Job preferences for medical and nursing students to work in rural Guizhou Province, China: a discrete choice experiment


 Background Maldistribution of health workers between urban and rural areas has been a critical difficulty in China. The shortage of health workers in disadvantaged areas reduces access to essential health services delivery, and adversely affects the population health. Policies on attracting health workers to locate in rural areas are needed to be explored. In order to identify the appropriate incentives, we conducted a discrete choice experiment to determine how specific job attributes might be valued by final year students in medical university in Guizhou Province, China. Methods Attributes of potential job were developed through literature review, in-depth semi-structured interviews and pretest. Salary, education opportunity, transportation, job location, workload, essential equipment, medical order, and identification ('bianzhi') were included. The questionnaire was formulated through a fractional factorial experiment design using %MktRuns macros of SAS 9.4. All medical and nursing students in the final year at Guizhou Medical University were invited to participate in the study. Mixed logit model was used to estimate stated preferences of attributes. Willingness to pay and uptake rates for a defined job were also calculated based on the mixed logit estimates. Results The final sample comprised 787 respondents, including 388 medical students and 399 nursing students. Attributes were statistically significant (with the exception of once every two years for education opportunity) and had expected signs. The results indicate that physical conflict between doctors and patients and identification ('bianzhi') were two of the most important non-monetary job characteristics for both medical and nursing students. And nursing students placed more value on identification ('bianzhi'). Policy simulation suggests that as for the individual incentive respondents were most sensitive to salary increasing. Incentive packages effects were stronger for students from rural background. Conclusions Strategies on medical order, identification ('bianzhi') and salary should be considered to attract final year medical and nursing students to work in rural areas. In addition, specific recruitment policy design tailored for subgroups should be taken into account.


Methods:
Attributes of potential job were developed through literature review, 23 in-depth semi-structured interviews and pretest. Salary, education opportunity, 24 transportation, job location, workload, essential equipment, medical order, and 25 identification ('bianzhi') were included. The questionnaire was formulated through a 26 fractional factorial experiment design using %MktRuns macros of SAS 9.4. All 27 medical and nursing students in the final year at Guizhou Medical University were 28 invited to participate in the study. Mixed logit model was used to estimate stated 29 preferences of attributes. Willingness to pay and uptake rates for a defined job were 30 also calculated based on the mixed logit estimates.

31
Results: The final sample comprised 787 respondents, including 388 medical students 32 and 399 nursing students. Attributes were statistically significant (with the exception 33 of once every two years for education opportunity) and had expected signs. The 34 results indicate that physical conflict between doctors and patients and identification 35 ('bianzhi') were two of the most important non-monetary job characteristics for both 36 3 / 41 medical and nursing students. And nursing students placed more value on 37 identification ('bianzhi'). Policy simulation suggests that as for the individual 38 incentive respondents were most sensitive to salary increasing. Incentive packages 39 effects were stronger for students from rural background.

40
Conclusions: Strategies on medical order, identification ('bianzhi') and salary should 41 be considered to attract final year medical and nursing students to work in rural areas. 42 In addition, specific recruitment policy design tailored for subgroups should be taken 43 into account. The uneven distribution of health workers reduces access to essential health services 48 delivery, and contributes to inequalities in health outcomes [1]. As one study found 49 that, reductions in under-five mortality rate were associated with increased health 50 workers [2]. Several studies have cast light on the systematic categories of imbalances 51 in the health workforce, including geographic imbalances, institutional imbalances, 52 public/private imbalance, and profession/specialty imbalance [3]. Among all these 53 imbalance categories, the lack of health workers in disadvantaged areas will impede 54 the universal health coverage and ultimately adversely affect the population health 55 living in remote and rural areas [4]. As one study found that, reductions in under-five 56 mortality rate were associated with increased health workers [2]. Disparity between 57 urban and rural areas in terms of health workforce has become a critical health policy 58 concern in many countries [5,6]. China is not an exception, since health workers are 59 inclined to serve in tertiary hospitals in urban areas rather than in primary care 60 facilities in rural areas because of working conditions, career development, etc. 61 China's higher education system reform has been launched in 1998, since then a 62 major expansion of universities has facilitated to a fast growth of health education [6]. 63 However, a big dilemma is that faculty numbers might not have kept pace with 64 numeric expansion of students [7]. Although China's total health workforce rose 65 remarkably from 6.7 million people in 2006 to 10.2 million in 2014 [8], a shortfall of 66 over 500 000 physicians exits in rural areas [9]. In addition, the low nurse to doctor 67 ratio (1.1:1) to some extent constrains the development of clinical practice [8]. Further, 68 the density of nurses in urban areas is much higher than in rural areas and the gap is 69 widening [10]. 70 Undergraduate health students' job preferences will influence the distribution of 71 health workers in the future. Thus, it is necessary to elicit policy incentives specific 72 for final year medical and nursing students, which can be tailored to make the posts in 73 rural areas more attractive, since posts in primary care facilities in rural areas are 74 usually considered less desirable because of high workloads, inconvenient traffic and 75 poorly infrastructure [11]. Incentives to attract health workers to locate in 76 disadvantaged areas have been well described by World Health Organization(WHO) 77 including education interventions, regulatory interventions, financial incentives, 78 personal and professional support [12]. WHO also proposed that countries need to 79 identify the appropriate interventions that are suitable for their local context, given 80 diverse health labor market and local demand [13]. Thus, the development of such 81 strategies really needs a precise insight into job preferences of health workers in 82 various counties. 83 The policy implication could be elicited from the use of discrete choice experiment 84 (DCE) method. DCEs have been commonly used in health economics, while recently 85 determining health workers job preferences in human resource research field can be 86 informed by the results from this methodology. In particular, a user guide on how to 87 conduct a DCE for health workforce recruitment and retention in rural areas was 88 developed by WHO and other two agencies, which mentioned that DCE could assess 89 the stated preferences of health workers for a job [14]. The choice experiment is a 90 combination of the characteristics theory of demand and random utility theory [15]. 91 The method assumes the utility associated with a good or service is made up of the 92 utilities of its attributes [16],which is well-established for identifying the relative value 93 that people place on factors (attributes) [17]. The benefit of DCE is that of 94 opportunity cost. Individuals want the best of everything, but when resources are 95 limited, the method gives a weighed relevance to distinguish which attributes are the 96 most highly valued. In addition, quantitative information on the relative strength of 97 the selected attributes could be provided, as well as the trade-offs between these 98 attributes and the probability of take-up of defined jobs [14,18]. 99 Under the circumstances of China's reform and opening-up, doctors has confronted 100 and undergone the challenges of ongoing economic reforms and medical reforms. 101 With the emergence of new problems such as violence against medical staff, previous 102 difficulties still exit, e.g. identification ('bianzhi'). The surge in hospital violence 103 seriously affects the normal medical order, which is a proper status where health 104 workers and patients in the medical relation share orderly interactions. And a 105 widespread concern among the health workers has been caused by the assaults on couple of studies to explore the job preferences of health workers using DCEs in 118 china [10,21], and which found that identification ('bianzhi') , as well as income, 119 benefits, equipment, career development, respect from the community, training 120 opportunity, education environment for children could be the push and pull factors 121 affecting Chinese health workers' decision on location in rural areas [21,22]. 122 However, the paucity of evidence on how incentive policies influence undergraduate' 123 decision to work in rural areas is cause for concern. And to our knowledge, this is the 124 first study to explore the impact of medical order on rural recruitment. More evidence 125 would be needed to inform practical policy decisions. In this paper, we aim to explore 126 a set of factors that are amenable to job preferences among final year students, and to 127 assist policy-makers in designing better-informed interventions for attracting them to 128 rural areas. In this study we examined the preferences of Undergraduate health students who 144 are seen as the fresh blood of health professionals. China has a vast and complex 145 system of health professional education. The most common education program for 146 medical students is 5-year curricular plans, meanwhile 3-year schools after senior 147 high school or junior high school and 7-8-year degree programs exist all at once. 148 Graduates can be awarded Bachelor's degree after 5-year curricular plans, while 149 graduates for 3-year schools cannot get bachelor's degree and will be assistant doctors.  Figure A1). A forced choice approach was employed for two reasons: 1) to elicit 242 more information on respondents' preferences for the attributes; 2) since students had 243 not empirical understanding of the job in reality, the opt-out option (to choose their 244 current job) was not included. We used an unlabeled DCE, because some evidence 245 suggested that labels may distract respondents from job attributes and thus diminish 246 the reliability of estimates of job preferences [29]. The questionnaire was presented in 247 three sections. Section one was an introductory script. The purpose of this part was to 248 acclimate respondents to the hypothetical nature of the DCE they were about to take. 249 The telephone number of research group member was also in this part for the 250 interviewees to connect while they had problems during the process of filling in 251 questionnaires. Section two included socio-demographic characteristics, such as Pilot-testing 265 Prior to the start of data collection, we piloted with 33 medical students. This process 266 provided an opportunity to determine if the presentation was conceptually clear. 267 Minor correction was made. Respondents did not think 20 choice sets were a 268 cognitive burden, and they completed the survey within 20 minutes.

328
Of 1168 eligible final year medical and nursing students in Guizhou medical 329 university, overall 879 (75%) agreed to participate in the survey: 438 were medical 330 students and 441 were nursing students. The response rates were 61% and 97%, 331 respectively. Questionnaires that were not completed were excluded from the sample, 332 thus the final sample comprised 787 respondents, including 388 medical students and 333 399 nursing students. The questionnaire qualification rates were 89% and 90%, 334 respectively. 335 The mean age of medical and nursing students were all 24 years old (SD=1.3), with 336 ages ranging from 21 to 29 years. Nursing students were predominantly female (92%), 337 while 58% of medical students were female. Medical students of rural origin were 338 52%, whereas 66% of nursing students had a rural background (Table 1)  In terms of essential equipment, medical students and nursing students were willing to 370 reduce the almost same monthly salary that was RMB 954 (US$ 141) to have a job 371 with adequate essential equipment. 372 Analysis for subgroups on sex for medical students suggests that women would be 373 compensated with more money for tuneless medical order than men. And men would 374 pay more money to obtain the other attributes than women with the exception of 375 adequate essential equipment. Medical students from rural background would pay or 376 be compensated with more money for all attributes than those from urban background 377 except the job location of city (Table A2). 378 379 Policy simulation 380 The relative impact of DCE attributes is often better appreciated to predict the impact 381 of different policies. Figure 1 showed the likely uptake of rural jobs to reflect the 382 relative effectiveness of different policies proposing improved rural jobs, for the 383 medical and nursing students, as well as the rural and urban background students. Not 384 surprisingly, urban jobs were more strongly preferred than rural ones, especially for 385 nursing students and urban background students (the total uptake rate for rural 386 postings at the baseline was 2% and 4%, respectively). It appears that of all the single 387 incentives that could be introduced to improve the current rural job offers, the most 388 effective would be to offer rural posts with increased salary for the four groups. As 389 salary level went up, the impact on rural recruitment would increase. As the 390 simulation result showed, if the salary increase from 5000RMB to 7000RMB, the 391 probability of accepting the job would increase by 19%. Intervention packages were 392 also explored (Table 4) that health workers pay more attention to earnings than the patients themselves. 454 Gradually, health workers in medical facilities lose the trust and respect from patients. 455 However, one study in China showed that, from the perspective of health workers, 456 respect from community would be of great importance for them [21]. The DCE 457 information presented here may be used to inform the development of specific policy 458 intervention. Given the high doctor-patient tensions in China and the results of our 459 studies, the introduction of humanistic practice and the inculcation of professional 460 norms in medical schools could be urgent and essential [8]. 'Physical conflict' was the 461 most disliked scenario for medical and nursing students, followed by 'suit'. The 462 findings suggest that these students place more value on their life safety, rather than 463 medical event that could be solved through suit in court. Thus, in order to attract final 464 year students to locate in positions in disadvantaged areas, regulations and institutions 465 need to be improved to protect the safety of them. 466 Not surprisingly, 'bianzhi' had a large effect on job preferences. This key finding 467 has important policy implications. The majority of health workers in China work at 468 state-owned health facilities, 'bianzhi' can attach them with not only identification 469 from the government, but also corresponding benefits. Nursing students were willing 470 to pay more money to obtain 'bianzhi' than medical students, partly because of a high 471 utilization of contract-based nurses as opposed to 'bianzhi' nurses in China and the 472 inequities between the two types of nurses in terms of wages and job-related benefits. 473 Further, study finds that the disparities may adversely affect both nurse and patient 474 satisfaction in hospitals [20]. In real life, such environments are hard to change, but its 475 substantial impact on job preferences should be considered in health policy 476 discussions. For example, 'equal pay for equal work' strategy could be emphasized by 477 government in order to eliminate the disparities between 'bianzhi' and contract-based. 478 Our findings confirm that financial incentives are very important in attracting final 479 year students to locate in a rural posting. Final year health students viewed a higher 480 salary as a very important attribute. This reinforces previous studies in China [33], 481 which might be due to the fact that they are less satisfied with their salaries. As 482 Chinese health students still value the most fundamental needs, the lowest level of 483 needs in Maslow's theory, that is basic needs for safety [21]. From a health policy 484 makers' perspective, future programs should focus on not only non-financial strategies, 485 like educational opportunities or adequate equipment, but also financial incentives. 486 Unlike several studies showing that financial incentives were not found to be most 487 powerful policy levers [13,26,28,31,33,36,38,44]. However, our findings 488 conformed that monthly income had a significant impact on the job choices of 489 undergraduate health students, similar with other studies [10,37,45]. 490 Willingness to pay was used as an estimate of the minimum compensation. The 491 WTP results from the sub-group analysis on sex for medical students showed that 492 women would pay more money for harmonious medical order than men. As a result, 493 female doctors could be considered to be absorbed in the clinical team to smooth the 494 relationship between doctor and patient. Policymakers could target interventions to 495 different sub-groups of students or at least consider the differential impact in their 496 planning. 497 Our studies found that medical students from rural background would pay more 498 money on the all attributes of the hypothesized work with the exception of rural job 499 location. This finding is in line with previous research [26], but is somewhat at odds 500 with study in India [46]. It follows that the rural shortage of health workers could be 501 mitigated to some extent by preferential admission of health students from rural areas. 502 On the other hand, we could infer that if students with rural background can obtain the 503 potential work, they maybe cherish the job opportunity. So these were all prompts that 504 need to be taken into account when planning an attraction strategy. This also suggests 505 that preferential selection of rural students by training institutions can be an effective 506 strategy, and it also lends support to claim that student selection policies are a key 507 important of human resource intervention packages. This could be considered, e.g., 508 through a scholarship or student loan scheme for students who are from a rural area 509 and who are willing to accept a job in a rural area upon completing their study. The results presented here should be considered by policy makers for its 548 implementation, while weighing carefully the labor market and policy cost.

Availability of data and materials 561
The datasets used and/or analysed during the current study are available from the corresponding 562 author on reasonable request. 563

Competing interests 564
The authors declare that they have no competing interests. 565 Funding 566 Not applicable. 567 Authors' contributions 568 MB designed the study and data collection tools, conducted the data analysis and drafted the 569 manuscript. CH contributed to the development of the research design, monitored the data 570 collection activities and contributed to preparing the draft manuscript. Both authors read and 571 approved the final manuscript. 572

Acknowledgments 573
We thank the medical and nursing students who participated in the study. We also appreciate the 574 health officials and health workers who gave their valuable time to participate in the study. 575