Geographic distribution of family physicians in Japan and the United States: a cross-sectional international comparative study

Background: Family physicians are known to distribute more equally among the population than other physicians. The maturity of family medicine, i.e. the length of its history as a part of healthcare system and the population of qualied family medicine experts, may affect the distribution, but this has not been shown in the literature. This study compares the geographic distribution of family physicians in Japan and the United States (U.S.), both of which are developed countries without a physician allocation system by the public sector, but the two countries differ greatly in the maturity of family medicine as a clinical specialty. Methods: This is a cross-sectional international comparative study using publicly available online database on the number of physicians in Japan (Board-certication Database of Japan Primary Care Association, and Survey of Physicians, Dentists and Pharmacists by Ministry of Health, Labour and Welfare) and the U.S. (Area Resource File by Health Resources and Services Administration). The municipalities in Japan and counties in the U.S. were divided into quintile groups according to population density. The number of family physicians per unit population in each group of areas was calculated. The geographic distribution of all physicians in Japan was simulated assuming that the proportion of family physicians among all physicians in Japan (0.16%) was increased to that in the U.S (11.8%). Results: The distribution of family physicians in Japan noticeably shifted to the areas with the lowest population density. In contrast, family physicians in the U.S. distributed equally across areas. The distribution of physicians with other specialties (general internists, pediatricians, surgeons and obstetricians/gynecologists) shifted heavily to the areas with highest population density in both countries. The simulation analysis showed the geographic maldistribution of all physicians improved substantially if the proportion of family physicians in Japan increases to that in the U.S. Conclusion: The distribution of family physicians is more equal than other medical specialists, and the immaturity of family medicine can even lead to a rural-biased distribution. In a country with emerging family medicine such as Japan, increasing the number of family physicians may effectively mitigate the urban-rural imbalance of physician supply.

conducted in some European countries. The other reason is that in both countries a major part of healthcare system is privately operated and owned. Practice of physicians in both countries is a personal business, or an employment by a for-pro t private or semi-private hospital. In such a society, unlike in societies whose healthcare system is dominated by publicly owned medical institutions, the distribution of physicians is driven largely by the market [10]. Thus, the distribution in both countries can be interpreted in the same dynamics, which is different from the government-led physician distribution in some European countries and Canada. Between Japan and the U.S. we can assess the relationship between the maturity of family medicine and the geographic distribution of family physicians excluding the impact of the intervention by the public sector on the distribution.
We thus evaluate the geographic distribution of family physicians and other major clinical specialists in Japan and the U.S., and compare them internationally. Family physicians should be distributed equally because family medicine improves health indicators for all community residents, and contributes to patient satisfaction and lower health care costs [11]. We also simulate the distribution of all physicians in Japan, assuming that the proportion of family physicians among all the physicians step-wisely increases to the level of that in the U.S. Based on the results, we will discuss the association between the maturity of family medicine and the geographic distribution of family physicians, and how important it is to increase the number of family physicians to reduce the urban-rural imbalance of physicians.

Methods
Study setting This is a cross-sectional international comparative study using publicly available online database including all municipalities in Japan, counties and the District of Columbia in the U.S.

De nition of family medicine
This study de nes family medicine as medical care provided by JPCA-certi ed family physicians (in Japan), and as medical care provided by American Board of Family Medicine-certi ed family physicians (in the U.S.).

Data of physicians in Japan
We collected an online database of JPCA-certi ed family physicians in 2018 who gave permission to the JPCA o ce to make their information available online [12]. The data disclosed each physician's name, workplace (municipality and medical institution) and area of special interest. The number of physicians in other specialties who worked as clinicians in a medical institution and were board-certi ed by the Japanese Society of Internal Medicine, Japan Pediatric Society, Japan Surgical Society, and Japan Society of Obstetrics and Gynecology was collected from the online database of the Survey of Physicians, Dentists and Pharmacists 2018 compiled by the Ministry of Health, Labour and Welfare [13].

Data on municipalities in Japan
Japan is divided into three administrative levels: national, prefectural and municipal. The number of municipalities (cities, towns and villages) was 1741, and we collected data on 1737. Population and population density in 2015 were collected for each municipality from online database of population census compiled by the Ministry of Internal Affairs and Communications [14]. We de ned the rurality of each municipality by its population density. We calculated the population density of a municipality by dividing the population (number of residents) by land area (square kilometers). We divided the municipalities into quintiles (i.e. one-fth of a set) according to population density so that the number of municipalities in each group was 20% of all the municipalities (Quintile 1 < = 43.4, Quintile 2 < = 140.6, Quintile 3 < = 403.9, Quintile 4 < = 1650.4, Quintile 5< = 22380.2 people per square kilometer). The municipality boundary data was obtained from the National Spatial Planning and Regional Policy Bureau of the Ministry of Land of Japan. pediatricians, general surgeons, and obstetricians and gynecologists) was used for the analysis. Population information in 2018 and area information in 2010 for each county contained in the database was also used for the analysis. To evaluate rurality, as in Japan, counties were divided into quintiles according to population density (Quintile 1 < = 4.6, Quintile 2 < = 11.9, Quintile 3 < = 24.3, Quintile 4 < = 67.4, Quintile 5 < = 18721.1 people per square kilometer). Census boundary data was obtained from the U.S. Census Bureau TIGER/Line Shape les.

Simulation analysis
We simulated the geographic distribution of all the physicians in Japan assuming that the proportion of family physicians among all the physicians in Japan increased stepwise to that in the U.S. This analysis was based on the following assumptions: 1) the number of family physicians increased and the number of the other physicians decreased, resulting in no change in the total number of physicians; 2) the distribution patterns of family physicians and the other physicians remained constant.
The proportion of family physicians in Japan was 0.16%, while the proportion of family physicians in the U.S. was 11.8%. To increase the proportion to the level of the U.S., the number of family physicians in Japan needed to increase by 72.6 times and the number of other specialists needed to decrease by 0.88 times, resulting in the total number of physicians holding steady. We multiplied family physicians and other physicians as such, while maintaining the distribution of each group of physicians. We then calculated the total number of physicians per population in each group of population density quintile, and compared it before and after the simulation. We simulated three patterns of increase of family physicians: 10, 30 and 72.6 times the current level. The magni cation of 10 times was derived from the assumption that the number of the entry to general practice (sougou-shinryo) training program triples (570) the current level (190) and this condition lasts for 20 years. Also, the 30 times assumes that the number increased by ve times (950) and lasts for 20 years. The distribution of family physicians can be in uenced by factors that change as the number of family physicians increases. For example, the size of the population accustomed to family physicians, the age distribution of family physicians, and the amount of money family physicians can save for starting their private practice could change as this number increases. To simplify the calculation, this simulation does not consider these factors.

Statistical analysis:
We used the chi-square test to compare the distribution of ordinal variables among ve groups of areas in both countries. We used the residual analysis to compare the real value of the number of family physicians per unit population and the expected value derived from the distribution of all family physicians in both countries. All statistical analyses were performed using Microsoft Excel and STATA/SE version 15 (Stata Crop 2017). We excluded data of four (0.2%) municipalities and nine (0.3%) counties because of a lack of information about the number of physicians or population density. A p-value, 0.05 (two-sided test) was considered statistically signi cant.

Results
The national population in Japan was 127,094,745 and that of the U.S. was 330,530,359. There were 320,084 physicians in Japan and 899,244 in the U.S. There were 519 (77.2%) family physicians in Japan and 105,999 (100%) in the U.S.; this data was available in an online database. Fig. 1 shows the number of family physicians per unit population in each quintile of population density in Japan and in the U.S. The concentration of family physicians in Japan was relatively low in the densely populated coastal areas. In the less densely populated inland mountain areas, the concentration of family physicians was relatively high. In contrast, family physicians in the U.S. did not concentrate in certain geographic locations. These patterns of distribution also appear in Fig.   2 which depicts the number of family physicians per unit population in each quintile of population density. The distribution of family physicians in Japan shifted substantially to Quintile 1 (the lowest population density) municipalities ( Fig. 2A). In contrast, family physicians were uniformly distributed across quintiles in the U.S. (Fig. 2B). Chi-square tests showed that the distribution of Japanese family physicians among the quintile groups differed signi cantly from that of American family physicians (p<0.001). The residual analysis showed that the number of family physicians in Quintile 1 in Japan (117 per 10,000,000) was signi cantly higher than the number expected from the distribution of all family physicians (the total of Japanese and U.S. family physicians) (73.5 per 10,000,000) (p < 0.001). The number of family physicians in Quintile 5 in Japan (31 per 10,000,000) was signi cantly lower than the number expected from the distribution of all family physicians (56.8 per 10,000,000) (p<0.001). Fig. 3 shows the distribution of other specialists per 100,000 population classi ed by population density quintile in Japan and the U.S. The distribution of other specialists shifted markedly to urban areas in both Japan and the U.S., unlike that of family physicians.
In the simulation analysis, assuming an increase in the number of family physicians in Japan, the geographic maldistribution of all physicians in Japan would improve stepwise if the number of family physicians increased from its current level to 10, 30 and 72.6 times greater (Fig. 4). For example, the number of physicians in Quintile 1 increased by 1.64 times from 2558 to 4199 if the number of family physicians increased to the same level of the U.S. (72.6 times)

Discussion
The distribution of family physicians in Japan signi cantly shifted to rural areas compared with the distribution of family physicians in the U.S.; in both countries other specialists concentrated heavily in urban areas. If the proportion of family physicians in Japan increased to that in the U.S., the geographic maldistribution of all physicians improved substantially. These results have two implications. One implication is that the distribution of family physicians is roughly equitable among the population in any society compared with that of other specialists, but the distribution can be biased substantially to rural areas in a society in which family medicine is emerging. The second implication is that, based on the rural-biased distribution of family physicians in Japan, increasing the number of family physicians can resolve the urban-rural imbalance in the supply of physicians.
Family physicians distribute equally among population in most societies with a mature primary care system. One reason for the equal distribution is that family physicians can make the most of their expertise in the rural healthcare environment [16,17]. In urban areas, there are many more medical facilities, including highly specialized ones, and the geographic accessibility to medical care is much higher than in rural areas [18]. This gives patients in urban areas more choices when it comes to their care. They can go to medical institutions that specialize in their health conditions [19]. In contrast, in rural areas, health care providers are required to provide holistic and comprehensive care [20]. The absence of nearby health care facilities also makes it easier for physicians to provide continuity of care to a single patient [16]. Therefore, family physicians can provide continuous and comprehensive services in rural areas, two of the core values of primary care [16,21,22]. In contrast, for other specialists, rural areas may not be an ideal place to apply their expertise. They need to treat a narrower spectrum of diseases, which is rare in rural areas, and thus need to cover a wider geographic area with a larger population to stay in business [23]. As a result, the distribution of other specialists tends to shift to urban areas [24].
In Japan however, the distribution of family physicians shifted to the rural areas. This can be linked to the immaturity of family medicine in Japan's healthcare system. In Japan only about 10 years have passed since family medicine was established as a clinical discipline with a board-certi cation system. Only 0.2% of all physicians in Japan are board-certi ed family physicians. Urban primary care providers were other specialists originally employed by a hospital, who then opened their own private clinics [5]. These former specialists treat some patients within their specialty, and also offer primary care services without being trained in primary care or certi cation as a family physician. It is di cult, if not impossible, for family physicians, all of whom were board-certi ed recently, to enter this specialoid-saturated market. Therefore, they might prefer to work in rural areas where competition with these specialoids is less erce. In a society like Japan's, with immature family medicine, family physicians are at a disadvantage in the health professional market.
Japan's unique system of healthcare provision may also have affected the distribution of family physicians. Almost all physicians who have completed postgraduate clinical training are employed by a hospital as a domain-speci c specialist early in their career. Even if these physicians, including younger family physicians, would prefer to work in a clinic, it is nancially impossible for them to do so, because most clinics in Japan are private solo practices that require approximately US$880,000 to establish [25]. Therefore, it is extremely di cult for younger family physicians to work in their own clinics. To continue practicing family medicine, a substantial proportion of them must be employed by a public clinic set up by the local government to provide medical care in Japan's rural, remote and underserved areas. These are, however, a small fraction of all clinics in Japan.
The simulation analysis showed that the geographic maldistribution of physicians in Japan improved with a larger number of family physicians. The geographic maldistribution of physicians has been a serious social problem in Japan [26]. To compensate for the shortage of doctors in rural areas, public medical institutions in the rural areas were established by the national and municipal governments. Jichi Medical University, established solely for the purpose of training rural physicians, was established in 1972, and a regional quota, a special admission quota for producing rural physicians, was spread among most of the medical schools in Japan for the last ten years [27,28]. Even with these ambitious national policies, the maldistribution of physicians has not changed. Instead, there is an even greater disparity in the supply of physicians between rural and urban areas [26,29,30]. The unique distribution of family physicians in Japan could contribute to ameliorating the maldistribution of physicians. Based on the results of this study, policies to increase the number of family physicians and general practitioners and reining in the number of other specialists is a plausible solution for national and local governments [31].
This study has several limitations. First, we could not obtain the data on all JPCA-certi ed family physicians because individual information of some of them was not available on the JPCA website. We considered, however, that selection bias for Japanese family physicians was minimal given the high (77.2%) covering proportion. Second, the simulation analysis assumed the distribution of family physicians did not change when it actually increased dramatically. If the number of family physicians increased to such an extent, it is possible that their geographic distribution substantially changed because of changes in government incentives, savings, supply-demand balance, and recognition of family physicians among the population. This change would enable family physicians to replace specialoids in urban areas and leads to a total distribution of physicians different from that shown by the simulation. Above all it was not realistic that the number of family physicians in Japan increased by 72 times in a short period. The simulation was a thought experiment under an extreme assumption that showed, in a simple manner, the magnitude of the effect of the policy to increase the number of family physicians.
Two other important issues are the comparability of Japan and the U.S. and the limited generalizability based on the two-country comparison. The healthcare systems of Japan and the U.S. are quite different, especially in terms of their medical insurance systems. The results should therefore should be interpreted cautiously. Adding another country or two to the comparison would make the conclusion sounder. For example, the maturity of family medicine is even higher in the United Kingdom and the Netherlands than in the U.S; the percentages of general practitioners (GPs) among all doctors are 26.2%, 45.4%, 11.7%, respectively [8]. These European countries have a system of population registration at each GP practice and seem to have a smaller disparity of the GP supply between urban and rural areas than Japan and the U.S. although there is still some di culty in recruiting physicians to work in rural areas [24,26,32,33,34]. Adding these countries to the international comparison would reveal the effect of such a socialized system of primary care provision.

Conclusion
The distribution of family physicians is more equal than other medical specialists and can be affected by the maturity of family medicine of the society. In countries like Japan in which family medicine has emerged recently, the distribution can be biased to rural areas. In such countries, increasing the number of family physicians and controlling the number of other specialists may be a way to rectify the urban-rural imbalance of physicians. This study needs no ethical approval because the data used for the analyses was all available online.

Consent for publication
Not applicable.

Availability of data and materials
Available online to anyone.

Competing interests
The authors declare that they have no competing interests.