Exploratory subgroup analyses further suggested that lack of institutional support for reporting (SR-WPVIs) was more strongly associated with both types of violence among male healthcare workers, nurses and those with high work intensity; however, these differences did not alter the overall direction or significance of the main findings.

Table 2. Multiple logistic regression analysis of the correlation between HCPs’ perceptions of HATWPV and the prevalence of WPV.

Variables		Workplace violence
		Physical violence		Verbal threat
		OR (95% CI)	AOR* (95% CI)	OR (95% CI)	AOR* (95% CI)
Gender	Male	1.50 (0.98 to 2.30)	1.72 (1.02 to 2.90)†	1.92 (1.58 to 2.34)‡	1.94 (1.52 to 2.47)‡
	Female (reference)
Age (years)	18–29	0.57 (0.33 to 0.99)†	2.34 (0.72 to 7.63)	0.46 (0.36 to 0.59)‡	0.89 (0.52 to 1.53)
	30–39	0.97 (0.61 to 1.56)	1.67 (0.73 to 3.83)	0.84 (0.68 to 1.05)	1.05 (0.74 to 1.49)
	40–65 (reference)
Marital status	Married	2.88 (0.39 to 21.07)	2.73 (0.35 to 21.06)	1.46 (0.74 to 2.89)	1.22 (0.59 to 2.54)
	Unmarried	1.25 (0.16 to 9.66)	1.34 (0.15 to 11.76)	0.84 (0.42 to 1.69)	1.24 (0.55 to 2.78)
	Other as referent
District	Western	1.41 (0.84 to 2.36)	1.32 (0.72 to 2.44)	0.91 (0.70 to 1.18)	0.87 (0.64 to 1.17)
	Central	1.56 (0.96 to 2.53)	0.70 (0.39 to 1.25)	1.55 (1.25 to 1.93)‡	0.98 (0.75 to 1.28)
	Eastern as referent
Level of medical institution	Primary	0.56 (0.35 to 0.89)†	0.75 (0.40 to 1.42)	0.56 (0.45 to 0.70)‡	0.76 (0.56 to 1.04)
	Secondary	0.49 (0.28 to 0.88)†	0.512 (0.25 to 1.03)	0.79 (0.62 to 1.00)	0.89 (0.65 to 1.21)
	Tertiary as referent
Profession	Clinician	2.75 (1.43 to 5.30)†	1.96 (0.96 to 3.99)	2.38 (1.81 to 3.13)‡	1.73 (1.28 to 2.35)‡
	Nurse	2.21 (1.08 to 4.48)†	3.45 (1.49 to 7.98)	1.37 (1.00 to 1.87)	1.73 (1.19 to 2.51)†
	Other medical profession (reference)
Education level	Associate’s degree or below	0.25 (0.11 to 0.54)‡	0.49 (0.16 to 1.48)	0.37 (0.27 to 0.51)‡	1.02 (0.64 to 1.65)
	Bachelor’s degree	0.84 (0.50 to 1.39)	0.99 (0.47 to 2.09)	0.68 (0.54 to 0.87)†	1.21 (0.85 to 1.71)
	Master’s degree or above (reference)
Professional title	Junior	0.28 (0.16 to 0.49)‡	0.34 (0.13 to 0.87)†	0.29 (0.22 to 0.37)‡	0.45 (0.29 to 0.69)‡
	Intermediate	0.72 (0.44 to 1.19)	0.77 (0.40 to 1.51)	0.54 (0.42 to 0.68)‡	0.65 (0.47 to 0.89)†
	Senior (reference)
Uncertainty stress	No	0.28 (0.19 to 0.43) ‡	0.52 (0.33 to 0.84)†	0.55 (0.45 to 0.68) ‡	0.90 (0.71 to 1.15)
	Yes (reference)
Years of work experience	<10 years	0.56 (0.34 to 0.91)†	0.45 (0.16 to 1.27)	0.57 (0.44 to 0.72)‡	0.85 (0.54 to 1.36)
	10–19 years	0.68 (0.40 to 1.18)	0.33 (0.14 to 0.77)†	0.99 (0.76 to 1.27)	0.91 (0.64 to 1.31)
	≥20 years (reference)
Occupational strain**	High	6.03 (3.38 to 10.76)‡	3.16 (1.65 to 6.03) ‡	3.11 (2.43 to 3.97)‡	1.67 (1.26 to 2.21)‡
	Medium	2.29 (1.20 to 4.40)†	1.87 (0.94 to 3.73)	2.05 (1.59 to 2.64) ‡	1.59 (1.21 to 2.09)‡
	Low (reference)
CAE-WPV	Extremely concerned	9.21 (2.18 to 38.95)†	3.95 (0.84 to 18.60)	10.15 (4.01 to 25.68)‡	5.85 (2.22 to 15.40)‡
	Concerned	2.57 (0.60 to 11.04)	1.35 (0.28 to 6.48)	6.98 (2.80 to 17.43)‡	4.20 (1.62 to 10.90)†
	Somewhat concerned	1.24 (0.29 to 5.30)	0.94 (0.20 to 4.42)	3.89 (1.57 to 9.65) †	2.80 (1.09 to 7.21)†
	Not concerned	0.44 (0.08 to 2.44)	0.36 (0.06 to 2.14)	1.06 (0.39 to 2.83)	0.90 (0.32 to 2.49)
	Not at all concerned (reference)
SR-WPVIs	Strongly opposed	4.30 (1.74 to 10.65)†	1.92 (0.64 to 5.78)	2.62 (1.54 to 4.47)‡	1.78 (0.94 to 3.36)
	Opposed	4.03 (1.82 to 8.94)‡	2.01 (0.79 to 5.07)	3.44 (2.26 to 5.22)‡	2.20 (1.34 to 3.59) †
	Neutral	2.54 (1.45 to 4.46)†	1.52 (0.79 to 2.95)	2.19 (1.67 to 2.88) ‡	1.53 (1.11 to 2.10)†
	Supported	0.73 (0.39 to 1.37)	0.73 (0.36 to 1.45)	1.21 (0.93 to 1.58)	1.16 (0.86 to 1.55)
	Strongly supported (reference)
SHPS-aWPV	Very dissatisfied	21.53 (7.15 to 64.80)‡	5.25 (1.43 to 19.23)†	12.42 (6.71 to 23.01)‡	4.48 (2.22 to 9.02)‡
	Dissatisfied	10.94 (3.98 to 30.001) ‡	2.89 (0.90 to 9.24)	7.53 (4.55 to 12.45)‡	2.56 (1.44 to 4.54) †
	Neutral	2.66 (1.04 to 6.82)†	1.48 (0.51 to 4.29)	3.03 (1.98 to 4.65)‡	1.66 (1.03 to 2.67)†
	Satisfied	1.12 (0.41 to 3.01)	1.04 (0.36 to 3.04)	1.63 (1.05 to 2.52) †	1.20 (0.75 to 1.92)
	Very satisfied (reference)

^*Models were adjusted for gender, age, marital status, district, level of medical institution, profession, education, professional title, uncertain pressure, years of work experience, occupational strain (the weighted average of standardised working hours and workload intensity indicators), CAE-WPV, SR-WPVIs and SHPS-aWPV.

^†p<0.05.

^‡p<0.001.

.AOR, adjusted OR; CAE-WPV, concern about experiencing WPV personally; HCPs, healthcare professionals; SHPS-aWPV, SHPS-aWPV, satisfaction with their hospitals’ procedures and systems for addressing WPV; SR-WPVIs, support HCPs reporting of WPV incidents; WPV, workplace violence.

Interaction effects of HCPs’ perceptions of HATWPV with hospital level and geographical region

Interaction analysis revealed that only a few specific interaction terms reached statistical significance (see online supplemental table S5 and online supplemental table S6). Regarding hospital level, dissatisfaction with procedures showed a stronger association with verbal threats in primary care institutions (interaction AOR=2.93, 95% CI 1.31 to 6.54), whereas a neutral attitude toward procedures was associated with a lower reporting rate of verbal threats in secondary hospitals (interaction AOR=0.53, 95% CI 0.29 to 0.94). Concerning geographical region, the association between dissatisfaction with procedures and physical violence was most pronounced in the eastern region (AOR=8.16, 95% CI 2.92 to 22.76). Thus, only for these specific dimensions did differences in HCPs’ perceptions of HATWPV vary across institution levels and geographical regions.

Attributable risk and population attributable fraction of key risk factors

Our population attributable fraction estimates indicate a strong association between hospital attitude variables and violence incidents, assuming a potential causal relationship and no unmeasured confounding given the cross-sectional design. CAE-WPV was associated with 64.9% of physical violence and 69.3% of verbal threats. Lack of SR-WPVIs was attributable to 29.9% of physical violence and 16.8% of verbal threats, while dis-SHPS-aWPV was associated with 28.1% of physical violence and 23.3% of verbal threats. The corresponding attributable risk values for each factor and violence type are presented in table 3.

Table 3. Attributable risk and population attributable fraction of key HCPs’ perceptions of hospital attitude factors for workplace violence.

Variables	Type of violence	AR (%)	PAF (%)
CAE-WPV	PV	2.58	64.87
	VT	14.76	69.25
SR-WPVIs	PV	2.86	29.85
	VT	8.62	16.75
SHPS-aWPV	PV	1.84	28.12
	VT	8.20	23.33

AR, attributable risk; CAE-WPV, concern about experiencing WPV personally; PAF, population attributable fraction; PV, physical violence; SHPS-aWPV, satisfaction with their hospitals’ procedures and systems for addressing WPV; SR-WPVIs, support HCPs reporting of WPV incidents; VTs, verbal threats.

Discussion

This multicentre cross-sectional study is the first to investigate the association between HCPs’ perceived HATWPV and the prevalence of WPV in Chinese medical institutions, considering three key dimensions: environmental safety concerns, institutional tolerance and organisational culture. The findings show that negative perceptions of HATWPV are associated with higher odds of WPV. Specifically, concerns regarding hospital environmental safety, perceived institutional tolerance of WPV and inadequate institutional support for reporting WPV were significantly associated with elevated risks of physical violence and verbal threats. These findings align with a recent nationwide online cross-sectional survey,⁶ and together suggest that healthcare institutional attitudes and systemic environments are associated with WPV occurrence.

Chinese HCPs in our study expressed widespread concern about WPV, with a reported prevalence of 79.6%. This proportion is notably higher than that reported in the National Emergency Medical Services Safety Study in the USA (25%),³⁰ and also higher than the perceptions of medical safety among Indian physicians (56.6%)³¹ and the 74.32% reported in another study of secondary medical institutions in China.³² This cross-national disparity may have three primary factors. First, methodological differences in study design may contribute to this variation. The US study partially relied on administrative databases for retrospective analysis, which may have underestimated the subjective perceptions of risk; additionally, it focused exclusively on emergency HCPs, whereas the present study encompassed a broader sample of healthcare workers. Second, structural disparities in healthcare systems contribute to these discrepancies. The Indian study highlighted challenges related to resource allocation between public and private hospitals,¹⁹ which contrasts with China’s predominantly public healthcare infrastructure.

A positive graded association exists between CAE-WPV and verbal threats, which remained robust after adjusting for pandemic-related factors such as uncertainty stress and workload. The population attributable fraction results indicate that CAE-WPV was associated with 64.9% of physical violence and 69.3% of verbal threats, consistent with a strong association between perceived environmental safety and the prevalence of WPV without considering causality or confounding factors. The strength of this association notably exceeds that observed among emergency HCPs in the USA.³³ This disparity may be attributed to the prolonged exposure of Chinese HCPs to high-pressure work environments.³⁴ High-intensity workloads can trigger amygdala-mediated stress responses and activate the hypothalamic-pituitary-adrenal axis. The consequent elevation in cortisol and norepinephrine levels may induce fear, anxiety and defensive aggression.³⁵ The correlation between hypothalamic-pituitary-adrenal axis activation and aggressive behaviour has been substantiated by numerous experimental studies.³⁶ Furthermore, concerns regarding the safety of the medical environment may threaten the long-term stability and sustainability of the healthcare workforce by undermining professional identity and the willingness for intergenerational transmission. Collectively, these findings underscore the necessity of adopting a bio-psycho-social integrative model for the prevention and management of WPV.³⁷

This study revealed that 48.3% of HCPs expressed dissatisfaction with the hospital’s system and procedures for handling WPV, a proportion lower than that reported in the USA (57.4–76.7%)³⁸ and India (76.7%).³⁹ Notably, physical violence was highly concentrated in the highly dissatisfied population, while the prevalence of verbal threats exhibited a marked dose-response increase with the level of dissatisfaction toward procedures and systems. This observation aligns with the findings of Xiao et al⁶ The perceived ‘high dissatisfaction–high tolerance’ pattern showed a graded association between negative views of institutional responses and the experience of violence. While causality cannot be inferred from cross-sectional data, this association suggests that perceptions and experiences may reinforce each other within the institutional context. Notably, in Chinese practice, despite regulatory improvements, a perceived ‘stability-first’ culture often favours conciliatory dispute resolution.⁴⁰ This may create a gap between formal norms and perceived enforcement, further entrenching the sense of inadequate support and its association with WPV risk. This perceived institutional approach resembles the ‘bureaucratic silence’ noted elsewhere,⁴¹ while in China, financial compensation is often perceived as substituting for deeper systemic reform.⁴²

This study found that after adjusting for pandemic-related factors, HCPs’ perceptions of inadequate institutional support for reporting WPV incidents remained significantly associated with the prevalence of WPV. Approximately 33% of HCPs reported that their institutions do not actively encourage the reporting of WPV incidents—a finding consistent with nursing research conducted in the USA and Canada^{43 44} as well as survey results from Israel.⁴⁵ Unlike studies focusing on reporting complexity, lack of follow-up or fear of retaliation,^{38 44 46 47} this and some domestic surveys reveal a notable perception: a considerable proportion of HCPs believe their institutions lack formal reporting systems or prioritise patient satisfaction over staff safety.⁴⁸ This perception may stem from the performance-based management model prevalent in China’s healthcare system, which emphasises service volume and patient satisfaction.^{49 50} Such a model may foster a widespread belief among HCPs that rapid conflict resolution and maintaining ‘stability’ are prioritised over transparent reporting of violence,⁵¹ creating a tension with WHO’s advocacy for transparent, non-punitive reporting mechanisms.^{52 53} To improve the reporting environment, medical institutions should systematically assess and address HCPs’ perceptions, mitigate potential reporting inhibitions and establish trusted, non-punitive reporting and support systems, reinforced by independent oversight to enhance credibility.

Our interaction analysis further indicated that the association between HCPs’ perceptions of HATWPV and WPV prevalence varied by hospital level and geographical region. In primary care institutions, staff’s perceived dissatisfaction with procedures showed a stronger association with verbal threats, possibly reflecting heightened sensitivity to procedural shortcomings in resource—constrained settings with weaker perceived institutional support. Although the eastern region is more economically developed, perceived dissatisfaction with procedures was most strongly associated with physical violence there. One possible explanation is that in contexts with higher patient expectations, greater population mobility or heavier workloads, perceived inadequacy in institutional responses may amplify HCPs’ own stress or unease and this perceptual gap could be linked to more intense forms of conflict.

Based on these findings, the following policy-level considerations may be offered. First, prior to implementing interventions, healthcare institutions could assess HCPs’ perceptions of safety, identify potential weaknesses and conduct adaptive pilot programmes tailored to different hospital levels and regional resources and cultures. Second, strengthened financial support might reduce institutional over-reliance on service volume,⁵⁴ while measures such as optimising shift scheduling, staggered rest and deploying auxiliary staff during peak hours may help lower workload. These adjustments are expected to alleviate work intensity and improve HCPs’ perceptions of institutional support. Third, building on the Health Workers’ Safety Charter,⁵⁵ the feasibility of establishing a mandatory reporting system (eg, the California model)⁵⁶ and a standardised response framework (eg, Canada’s SEIPS V.3.0)^{57 58} could be explored within the local context, alongside ongoing training and communication. These measures may be associated with HCPs’ awareness, trust and willingness to use such mechanisms, and might also relate to perceived institutional support. Fourth, incorporating HCPs’ safety perception and violence management indicators into hospital evaluation systems could be further examined to explore their association with policy implementation and actual outcomes.^{9 59} Finally, establishing a psychological support system that includes screening, training and counselling might help convey organisational care, which may in turn be associated with HCPs’ overall sense of safety and work environment.⁶⁰

Limitations of the study

This nationwide cross-sectional study has several limitations. First, the cross-sectional design precludes causal inference between HCPs’ perceived HATWPV and WPV prevalence; longitudinal or intervention studies are needed to explore potential causal pathways. Second, although this study classified participants who experienced both physical violence and verbal threats into the physical violence category based on the clinical and legal priority of physical violence, this may underestimate the prevalence of verbal threats. Third, self-reported data may introduce recall, social desirability and self-selection biases, potentially underestimating true prevalence. Fourth, participants from the eastern region were slightly over-represented, which aligns with the macro-distribution trend of China’s health workforce. Nonetheless, future studies should include more provinces and collect regional-level covariates to better elucidate the role of regional contextual factors. Fifth, perceived HATWPV cannot distinguish whether negative attitudes stem from institutional resource shortages or policy implementation failures; future research should integrate objective data (eg, security facilities, patient volume, policy records). Sixth, data were collected after the relaxation of COVID-19 controls. Although pandemic-related covariates were adjusted for, residual confounding (eg, department type, patient volume) may remain. However, the clear dose-response relationship between HCPs’ perceived HATWPV and WPV prevalence suggests a structural association rather than short-term disturbance.

Conclusion

This cross-sectional study indicates an independent association between HCPs’ perceived institutional tolerance and WPV prevalence. Although causality cannot be inferred, the strength and consistency of this association suggest that healthcare institutions may need to evaluate their current attitudes and systems regarding violence prevention. Future longitudinal studies could help explore the underlying mechanisms and causal pathways. Based on these findings, institutions might consider assessing HCPs’ perceptions of institutional attitudes and, in response, explore more transparent support mechanisms, a non-punitive organisational culture and multitiered psychological support systems. Systematically addressing and improving HCPs’ perceived experiences and the institutional environment may represent one approach toward a safer healthcare workplace.

Data availability statement

Data are available upon reasonable request.