Relationship between medical students’ career priority and specialty choice: A nationwide multicenter survey

Abstract Background The shortage of physicians in several specialties has been brought to public attention in several countries. However, little is known about factors affecting medical students’ specialty choice. The objectives of our study were to illustrate medical students’ career priority clusters and to assess their association with specialty preference. Methods We conducted a nationwide multicenter survey in 2015 at 17 medical schools. The study participants were asked their top three specialty preferences, demographic characteristics, and 14 career priority questions. Multilevel logistic regression models were used to determine the effect of each variable on student career choice. Results A total of 1264 responses were included in the analyses. The top five specialty choices were internal medicine: 833, general practice: 408, pediatrics: 372, surgery: 344, and emergency medicine: 244. An exploratory factor analysis mapped the 14 career priorities into 3‐factor solution: “primary care orientation,” “advanced and specific care,” and “personal life orientation.” Multilevel logistic regression models yielded satisfactory accuracy with the highest ROC curve (AUROC) noted in surgery (0.818), general practice (0.769), and emergency medicine (0.744). The career priorities under “primary care orientation” had positive association with choosing general practice, emergency medicine, internal medicine, and pediatrics. The “advanced and specific care” career priorities facilitated surgery and emergency medicine choice, while reducing the likelihood of choosing less procedure‐oriented specialties, such as internal medicine, general practice, and pediatrics. Conclusions Our results demonstrated medical students’ career priorities and their association with specialty preference. Individualized career support may be beneficial for both medical students and each specialty fields.

included in the original JMECS study were students' demographics (age, gender, birthplace, other academic or professional experiences prior to medical school, physician parent, and plan to inherit an existing practice) and 14 career priority questions with 6-point Likert scale answers ranging from 1 (strongly disagree) to 6 (strongly agree).

| Statistical analysis
The primary outcome variable was dichotomized based on whether or not each specialty field was included in up to three career choices.
Explanatory variables excluding age were treated as nominal variables. To investigate the structure of career priorities, an exploratory factor analysis with minimum residual solution and Harris-Kaiser's orthoblique rotation were conducted. The number of factors was determined using a scree plot, which showed the eigenvalues on the y-axis and the number of factors on the x-axis. Items with factor loading below 0.4 or above 0.4 for two or more domains were eliminated. A series of multilevel logistic regression analyses were conducted to examine the effect of each demographic and career priority variable on student career choice. School ID was set as level 2 and subject ID as level 1. Odds ratios and 95% confidential intervals were calculated from estimates and corresponding standard errors. Each model for five specialties was validated using a 2 × 2 table and AUROC. All the analyses were conducted with R 3.5.3 in RStudio 1.2.1335, and with following packages: haven, 11 tidyverse, 12 psych, 13 GPArotation, 14 lme4, 15 pROC, 16 and tableone. 17   Students who chose surgery and emergency medicine gave higher priority to "mastering advanced procedures" and "acute care rather than chronic care," while those who chose general practice and pediatrics generally gave higher scores to "frequent patient communication." Exploratory factor analysis revealed three major factors: "primary care orientation," "advanced and specific care," and "personal life orientation," while two items with factor loading below 0.4 were eliminated. Table 2 presents factor loadings of career priorities based on Harris-Kaiser's orthoblique rotation.

| General practice
In contrast to internal medicine, while general practice students

| Pediatrics
Students who chose pediatrics showed interest in frequent patient

| Surgery
Female students were less likely to consider surgery as their specialty choice as compared to male students (OR 0.55 [0.39-0.78]).
Differences between surgery students and internal medicine/general practice students were largely driven by career priority variables in "primary care orientation" and "advanced and specific care." For instance, surgery students were less interested in clinical diagnos-

| Emergency medicine
The strongest driver for choosing emergency medicine was a preference in acute care (OR 2.14 [1.79-2.57]). In addition, students who chose emergency medicine shared several common features with surgery students. First, a significantly lower number of female

| D ISCUSS I ON
In 2020, a total of 9082 residents began residency training under the new board certification system in Japan. The number of new residents for each specialty in 2020 were as follows: internal medicine:   were not as important to them. 20 Vaidya et al found that surgery, emergency medicine, and gynecology and obstetrics students demonstrated a higher "novelty-seeking" tendency. They also reported that surgery students had lower "harm avoidance" and "reward dependency" scores. In contrast, students who chose primary care, emergency care, and gynecology and obstetrics had a high "reward dependency". 8 As aforementioned, the concept by Weiss et al 10  Our results may imply that individualized career support based on student's preference in three simple factors, "primary care F I G U R E 1 Adjusted odds ratios for specialty preference. The central points of each horizontal line represent the adjusted odds ratios for each item, and the lines demonstrate 95% confidence intervals orientation," "advanced and specific care," and "personal life orientation," in addition to knowing key differences among the cluster of related specialties, may be beneficial to facilitate the recruitment process in each specialty field.
This study has several limitations. First, the cross-sectional data might not reflect medical students' actual career choice.
Thus, our models need to be validated in the future using longitudinal cohort of medical school graduates. Secondly, social desirability response bias may have led to ceiling effects on several career priority variables, which could undermine the discrimination capacity of the models. The relatively large sample size may have amplified the small effects. In addition, our results may not be applicable to junior residents since the study was limited to undergraduate medical education.
Our results demonstrated medical students' career priorities and their association with specialty preference, using the concept of a cluster of related specialties and a cluster of career priority features. Since the majority of students have several career options, using the cluster of career priority features in addition to knowing key differences among the cluster of related specialties may be beneficial for both medical students and each specialty field, as well as for those considering interventions to manage the medical workforce.

ACK N OWLED G EM ENTS
We would like to thank the respondents of our survey for their support. We would also like to thank the faculty and staff in the

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interests for this article.