Hospital characteristics preferred by medical students for their residency programs: A nationwide matching data analysis

Abstract Background In 2004, Japan introduced a mandatory 2‐year postgraduate training program for graduating medical students with a super‐rotation curriculum. A national matching system was established to determine the hospital residency programs best suited for the students. We examined the hospital characteristics preferred by applicants for residencies. Methods A nationwide cross‐sectional study was conducted. Data on salaries, bonuses, and number of accepted ambulances were compiled from the Residency Electronic Information System. Information on the prefectural population, urban area, and number of senior residents (postgraduate years 3–5) for specialty training was extracted from data published on the web page. The ratio of the number of first‐choice applicants to recruitment capacity (matching ratio) for each program was compared between the characteristics of the hospitals and prefectures. Results A strong linear relationship was observed between the number of first‐choice applications and the allocated number of resident positions (correlation coefficient, .72). The matching ratio was greater in community hospitals (2.10 times compared with university hospitals; 95% confidence interval [CI], 1.75–2.53), in hospitals with higher numbers of accepted ambulance cases (1.05 times per 1000 annually; 95% CI, 1.03–1.08), and in hospitals that served a larger prefectural population (1.05 times per million; 95% CI, 1.02–1.08). Conclusions Financial incentives do not seem to attract residency applicants. Applicants prefer non‐university hospitals located in populous areas and those that accept larger number of ambulance cases. To recruit junior residents, an emergency department may need to have higher activity with larger numbers and variety of cases.


| INTRODUC TI ON
In 2004, Japan introduced a mandatory 2-year postgraduate training program for graduating medical students with a super-rotation curriculum. A national matching system was established to determine the hospital residency programs best suited for the students. Teaching hospitals make efforts to recruit prospective residents who provide patient care, thereby representing valuable healthcare providers. Because the allocated number of resident positions is greater than the total number of applicants, there are many hospitals with smaller numbers of residents, and these hospitals are at risk of having a reduced number of healthcare providers and consequently a lower quality of care. Thus, it is important for all teaching hospitals to recruit an adequate number of residents annually.
Numerous studies have reported associations between financial incentives and specialty choices in physicians' career. [2][3][4][5][6][7][8] Enari and Hashimoto analyzed data in 2006 and 2009 and reported that financial incentives affected the choice of training hospital among Japanese medical students who chose non-university settings. 9 In a 2006 survey, Nomura et al 10 found that resident physicians were more satisfied with their residencies at city hospitals in terms of income, residency systems, and clinical skill training than their counterparts who performed their residencies at university hospitals.
However, the findings of these studies rely on data from more than 10 years ago, calling for new analyses based on the most recent data.
The lifestyles and environment surrounding medical students have changed dramatically during the last 10 years. The most notable change in lifestyle likely is the widespread use of smartphones.
Smartphones have enabled medical students to access the Internet anytime and from anywhere. Medical students now are able to collect detailed information about the clinical training hospital and amenities in surrounding areas in real time.
Since we hypothesized that young people would have a preference for large cities, such as Tokyo, Osaka, and Fukuoka, which all are easily accessible, we also surveyed whether hospitals located near a shinkansen (super express train) station were more popular; this point increases the novelty of this study.
Thus, we used the most updated matching data reported by the Japan Residency Matching Program (JRMP) to assess the characteristics of popular hospitals named as first choice by a greater number of graduating medical students during the matching. 11

| Study design
This was a cross-sectional study using nationwide matching data.

| Resident matching system in Japan
The resident matching system matches candidates for postgraduate clinical training with teaching hospitals that conduct postgraduate clinical training. The combination of candidate and hospital is determined by a computer according to a certain rule (algorithm) based on the wishes of both the candidate and the hospital. 11

| Japan Residency Matching Program
The JRMP is composed of the Japan Medical Association, 12 the Foundation for Promotion of Medical Training, 13 the Association of Japan Medical Colleges, 14 and the Association of Clinical Training. 15 The JRMP fulfills various roles in operations related to resident matching. A website related to resident matching has been launched and provides information to participants and hospitals related to recruitment, such as the number of recruited residents for participating hospitals provided by the Ministry of Health, Labour and Welfare.
A helpline has been established to answer queries about matching from participants and participating hospitals and to publicize information related to the inquiries as necessary. Schedule management, algorithm publication, and interim reports are also performed. After residency matching, surveys are conducted to improve the resident matching system. 11

| Data collection
Data on the numbers of applicants naming a hospital as their first choice were collected from the 2019 JRMP report on clinical training matching for junior doctors. This matching system has an interim report and a final report. Data from the interim report were used because they represented more precisely the hospitals' popularity, according to the choices of medical students in the final (6th) year.
Data on monthly salaries, bonuses, number of beds, number of accepted ambulance cases per year, and certification as a tertiary emergency center for each hospital were compiled from the Residency Electronic Information System (REIS). 16 Information on the prefectural population, urban area (200 000 people or greater), the presence of stations for shinkansen (super express trains), and the number of senior residents (postgraduate years 3-5) for specialty training was extracted from data published on the web page.

| Statistical analyses
The ratio of the number of first-choice applicants to recruitment capacity (matching ratio) for each program was compared between the characteristics of the hospitals and prefectures. The mean numbers of first-choice applicants were modeled with negative-binomial model with log-link function, including the characteristics as covariates and the log of the allocated number of residency positions as an offset variable. Correlations between multiple programs within the same hospital were adjusted through generalized estimating equations. In addition, we summarized Pearson's correlation coefficients for program-level variables with program-, hospital-, and prefecturelevel variables at the program-level data (n = 1363); correlations for hospital-level variables with hospital-and prefecture-level variables at the hospital-level data (n = 1020); and correlations between prefecture-level variables at the prefecture-level data (n = 47). All analyses were conducted by SAS version 9.4 (Cary, NC, USA).

| Ethical approval
This study was based on analysis of public data, and therefore, approval of the ethics committee was not required.

| RE SULTS
There were 1363 residency programs in 1020 hospitals (907 community and 113 university hospitals, including affiliated hospitals).  Figure 1 shows the relationship between the mean number of first-choice applications and the number of accepted ambulance cases. In addition, the matching ratio was greater in hospitals that served a larger prefectural population (1.05 times per million; 95% CI, 1.02-1.08) ( Table 2). There were no significant associations between the number of first-choice applications and the location of the hospital near stations for shinkansen (super express trains), residents' salaries, or number of hospital beds.
Despite the strong correlation (.9) between two prefecture-level variables (Table 3), we confirmed that including or excluding these variables from the regressors did not affect the coefficients of other variables. We also assessed the dependence between the variables using a variance inflation factor (without considering the nonlinear model form for matching ratio and the multilevel structure) at the program-level model fit. The values ranged from 1.3 to 2.5 for the program-and hospital-level variables and approximately 6.5 for the prefecture-level variables.

| D ISCUSS I ON
This study is the report from Japan to examine the characteristics of teaching hospitals that are popular choices by medical students using the most updated matching data reported by the JRMP.
Community hospitals (vs university hospitals), increased acceptance of ambulance cases, and larger prefectural populations were significant attributes that appealed to medical students in the final (6th) year who were selecting a teaching hospital.  The matching ratio is defined as the ratio of the number of first-choice applications to the recruitment capacity number. In the negative-binomial log-link regression models for first-choice application number, including log (recruitment capacity number) as an offset variable whose coefficient was set at 1, each exponentiated coefficient is interpretable as the ratio of matching ratios between distinct levels of that variable. The model was fitted by generalized estimating equations using hospitals as clusters factor for residents who wish to do their residency under the mentorship of foreign-national physicians to develop clinical skills with global standards. Fourth, we focused on data from 2019 because we specifically wanted the results of our analysis to reflect the most recent data. This not only allowed us to analyze the newest data, but also limited our study to analysis of the data from a single year.
Finally, we primarily selected the factors that we deemed important among the data on clinical training hospitals accessible within the REIS. However, factors other than these may influence popularity among medical students, such as the number of supervising physicians at each training hospital, online medical resource accessibility, and frequency of lectures related to the primary care. It is another limitation of our study that we did not assess.
In conclusion, financial incentives do not seem to attract residency applicants. Applicants prefer non-university hospitals located in populous areas and those that accept larger numbers of ambulance cases.
To recruit junior residents, an emergency department may need to have higher activity with larger numbers and variety of cases.

ACK N OWLED G EM ENTS
The authors would like to thank Enago (www.enago.jp) for the English language review. This study has not received any funding and has not been presented.

CO N FLI C T O F I NTE R E S T
None.