Study Design and Participants
Our investigation utilized a cross-sectional survey involving a self-administered questionnaire. This survey was conducted at the 10 medical schools in the Riyadh region of Saudi Arabia. The Riyadh region was chosen because it has several advantages that are not available in other locations within Saudi Arabia. First, it contains one third of all medical schools in Saudi Arabia. Second, it includes both public and private medical schools, and finally, the medical schools are within both urban and rural settings. The study population was comprised of all fourth- and fifth-year medical students at each medical school.
Data Collection Protocol and Sample Size
Data were collected through a paper, self-administered questionnaire. Data collection occurred six weeks after the beginning of 2019 the Fall semester, starting in October and lasted for four weeks. The questionnaire was pilot tested for understandability using a small subsample of those who shared the major characteristics of the target study population.
All 1,600 fourth- and fifth-year medical students of both genders were eligible to participate in the survey. To ensure a higher response rate, the paper-and-pencil questionnaire was group-administered to these students. We arranged in advance to have the students remain in their classroom past the end of class so that we could ask for study participation. Prior to the distribution of the questionnaire, informed consent was obtained from the potential participants about what participation in the study involved, so that they could make the decision about whether to engage in the research. The approval of the Institutional Review Boards (IRB) was obtained for the purpose of the study.
Variables and Data Sources/Measurement
The dependent variable for this study was choice of specialty: primary care or otherwise. Using previously established definitions per the Saudi Arabian Ministry of Health, we classified family medicine and preventive medicine as primary care specialties.15,16,17 Respondent were asked to indicate their first specialty choice in general. To better categorize levels of interest in the various specialties, students were also asked to rank three additional specialties in which they had the most interest and to identify three specialties that they were least interest in. Students then were asked directly about their desire to choose primary care as a career after they graduate.
We had two independent variables of interest: lifestyle and perceived income. The students were given a list of 26 specialties to rate according to a 3-point scale, with options of lifestyle friendly, lifestyle intermediate, and lifestyle unfriendly. Prior to rating the specialties, students were asked to rate the importance to their career choice of the following criteria: financial compensation, control of the work schedule, workload, enjoyment of the work environment, and enjoyment of this type of work. After that, we asked the students specifically about how they evaluate primary care careers in terms of these five criteria. Then, using a 3-level scale, students were asked to evaluate each medical specialty regarding lifestyle category.
The second independent variable was income disparity. Adapting Newton7 and his colleagues’ survey question, a 4-point scale was used to rate the influence of each attribute on the students’ selection of career specialties, with verbal anchors of “essential,” “very important,” “somewhat important,” and “not important.” Students were asked whether a primary care career provides an income that will allow them to live comfortably, provides an income sufficient to provide adequately for their family, and provides an adequate financial reward for the years of training required. Also, students were asked if the opportunity to work in the private sector is an incentive for excluding the primary care specialty in the career choice. Finally, several covariates were also collected including gender, age, marital status, the type of medical school, and geographical background.
Analysis Methods
Comparisons were made between the selection of students’ two factors (lifestyle and income) to primary care with other specialties. For this report, the only dependent variable we utilized was choice of specialty: primary care. Based on previously established definitions, we classified family medicine and preventive medicine, as primary care careers. The variable was operationalized to reflect one of two possible mutually exclusive occurrences: (0) no primary care selected, meaning that family medicine, or preventive medicine were not indicated as the career choice; or (1) primary care selected, meaning that family medicine, or preventive medicine was indicated as the career choice.
Binary logistic regression was used to model the data, which was conducted in two ways. The first involved including all the variables in the model for each of the two factors: lifestyle and income. Second, we grouped the variables of each factor using the mean method, and then add them to the model. For the lifestyle factor, we summed the students’ rating of each variable in the lifestyle part of the questionnaire, and then divided it by the number of variables, yielding an average score across the various dimensions of lifestyle with a lower score indicating a greater preference for primary care. A similar approach was taken for the income variable. We used this approach to be able to examine the overall importance of the set of dimensions defining each factor in selection of a career specialty. Statistical significance was assessed at the p ≤ 0.05 level. We used the Hosmer-Lemeshow test for postestimation goodness-of-fit test. The result was insignificant with p-value of 0.963, indicating that our model fit the data well. Also, Cronbach’s alpha test was utilized to measure of internal consistency across the dimensions defining each factor, that is, how closely related a set of items are as a group. Scale reliability coefficient was 0.7285, where r=0.7 or greater considered as sufficiently reliable. All analyses were performed using STATA version 14.1.