Working intentions of medical students in response to healthcare workplace violence and descending resources reform in China.

BACKGROUND: In order to curb healthcare workplace violence (WPV) and better allocate healthcare resources, China launched the descending resources reform in 2013 and tightened the anti-violence legal environment simultaneously. Medical students are expected to reconsider their working intentions of entering the medical market (inter-market effect) and choosing high- or low-level hospitals (intra-market effect) in response to the evolving WPV. The goal of this study was to explore the link between the perceived WPV incidence and medical students' willingness to practice medicine in the context of China's descending resources reform.
METHOD: Medical students were selected with cluster sampling from 8 medical colleges in Zhejiang Province, China, and 1497 valid questionnaires were collected by using a five-point unbalanced scale, to perform cross-sectional empirical research using the ordered logit model (OLM).
RESULTS: The perceived WPV incidence negatively correlate with the willingness of medical students to practice medicine but positively correlate with their willingness to practice in low-level hospitals, indicating the existence of inter- and intra-market effects. The anti-violence legal environment has no direct link with working intention but contributes to the perceived decline in the incidence of violence. Descending resources reform has simultaneous opposite effects on medical students, with the coexistence of prudent motives driven by reform costs and optimistic expectations of sharing external benefits.
CONCLUSIONS: Safety needs and risk aversion motive play an important role in medical students' career choice when facing severe WPV. Tightening of the anti-violence legal environment and the descending resources reform could drive medical students to low-level hospitals.

Introduction

Over the past two decades, the uneven allocation of resources in China’s medical market and resulting healthcare workplace violence (WPV) have received widespread attention in academia and politics [1-3]. Since 2013, China has promoted descending resources reform to reallocate health resources from high-level (tertiary) to low-level (primary) hospitals, i.e., the government forces and/or subsidizes doctors from high-level hospitals to move down to low-level hospitals, to improve the latter’s diagnostic and treatment capabilities through human capital spillover, thus attracting more patients [4]. In addition, since 2014, the high incidence of WPV has impelled the Chinese government to enact legislation and enforcement efforts to curb such behavior [5]. WPV would drive healthcare workers and medical students to leave health labor market, or seek jobs with lower risk and/or higher income among different level hospitals. Meanwhile, descending resources reform is designed to rebalance the healthcare market [6], and improvement in the legal environment is expected to curb WPV, thus influencing the willingness of medical students to practice medicine. This paper is the first to explore the link between WPV and medical students’ willingness to practice medicine in the context of China’s descending resources reform.

Recruiting and retaining medical students in areas of workforce need is of international concern. One branch of research tends to focus on the perspectives of employers [7, 8], i.e., approaches for selecting and recruiting medical students, while another branch tends to concentrate on factors influencing career choice from the perspective of medical students. In addition to factors such as parental influence, personality, medical education, and career advice and guidance [9-11], WPV and working intention responses of medical students have received increasing attention [12-14]. WPV is a widespread phenomenon that affects medical students and impedes future retention rates [12], resulting in negative career choice satisfaction and exit from the healthcare profession [13, 15]. Research suggests that there may be a direct link between episodes of violence and poor recruitment and retention rates [10], and that intense doctor-patient relationships can impact career choice among medical students, leading to an increase in students switching to other majors with lower violence risk, e.g., auxiliary medical departments [16]. However, more empirical evidence is needed to better understand the responses of medical students to WPV.

In a two-stage career selection process, i.e., high-school student--medical student--healthcare worker, practicing medicine results in higher factor specificity compared with other occupations [17]. In the context of deteriorating doctor-patient relationships and increasing WPV incidence in China, a decline in the willingness to study medicine has been reported in high-school students, leading to the lowering of admission scores and quality of students entering medical schools [18, 19]. Hu and Ye et al. separately reported that 94.61% of respondents and > 70% of doctors do not want their children to study medicine [20, 21]. WPV not only negatively impacts the intention of students to practice medicine, but also increases the intention of doctors to leave their jobs [22, 23]. If medical students decide to enter healthcare labor market, their choices are impacted by working conditions, salary, and other factors [24]. Medical students are more inclined to work in cities [25-28] and at higher-level hospitals that have better working conditions and higher salaries [29, 30]. Nevertheless, doctor-patient relationships and safety conditions are much better in low-level hospitals due to the lower incidence of violence [31, 32]. To date, however, very limited empirical studies have discussed medical students’ risk aversion effects of job-seeking.

The core research question is the working intention responses of medical students to WPV. Given other conditions, it could be assumed that, for the sake of personal safety, medical students may consider leaving the healthcare labor market or seeking employment in workplaces with lower safety risk. However, to identify the marginal effects of WPV on career choice, we must control for the effects of other policy variables that may influence medical student career choice. In this paper, we explored two main policy variables, i.e., changes in anti-violence laws and introduction of descending resources reform. Notably, changes in the anti-violence legal environment may affect medical student perception of the incidence of WPV and the degree to which their safety needs are fulfilled. Furthermore, the expected costs and benefits of descending resources reform in China, which asks young doctors from high-level hospitals to move to low-level or rural hospitals for several months or years, may also impact the career choices of medical students.

In this paper, we recognized that medical students will choose if they would enter healthcare labor market or not, and decide which level of hospitals they will choose to work as the response of WPV. Compared with previous research, this paper used questionnaire data to estimate the link between WPV and the working intentions of medical students in the context of the latest health resource reallocation policies. We also investigated WPV and the legal environment to estimate their impacts on the intention of medical students to practice medicine. This research provides evidence on medical students’ response to the descending resources reform, healthcare workplace violence, and the anti-violence legal environment of China in recent years.

Background

Descending resources reform

China’s public hospitals are divided into three grades. Grade one and two hospitals are usually owned by the government at or below the county level, and are defined as low-level hospitals; while tertiary grade hospitals are usually located in major cities, and are defined as high-level hospitals [4]. Prior to 2009, since the Chinese government reduced its investment and medical insurance coverage, the unbalanced allocation of health resources led to serious inequality, doctor-patient conflicts, and harmful social consequences. In 2009, new healthcare reform was established with a focus on abolishing medicine markups and adjusting medical service fees to correct distorted incentives for over-prescription and over-examination in response to public complaints about excessive medical costs [33].

However, the 2009 reform did not affect patient preference for high-level hospitals, and thus high-level hospital congestion and low-level hospital idleness remained problematic [34]. There are two main explanations: The first is that past investment focused on infrastructure rather than human capital; the treatment ability of low-level hospitals was not improved [3, 33], so low-level hospitals failed to address the discrepancy in human capital with high-level hospitals and win back patient trust. The second is that past reforms mainly impacted patient affordability from the perspective of demand, but not biased resource allocation or choice of care providers from the perspective of supply, thus medical students were reluctant to seek jobs at low-level hospitals. Therefore, the under-utilization and reduced capacity of low-level hospitals remain important challenges in China [34].

In 2013, the descending health resources reform was launched in Zhejiang, a pilot province in China undergoing comprehensive medical improvement. In 2014, the Chinese President, Xi Jinping, bolstered the measures implemented in this reform [35]. In subsequent years, this reform has gradually expanded to other pilot provinces and has become one of the core health policies of China for the reallocation of medical resources between rural (county and town) and urban (major city) areas. As provincial governments in China own both high-level and low-level hospitals, this reform mandatorily requests that doctors from high-level hospitals move to low-level hospitals for a certain period (usually a few months to 1 ~ 2 years), and asks for high-level hospitals to establish inter-hospital cooperative ties with low-level hospitals. At the same time, the government encourages medical students to work at low-level hospitals through targeted-area student policies and employment subsidies (Fig.1). The government also provides subsidies for all hospitals involved in the reform of medical resources reallocation.

Fig. 1

Descending resources reform and intended incentive structure

According to the Zhejiang government [36], descending health resources reform can be summarized as follows. Firstly, for high-level hospitals, all province-level general hospitals/specialty hospitals should establish comprehensive trusteeships and ties with 2-4 county-level hospitals. No less than 5% of qualified doctors with intermediate or above professional titles should move to low-level hospitals. More than 80% of reallocated doctors should have intermediate or above professional titles. Secondly, for medical students, no less than 1000 targeted-area students should be trained to work in low-level hospitals each year. Low-level hospitals should recruit 10,000 medical students per year. Lastly, the government would provide financial subsidies for all level hospitals involved in the reform to (at least partially) compensate for their reform costs, and also provide tuition compensation for medical students who join low-level hospitals and facilitate income increase and job promotion.

Congestion in China’s medical market and healthcare workplace violence

Sun et al. provided a detailed description of the congestion in the Chinese medical market [17], describing the coexistence of overcrowding in high-level city hospitals and underutilization of resources in low-level hospitals. This structural congestion has contributed to the long-term decline in treatment satisfaction and deterioration of the doctor-patient relationship, which have, in turn, evolved into an escalation in medical conflicts and frequent violence against healthcare workers. However, doctor-patient conflicts and healthcare workplace violence are unevenly distributed in the medical market: high-level hospitals with congestion suffer from higher dissatisfaction and incidences of disputes and violence [31], whereas low-level hospitals experience low resource utilization and limited violence because they can transfer high-risk patients to high-level hospitals.

Healthcare workplace violence is a global concern and affects healthcare workers and even medical students in nearly all health-care settings [37-39], however, the scale, frequency, and viciousness of attacks on healthcare workers in China are particularly severe [2]. One survey found that more than 67% of violent incidences reported by the media from 2000 to 2015 were related to tertiary (high-level) hospitals, with low-level hospitals experiencing lower incidences of violence [32]. Furthermore, more than 50% of violent events resulting in casualties in 2015-2016 occurred in tertiary hospitals and more than 40% occurred in emergency services, which were mainly caused by congestion rather than triggered by therapeutic effects [31]. In addition, previous research has stated that violent events in hospitals are often unreported, with tertiary hospitals having the highest tolerance rate for unreported violence [40]. It is also reported that 18.5% medical students have witnessed healthcare workplace violence [41]; most medical students receive violence-related information from hospital and news media and give negative responses [42, 43].

The serious healthcare workplace violence disclosed by the media impelled the Chinese government to issue the Opinions on Punishing Medical-related Crimes and Maintaining Order in Hospitals in 2014, which demanded the strengthening of violence prevention and control, as well as more severe punishment for perpetrators. In 2020, China’s top legislator approved comprehensive laws for protecting healthcare workers, which ban any organization or individual from threatening or harming healthcare workers. However, whether China has experienced an improvement in doctor-patient relationships and the incidence of healthcare workplace violence in the context of the current descending resources reform and anti-violence legal environment remains unclear. Thus, in the current study, we explored the link between healthcare workplace violence and medical students’ working intentions in the context of the latest reform and anti-violence legal environment.

Methods

Theory and hypotheses

We used Maslow’s hierarchy of needs theory as the core theory of this paper. According to this concept, humans have five basic needs in a hierarchical form from low to high, i.e., physiology, security, society, esteem, and self-actualization [44]. Medical students’ willingness to practice medicine is a labor market choice behavior associated with their expected cost-benefit analysis when making career choices. This analysis is determined by the expected satisfaction of different levels of needs brought by medical practitioners [45].

The conceptual models and hypotheses of this study are shown in Fig. 2. Based on China’s long-term policies in medical-price control and the government’s dominant role in hospital systems, certain degrees of satisfaction of Maslow’s hierarchy of needs (physiological, social, esteem, and self-actualization) can be treated as given. Since the incidence of violence regarding safety needs changes dynamically, medical students would reconsider their labor market choice by using all available violence-related information. As income cannot compensate for high violence risk, if the incidence of healthcare workplace violence is low or similar to the average experienced in different workplaces, it should not affect the willingness to practice medicine; however, if it is far beyond the average level, it should have a negative impact on the willingness to practice medicine (Hypothesis 1a, H1a; shown in Fig. 2). This negative impact can worsen due to the specificity of factors [4], which refers to the inflexibility to switch to other industries or occupations because of occupational lock-up after medical training. In this paper, we divided employers into two categories, i.e., high-level and low-level hospitals: if students realize that violence mainly occurs in high-level hospitals, their marginal willingness to work in low-level hospitals may increase compared to their willingness to work in high-level hospitals (Hypothesis 1b, H1b).

Fig. 2

Conceptual model of medical students’ working intention

Hypothesis 1a: Increased incidence of healthcare workplace violence reduces medical students’ willingness to practice medicine (inter-market effect).

Hypothesis 1b: Increased incidence of healthcare workplace violence increases medical students’ willingness to work in low-level hospitals (intra-market effect).

The anti-violence legal environment involves ex ante and ex post levels and includes government, legislation, law enforcement, and hospital-related factors. The function of government in dealing with healthcare workplace violence involves two aspects, i.e., ex-ante prevention/protection and ex-post response. The role of hospitals is limited to ex-ante prevention. Regarding legislation and law enforcement, practitioners can evaluate whether current punishments for violent perpetrators are appropriate or not, whereas law enforcement is usually involved in ex-post responses after events. However, the above anti-violence legal environment is more likely to be an indirect indicator and intermediate variable affecting medical students’ willingness to practice medicine (Hypothesis 2a, H2a). The direct indicator is the effect of the existing legal environment, i.e., incidence of and trends in healthcare workplace violence, which directly affect medical student satisfaction with safety needs (Hypothesis 2b, H2b).

Hypothesis 2a: Anti-violence legal environment does not affect medical students’ willingness to practice medicine.

Hypothesis 2b: Anti-violence legal environment reduces incidence of healthcare workplace violence.

Healthcare workplace violence and the anti-violence legal environment interact with the exogenous constraints of existing health policies. Although concern may arise whether medical students are aware of the reform or not, we believe that students will use all available information to understand and respond to the reform. So, descending resources reform may involve two policy effects. The first is a negative effect on medical students who wish to practice medicine (Hypothesis 3a, H3a). The main reason is that doctors from high-level hospitals are required to move to low-level hospitals and to bear the reform costs. The usual practice is that they must leave in-service hospitals and workplaces for a few months to 1 year and work in low-level hospitals far from the city; young doctors are also often the ones that are reallocated, thus increasing their cost of living and causing inconvenience. Student awareness of this policy may weaken their willingness to practice medicine. As reform could help to alleviate congestion in high-level hospitals and idle resources in low-level hospitals, the external benefit of reform could be shared by students and trigger their willingness to practice medicine (Hypothesis 3b, H3b).

Hypothesis 3a: Awareness of descending resources reform has a negative impact on medical students’ willingness to practice medicine.

Hypothesis 3b: The realization of the goal of descending resources reform has a positive impact on medical students’ willingness to practice medicine.

Questionnaire design and variables

Based on Maslow’s hierarchy of needs theory, a five-point unbalanced scale was used in this paper. In addition to demographic items, variables on working intention, reform cognition, working environment, and policy evaluation were included in the questionnaire (Table 1). In our research, working intention included the desire to practice medicine or to practice medicine in low-level hospitals after reform, which were used to measure inter- and intra-market effects, respectively.

Table 1

Questionnaire design, variables, and definitions

Category	Question item	Variable	Definition
Working intention	Your desire to practice medicine after reform	Y	1 decrease, 2 no change, 3–5 slight, significant, and very significant increase, respectively
	Your desire to practice medicine in low-level hospitals after reform	Y1	1–5 variables defined as Y
Demography	Your education level	Education	1 technical degree, 2–5 college, bachelor, Master, and PhD degree, respectively
	Your grade	Grade	1–5 variables for different grades
	Are you a targeted-area student	Target	1 for Yes, 0 for No
Working environment	Your perceived level of respect for doctors	Respect	1–5 very low, low, moderate, high, and very high, respectively
	Your frequency of concern about violence	Concern	1 never, 2 rarely, 3–5 occasional, frequent, and high, respectively
	Your perceived incidence of violence	Violence	1–5 very low, low, moderate, high, and very high, respectively
	Your perceived improvement in incidence of violence after reform	Y2	1–5 very low, low, moderate, high, and very high, respectively
Policy	Your degree of recognition of reform	Recogn	1–5 very low, low, moderate, high, and very high, respectively
	Perceived policy effect	Policy	1-5 negative, no effect, positive, good, and very good, respectively
	health policy or anti-violence lesson	Lesson	1 for Yes, 0 for No
Legal environment	Government’s violence prevention	Legal_A	1–5 very weak, weak, moderate, strong, and very strong, respectively
	Government’s response to violence	Legal_B	1–5 variables defined as Legal_A
	Legislative punishment for violence	Legal_C	1–5 variables defined as Legal_A
	Anti-violence law enforcement	Legal_D	1–5 variables defined as Legal_A
	Hospitals’ protection for medical staff	Legal_E	1–5 variables defined as Legal_A

The working environment was first measured based on the perceived level of respect for doctors and then violence-related items, which included frequency of concern about violence, perceived incidence of violence, and perceived improvement in incidence of violence after reform. For reform policy, four items were included: degree of recognition of reform, perceived policy effect, health policy, and anti-violence education. The anti-violence legal environment was measured based on government efforts, legislative punishment, law enforcement, and hospital protection.

Before the survey, the questionnaire used in this paper has experienced pre-release, expert review and revision. The survey was performed from August 2018 to September 2019. Medical students were selected with cluster sampling from 8 medical schools in different cities of Zhejiang Province, China. And all respondents provided informed consent to participate. We obtained 1641 questionnaires, of which 1497 were valid (effective rate of 97.33%).

Due to the differences among question items, we first set the different evaluation items from low to high and then converted them to 1–5 variables and converted questions with responses of only Yes/No to 1–0 variables. The correlation coefficients between measurement variables were all lower than 0.60. By using SPSS 23.0 and reliability tests, the Cronbach’s α coefficient was 0.804; and the Kaiser-Meyer-Olkin (KMO) value was 0.925, with Sig. = 0.000 for the Bartlett spheroid test, indicating that the scale and data were suitable for empirical analysis.

Empirical model

Based on the above theory and hypotheses, we used Violence and Legal variables to depict impact on safety needs and corresponding legal environment changes, respectively. The variable Respect was used to measure the effect of esteem needs. The variable Policy was used to incorporate other needs in Maslow’s theory. As targeted-area students all come from rural (county and town) areas, resulting in a high correlation between the student origin (urban/rural) and Target variables; meanwhile, grade and age also has high correlation, we did not include student origin and age in the empirical model. The demographic variables (Demographic) included Target, Education, and Grade. Health education, Recog, and Concern were also added to the model as control variables (Control). The empirical model is as follows:

As the questionnaire data were discrete and most variables were ordered one, the ordered logit model (OLM) was utilized to estimate the model [46]. To test the robustness of the regression results, we used stepwise regression technology, and also chose different legal environment variables for empirical testing. All methods were carried out in accordance with relevant guidelines and regulations.

Results

Descriptive analysis

Table 2 shows the demographic distribution of samples. Undergraduate students and non-targeted students constituted the largest proportions, accounting for 79.09 and 94.99%, respectively. Students from urban areas and non-Zhejiang provinces accounted for 39.21 and 33.27%, respectively. Student participants ranged from 14.83% in the first year to 24.25% in the fourth year.

Table 2

Demographic distribution (sample size = 1497)

Student origin	Number (%)	Education level	Number (%)	Grade	Number (%)	Targeted-area student	Number (%)
Urban area	587 (39.21)	Technical	1 (0.07)	One	222 (14.83)	Yes	75 (5.01)
Rural area	910 (60.79)	College	18 (1.20)	Two	308 (20.57)	No	1422 (94.99)
		Bachelor	1184 (79.09)	Three	357 (23.85)
Zhejiang Province	999 (66.73)	Master	251 (16.77)	Four	363 (24.25)
Other provinces	498 (33.27)	Ph.D.	46 (3.07)	Five	247 (16.50)

As shown in Table 3, the perceived incidence of violence was high, with scores of very high (5) and high (4) accounting for 64.67% of the total sample. In total, 41.82% of respondents chose very high as their perceived incidence of violence, whereas only 11.55% believed the incidence of violence to be low. However, respondent Concern was not high, with only 31.93% of participants paying high or frequent attention to violence and 50.43% of participants paying occasional attention to violence. Furthermore, 41.35% of respondents chose the very low (score of 1) recognition option for reform, with 37.74% of respondents believing that they knew something about this issue; 47.53% of respondents did not have any health policy course/knowledge and 60.39% of respondents received no anti-violence education.

Table 3

Evaluation and health education of violence and reform (%)

Evaluation	Violence	Concern	Recogn	Education	Policy	Anti-violence
5	41.82	11.56	6.28	Course	8.45	7.01
4	22.85	20.37	12.42	Teacher	19.02	13.05
3	24.11	50.43	19.04	Lecture	19.37	20.05
2	7.01	13.63	21.24	Self-learn	5.63	6.46
1	4.54	4.34	41.35	No	47.53	53.43

5 is highest score of evaluation and 1 is lowest

According to Table 4, we divided legal environment and policy evaluations into positive (score of 4-5), neutral (score of 3), and negative (1-2). The positive and negative evaluations of the legal environment were 18–20% and 46–54%, respectively. For reform, 56.24% of respondents gave a positive evaluation (3-5), with the remaining respondents giving a neutral (score of 2, 39.01%) or negative (1) evaluation.

Table 4

Evaluation of anti-violence legal environment and reform (%)

Evaluation	Legal_A	Legal_B	Legal_C	Legal_D	Legal_E	Policy
5	5.34	6.95	7.55	7.15	5.68	8.28
4	12.89	11.22	10.62	11.22	14.76	17.43
3	28.39	30.33	29.93	33.27	33.27	31.20
2	45.29	45.49	44.36	41.75	40.41	39.01
1	8.08	6.01	7.55	6.48	5.88	4.07

5 is highest score of evaluation and 1 is lowest

Working intention response to health labor market

Test result of H1a. As seen in Table 5, Violence negatively affected willingness to practice medicine at the 1% significance level with robust results, thus confirming Hypothesis 1a. Results showed that healthcare workplace violence, as an exogenous variable that negatively influences safety needs, was not conducive to the choice to practice medicine. The variable Respect, which represents the esteem needs of medical students, did not significantly impact the willingness to practice medicine. This indicates that respect for healthcare workers is not essential for working intention.

Table 5

Estimation results of OLM regression of medical students’ working intention

Dependent variable: working intention Y (1 for decrease, 2 for no change, 3-5 for slight increase, significant increase, and great increase)
Variable	Model 1: Legal_A	Model 2: Legal_B	Model 3: Legal_C	Model 4: Legal_D	Model 5: Legal_E
Target	0.8102*** (0.2150)	0.7876*** (0.2154)	0.7709*** (0.2156)	0.7776*** (0.2149)	0.7932*** (0.2146)
Education	0.6571*** (0.1031)	0.6532*** (0.1031)	0.6572*** (0.1029)	0.6499*** (0.1032)	0.6327*** (0.1032)
Grade	−0.0934** (0.0373)	− 0.0890** (0.0373)	− 0.0864** (0.0373)	− 0.0878** (0.0372)	− 0.0870** (0.0372)
Violence	− 0.4389*** (0.0456)	− 0.4364*** (0.0457)	− 0.4342*** (0.0457)	− 0.4355*** (0.0456)	− 0.4320*** (0.0457)
Concern	0.2329*** (0.0558)	0.2306*** (0.0559)	0.2291*** (0.0559)	0.2273*** (0.0560)	0.2160*** (0.0562)
Lesson	−0.1518 (0.1060)	− 0.1560 (0.1059)	− 0.1608 (0.1060)	−0.1683 (0.1064)	− 0.1733 (0.1062)
Respect	0.0186 (0.0622)	0.0071 (0.0613)	−0.0038 (0.0617)	0.0005 (0.0614)	−0.0213 (0.0621)
Legal	−0.0166 (0.0548)	0.0434 (0.0527)	0.0835 (0.0535)	0.0808 (0.0545)	0.1537*** (0.0575)
Recogn	−0.1322*** (0.0499)	−0.1345*** (0.0424)	− 0.1392*** (0.0425)	−0.1382*** (0.0425)	− 0.1453*** (0.0426)
Policy	0.2269*** (0.0545)	0.2202*** (0.0540)	0.2103*** (0.0545)	0.2090** (0.0548)	0.2055** (0.0542)
LR stat	320.6237***	321.2077***	322.9627***	322.7065***	327.6668***
Obs	1497	1497	1497	1497	1497

[1] ***,** and* denote significance level of 1, 5 and 10%, respectively; [2] Clustered standard error in parentheses

Test result of H2a. The variable Legal also had no significant impact on the willingness to practice medicine. These results provide evidence for Hypothesis 2a, i.e., that medical students treat the incidence of violence as the determinant factor affecting working intention, with the legal environment being less important. To test the robustness of the above results, we introduced five different legal environment variables to model 1-5, respectively. It is found that Hypothesis 2a could be supported consistently and robustly.

Test results of H3a and H3b. The variable Policy had a significantly positive effect on the intention to practice medicine; however, Recogn had a significantly negative impact, fully supporting Hypotheses 3a and 3b. These results demonstrate that the direct costs and external benefits of reform coexist and are recognized by medical students. Therefore, improvement in recognition failed to enhance the willingness to practice medicine.

For health policy education and demographic variables, our results implied that targeted-area medical students had significantly stronger willingness than that of non-targeted students. That is, cost-locking based on contracts had an incentive effect on the desire to practice medicine. We also found that the higher the level of education, the stronger was the willingness to practice medicine. This result shows that higher level education increases factor specificity and opportunity costs, and thus a higher intention to practice medicine would be rational. The variable Grade was significantly negative, suggesting that, under a certain education level, the higher the grade, the higher the risk of the choice behavior.

We reported two different regression results to test the robustness. Five different legal environment variables are introduced into the empirical models in Table 5, and then stepwise estimation results were reported in Table A1. All results verified the robustness of the above test results. Besides, the demographic variables Target, Education, and Grade all had the same significant impact. The results of health policy education were non-significant, indicating that health policy education does not link to the working intention directly.

Working intention response to low-level hospitals

Test result of H1b. As shown in Table 6, the perceived incidence of violence had a positive impact on the desire to work in low-level hospitals, fully supporting the existence of intra-market effect shown in Hypothesis 1b. That is, considering the constraints of factor specificity and differences in the incidence of violence among different level hospitals, it is rational to choose to work in low-level hospitals to better satisfy safety needs.

Table 6

OLM estimation results of medical students’ intention to work in low-level hospitals

Dependent variable: working intention for low-level hospital Y1 (1 for decrease, 2 for no change, 3-5 for slight increase, significant increase, and great increase)
Variable	Model 1: Legal_A	Model 2: Legal_B	Model 3: Legal_C	Model 4: Legal_D	Model 5: Legal_E
Target	0.7949*** (0.2242)	0.7826*** (0.2235)	0.7884*** (0.2244)	0.8199*** (0.2247)	0.8796*** (0.2242)
Education	0.4278*** (0.1062)	0.4082*** (0.1062)	0.4178*** (0.1066)	0.4109*** (0.1061)	0.3736*** (0.1071)
Grade	0.0165 (0.0384)	0.0084 (0.0383)	0.0038 (0.0383)	0.0031 (0.0382)	−0.0127 (0.0380)
Violence	0.1285*** (0.0449)	0.1230*** (0.0448)	0.1223***(0.0448)	0.1195*** (0.0449)	0.1185*** (0.0448)
Concern	0.1693*** (0.0570)	0.1539*** (0.0572)	0.1561*** (0.0569)	0.1447** (0.0569)	0.1543*** (0.0571)
Lesson	0.1329 (0.1052)	0.1371 (0.1052)	0.1466 (0.1049)	0.1238 (0.1052)	0.1579 (0.1048)
Respect	0.3058*** (0.0645)	0.3583*** (0.0637)	0.3479*** (0.0641)	0.3572*** (0.0639)	0.3534*** (0.0645)
Legal	0.4777*** (0.0568)	0.4356*** (0.0540)	0.3581*** (0.0537)	0.4115*** (0.0561)	0.2945*** (0.0581)
Recogn	0.0693* (0.0418)	0.1071** (0.0416)	0.1009** (0.0415)	0.0944** (0.0415)	0.0914*** (0.0416)
Policy	0.1173** (0.0537)	0.1359** (0.0536)	0.1245** (0.0537)	0.0959* (0.0541)	0.1381*** (0.0534)
LR stat	282.0510***	275.9580***	254.5599***	264.4470***	235.6958***
Obs	1497	1497	1497	1497	1497

[1] ***,** and* denote significance level of 1, 5 and 10%, respectively; [2] Clustered standard error in parentheses

By introducing different legal environment variables and using stepwise regression technology in Table 6 and Table A2, the test results of H1b could be proved robust. For the legal environment in Table 6, the five variables had significant positive impact. Thus, although the legal environment does not improve the overall intention to practice medicine, it plays a role in shifting the intention to low-level hospitals in the medical market.

Regarding the policy effect of reform, Table 6 provides obvious evidence of positive correlation between the intention to practice in low-level hospitals and perceived reform effect. And Recogn also has a positive link with intention significantly, showing that students need not bear the reform costs if choosing to work at low-level hospitals.

Impact of legal environment on reducing violence

Test result of H2b. As seen in Table 7, all legal environment variables caused a significant perceived decline in violence, confirming Hypothesis 2b. As shown in Table 5, Violence negatively impacted the willingness to practice medicine, and thus we inferred that an improvement in the legal environment could contribute to a decline in violence, and indirectly affect the working intention of medical students.

Table 7

OLM estimation results of perceived improvement in healthcare workplace violence

Dependent variable: perceived improvement in incidence of violence Y2 (1–5 variables from Very low, Low, Moderate, High, Very high)
Variable	Model 1: Legal_A	Model 2: Legal_B	Model 3: Legal_C	Model 4: Legal_D	Model 5: Legal_E
Target	0.8491*** (0.2246)	0.8391*** (0.2250)	0.7442*** (0.2266)	0.8335*** (0.2267)	0.8821*** (0.2272)
Education	0.4095*** (0.1038)	0.4004*** (0.1046)	0.4155*** (0.1041)	0.3934*** (0.1043)	0.3163** (0.1046)
Grade	−0.1128*** (0.0386)	−0.1264*** (0.0386)	− 0.1297*** (0.0385)	− 0.1230** (0.0386)	− 0.1383*** (0.0385)
Concern	0.2882*** (0.0574)	0.2524*** (0.0579)	0.2555*** (0.0573)	0.2562*** (0.0578)	0.2492*** (0.0578)
Legal	0.4975*** (0.0577)	0.4134*** (0.0548)	0.5315*** (0.0555)	0.5568*** (0.0574)	0.5193*** (0.0596)
Lesson	0.1871* (0.1061)	0.2154** (0.1059)	0.2061* (0.1062)	0.1652 (0.1065)	0.1935 (0.1061)
Respect	0.1751*** (0.0637)	0.2246*** (0.0634)	0.1882*** (0.0632)	0.1952*** (0.0633)	0.1739** (0.0640)
Recogn	0.0787* (0.0423)	0.1148*** (0.0421)	0.1073** (0.0422)	0.1033** (0.0423)	0.0963** (0.0422)
Policy	0.2787*** (0.0547)	0.3077*** (0.0545)	0.2672*** (0.0549)	0.2429*** (0.0553)	0.2794*** (0.0546)
LR stat	393.4340***	375.2681***	411.7182***	414.2574***	395.4193***
Obs	1497	1497	1497	1497	1497

[1] ***,** and* denote significance level of 1, 5 and 10%, respectively; [2] Clustered standard error in parentheses

Compared with our previous results (Tables 5 and 6), health policy education variable now significantly generates a positive link to perceived decline in violence. The reason could be that, although medical education itself cannot influence violent perpetrators, it can help students to truly understand the incidence of WPV and medical environment. Besides, it should be noted that the policy effect of reform also contributed to the perceived decline in the incidence of violence, indicating that descending resources reform has helped to rebalance patient trust, reduce congestion in the medical market, and decrease the WPV incidence.

Discussion

The goal of medical education is to provide sustainable human capital supply for the medical market and different levels of hospitals. However, medical students are facing a variety of career choices other than practicing medicine, and their career choices depend on the satisfaction (utility) of different demand levels brought by practicing medicine compared with other occupations. If medical students expect that being a healthcare worker means higher safety risks and higher practice costs, their willingness to practice medicine will be weakened [47, 48], which will result in medium and long-term insufficient supply of medical services because of shortage of human capital. In the context of ongoing descending resources reform and high incidence of healthcare workplace violence, China’s dominant public hospital system and hospital classification system enable us to observe how medical students will respond to the reform and workplace violence, and help us understand the inter- and intra-market effect of exogenous shock on medical students’ willingness to practice medicine.

Our study confirmed that, the perceived violence incidence of medical students has a negative impact on their working intentions to practice medicine and a positive impact on their intentions to practice in low-level hospitals. The negative impact means the existence of inter-market effect, that is, the higher incidence of perceived violence means the expected higher practice risk and the lower satisfaction of safety needs compared with other occupations. Then, it will inevitably weaken the working intentions to practice medicine for medical students [4, 12]. At the same time, there exist obvious differences in the WPV incidence between different levels of hospitals, and high-level hospitals usually experience much higher incidence than that of low-level hospitals [31, 32]. So, medical students with factor specificity will more choose low-level hospital so as to receive higher satisfaction of safety needs in the medical market. Therefore, the positive effect of the perceived incidence of workplace violence reflects this intra-market effect, that is, the high WPV incidence in high-level hospitals will drive medical students to choose low-level hospitals.

China’s recent high WPV incidence is accompanied by the ongoing descending health resources reform. This reform is characterized by the fact that young doctors in high-level hospitals need to bear the reform cost by leaving cities where they work and descending to nonlocal low-level hospitals [49]. But at the same time, the goal of this reform is to alleviate the congestion of high-level hospitals and make better use of the idle resources of low-level hospitals, so as to correct the resource mismatch in China’s medical market. It will also help to improve the practice environment of medical students in the future and improve their utility level. Therefore, the effect of reform policy perceived by medical students will positively impact their willingness to practice medicine, but their cognition of reform means that their choice to practice medicine will make them bear the reform cost in the near future, which could exert a negative effect. Our estimation results support the coexistence of the above two opposite reform effects. However, since working in low-level hospitals means no need to bear the reform cost, the above negative effect of reform cognition will tend to disappear, which is also confirmed by our empirical results.

There may be concerns that the cross-sectional questionnaire data used in this paper cannot fully reflect students’ real career choice behavior after graduation. It should be noted that, medical students would not make decision on their career choice only around graduation period, their working intentions to practice medicine depend on the collection and evaluation of all available information in the whole process of medical education. Even if medical students’ cognition on the incidence of workplace violence and health reform seriously deviates from the real situation, they will make career choices based on it. However, in the case of biased information, medical students tend to hold a low willingness to practice medicine, which is against the goal of medical education and requires intervention in the early stage of medical education.

This study has several limitations. Firstly, due to the nature of the cross-sectional data and models used, our empirical results should not be interpreted in terms of causality, but rather as correlations. In addition, working intention does not fully determine real career choice when students graduate and enter the labor market. If follow-up investigation could be performed, we can combine the real career choice data with the working intention response data to the model, which would be helpful for us to better understand the dynamic impact of exogenous shock on medical students’ career choice.

Conclusions

Workplace violence against healthcare workers is a global problem [37, 50], and some reports have found that such violence occurs in medical students and impact their working intention to practice medicine [48, 51]. This paper focused on the impact of WPV and the anti-violence legal environment on medical students’ working intentions in the context of China’s descending resources reform launched in 2013. We identified inter- and intra-market effects of WPV on medical students’ career choice, emphasizing risk aversion motives and the importance of medical student safety when facing severe WPV risk. This research provides new insights into the factors influencing medical students’ career choices beyond family, personality, and academic achievement.

Based on our results, the anti-violence legal environment has no direct link with medical students’ working intentions but may contribute to the perceived decline in the incidence of violence. Our study also showed that the perceived costs and benefits of reform have the opposite effects on driving students’ desire to practice medicine, thereby highlighting the complexity of reform. Thus, the anti-violence environment and reform policies may impact medical students, and generate an intra-market effect, i.e., push medical students to work in low-level hospitals. To incentivize constant inflow and balanced allocation of human capital in the health market, urgent action is needed to reduce the incidence of healthcare WPV with the cooperation of the government, legislation, law enforcement, and medical institutions. Improving student awareness of WPV and reform could also encourage more rational and well-informed career choices.

Additional file 1.