Study design and data collection
The Cohort Study of Medical Graduates with Compulsory Services in Rural Areas Study is a prospective cohort that investigates the medical study, residency training, employment, and career development of medical graduates, in order to contribute to the development of health workforce in rural and remote areas in China. The study was approved by the institutional review board (IRB) at Peking University Health Science Center (IRB00001052-15027). All participants provided informed consent. In 2015, the first batch of medical graduates trained by CSP graduated after receiving five-year undergraduate study, so we established the first sub-cohort and collected baseline survey at four medical universities since then [15,16,17]. All CSP classes in the four universities were included in the cohort, and the participation rate was 100% at recruitment. CSP classes were matched with corresponding same-year NCSP classes at 1:1 ratio. We continued establishing sub-cohorts among graduates of 2016, 2017, 2018 and 2019, and conducted follow-up surveys annually. The study included 3,620 medical graduates from the Medical College of Qinghai University (Qinghai province), Jiujiang University (Jiangxi province), Gannan Medical University (Jiangxi province), and Guangxi Medical University (Guangxi Zhuang autonomous region). Participants completed a paper questionnaire at baseline before they finished the undergraduate study. In the baseline survey, we created a WeChat group to include all participants within each school and each sub-cohort (WeChat is a widely used instant messaging app in China. The WeChat group enables to exchange communication between investigators and participants). The links of online self-reported questionnaires were sent each year with notification via WeChat groups, text messages of mobile phones, and E-mails. Till now, we have collected follow-up data in 2016, 2017, 2018, and 2020.
There are two types of medical graduates in this study, those who are required to practice in rural and remote areas after graduation (compulsory services program, CSP) and common medical graduates (non-compulsory services program, NCSP). Baseline survey covered demographic characteristics, attitudes towards medical study, and preferences of career development for the two types of graduates. Response behaviors of whether participants provided some sensitive information were also recorded. A flow diagram that summarizes the process, baseline sample size, and follow-ups is presented in Fig. 1.
In the follow-up survey, the online questionnaire took 5-10 min to finish. Follow-up surveys mainly collected the following information: (1) current jobs: working location, professional title, job income, formally funded positions, especially for CSP graduates etc.; (2) career mobility: whether change jobs and reasons, and information about new jobs etc.; and (3) career development: promotion, residency training, and job performance etc. The content of each follow-up survey was similar. Whether breaking the contracts for CSP graduates and job income for both types of graduates could be invasive. However, these questions were asked repeatedly in each follow-up survey, the content of follow-up surveys was unlikely to have a differential impact on attrition.
Notes: the 2019 follow-up survey was not conducted for graduates of 2015-2018 due to logistic reasons.
Strategies for increasing response
To maintain the participation of medical graduates in the cohort, we mainly adopted the following strategies. First, in between scheduled follow-ups, regular newsletters were sent to the WeChat groups by project members, including national or regional policies related to CSP, newly published research concerning CSP, and greetings on important holidays. Second, we organized a writing competition each year to encourage graduates who have become GPs in rural and remote areas to share their experiences. Third, we built a liaison team composed of student leaders in the baseline survey, and asked them to send the questionnaire to graduates who did not respond.
To increase response rate, we offered ten CNY as allowances for those who completed questionnaires. If participants did not fill in the questionnaire in time, we would add them as WeChat friends, call them, or send E-mails to remind them. Besides, we would also ask liaison members to contact and remind them to fill in the questionnaire.
1. Follow-up status
Participants were categorized into two mutually exclusive groups: (1) complete follow-up refers to those who finished all follow-up surveys; (2) incomplete follow-up refers to those who did not finish at least one follow-up survey.
Participants with incomplete follow-up were further categorized into three mutually exclusive sub-groups: (1) “always-out” refers to those who did not respond to any follow-up surveys after the baseline visit, and those who could no longer be contacted through their phone, WeChat, E-mail, or other contact ways; (3) “rejoin” refers to those who finished the 2020 follow-up survey but was lost at least once in previous follow-up surveys; (4) “other” refers to those who did not respond to the 2020 follow-up survey but finished at least one previous follow-up surveys.
2. Predictors of attrition and outcome
For potential predictors of attrition, we tried to cover variables that were found to have impact on attrition in previous literature, and also explored some variables related to medical study. The variables were categorized into sociodemographic characteristics [18, 19], attitudes towards medical study, preferences towards career development , and response behaviors in the baseline survey . Sociodemographic information included types of students (whether CSP or NCSP graduate), medical university (Qinghai, Guangxi, Jiujiang, Gannan), graduation year (2015-2019), gender, family background (rural or urban) , highest completed education level of theirs parents (low middle school and below, or high middle school and above) , occupation of father (farmer or non-farmer), pressure from tuition and other fee at school (having pressure or no pressure), and household members (four family members or above, three family members or below) . Attitudes towards medical study comprised of whether studying medicine was the first choice, whether planning to pursue a postgraduate degree, whether willing to participate in residency training, whether satisfied with medical education received, and whether understanding compulsory services program policy for CSP graduates.
Preferences of career development included whether desired to work in public hospitals above county level after graduation, whether believing they can pass the China National Medical Licensing Examination within one year, whether income was the primary consideration when looking for jobs, and whether the contract-signing place was hometown for CSP graduates.
Response behaviors included whether provide the score of university entrance examination, whether provide contact information (we collected WeChat, QQ, E-mail, and cellphone, providing at least one way of contact was considered “provide”. Providing none of the aforementioned contact was considered “not provide”). It should be noted that although we established WeChat groups in the baseline survey, students can still withdraw anytime from the group. Whether reporting household income was also included as a predictor of response behavior.
To answer question 1, descriptive analyses were used to present the response rate of five sub-cohorts for each follow-up survey and the cumulative follow-up status ratio.
To answer question 2, firstly, the baseline characteristics between complete and incomplete follow-up of medical graduates, and among three types of attrition groups were presented. Chi-square tests were used to compare the differences between different groups of follow-up status. The P-value below 0.05 was considered statistically significant. Secondly, logistic regression was performed to investigate the factors associated with attrition of the cohort (the dependent variable was 1 for completing all follow-up surveys and 0 for incomplete) in the total sample and by types of graduates. Demographic characteristics, attitudes towards medical study, preferences towards career development, and response behaviors in baseline were included as explanatory factors in the model. In order to examine the predictors associated with rejoining, we conducted logistic regression in the incomplete sample (the dependent variable was 1 for rejoin and 0 for always-out or other). Thirdly, to rule out the effect of time, the Anderson and Gill model (AG model) was used to conduct a time-to-event analysis with recurrent events . The detailed process and results was presented in the additional file.
To answer question 3, we adopted a multiple imputation method which has been widely used in dealing with missing data . We performed a multiple imputation with chained equations (MICE), and all baseline variables were used to impute the interested outcome measures. For outcome measures, we chose three indicators that can reflect the career development of medical graduates, including income of the participant’s current job in the 2020 wave (income was collected by the question: “your income of the current job was___CNY per month”), whether passing the China National Medical Licensing Examination (NMLE) till 2020 wave, and obtaining a professional promotion till 2020 wave. The package “ice” was used to conduct the multiple imputation process in Stata.. For the three outcome measures, comparisons were made between the imputed data and the data before imputation. T-tests were used for job income, and Chi-square tests were used for the other two outcome measures. Stata 16.0 (Stata Corp LP, College Station, TX, USA) and R version 4.0.2 were used to perform the analysis.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.