Job preferences among traditional Chinese medicine clinical graduates in China: a discrete choice experiment | BMC Medical Education
Sampling
This study uses a stratified sampling method based on geographical divisions, surveying students from 16 out of 25 higher education institutions specializing in Traditional Chinese Medicine (TCM) across China. This study was a full-sample survey, including all eligible participants, who were all students in their fifth year of undergraduate study and third year of Master’s study in TCM and integrated Chinese-Western medicine from 16 universities. The selected institutions were distributed across three major regions, namely Eastern China (7/13 institutions), Central China (4/6 institutions), and Western China (5/12 institutions), ensuring broad geographic representation. The survey was conducted between April 22 and May 22, 2024, using a professional online survey platform. We contacted relevant faculty members, such as counselors, from the respective universities, who distributed the survey link to the target population via WeChat, a widely used social media platform in China. This research was conducted as an anonymous online survey, and all participants provided informed consent before participating.
Discrete choice experiment (DCE)
The discrete choice experiment (DCE) is currently the most powerful tool for the measurement of individual preferences, allowing for the assessment of healthcare workers’ preferences for different combinations of job attributes. It has been widely used in market research, public policymaking, healthcare, and natural resource management [20, 21]. In this study, simulated TCM job scenarios were described in terms of various attributes and their corresponding levels, and students were required to obtain trade-offs in a series of hypothetical TCM job settings.
Selection of attributes and their levels
According to Emily Lancsar and colleagues [22], the selection of attributes should consider the following aspects: firstly, the attribute should be based on government or clinical policies relevant to the research context; secondly, each attribute should be significant to the respondents. Attributes can be either quantitative (e.g., waiting time) or qualitative (e.g., whether a service is provided) and are typically identified through literature reviews or qualitative studies, such as semi-structured interviews or focus group discussions [23]. Following the DCE design guidelines [24], we identified the attributes through a literature review [15, 17, 18,20, 25,26,27,28] and subsequent qualitative research (focus group discussions and in-depth interviews).
Initially, ten job-related attributes were identified, including the monthly income [14,15,16,17, 19, 26], housing benefits [17], job location [14,15,16,17, 19, 26], work environment [15,16,17, 19], work intensity [14, 16, 19], on-the-job training and career development opportunities [15, 16, 19], employment status(tenure) [14, 15, 16, 17], social reputation [28,29,30], transportation [14], and children’s educational opportunities [17]. An iterative qualitative process was used to refine the attributes and levels. Among these, “tenure” is a job attribute that is unique to China, where a tenured position guarantees lifelong employment, and the employee cannot be dismissed by the employer. Additionally, tenured positions are considered “iron rice bowls” in China due to the job security that they offer through benefits such as health insurance and pensions [31, 32]. The work intensity included aspects such as shift work, being on-call, rest periods, emergency or on-call time arrangements, the work–life balance, the daily workload (whether they have enough time to complete tasks), and overtime [19].
We conducted online or offline in-depth interviews with 16 undergraduate and graduate students from Shanghai University of Traditional Chinese Medicine, Henan University of Chinese Medicine, and Guizhou University of Traditional Chinese Medicine. The results indicated that “housing benefits,” “social reputation,” “transportation,” and “children’s educational opportunities” were not among the most important attributes among them. Therefore, these attributes were excluded from the subsequent analysis so as to better focus on the issues that were considered the most important among the respondents.
Additionally, we consulted two experts in the field of DCE and six experts working in related TCM fields to finalize the attribute levels. Based on their feedback, and considering TCM as a long-standing traditional medical practice in China, we added the attribute “mentorship opportunities” to reflect graduates’ emphasis on apprenticeship education. We also retained the attribute “career development opportunities” and added “hospital level” to capture the graduates’ considerations regarding career development paths and work environments. We decided to remove the attribute “work environment” because it was partially reflected in other attributes or had a relatively minor impact on the graduates’ choices. The level of monthly income was adjusted to CNY 3,000–8,000, equivalent to 411.0–1,096.0 USD (based on an exchange rate of 1 USD = CNY 7.299 as per OECD data, 2024). This income range aligned with the respondents’ educational levels and professional expertise, the real-world conditions, and their expectations. The final attributes, their definitions, and their corresponding levels are shown in Table 1.
Questionnaire design
The survey included demographic information, choices regarding clinical work related to Traditional Chinese Medicine (TCM), and the discrete choice experiment (DCE) choice sets. This study was divided into a pilot study and a formal study, and we followed standard approaches for the design of the DCE in order to achieve unbiased, statistical response efficiency [33]. The DCE was based on seven attributes, with six attributes having three levels each and one attribute having two levels. The design method provided by Huber and Zwerina [34] was used, and the %MktRuns macro command in the SAS 9.4 software was employed for efficient design. This resulted in 18 choice sets, each containing two alternatives (Job A and Job B). To mitigate respondent fatigue, the choice sets were further randomly divided into two versions, with each version containing nine choice sets.
To verify that the respondents answered carefully and clearly understood the questionnaire, each version of the questionnaire included one validation choice set. In this choice set, all attribute levels of Job A were clearly superior to those of Job B. If a respondent chose Job B (the disadvantageous option), it indicated that they might not have properly understood the questionnaire, and their response was excluded. At the beginning of the survey, we first asked the TCM clinical graduates whether they were willing to engage in TCM-related work. The questionnaires from respondents who answered “yes” were included in the analysis, while those who answered “no” were excluded. This screening step ensured that the sample was focused on individuals willing to work in TCM, thereby enhancing the relevance and effectiveness of the research results. Each version of the final questionnaire included ten choice sets. To maximize the accuracy of the information collected on the employment preferences of TCM clinical graduates, this study did not include an opt-out option in the experimental design. Additionally, opting out may only introduce minor differences in the estimates [35], while a forced choice method provides more comprehensive responses and higher-quality data [36].
The survey began with an informed consent statement and an explanation of the research, emphasizing that the respondents’ participation was voluntary and involved no risks or costs and that their responses would be kept confidential. The respondents proceeded with the questionnaire only after selecting the option, “I agree to participate in the survey.”
The complete questionnaire can be found in Attachment 4. The distribution and collection of the questionnaires were conducted through an online survey platform (Wenjuanxing). One advantage of this platform is that it can produce preliminary classification statistics for the collected data according to the questionnaire items. Once the sample size reaches the target, the data can be directly downloaded from the platform, eliminating the need for data entry. After downloading the platform data in Microsoft Excel format, they were specifically converted into the required format for DCE data.
Pilot study
Following the pilot study, the preliminary analysis confirmed that all attributes aligned with our expected effects. However, we noticed that the “master inheritance” attribute in the original setting might have been difficult for some respondents to understand. Therefore, we revised it to “opportunity for master inheritance,” with the attribute levels changed to “opportunity available,” “uncertain opportunity,” and “no opportunity.” After modifying the language and format of the questionnaire, a formal survey was conducted using stratified sampling across 16 TCM higher education institutions.
Data analysis
Descriptive statistical methods were used to describe the demographic characteristics, family annual income, etc. Continuous variables were described using means and standard deviations, while categorical variables were described using frequencies and proportions. Statistical significance was set at 0.05, with p < 0.05 indicating statistical significance.
The data analysis was conducted using Stata 16.1. Under the random utility theory framework, it was assumed that student 𝑛 would choose the option that provided the highest utility 𝑈. The utility 𝑈 was composed of a deterministic component 𝑉 and a random component 𝐸(𝜖). The deterministic component 𝑉 could be explained by observable factors (job attributes) 𝑋, such as mo(𝛽1,𝛽2,…,𝛽𝑚). 𝜖 represented unobserved utility and error effects. In this study, if student 𝑛 preferred option 𝑖 over option 𝑗, then 𝑈𝑛𝑖>𝑈𝑛𝑗, and the utility from option 𝑖 for student 𝑛 could be expressed as.
$$ \begin{array}{*{20}{c}}{{U_{ni}} = {V_{ni}} + { \in _{ni}} = {\beta _1}{X_{1ni}} + {\beta _2}{X_{2ni}} + \ldots + {\beta _m}{X_{mni}} + { \in _{ni}} = {\beta _1}{\rm{income}} + }\\{{\beta _2}{\rm{location}}\_{\rm{second}} + {\beta _3}{\rm{location}}\_{\rm{first}} + {\beta _4}{\rm{hospita}}\_{\rm{second}} + }\\{{\beta _5}{\rm{location}}\_{\rm{third}} + {\beta _6}{\rm{Bianzhi}} + {\beta _7}{\rm{workload}}\_{\rm{normal}} + }\\{{\beta _8}{\rm{workload}}\_{\rm{easy}} + {\beta _9}{\rm{generation}}\_{\rm{maybe}} + }\\{{\beta _1}{\rm{generation}}\_{\rm{maybe}} + {\beta _{11}}{\rm{career}}\_{\rm{common}} + }\\{{\beta _{12}}{\rm{career}}\_{\rm{high}} + { \in _{ni}}}\end{array} $$
By analyzing the significance, direction, and magnitude of the regression coefficients 𝛽, the importance, influence direction, and emphasis of each attribute level could be determined. The statistical significance of the regression coefficients indicated the importance of the attribute levels to the respondents: a positive coefficient indicated a positive preference for the attribute level, while a negative coefficient indicated a negative preference. The size of the coefficient indicated the degree of importance of the attribute level to the respondents.
When estimating the mixed logit model (MXL), the willingness to pay (WTP) [33, 37, 38] was calculated to determine the relative monetary value of each aspect of the job. The WTP estimate was the ratio of the coefficient of each attribute level to the income attribute. This ratio theoretically reflected the degree to which the participants were willing to pay more or require compensation for changes in an attribute level. In other words, a positive value indicated that the participants were willing to pay more for an attribute level, while a negative value indicated that the participants required more compensation to forego this attribute level. Finally, the marginal effects of each job attribute were calculated to analyze the job preferences under the simulated scenarios. The marginal effects analysis helped us to understand how changes in a specific job attribute affect the probability of accepting the corresponding job and simulate jobs with high acceptance rates among the respondents.
Ethical approval
This study received ethical approval from the Shanghai Health and Development Research Center Ethics Committee (Reference No. 2023008) and adhered to the principles of the Declaration of Helsinki.
link
