^{1}

^{2}

^{3}

^{*}

^{4}

^{5}

^{6}

^{7}

^{8}

^{9}

^{10}

^{11}

^{12}

The authors have declared that no competing interests exist.

Conceived and designed the experiments: RS IK HT. Analyzed the data: RS RG. Contributed reagents/materials/analysis tools: RS WW NY. Wrote the paper: RS. Helped draft the manuscript: NY HT RG.

Inequity in physician distribution poses a challenge to many health systems. In Japan, a new postgraduate training program for all new medical graduates was introduced in 2004, and researchers have argued that this program has increased inequalities in physician distribution. We examined the trends in the geographic distribution of pediatricians as well as all physicians from 1996 to 2010 to identify the impact of the launch of the new training program.

The Gini coefficient was calculated using municipalities as the study unit within each prefecture to assess whether there were significant changes in the intra-prefectural distribution of all physicians and pediatricians before and after the launch of the new training program. The effect of the new program was quantified by estimating the difference in the slope in the time trend of the Gini coefficients before and after 2004 using a linear change-point regression design. We categorized 47 prefectures in Japan into two groups: 1) predominantly urban and 2) others by the definition from OECD to conduct stratified analyses by urban-rural status.

The trends in physician distribution worsened after 2004 for all physicians (p value<.0001) and pediatricians (p value = 0.0057). For all physicians, the trends worsened after 2004 both in predominantly urban prefectures (p value = 0.0012) and others (p value<0.0001), whereas, for pediatricians, the distribution worsened in others (p value = 0.0343), but not in predominantly urban prefectures (p value = 0.0584).

The intra-prefectural distribution of physicians worsened after the launch of the new training program, which may reflect the impact of the new postgraduate program. In pediatrics, changes in the Gini trend differed significantly before and after the launch of the new training program in others, but not in predominantly urban prefectures. Further observation is needed to explore how this difference in trends affects the health status of the child population.

Optimizing the distribution of physicians poses a challenge to the health systems of many countries

In Japan, attempts have been made to increase medical student quotas and the number of medical schools in order to increase the numbers of physicians. Thus far, however, these efforts have not resolved the disparity in the distribution of physicians

The Ministry of Health, Labor and Welfare (MHLW) introduced a new postgraduate medical education program to improve residency training in 2004. The new scheme was introduced to address deficiencies in clinical training in the country's 6-year undergraduate medical program. The generally accepted view in the Japanese media is that access to medical care has become more unequal, particularly since the advent of the new postgraduate training program

The situation in pediatrics in Japan is particularly serious. There have been reports of pediatricians' deaths due to overwork and suicides resulting from depression due to overwork

There is a need to address the distribution of physicians according to geographic and specialization needs

In this study, trends in the distribution of pediatricians as well as total numbers of physicians in Japan from 1996 to 2010 were examined to identify the impact of the launch of the new training program in 2004 on physician, especially pediatrician, distribution in Japan.

The Gini coefficient was used to assess the distribution of physicians, as in several previous studies

Japan has three levels of government: municipal, prefectural and national. Municipalities are the basic geographical units of administration. Prefectures and municipalities in Japan are roughly comparable to states and counties in the United States. There are 47 prefectures in Japan. Data on the number of all physicians and pediatricians by municipality level were obtained from the Survey of Physicians, Dentists, and Pharmacologists

Data on municipal populations were obtained from the Basic Resident Registers, which are collected and compiled by the Ministry of Internal Affairs and Communications in March of each year

To calculate the Gini coefficient for all physicians, the general population of each locality was used (i.e. plotting the Lorenz curve which is based on the cumulative proportion of the total population served by physicians within each locality). The x-axis of the Lorenz curve represents the cumulative proportion of total population ranked by physician-to-population ratio and the y-axis represents that of total physicians. To calculate the Gini coefficient for pediatricians, the child population in each locality was used. The x-axis of the Lorenz curve represents the cumulative proportion of child population ranked by pediatrician-to-child population ratio and the y-axis represents that of pediatricians. The child population is defined as the population under 15 years old because children in this age group are treated by pediatricians in Japan.

This study is composed of the following four steps. First, the trends in the Gini coefficient are shown for all of Japan using prefectures as the study unit. Second, the trends in the Gini coefficient are analyzed for all of Japan using municipalities as the study unit. Third, the Gini coefficient is calculated using municipalities as the study unit within each prefecture to assess whether there are significant changes in inequity in the distribution of all physicians and pediatricians before and after the launch of the new training program. The effect of the 2004 postgraduate training program is quantified by estimating the difference in the slope in the time trend of the physician Gini using a linear change-point regression procedure. We fit the following model for the study outcome:_{ij} ≥2004 and is otherwise equal to 0. When expressed in terms of the mean response prior to and after 2004,

Thus, β_{3} provides a measure of the difference in the trend in E (Y) prior to and after the year 2004 and can be interpreted as the effect of the new postgraduate medical education program. Hereafter, the ^{2} are defined as urban, and others are defined as rural. Japan underwent administrative re-organization by a large-scale merging of municipalities. Therefore, the total number of municipalities dramatically decreased during the study period. The number of physicians and the population of the municipality in every data set were adjusted for the new municipal boundaries, by merging former smaller municipality into later larger ones. To examine the trend in the geographic distribution of physicians using the Gini coefficient, the number and boundaries of geographic units must be fixed. Therefore, the 2010 boundaries (n = 1,750) were used for all time points.

A two-tailed p value of less than 0.05 was considered statistically significant. All analyses were performed using SAS software 9.2 (SAS Institute, Inc., Cary, NC).

1996 | 1998 | 2000 | 2002 | 2004 | 2006 | 2008 | 2010 | |

Total population (*1,000,000) | 124.9 | 125.6 | 126.1 | 126.5 | 126.8 | 127.1 | 127.1 | 127.1 |

Child population (*1,000,000) | 19.7 | 19.1 | 18.6 | 18.1 | 17.8 | 17.5 | 17.3 | 17.1 |

Number of total physicians | 230,297 | 236,933 | 243,201 | 249,574 | 256,668 | 263,540 | 271,897 | 280,431 |

Number of pediatricians | 13,781 | 13,989 | 14,160 | 14,481 | 14,677 | 14,700 | 15,236 | 15,870 |

Per capita total physicians | 184.4 | 188.7 | 192.9 | 197.3 | 202.4 | 207.4 | 214.0 | 220.7 |

Per child capita pediatricians | 70.3 | 73.2 | 76.3 | 79.9 | 82.5 | 83.8 | 88.1 | 93.0 |

1996 | 1998 | 2000 | 2002 | 2004 | 2006 | 2008 | 2010 | |

Physician totals | ||||||||

Gini coefficient by prefecture | 0.12 | 0.12 | 0.12 | 0.12 | 0.12 | 0.11 | 0.11 | 0.11 |

Gini coefficient by municipality | 0.34 | 0.34 | 0.33 | 0.33 | 0.33 | 0.33 | 0.33 | 0.33 |

Pediatrician | ||||||||

Gini coefficient by prefecture | 0.13 | 0.12 | 0.12 | 0.12 | 0.12 | 0.12 | 0.11 | 0.11 |

Gini coefficient by municipality | 0.38 | 0.37 | 0.37 | 0.37 | 0.36 | 0.36 | 0.36 | 0.37 |

Effect | Estimate | SE^{a} |
p value | |

all physicians | ||||

β0 | intercept | 0.2838 | 0.00916 | <.0001 |

β1 | year | −0.0028 | 0.00073 | 0.0002 |

β2 | z^{b} |
−0.0185 | 0.0043 | <.0001 |

β3 | z ^{b}•year |
0.00625 | 0.00103 | <.0001 |

pediatrician | ||||

β0 | intercept | 0.3444 | 0.01035 | <.0001 |

β1 | year | −0.0042 | 0.00146 | 0.0042 |

β2 | z^{b} |
−0.0248 | 0.00865 | 0.045 |

β3 | z ^{b}•year |
0.00575 | 0.00207 | 0.0057 |

A total of 13 prefectures (Miyagi, Saitama, Chiba, Tokyo, Kanagawa, Shizuoka, Aichi, Kyoto, Osaka, Hyogo, Nara, Hiroshima, and Fukuoka) were defined as predominantly urban prefectures and the remaining 34 are defined as others. In 2010, 58.4% of the total population of Japan lived in predominantly urban prefectures, and the remaining of 41.6% lived in others. _{1}_{1}_{3}_{3}_{1}_{3}

Predominantly urban prefectures^{a} |
Others | |||||||

(n = 13) | (n = 34) | |||||||

Effect | Estimate | SE^{b} |
p value | Estimate | SE^{b} |
p value | ||

β0 | intercept | 0.3192 | 0.02013 | <.0001 | 0.2702 | 0.00968 | <.0001 | |

β1 | year | −0.00510 | 0.00136 | 0.0003 | −0.0019 | 0.00079 | 0.0201 | |

β2 | z^{c} |
−0.0206 | 0.00806 | 0.0124 | −0.0177 | 0.00468 | 0.0002 | |

β3 | z^{c} •year |
0.00645 | 0.00193 | 0.0012 | 0.00617 | 0.00112 | <.0001 | |

β0 | intercept | 0.3542 | 0.01746 | <.0001 | 0.3406 | 0.0127 | <.0001 | |

β1 | year | −0.0042 | 0.00249 | 0.0943 | −0.0042 | 0.00178 | 0.0187 | |

β2 | z^{c} |
−0.0245 | 0.01476 | 0.1008 | −0.0249 | 0.01054 | 0.0192 | |

β3 | z^{c} •year |
0.00677 | 0.00353 | 0.0584 | 0.00536 | 0.00252 | 0.0343 |

Prefecture with central cities of major metropolitan areas ^{a} |
Other prefectures | |||||||

(n = 14) | (n = 33) | |||||||

Effect | Estimate | SE^{b} |
p value | Estimate | SE^{b} |
p value | ||

All physicians | ||||||||

β0 | intercept | 0.3102 | 0.02038 | <.0001 | 0.2726 | 0.009602 | <.0001 | |

β1 | year | −0.00419 | 0.001263 | 0.0013 | −0.00215 | 0.000845 | 0.0117 | |

β2 | z^{c} |
−0.01943 | 0.007474 | 0.0108 | −0.01809 | 0.004998 | 0.0004 | |

β3 | z^{c} •year |
0.006029 | 0.001787 | 0.0011 | 0.006344 | 0.001195 | <.0001 | |

Pediatricians | ||||||||

β0 | intercept | 0.3489 | 0.0187 | <.0001 | 0.3424 | 0.01257 | <.0001 | |

β1 | year | −0.00291 | 0.002268 | 0.2029 | −0.00477 | 0.001853 | 0.0106 | |

β2 | z^{c} |
−0.02027 | 0.01342 | 0.1342 | −0.02665 | 0.01096 | 0.0158 | |

β3 | z^{c} •year |
0.004256 | 0.003208 | 0.1877 | 0.006385 | 0.002621 | 0.0156 |

When population density was used as the urban/rural definition, a total of 7 prefectures (Saitama, Chiba, Tokyo, Kanagawa, Aichi, Osaka, and Fukuoka) are defined as urban and the remaining 40 are defined as rural. In 2010, 44.5% of the total population of Japan lived in urban areas, and the remaining 55.5% lived in rural areas. Namely, almost half of Japanese population lived in the above-mentioned 7 prefectures. The pattern of results was similar under this definition as well. The mean of the Gini coefficient in urban areas is higher for all physicians and pediatricians during study period (

Prefectures with population density | Prefectures with population density | ||||||

> = 1000/km^{2} ^{a} (n = 7) |
<1000/km2 (n = 40) | ||||||

Effect | Estimate | SE^{b} |
p value | Estimate | SE^{b} |
p value | |

β0 | intercept | 0.329 | 0.026 | <.0001 | 0.2759 | 0.009609 | <.0001 |

β1 | year | −0.00587 | 0.001322 | <.0001 | −0.00221 | 0.000754 | 0.0036 |

β2 | z^{c} |
−0.01104 | 0.007818 | 0.1646 | −0.01979 | 0.004462 | <.0001 |

β3 | z^{c} •year |
0.004126 | 0.001869 | 0.0323 | 0.006622 | 0.001067 | <.0001 |

β0 | intercept | 0.3516 | 0.01839 | <.0001 | 0.3431 | 0.01179 | <.0001 |

β1 | year | −0.0023 | 0.002569 | 0.3752 | −0.00455 | 0.00166 | 0.0065 |

β2 | z^{c} |
−0.01255 | 0.0152 | 0.4131 | −0.02689 | 0.009821 | 0.0066 |

β3 | z^{c} •year |
0.001765 | 0.003634 | 0.6295 | 0.006449 | 0.002348 | 0.0064 |

In this study, distributions of all physicians and pediatricians were examined over the period straddling the introduction of the new postgraduate medical training scheme in 2004. We note the following two major findings. First, the changes in trends in the geographic intra-prefectural distribution both all physicians and pediatricians differed significantly between the

Previous quantitative analyses of physician distribution in Japan

Our study improves on the previous four studies because we used a longer period of observation after the launch of the new training system, which allowed us to have more power to detect the differences. We also showed statistically significant differences by employing proper statistical analysis. We believe that this study adds new evidence to the existing literature and heightens the debate about the impact of the postgraduate training program on physician distribution. The results of our study show that the intra-prefectural distribution of physicians worsened the

It is noteworthy that there was a significant change in trends of pediatrician intra-prefectural distribution between the

Our study has the following limitations. First, the Survey of Physicians, Dentists, and Pharmacologists does not include data on whether a physician works full time or part time. As a result, this analysis was based on an overall headcount, which might overestimate the number of physicians. In particular, the percentage of female physicians in pediatrics is high

Despite these limitations, we believe that our study contributes to the debate about the impact of the new postgraduate training program on inequality in the physician supply for the following reasons. First, we used time series data over a comparatively long period. Second, this is the first study to show statistically significant differences in the trends of physician distribution before and after the launch of the new training program. Third, our analyses reveal that a detailed and disaggregated approach by specialties is needed although previous studies considered only the overall number of physicians

The changes in trends in the intra-prefectural distribution of physicians and pediatricians differed significantly before and after the launch of the new postgraduate training program, and our findings suggest an adverse impact of the new postgraduate training system. In pediatrics, changes in the Gini trend differed significantly before and after the launch of the new postgraduate training program in other prefectures, but not in predominantly urban prefectures. Further observation is needed to explore how this difference in trends affects the health status of the child population.

(DOCX)

(DOCX)

(DOCX)

(DOCX)

This study was conducted with the support of Harvard Catalyst and the Takemi Program in International Health at Department of Global Health and Population, Harvard School of Public Health. We would like to thank Usheer Kanjee,