Abstract

Background

In the context of public health emergencies, the presence of medical and health talents (MHT) is critically important for support in any country or region. This study aims to analyze the spatial and temporal distributions and evolution of MHT in China and propose strategies and recommendations for promoting a balanced distribution.

Methods

This research used data from 31 provinces in China to construct a multidimensional index system for measuring the agglomeration level of MHT. The indices include talent agglomeration density (TAD), talent agglomeration scale (TAS), talent agglomeration intensity (TAI), and talent agglomeration equilibrium (TAE). Using provincial data from the years 1982, 1990, 2000, 2010, and 2020, a spatiotemporal analysis of the MHT agglomeration levels was conducted. Furthermore, the regional dynamic distribution of MHT was analyzed using kernel density estimation diagrams. The spatial autocorrelation of MHT was assessed through global and local Moran鈥檚 I, and the spatial gap and decomposition of MHT were analyzed using the Dagum Gini coefficient.

Results

From the temporal level, the TAD and TAI of MHT showed an increasing trend over the studied period, whereas TAS decreased and TAE first increased and then decreased from 1982 to 2020. At the spatial level, the TAD, TAS, TAI, and TAE of MHT exhibited varied patterns among the eastern, central, and western regions of China, showing significant geographical disparities, generally demarcated by the Hu Huanyong Line. The regional dynamic distribution level of MHT in the country and the three regions were expanding. Spatial autocorrelation analysis using global and local Moran鈥檚 I for TAD, TAS, TAI, and TAE demonstrated significant regional differences. The Dagum Gini coefficient of TAD, TAS, TAI, and TAE revealed divergent trends in regional disparities, with overall declines in disparities for TAD and TAI, a slight increase for TAS, and fluctuating patterns for TAE.

Conclusions

From a temporal perspective, the overall number of MHT in China has been increasing annually at the national and provincial levels. From the spatial perspective, TAD, TAS, TAI, and TAE exhibit significant differences among the three regions. Kernel analysis reveals that the distribution differences are gradually expanding in national level and varying in regional level. Moreover, the global and local Moran鈥檚 I indices reveal varying spatial autocorrelation for TAD, TAS, TAI, and TAE. The Dagum Gini coefficients of TAD, TAS, TAI, and TAE show different patterns of decomposition.

Introduction

The outbreak of COVID-19 at the end of 2019 significantly affected the development of countries worldwide and posed a major test for China鈥檚 medical and health systems under the 鈥淗ealthy China 2030鈥� development strategy [1]. As the country with the largest population, China鈥檚 regions and provinces exhibited varying speeds and accuracy in fighting the epidemic. Medical and health talents (MHT), as the frontline personnel in this battle form the main force for effective anti-epidemic efforts in all regions and provinces and play a vital role in public health. In response, China鈥檚 State Council released the 鈥�14th Five-Year Health and Talent Development Plan鈥� in 2022 [2], which aims to accelerate the expansion of high-quality medical and health resources; ensure balanced regional distribution; narrow inequalities in resource allocation, serviceability, and health levels between urban and rural areas, regions, and group of populations; and strengthen the development of MHT.

MHT is healthcare professionals, including doctors (physicians), nurses, pharmacists, laboratory technicians, and other similar healthcare personnel, who are a crucial factor in responding to public health emergencies [3,4,5]. The total number of medical and health talents in China is growing steadily, from 3.14听million in 1982 to 10.67听million in 2020. Despite China鈥檚 remarkable achievements in combating the epidemic, the unbalanced distribution of health resources remains a significant problem, arising not only from internal structural inequalities in resource allocation but also from multidimensional external inequalities between urban and rural areas and various regions [6]. The average annual growth rate of the health resource density index of public health personnel in China from 2012 to 2018 was 4.45%. The Gini coefficient of public health personnel allocation per population is 0.108鈥墌鈥�0.138. Although the allocation level of MHT in China is constantly increasing, the growth rate is slow, and clear disparities exist in resource allocation among regions with different levels of economic development [7]. The growth rate of the number of public health personnel per 1,000 population is significantly polarized among provinces, with 17 provinces experiencing positive growth and 14 provinces experiencing negative growth [8]. In 2018, the total number of specialized public health institutions in China reached 18,033, but there are great differences among provinces [9]. The main reason for these imbalances lies in the disparity of material and financial resources [10]. Since the COVID-19 outbreak, China has mobilized health personnel from all regions and provinces where the epidemic was concentrated, enhancing public trust in medical and health experts [11]. However, the pandemic has demanded substantial work and effort from MHT, increasing the vulnerability of the healthcare workforce [12], particularly in areas with fewer healthcare workers, such as rural China [13]. Therefore, as the most critical and mobile resource among health resources, MHT is indispensable. A comprehensive study of the spatial and temporal distributions and evolution of MHT will help populous countries such as China in better responding to public health emergencies.

The inequitable allocation of health resources is a common issue in many countries. For instance, in the Philippines, significant health inequalities exist, with poorer health outcomes and fewer physicians in rural areas compared to urban areas [14]. In India, emerging health needs and uneven distribution of health human resources result in poor access to quality healthcare in rural areas [15]. Poland has developed a health needs map to help policymakers effectively manage current health resources [16].

To address the inequitable distribution of MHT, the distribution and evolution of MHT must be identified. In China, some research have four indicators- the number of public health institutions, the amount of public health technical personnel, the number of beds, and equipment in public health institutions [17]. A large number of studies [18, 19] in China often regard MHT merely as part of overall health resources, however, there has been limited in-depth analysis of MHT as a distinct subset. Some studies address the regional distribution of physicians, but physicians represent only one type of medical and health talents [20]. Focusing solely on physicians is insufficient, as their effective functioning depends on collaboration with nurses, technicians, and other medical and health personnel, which is particularly critical during public health emergencies. Other studies focus on the primary healthcare (PHC) workforce, using three groups of indicators: health professionals, physicians and registered nurses [21]. These studies focus on the total number of health care workforce, without considering the relative indicators. Some research uses relative metrics, such as the number of doctors per million population [19] or health resource density index of physician and nurse [22]. These researches rely on single relative index of health resource intensity or density, which can鈥檛 fully reflect the relative level of indicators. Furthermore, a notable amount of studies focus on static spatial analyses at specific time points to assess regional disparities, while neglecting long-term dynamic evolution trends [23, 24].

The application of spatial-temporal analysis methods in public health has made significant progress, particularly in analyzing disease distribution and public health equity. Methods such as spatial-temporal regression analysis have been used to identify spatial-temporal clustering of diseases and evaluate the spatial-temporal evolution of healthcare resources [25, 26]. However, these methods face limitations in handling complex nonlinear relationships, making it challenging to capture intricate dynamic patterns. In recent years, with advancements in computational power, machine learning-based spatial-temporal analysis methods have emerged [27]. However, these models often require large amounts of training data [28], limiting their applicability in public health research.

MHT is the core resource for developing medical and health infrastructure in any city, region, or country. This study aims to address these research gaps by examining the spatial-temporal distribution and evolution of MHT in China. First, leveraging a long-term data set spanning from 1982 to 2020, we constructed a multidimensional index system encompassing Talent Agglomeration Density (TAD), Talent Agglomeration Scale (TAS), Talent Agglomeration Intensity (TAI), and Talent Agglomeration Equilibrium (TAE) to comprehensively quantify the distributional characteristics of MHT. Second, this study employs a multi-method approach that integrates kernel density estimation, Moran鈥檚 I index, and Dagum Gini coefficients to analyze the spatial and temporal evolution of MHT in China. This framework not only captures the spatial agglomeration and disparities of MHT but also reveals dynamic trends over nearly four decades (1982鈥�2020). Finally, the findings identify specific sources of regional disparities and provide empirical evidence to support targeted policy recommendations, contributing to the rational allocation of healthcare resources and the promotion of health equity in China.

Research design

Division of regions

In this study, China is divided into eastern, central, and western regions based on geographic, economic, and policy factors. This classification is widely used in research [29] and follows the framework established by the Western Development Strategy (2000) [30]. The eastern region encompasses 11 provincial-level administrative areas, including Beijing, Tianjin, Hebei, Liaoning, Shanghai, Jiangsu, Zhejiang, Fujian, Shandong, Guangdong, and Hainan. The central region includes eight provincial administrative areas: Shanxi, Jilin, Heilongjiang, Anhui, Jiangxi, Henan, Hubei, and Hunan. The western region comprises 12 provincial-level administrative areas: Inner Mongolia, Guangxi, Chongqing, Sichuan, Guizhou, Yunnan, Tibet, Shaanxi, Gansu, Qinghai, Ningxia, and Xinjiang. In total, the Chinese Mainland is divided into 31 provinces, autonomous regions, and municipalities. Figure听1 shows the map of provincial distribution in China and the composition of the eastern, central, and western regions.

Fig. 1

Three regions in China

Data sources

For data analysis in this paper, five time points were selected: 1982, 1990, 2000, 2010, and 2020, corresponding to the third, fourth, fifth, sixth, and seventh national censuses in China, respectively. The data on MHT were derived from various statistical yearbooks 1983 [31] and 1991 [19, 32] China Statistical Yearbooks 2001 [33] and 2011 [34], and China Health Statistics Yearbooks 2021 [35]. These yearbooks provide information on the number of healthcare professionals, including doctors (physicians), nurses, pharmacists, laboratory technicians, and other similar healthcare personnel. The population and administrative area data for the regions involved in this study were gathered from the official website of the National Bureau of Statistics of China. Since Hainan was established as a province in April 1988 and Chongqing became a municipality directly under the central government on June 18, 1997, the data for Hainan in 1982 were included in Guangdong Province, and the data for Chongqing in 1982 and 1990 were included in Sichuan Province.

Methods

Index of MHT

MHT is healthcare professionals, including doctors (physicians), nurses, pharmacists, laboratory technicians, and other similar healthcare personnel. Previous studies have measured MHT levels more in terms of the total number of medical and health talents [36]. However, due to the large differences in geographical area and population size of each province in China, only the total number of MHT can鈥檛 fully reflect the relative level of this indicator [22], it is necessary to use the relative indicator to better measure the level of MHT. Therefore, this paper constructs a comprehensive and multi-dimensional index system to measure the agglomeration level of MHT in China, utilizing four key indicators: talent agglomeration density (TAD), talent agglomeration scale (TAS), talent agglomeration intensity (TAI), and talent agglomeration equilibrium (TAE).

TAD

Based on existing research literature [37, 38], TAD is defined as the proportion of MHT relative to the total area of a region, indicating the MHT agglomeration degree per unit area. A higher proportion signifies greater regional MHT density, whereas a lower proportion indicates smaller density. The calculation formula for TAD is as follows:

$$TA{D_{it}} = {{MH{T_{it}}} \over {{A_{it}}}}$$

(1)

where i represents the province, t represents the year, MHT_it is the number of medical and health talents of province i in year t, and A_it is the area of province i in year t.

TAS

Based on the literature [39,40,41], TAS is described as the proportion of a province鈥檚 MHT in comparison to the total MHT in China, reflecting the agglomeration scale of provincial MHT compared with the whole country. A larger proportion reflects a larger agglomeration scale of MHT in the province, and a smaller proportion indicates a smaller scale. The formula for TAS is expressed as follows:

$$TA{S_{it}} = {{MH{T_{it}}} \over {MH{T_t}}}$$

(2)

where i represents the province, t represents the year, MHT_it is the number of medical and health talents of province i in year t, and MHT_t is the whole number of medical and health talents in year t.

TAI

Based on the literature [39, 42] and official indicators from China鈥檚 National Health Commission [2], TAI is defined as the proportion of a province鈥檚 MHT relative to its population. This indicator reflects the agglomeration intensity of provincial MHT compared to its population size. A larger proportion denotes larger agglomeration intensity of MHT in the province. The TAI calculation formula is as follows:

$$TA{I_{it}} = {{MH{T_{it}}} \over {PO{P_{it}}}}$$

(3)

where i represents the province, t represents the year, MHT_it is the number of medical and health talents of province i in year t, and POP_it is the population in province i in year t.

TAE

Based on the research literature [39, 43], TAE is constructed using the location entropy index to measure the equilibrium of MHT agglomeration degree. A higher TAE suggests a higher provincial MHT agglomeration degree. When TAE鈥�>鈥�1, the provincial MHT converging equilibrium degree exceeds the national average; when TAE鈥�<鈥�1, it is below the national average; and when TAE鈥�=鈥�1, it equals the national average. The TAE formula is as follows:

$$TA{E_{it}} = {{MH{T_{it}}/PO{P_{it}}} \over {MH{T_t}/PO{P_t}}}$$

(4)

where i represents the province, t represents the year, MHT_it is the number of medical and health talents of province i in year t, POP_it is the population in province i in year t, MHT_t is the whole number of medical and health talents in year t, and POP_t is the population in year t.

Research methods

In this paper, kernel density was used to analyze the dynamic evolution of MHT, Moran鈥� I index was used to analyze the spatial autocorrelation of MHT, and Dagum Gini Coefficient was used to analyze the regional difference of MHT. The combination of these three methods can more comprehensively analyze the spatial and temporal distribution and dynamic evolution of MHT.

Kernel density estimation

Kernel density estimation describes the spatial evolution trend by comparing the position, pattern, differentiation, and curvature of density distribution curves over different time periods. It uses the Gaussian kernel function [36, 44] to analyze regional differences in the distribution level of MHT in China and determine the existence of any polarization trends. The formula for this estimation is as follows:

$$f(x) = {1 \over {Nh}}\sum\limits_{i = 1}^N {K({{{X_i} - x} \over h})}$$

(5)

$$K(x) = {1 \over {\sqrt {2\pi } }} \times \exp ( - {{{x^2}} \over 2})$$

(6)

where N is the number of observations, Xi is the independent and identically distributed observations, x is the average value, K is the kernel function, and h is the bandwidth.

Moran鈥� I index

The exploratory spatial data analysis method is utilized to detect spatial agglomeration, including global and local spatial autocorrelation [22, 45, 46]. Global and local Moran鈥檚 indices are frequently used to test the spatial correlation of variables.

The global Moran鈥檚 I index aims to estimate the spatial agglomeration and distribution divergence of observations across the entire study area. The calculation formula for the global Moran鈥檚 I index is as follows:

$$I(g) = {{n\sum\limits_{i = 1}^n {\sum\limits_{j = 1}^n {{W_{ij}}({Z_i} - \bar X)({X_j} - \bar X)} } } \over {\sum\limits_{i = 1}^n {\sum\limits_{j = 1}^n {{W_{ij}}\sum\limits_{i = 1}^n {{{({X_i} - \bar X)}^2}} } } }}$$

(7)

where X_i and X_j represent the TAD, TAS, TAI, and TAE of provinces i and j, respectively; $\bar X$ is the average value; and W_ij represents the spatial weight matrix between units i and j. The value of the global Moran鈥檚 I index is in the range of [鈭掆��1, 1]. I(g)鈥�>鈥�0 indicates positive spatial agglomeration, 滨(驳)鈥�&濒迟;鈥�0 indicates negative spatial correlation, and I(g)鈥�=鈥�0 indicates random spatial distribution.

The local Moran鈥檚 I index aims to explore the spatial distribution between the local unit and its adjacent neighbors. The index is calculated as follows:

$${I_i} = {Z_i}\sum\limits_{j - 1}^n {{W_{ij}}{Z_j}}$$

(8)

where Z_i and Z_j are the standardized values of the observations of spatial units i and j, respectively. The value of the local Moran鈥檚 I index is the range of [鈭掆��1, 1]. I_i > 0 indicates a positive correlation between the observation and its neighbors, that is to say, a higher value is surrounded by higher values, or a lower value is surrounded by lower values; and I_i < 0 indicates negative spatial dependence, including higher鈥搇ower or lower鈥揾igher value cluster.

Dagum Gini coefficient

The Dagum Gini coefficient [36, 47, 48] is used to decompose the regional difference of TAD, TAS, TAI, and TAE in eastern, central, and western regions in China. The total Gini coefficient G is calculated as follows:

$$G = {1 \over {2{n^2}\bar Y}}\sum\limits_{i = 1}^n {\sum\limits_{i = 1}^n {\left| {{Y_i} - {Y_r}} \right|} } = {1 \over {2{n^2}\bar y}}\sum\limits_{j = 1}^k {\sum\limits_{h = 1}^k {\sum\limits_{i = 1}^{{n_j}} {\sum\limits_{r = 1}^{{n_h}} {\left| {{y_{ji}} - {y_{hr}}} \right|} } } }$$

(9)

where G represents the total Gini coefficient, n is the number of provinces; k is the number of regions; n_j (n_h) represents the number of provinces in region j(h); j and h denote different regions; i and r represent different provinces in region j(h); Y_i represents the TAD, TAS, TAI, and TAE of province i; y_ji (y_hr) represents the TAD, TAS, TAI, and TAE of provinces in region j(h); and $\bar Y$ and $\bar y$ represent the mean value. A higher Dagum Gini coefficient indicates a greater degree of imbalance in the index.

According to Dagum Gini coefficient method, the total Gini ratio can be decomposed as follows:

$$G = {G_w} + {G_{nb}} + {G_t}$$

(10)

where G_w measures the contribution of the Gini inequality within regions to the total Gini ratio G, G_nb measures the net contribution of the extended Gini inequality between regions to the total Gini ratio G, and G_t measures the contribution of trans variation between regions to the total Gini ratio G.

Results

Spatial and temporal distribution

To investigate the spatial and temporal agglomeration levels of MHT in China, this paper examined MHT agglomeration in terms of density, scale, intensity, and equilibrium degree from 1982 to 2020. Using ArcMap 10.2 software for mapping, the spatial and temporal distributions were analyzed at the national, regional, and provincial levels.

National level spatial and temporal analysis

The national level of the MHT agglomeration index is presented in Table听1, it shows the mean, variance, minimum and maximum values of TAD, TAS, TAI and TAE at the national level for five years in 1982, 1990, 2000, 2010, and 2020. The average TAD had increased from 1.409 in 1982 to 3.8719 in 2020, and the variance of TAD values had increased over time, with TAD minimum and maximum values of 0.0057 and 15.6151 in 1982 and 0.0185 and 33.812 in 2020 respectively. The data revealed that TAD increased significantly, indicating that the agglomeration density of national MHT expanded year by year. The average TAS had decreased from 0.0345 in 1982 to 0.0323 in 2020. TAS decreased suggesting a reduction in the overall scale of national MHT agglomeration. The average TAI had increased from 0.0038 in 1982 to 0.0078 in 2020. TAI showed a general upward trend, except for a slight decline in 2000, indicating an increasing intensity in attracting talents nationally. The average TAE was 1.0259 in 1982, kept rising, to 1.1296 in 2000, and then began to decline, to 1.0576 in 2020. TAE initially increased before decreasing, reaching its optimal equilibrium level nationally in 2000.

Table 1 Agglomeration Level of MHT in China (1982鈥�2020)

Three regional spatial and temporal analysis

Table听2 shows the mean levels of TAD, TAS, TAI and TAE in the eastern, central and western regions in 1982, 1990, 2000, 2010, and 2020, respectively. Furthermore, to provide a clearer representation of the dynamic trends in the four key indicators of MHT鈥擳AD, TAS, TAI, and TAE鈥攁t the national level, the eastern, central, and western regions, we have utilized graphical illustrations to depict the changes in these indicators over the period from 1982 to 2020, as shown in Fig.听2.

The details are as follows:

(1) The TAD index in the eastern region was significantly higher than in the central and western regions. In 1982, the TAD for the eastern region was 3.2945, whereas the central and western regions had TAD values of 0.6869 and 0.2200, respectively. By 2020, the TAD for the eastern region rose to 8.3079, whereas the central and western regions reached 2.1148 and 0.9770, respectively. Although the TAD index increased over time in all three regions, the differences among them also widened.

(2) The TAS index showed little overall change across the eastern, central, and western regions. The TAS for MHT in the eastern and central regions was comparable and both were higher than that of the western region. The TAS index for the eastern region remained around 0.04, changing from 0.0403 in 1982 to 0.0396 in 2020. For the central region, the TAS index slightly decreased from 0.0414 in 1982 to 0.0358 in 2020. The TAS index for the western region remained around 0.02, moving from 0.0241 in 1982 to 0.0232 in 2020.

(3) The TAI index in the eastern region generally rose, except for a decline in 2000. The central and western regions exhibited a continuous increase. Since 2010, the TAI index has surged rapidly in all three regions. In 2020, the TAI index for the eastern, central, and western regions were 0.0081, 0.0074, and 0.0077, respectively.

(4) The TAE index demonstrated varied patterns across the regions. The eastern region鈥檚 TAE initially increased and then decreased. The central region showed minimal change, maintaining a level around 1, whereas the western region initially decreased and then increased. The TAE in the eastern region was found to be higher than the national average, except in 1982. In the central region, it remained above the national average from 1982 to 2020, whereas in the western region, it was higher than the national average, except in 2000.

Table 2 Regional analysis of the agglomeration level of MHT (1982鈥�2020)

Fig. 2

Time-series evolution of TAD, TAS, TAI, and TAE. (A) TAD; (B) TAS; (C) TAI; (D) TAE

Provincial level spatial and temporal analysis

Table听3 shows the echelon distribution at the provincial level in 1982, 1990, 2000, 2010, and 2020 according to the total number of MHT. The spatial-temporal distribution of MHT was analyzed the provincial level of MHT revealed several characteristics.

Table 3 Echelon of provincial MHT (1982鈥�2020)

(1) MHT was highly concentrated in Guangdong, Shandong, Henan, Jiangsu, and Sichuan Provinces. Data showed that the top five provinces with the number of MHT in China changed from 1982 to 2020. In 1982, the top five provinces with the number of MHT were Sichuan, Shandong, Guangdong, Liaoning, and Hubei. By 1990, the top five shifted to Sichuan, Shandong, Liaoning, Henan, and Jiangsu. In 2000, Liaoning was replaced by Guangdong in the top five. From 2000 to 2020, the top provinces remained Guangdong, Shandong, Henan, Jiangsu, and Sichuan, although with slight ranking changes.

(2) Xinjiang, Gansu, Tianjin, Hainan, Qinghai, Ningxia, and Tibet consistently remained in the third echelon from 1982 to 2020. Qinghai, Ningxia, and Tibet, all in the western region, consistently had the lowest numbers of MHT. Fujian and Guizhou changed echelons; Fujian moved from the third echelon in 1982鈥�2000 to the second echelon from 2010 to 2020, whereas Guizhou moved from the third echelon in 1982鈥�2010 to the second echelon in 2020.

(3) Liaoning and Heilongjiang witnessed a drop from the first echelon in 1982 to the second echelon by 2020. Jilin decreased from the second echelon in 1982 to the third echelon in 2020. Shanxi, Shanghai, Jiangxi, Shaanxi, Beijing, and Yunnan have consistently remained in the second echelon.

(4) Fig.听3 shows the evolution of the total number of MHT at the provincial level from 1982 to 2020, where the black line is the Hu Huanyong line. The population distribution map of China, published by Hu Huanyong in 1935, marked the maturity of China鈥檚 population geography [49]. Originally called the Aihui鈥揟engchong Line, this boundary became a significant geographical demarcation. Repeatedly verified by census data and explained through various scientific principles, including natural, economic, and sociocultural factors, it has gained recognition in academic circles [50,51,52]. The Hu Huanyong Line has become increasingly important for understanding and analyzing China鈥檚 national conditions [53, 54]. The spatial distribution of MHT aligns with this population line, showcasing significant spatial heterogeneity. MHT is mainly concentrated in the southeastern regions, whereas Tibet, Qinghai, Ningxia, and Gansu Provinces in the northwest form a 鈥渄epression of MHT.鈥�

Fig. 3

Provincial Spatial distribution of MHT. (A) 1982; (B) 1990; (C) 2000; (D) 2010; (E) 2020. Note: The solid black line represents the Hu Huanyong Line

Regional dynamic analysis

Kernel density estimation was employed to analyze the regional dynamic in MHT [55]. By utilizing Stata 17.0 software, a kernel density estimation map was created to analyze the regional dynamics in the MHT distribution level at the national and regional levels, focusing on four key aspects: distribution position, distribution pattern, differentiation phenomenon, and curve extension. Figure听4 shows the Kernel density distribution of MHT, Fig.听4A is the overall national level Kernel density distribution of MHT, Fig.听4B, C and D is Kernel density distribution of MHT in eastern, central and western region respectively. Changes in distribution position help investigate the MHT distribution level; changes in distribution pattern reflect the degree of regional distribution discreteness; the differentiation phenomenon, indicated by curve kurtosis changes, reflects the polarization of regional distribution; and curve extensibility is used to examine interregional differences.

Fig. 4

Kernel density estimation of MHT. (A) Overall; (B) Eastern; (C) Central; (D) Western

National level Kernel density analysis

Figure听4A shows the national distribution trend of MHT for the years 1982, 1990, 2000, 2010, and 2020. Observations are as follows. First, regarding distribution position, the center position of the curve shifted rightward over these five time points, indicating that the national distribution level of MHT was consistently improving. Second, in terms of distribution pattern, the peak heights of the curves gradually declined, whereas the width of the curves loosened, indicating a rise in the overall discrete degree of MHT nationally. Third, considering the differentiation phenomenon, the multipeak curve transitioned to a unimodal curve, showing a reduction in polarization. The flatter and wider kernel density curves over time suggest an increasing disparity in the distribution level of MHT among provinces. Fourth, from the curve extension perspective, the distribution curve鈥檚 right tail lengthened each year, signifying an expanding regional gap in MHT nationwide.

Regional level Kernel density analysis

Figure听4B and C, and 4D show the MHT distribution trends in the eastern, central, and western regions from 1982 to 2020, respectively. Key observations are first, concerning distribution position, the center positions of the curves in all three regions shifted rightward annually, indicating an overall upward trend in MHT distribution levels. Second, regarding distribution pattern, the peak heights decreased and the curve widths expanded across all three regions. Third, from the perspective of differentiation phenomenon, the eastern region鈥檚 curve initially exhibited multiple peaks, which gradually evolved into a single peak; the central region鈥檚 curve showed no clear multipeak; the western region鈥檚 curve initially had a bimodal shape, which also transitioned to a single peak. Generally, the eastern and western regions displayed a shift from bimodal to single-peak curves, indicating a weakening polarization phenomenon. Fourth, in terms of curve extension, the eastern region showed an evident right tail and the curve became flatter and wider; the central region also showed a right tail with small increments and a relatively narrow curve; the western region鈥檚 curve widened significantly with a right tail showing small increments. These patterns suggest clustered regional agglomeration in the eastern and western regions, whereas the central region exhibited discrete regional agglomeration.

Spatial autocorrelation analysis

To evaluate the spatial autocorrelation of MHT, spatial autocorrelation tests were conducted using global and local Moran鈥檚 I indices for TAD, TAS, TAI, and TAE.

Global Moran鈥檚 I

Table听4 shows the Global Moran 鈥榠 index of TAD, TAS, TAI and TAE from 1982鈥�2020.

Table 4 Global Moran鈥檌 of TAD, TAS, TAI and TAE

(1) The Z-values for TAD鈥檚 global Moran鈥檚 I index in 1982 and 1990 were above 1.96, indicating significance at the 0.05 level, whereas the Z-values in 2000, 2010, and 2020 were above 2.58, indicating significance at the 0.01 level. This demonstrates that TAD exhibited spatial autocorrelation. Moreover, all Moran鈥檚 I values for TAD were greater than zero and showed a gradual increase over time, indicating that the positive correlation of TAD increased year by year.

(2) In 1982, the Z-value for TAS鈥檚 global Moran鈥檚 I index was below 1.65, indicating significance at the 0.1 level, with a Moran鈥檚 I value of 0.149, suggesting some spatial positive correlation for TAS in 1982. However, the Z-values for subsequent years were insignificant, and the global Moran鈥檚 I values for TAS gradually decreased to zero, indicating that from 1990 to 2020, TAS exhibited characteristics of a random distribution without significant spatial autocorrelation.

(3) The Z-values for TAI鈥檚 global Moran鈥檚 I index were above 1.96 in 1982 and 2010, indicating significance at the 0.05 level. In 1990 and 2000, the Z-values exceeded 2.58, indicating significance at the 0.01 level, but the 2020 data was insignificant. This indicates that TAI exhibited spatial autocorrelation for all years except 2020. Moran鈥檚 I values for TAI showed an initial increase followed by a decrease, with all values remaining above zero, indicating that TAI鈥檚 positive correlation initially increased and then decreased over time.

(4) From 1982 to 2020, the Z-values for TAE鈥檚 global Moran鈥檚 I index were all above 2.58, indicating significance at the 0.01 level, confirming spatial autocorrelation for TAE. Moran鈥檚 I values for TAE were greater than zero, initially decreasing and then increasing, indicating that TAE鈥檚 positive correlation decreased initially before increasing over time.

Local Moran鈥檚 I

The scatter plot of Local Moran鈥檚 I are displayed in Fig.听5. Figure听5A show scatter plot of local Moran鈥橧 for TAD from 1982 to 2020, Fig.听5B show scatter plot of local Moran鈥橧 for TAS from 1982 to 2020, Fig.听5C show scatter plot of local Moran鈥橧 for TAI from 1982 to 2020, Fig.听5D show scatter plot of local Moran鈥橧 for TAE from 1982 to 2020.

Fig. 5

Moran scatter plot in 1982,听1990,听2000,听2010 and 2020. (A) TAD, (B) TAS, (C) TAI, (D)TAE

(1) The Moran鈥檚 local scatter plot of TAD from 1982 to 2020 is shown in Fig.听5A. Most provinces and cities are distributed in the first and third quadrants. From 1982 to 2020, Beijing, Tianjin, and Jiangsu were high鈥揾igh agglomeration areas, whereas Hebei was designated as a low鈥揾igh agglomeration area. Zhejiang transitioned from a low鈥揾igh agglomeration area from 1982 to 2000 to a high鈥揾igh agglomeration area from 2010 to 2020. Shanghai evolved from a high鈥搇ow agglomeration area in 1982 to a high鈥揾igh agglomeration area in 1990, 2000, 2010, and 2020. Shandong shifted from a low鈥搇ow agglomeration area to a high鈥搇ow agglomeration area in 2010, and later to a high鈥揾igh agglomeration area in 2020. The remaining 20 provinces and cities remained low鈥搇ow agglomeration areas in the third quadrant from 1982 to 2020.

(2) The Moran鈥檚 local scatter plot of TAS from 1982 to 2020 is shown in Fig.听5B. Most provinces and cities are distributed in the first, second, and third quadrants. From 1982 to 2020, Anhui, Hubei, Hunan, Hebei, Henan, Shandong, and Jiangsu were high鈥揾igh agglomeration areas, whereas Shanghai, Fujian, Jiangxi, and Guizhou remained low鈥揾igh agglomeration areas. Xinjiang, Qinghai, Gansu, Ningxia, Beijing, and Tibet were recorded as low鈥搇ow agglomeration areas, with Guangdong, Heilongjiang, and Sichuan as high鈥搇ow agglomeration areas. Other provinces exhibited significant changes over time.

(3) The Moran鈥檚 local scatter plot of TAI from 1982 to 2020 is shown in Fig.听5C. Most provinces and cities are located in the first and third quadrants. During this period, Tianjin and Jilin were consistently high鈥揾igh agglomeration areas, whereas Hebei and Gansu persisted as low鈥揾igh agglomeration areas. Fujian, Guangdong, Anhui, Jiangxi, Henan, Hubei, Hunan, Sichuan, Guizhou, and Tibet were identified as low鈥搇ow agglomeration areas. Jiangsu changed from a low鈥揾igh agglomeration area to a high鈥揾igh agglomeration area in 2020. Shanghai changed from a high鈥搇ow agglomeration area to a high鈥揾igh agglomeration area in 2010 and 2020. Inner Mongolia transitioned from a high鈥搇ow area in 1982 to a low鈥搇ow area in 1990, and then to a high鈥揾igh area in 2000, 2010, and 2020.

(4) The Moran鈥檚 local scatter plot of TAE from 1982 to 2020 is displayed in Fig.听5D. Most provinces and cities are situated in the first and third quadrants. From 1982 to 2020, Zhejiang, Fujian, and Guangdong remained low鈥搇ow agglomeration areas. Anhui, Jiangxi, Hunan, Guangxi, and Chongqing transitioned from high鈥揾igh agglomeration areas to low鈥搇ow agglomeration areas since 1990. Since 1990, Qinghai and Gansu changed from high鈥揾igh agglomeration areas to high鈥搇ow agglomeration and low鈥揾igh agglomeration areas, respectively. Beijing, Tianjin, Jilin, Heilongjiang, and Inner Mongolia shifted from low鈥搇ow agglomeration areas to high鈥揾igh agglomeration areas since 1990.

Spatial gap measurement and decomposition

The Dagum Gini coefficient and decomposition method were used to measure the overall Gini coefficient, intraregional Gini coefficient, interregional Gini coefficient, and the contribution rates of TAD, TAS, TAI, and TAE across 31 provinces and three regions in China for the years 1982, 1990, 2000, 2010, and 2020. The results are presented in Table听5.

Table 5 Decomposition of the spatial gap in TAD, TAS, TAI, and TAE

(1) Spatial difference of TAD

The overall Gini coefficient for TAD displayed a decreasing trend during 1982鈥�2020, dropping from 0.661 to 0.611, with a decline rate of only 7.56%.

The average Gini coefficients for the eastern, central, and western regions during this period were 0.5346, 0.2154, and 0.4598, respectively, highlighting that the central region had the smallest intraregional differences, followed by the western and eastern regions. The eastern region showed a downward trend, whereas the central and western regions exhibited upward trends.

For the interregional Gini coefficients, the averages were 0.5565, 0.6847, and 0.4127 for the east鈥揷entral, east鈥搘est, and central鈥搘est regions, respectively, with the east鈥搘est region demonstrating the greatest disparities. Although the east鈥揷entral and east鈥搘est regions experienced a downward trend, the central鈥搘est region showed an upward trend; however, the changes in all three regions were small.

Table听4 presents the Gini coefficients of each decomposition item and its contribution to the overall Gini coefficient during 1982鈥�2020. The intraregional and interregional Gini coefficients showed a decreasing trend, fluctuating from 27.08% to 72.25% in 1982 to 26.31% and 71.32% in 2020. Conversely, the Gini coefficient for trans variation between regions exhibited a fluctuating increase from 0.07% in 1982 to 2.37% in 2020. The primary contributor to the inequality of TAD was the interregional Gini coefficient.

(2) Spatial difference of TAS

The findings indicated that the overall Gini coefficient for TAS exhibited an upward trend from 0.312 in 1982 to 0.346 in 2020, marking a small increase of 10.90%.

The average Gini coefficients for the eastern, central, and western regions were 0.2852, 0.1608, and 0.3904, respectively, indicating that intraregional variation was most pronounced in the western region, followed by the eastern and central regions. Fluctuations in the eastern and central regions increased, whereas they decreased in the western region.

The average Gini coefficient between the eastern and western regions was the largest, starting at 0.36032 in 1982 and rising to 0.38476 in 2020, with an average of 0.3712. This was followed by the central鈥搘est region, which decreased from 0.3565 in 1982 to 0.32016 in 2020.

Table听4 presents the Gini coefficients for each decomposition item and their contribution to the overall Gini coefficient during 1982鈥�2020. The intraregional and interregional Gini coefficients showed a fluctuating upward trend, increasing from 29.90% to 26.02% in 1982 to 31.24% and 34.68% in 2020, respectively. Conversely, trans variations between regions showed a fluctuating downward trend, declining from 44.09% in 1982 to 34.08% in 2020.

(3) Spatial difference of TAI

The results indicated a significant decline in the overall Gini coefficient of TAI, dropping from 0.215 in 1982 to 0.068 in 2020, which represents a decrease of 68.37%.

For the period 1982鈥�2020, the average Gini coefficients for the eastern, central, and western regions were 0.1922, 0.1110, and 0.1106, respectively. This suggests that intraregional variation was most pronounced in the eastern region, followed by the central and western regions. Each of the three regions showed a steady downward trend in their Gini coefficients, with respective decrease rates of 67.15%, 64.29%, and 64.75%.

The average interregional Gini coefficients were 0.1747, 0.1680, and 0.1131 for east鈥揷entral, east鈥搘est, and west鈥揷entral regions, all showing a stable declining trend. The smallest differences were observed between the west鈥揷entral regions.

Table听4 presents the Gini coefficients for each decomposition item and their contribution rates to the overall Gini coefficient during 1982鈥�2020. The Gini coefficients for intraregional, interregional, and trans variations all showed a decreasing trend over this period, with decreasing rates of 67.65%, 73.53%, and 65.82%, respectively. The contribution rates of intraregional variation and trans variation fluctuated from 31.87% to 36.64% in 1982 to 32.93% and 39.93% in 2020, respectively. The contribution rate of interregional differences fluctuated from 31.49% in 1982 to 27.14% in 2020. Overall, these changes were relatively small, regardless of whether they were increases or decreases.

(4) Spatial difference of TAE

The overall Gini coefficient of TAE was 0.083 in 1982. It fluctuated over time, peaking at 0.246 in 2000, before returning to 0.078, forming an inverted U-shaped curve.

During the period of 1982鈥�2020, the average Gini coefficients for the eastern, central, and western regions were 0.1684, 0.1302, and 0.1332, respectively. This indicates that intraregional variation was greatest in the eastern region, followed by the western and central regions. The Gini coefficient for each region also demonstrated an inverted U-shaped trend.

For interregional comparisons, the average Gini coefficients were 0.1638, 0.1668, and 0.1364 for the east鈥揷entral, east鈥搘est, and west鈥揷entral regions, with the largest difference between the east鈥搘est regions. These interregional Gini coefficients similarly followed an inverted U-shape.

Table听4 illustrates the changing trends of the Gini coefficient for each decomposition item and their contribution rates to the overall Gini coefficient during 1982鈥�2020. The intraregional Gini coefficient exhibited an inverted U-shaped trend. The interregional Gini coefficient decreased from 0.046 in 1982 to 0.008 in 2020, whereas trans variation fluctuated between 0.016 and 0.045. The contribution rates for intraregional differences and trans variation fluctuated from 25.94% to 18.86% in 1982 to 32.58% and 57.56% in 2020, respectively. Conversely, the contribution rate for interregional differences decreased sharply from 55.20% in 1982 to 9.85% in 2020.

Discussion

By using provincial panel data of MHT in China from 1982 to 2020, this study reveals the spatial-temporal distribution differences and regional evolution characteristics of MHT in China.

The paper constructs a multidimensional index system to measure the agglomeration level of MHT, including TAD, TAS, TAI, and TAE. At the national level, TAD and TAI showed favorable development trends from 1982 to 2020, which are closely related to a series of policies introduced by the Chinese government, such as the 鈥淧rimary Healthcare Promotion Policy鈥� [56, 57], 鈥淢edical Consortium Construction Policy鈥� [58], and the 鈥淗ealthy China 2030 Plan鈥� [1, 59]. The study finds that the overall number of MHT in China has increased year by year, consistent with the findings of Lin Xiaodan et al. (2021) [60]. In the eastern, central, and western regions, these four indicators exhibited different upward or downward trends during 1982鈥�2020, with significant differences among these regions. Health resources tend to be concentrated in wealthier areas, showing an overall trend of increasing from west to east [61, 62]. However, the TAE index, which represents equilibrium (a value closer to 1 indicates greater equilibrium), exhibited a fluctuating pattern. It reached its highest value of 1.1296 in 2000, suggesting that the imbalance of MHT was greatest in that year. After 2000, under the influence of a series of policies implemented by the Chinese government [1, 57, 58], the TAE value dropped to 1.0576 by 2020, indicating a more balanced distribution of MHT.

A detailed analysis of the dynamic evolution of MHT at the provincial level shows that MHT in China is mainly concentrated in Guangdong, Shandong, Henan, Jiangsu, and Sichuan Provinces, with these provinces consistently ranking in the first echelon during 1982鈥�2020. However, Tibet, Qinghai, and Ningxia lagged behind, reflecting east鈥搘est disparity in the number of MHT. This finding aligns with the highest health resource density index of doctors and nurses being in the eastern region and the lowest in the western provinces, such as Tibet, Qinghai, and Xinjiang, with the gap intensifying in the past decade [22, 63, 64]. The study also notes that Fujian and Guizhou Provinces showed significant improvement, moving from the third to the second echelon. By contrast, MHT in Liaoning and Heilongjiang dropped from the first to the second echelon, and Jilin fell from the second to the third echelon, likely due to ongoing economic decline and population migration in these provinces. Moreover, the study finds that the distribution of MHT in China is essentially demarcated by the Hu Huanyong Line, indicating a significant geographical separation in MHT levels between the densely populated southeastern regions and the sparsely populated northwestern areas of China.

The regional dynamics of MHT in China, as estimated by kernel density, show that the spatial gap in its distribution level is gradually widening at the national level. Regionally, the kernel density estimation map presents different characteristics in terms of distribution location, shape, differentiation phenomenon, and curve extensibility. Generally, the eastern and western regions exhibited clustered regional agglomeration characteristics, whereas the central region demonstrated discrete agglomeration. The disparity in MHT distribution levels among the eastern, central, and western regions is also gradually expanding. This finding contradicts previous studies, which reveal that the eastern region is more equitable, followed by the central region, with the western region being the least equitable [21]. Another study indicates that the unequal allocation of health resources is increasing year by year, especially in the western region [65]. These differences may stem from our study鈥檚 utilization of kernel density analysis, which focuses on changes within regions rather than comparing them, leading to different conclusions.

The spatial autocorrelation analysis using global and local Moran鈥檚 I for TAD, TAS, TAI, and TAE reveals several key insights. From the global Moran鈥檚 I index, TAD showed an increasing trend in spatial correlation, whereas TAS showed a decline over time, and TAI and TAE displayed fluctuating correlations. From the local Moran鈥檚 I index, persistent agglomeration in regions like Beijing, Tianjin, and Jiangsu (high鈥揾igh for TAD) highlights areas of sustained advantage, whereas shifts in agglomeration for TAI and TAE occur across various regions.

Quantitative analyses of regional differences were further conducted using Dagum Gini analysis. The Dagum Gini coefficients for TAD, TAS, TAI, and TAE revealed distinct trends in regional inequality. TAD and TAI showed a clear overall reduction in disparities, while TAS experienced a modest yet steady increase. In contrast, TAE followed a fluctuating pattern over the observed period. This fluctuating pattern of TAE reflects that from 1982 to 2000, the equilibrium of MHT exhibited increasing divergence, whereas from 2000 to 2020, it demonstrated a trend toward convergence. Our analysis attributes this fluctuating pattern to both economic influences and the impact of policy interventions. The economic development that began with the reform and opening-up in 1978, along with increasing regional disparities鈥攑articularly the faster economic growth in the eastern regions compared to the central and western regions [66, 67]鈥攍ikely contributed to the growing imbalance in the distribution of MHT across regions from 1982 to 2000. From 2000 to 2020, this trend reversed as the Chinese government, especially after SARS, recognized that a more balanced distribution of MHT was crucial for effectively responding to public health emergencies, and introduced policies to address the unequal distribution of MHT. Key policies such as 鈥渢he New Rural Cooperative Medical Scheme (2003)鈥� [68] and 鈥渢he Opinions on Deepening the Reform of the Medical and Health System (2009)鈥� [69, 70] placed emphasis on ensuring more equitable healthcare services. These policies emphasize attracting and retaining MHT in central-western regions or rural primary healthcare institutions by increasing salaries, implementing incentive measures, and improving working conditions, in order to alleviate the imbalance of medical resources between regions. Furthermore, the Guidance on Promoting the Construction and Development of Medical Consortiums (2017) [58] promoted better integration of healthcare resources, facilitating the collaboration of MHT between large urban hospitals, county-level hospitals, and rural health institutions. We believe that these policies have played a positive role in promoting the gradual decline of the overall Dagum Gini coefficient of TAE from 2000 to 2020. In terms of intraregional differences, the eastern and western regions have significant disparities, whereas the central region exhibits smaller differences, likely due to economic development variations among provinces and cities. The central region is more balanced in terms of talent agglomeration. Interregionally, the imbalance between the eastern and western regions is the highest, although it is declining. The eastern region, benefiting from better economic strength, infrastructure, industrial layout, and other factors, is significantly more advanced than the central and western regions. However, with policy support and provincial-level assistance, the gap among these regions is gradually being reduced.

To better address the imbalance in the distribution of MHT and promote the overall level of healthcare, the Chinese government and research scholars have proposed policies or suggestions, such as enhancing the practice environment of the medical industry [71] or improving the public health emergency management system [72, 73]. Our study provides actionable insights for optimizing public health policies to address medical and health talent disparities in China. The identification of persistent regional imbalances in TAD, TAS, TAI, and TAE provide critical evidence to support targeted policy interventions. Based on our analysis, this research puts forward several policy recommendations. First, Provinces with consistently low levels of TAD, TAS, TAI and TAE, such as western provinces like Ningxia, Qinghai, and Tibet, should receive continued fiscal support and incentives to retain health professionals. The dispersion degree of the MHT distribution in the country is increasing, and the spatial difference of the distribution level is expanding. The dynamic trends observed in this study emphasize the need for adaptive policy frameworks that can respond to evolving demographic and regional conditions. To alleviate the excessive agglomeration of medical and health resources and ensure rational geographical distribution, targeted financial allocations should strengthen support for provinces like Liaoning and Heilongjiang, which exhibit slow growth or even decline in MHT levels. Second, the use of digital measures should be accelerated to address the shortage of MHT in regions. The continuous development of digitalization offers robust support in responding to public health emergencies. For example, during the COVID-19 epidemic, digital tools such as the Hangzhou Health Code and Wuxi Mobile鈥檚 digital Sentinel were instrumental in tracking individual movements and close contact, facilitating scientific and accurate epidemic prevention and control. Technologies such as telemedicine and digital health can effectively alleviate MHT shortages. By aligning resource allocation strategies with spatial-temporal evidence, policymakers can enhance the equity and efficiency of healthcare delivery, ultimately improving health outcomes across all regions.

Due to the large differences in population distribution between regions, the granularity of MHT studies at the provincial level is still not fine enough. Future studies can carry out more detailed analysis of the spatial and temporal distribution of MHT at the urban level in China. Given that the migration and flow of talents is a universal phenomenon, different types of health professionals, such as doctors and nurses, may exhibit varying characteristics. Further exploration is needed to understand the spatial and temporal evolution of different types of MHT. Moreover, since this study only analyzes the indicators of medical and health talents, and lacks the analysis of its influencing factors, it cannot conduct the regression analysis of time series. The regression analysis of such time series can be added in future studies to explore the influencing factors of MHT.

Conclusions

Since the Chinese government implemented reform and opening-up policies in 1978, population mobility and MHT in China have progressively increased. This study is academically valuable as it systematically examines the spatial and temporal distributions and evolution of MHT in China. The systematic analysis is highlighted in three main aspects. First, it uses comprehensive data on MHT spanning nearly 40 years, from 1982 to 2020. Second, it develops a more extensive index system for measuring MHT agglomeration, encompassing four dimensions: TAD, TAS, TAI, and TAE. Third, the research employs multiple methodologies to systematically analyze the spatial-temporal distribution of MHT, including the Kernel coefficient, Moran鈥檚 I analysis, and the Dagum Gini coefficient.

The study found several key trends: (1) From 1982 to 2020, TAD and TAI of MHT in China showed an upward trend, whereas TAS decreased. TAE initially increased and then decreased. (2) At the provincial level, MHT numbers grew annually from 1982 to 2020, revealing significant geographical disparities. Higher concentrations were noted in the densely populated southeastern regions compared to the sparsely populated northwestern areas, with the eastern regions鈥攇enerally demacrated by the Hu Huanyong Line. (3) From 1982 to 2020, the eastern and western regions showed agglomeration, whereas the central region displayed discrete agglomeration. (4) Spatial autocorrelation analysis using Moran鈥檚 I for TAD, TAS, TAI, and TAE over multiple decades highlighted significant trends in regional interdependencies and disparities. (5) Spatial gap analysis using the Dagum Gini coefficient and its decomposition across TAD, TAS, TAI, and TAE from 1982 to 2020 in China鈥檚 31 provinces and three regions reveals varied intraregional and interregional disparity trends.