Unequal roles of cities in the intercity healthcare system

Three types of patient mobility interweave to form an IHS

We proposed a method to more fully identify cross-city healthcare utilization (Methods) and found that three types of patient mobility interweave to form an IHS (Fig. 2a). Primarily, mobility for better treatment and diagnosis accounts for 76% of all cross-city healthcare utilization, and only 5% is for the best treatment and diagnosis. The pseudo mobility of migrant workers accounts for 19%; they return to their healthcare-deprived home cities for treatment and diagnosis owing to being excluded from the healthcare welfare system in the megacities²⁶ or a lack of family support²⁷ and familiarity with healthcare providers in the megacities²⁸. In China, the hukou (household registration) system and the multilayered and municipally administered healthcare insurance system deny migrant workers’ full rights to local healthcare welfare^12,29 (Supplementary Note 4), and over 380 million of the population are migrant workers who often live in the megacities separately from their families³⁰.

a, The percentage of three types of patient mobility in total, and the percentage of cities that have different types of patient mobility as origins and destinations. b, The percentage of pseudo patient mobility in outflowing patients of origin cities. c, The percentage of pseudo patient mobility in inflowing patients of destination cities.

Cities are not equally involved in the IHS as the origins and destinations of patient mobility (Fig. 2a,b); 92% of 364 Chinese cities (336) are the origin only of the mobility for better treatment and diagnosis, while 13 cities are the origin of both the mobility for better treatment and diagnosis and pseudo mobility. Only 15 cities (4%) are the origin of all three types of mobility; these are often China’s most developed cities, including Beijing, Shanghai and Shenzhen.

Only 273 cities act as destinations of patient mobility, among which 24 cities host only the mobility for better treatment and diagnosis. Two cities host only the pseudo patient mobility, 15 cities (5%) host both the mobility for the best and for better treatment and diagnosis, while the other 232 cities (85%) host both the mobility for better treatment and diagnosis and pseudo patient mobility. Notably, cities with over 50% of hosted mobility as the pseudo patient mobility are mainly provinces in the middle of China, including Hubei, Hunan and Jiangxi (Fig. 2c), where the amount of outflowing migrant workers is among the highest in China.

Role of cities in intercity healthcare provision

To discern the role of cities in intercity healthcare provision, we cluster the 273 cities into four types according to the size and type of patients they host and other city attributes related to healthcare provision to nonlocal patients (Methods). Type A cities exhibit the highest number of nonlocal patient origins (mean 235, P < 1%; Fig. 3b), generating the largest catchment area and serving the highest number of nonlocal patients (mean 53,362; Fig. 3c). The 13 type A cities provide 35% of all the intercity healthcare resources, meeting the needs for the best (mean ratio 15%) and better (mean ratio 85%) treatment and diagnosis (Fig. 3g). Both the abundance in quality healthcare and high transport accessibility (Fig. 3d,e) determine type A cities’ uppermost attractiveness to nonlocal patients. Hence, we refer to type A cities as national healthcare hubs, as exemplified by Beijing, Shanghai and Chengdu.

a, The spatial distribution of cities with types A and B mostly located in East China. The four conceptual graphs: the angle of each fan with an inward arrow indicates the average volume of inflowing nonlocal patients, and the number on each circle indicates the average number of served cities at the corresponding spatial scale. The number of cities is 13 for type A, 30 for type B, 141 for type C, 89 for type D and 91 for type O. b, The number of served cities. c, The number of inflowing patients. d, Transport accessibility measured by indegree centrality of cities in national railway networks. e, The capability of quality healthcare measured by the number of tertiary-A hospital beds per 1,000 persons. f, The relationship between the number of served cities and the number of inflowing patients. g, The average ratio of the three types of patient mobility in the four types of cities (‘t/d’ refers to ‘treatment and diagnosis’). A Shapiro–Wilk test is used for normality testing, and not all the variables for each type of cities are normally distributed. In b–e, the included P values correspond to two-sided Mann–Whitney tests for each comparison between different types. The error bars represent the standard deviation, and the center values correspond to the means, which are the height of the bars.

Type B cities exhibit the second-highest diversification of nonlocal patient origin cities (mean 132). They serve a considerable but much lower number of nonlocal patients than do type A cities (mean 20,387, P < 1%). Having fewer quality healthcare services than type A cities (P < 1%), type B cities attract a lower ratio of nonlocal patients from nonadjacent provinces (Supplementary Fig. 5a). Rather than serving a nationwide catchment area, type B cities serve the nearby cities of their own and adjacent provinces and mainly meet the needs for better treatment and diagnosis (mean ratio 77%). Hence, we refer to type B cities as regional healthcare hubs, as exemplified by Kunming, Shenyang and Fuzhou.

Type C cities exhibit the lowest number of nonlocal patients (mean 1,268) and the fewest patient origin cities (mean 18; Fig. 3f). With the least richness of quality healthcare, type C cities attract few nonlocal patients, and the majority of the nonlocal patients they do serve are returning migrant workers (mean ratio 52%) from the province’s capital city. After excluding pseudo patient mobility, the number of origin cities of nonlocal patients served by type C cities is minimal (Supplementary Fig. 6a,c). Hence, we refer to type C cities as pseudo patient destinations.

Type D cities differ from type C cities by having a higher number of nonlocal patients (mean 5,623, P < 1%) and origin cities (mean 60, P < 1%). They provide healthcare to meet the needs for better treatment and diagnosis (mean ratio 56%) and of returning migrant workers (mean ratio 44%). In contrast to type C cities, returning patients in type D cities come from a variety of megacities rather than mainly from their own provincial capital city (Supplementary Fig. 6c,d). After excluding pseudo patient mobility, type D cities serve a much smaller number of nonlocal patients than type B cities (Supplementary Fig. 6b). Hence, we refer to type D cities as occasional destinations with a very limited function of serving nonlocal patients.

Type O cities do not serve any nonlocal patients. The poorest transport accessibility and the lack of quality healthcare mean that such cities are rarely a destination for nonlocal patients, even if they have outflowing migrant workers. Type O cities are often remote cities near border regions (Fig. 3a).

Figure 3a conceptualizes the four types of cities regarding the function of intercity healthcare provision. With substantial catchment areas constitutive of a wide range of cities, type A and B cities have a real function of serving other cities that lack quality healthcare; they account for 12% of Chinese cities (13 and 30, respectively). Accounting for 63% of Chinese cities (141 and 89, respectively), type C and D cities host mainly the returning migrants and occasional visits from nonlocal patients with few functions of serving nearby cities.

Role of cities in intercity healthcare demand

We use the same clustering method to discern the role of cities in intercity healthcare demand (Methods), with four types of cities classified. Type 1 cities exhibit the highest number of outflowing patients (mean 33,097, P < 1%; Supplementary Fig. 7a), and the patients go the farthest (mean 660 km; Fig. 4c) to reach the greatest diversification of destinations for healthcare (mean 176, P < 1%; Fig. 4b). On average, 59% of the patients go outside their province (Supplementary Fig. 8a). Type 1 cities are often the developed megacities, which are the richest in quality healthcare (Fig. 4e), and the populations’ education conditions and social networks are also the best, allowing the residents to know the best physicians and hospitals in other cities for treating their illnesses^10,31. After excluding pseudo mobility, almost half of their outflowing patients (mean ratio 49%) travel to obtain the best treatment and diagnosis (Supplementary Fig. 9a). This suggests that the residents in type 1 cities are mobile nationwide to obtain the best diagnosis and treatment.

a, The spatial distribution of cities with types 1 and 2 mostly located in East China. The four conceptual graphs: the angle of each fan with an outward arrow indicates the average volume of outflowing nonlocal patients, and the number on each circle indicates the average number of destination cities at the corresponding spatial scale. The number of cities is 16 for type 1, 33 for type 2, 203 for type 3 and 112 for type 4. b, The number of destination cities. c, The weighted travel distance of outflowing patients. d, The GDP per capita. RMB, renminbi. e, The number of tertiary-A hospital beds per 1,000 persons. f, The relationship between the number of destination cities and the number of outflowing patients. g, The average ratio of the three types of patient mobility in the four types of cities (‘t/d’ refers to ‘treatment and diagnosis’). A Shapiro–Wilk test is used for normality testing, and not all the variables for each type of cities are normally distributed. In b–e, the included P values correspond to two-sided Mann–Whitney tests for each comparison between different types. The error bars represent the standard deviation, and the center values correspond to the means, which are the height of the bars.

Type 2 cities exhibit the second-highest number of outflowing patients, but this number is still much lower than for type 1 cities (mean 12,610, P < 1%). The same is true of the diversification of destinations (mean 103). On average, about 42% of the outflowing patients from type 2 cities go outside their province to seek quality healthcare. Even though type 2 cities are relatively rich in local healthcare (Fig. 4e), the majority of their outflowing patients pursue better treatment and diagnosis outward (mean ratio 78%). Only one-third of type 2 cities have pseudo patient outflows, as they are not the major host cities of migrant workers. Type 2 cities are often medium-sized cities in the developed eastern coastal region that have good transport accessibility and economic conditions, so their residents can travel to reach nearby healthcare-rich cities (Fig. 4d and Supplementary Fig. 7b). This suggests that the residents in type 2 cities are mobile regionwide for better diagnosis and treatment.

Type 3 cities exhibit a minimal number of outflowing patients (mean 1,858) and destinations (mean 16; Fig. 4f), and only 25% of the patients go outside their province (Supplementary Fig. 8a) to several cities in nearby provinces (mean 11). Type 3 cities are the most deprived of quality healthcare (Fig. 4e), while the worst transport accessibility and economic conditions limit the residents’ ability to travel far and wide to obtain healthcare. Type 3 cities are often small cities in the less developed western inland region (Fig. 4a). This indicates that the residents in type 3 cities are immobile regarding obtaining better diagnosis and treatment.

Type 4 cities differ from type 3 cities in having a higher percentage of outflowing patients going beyond the province boundary (mean ratio 35%) and a higher number of destinations (mean 48, P < 1%). The much better transport accessibility (P < 1%) enables the residents of type 4 cities to travel to more nearby healthcare-rich cities, although they suffer from poor economic conditions and a shortage of quality healthcare similar to type 3 cities. They are often small cities in central and coastal China. This indicates that the residents in type 4 cities are mobile locally to obtain better diagnosis and treatment (Fig. 4g).

Figure 4a conceptualizes the four types of cities regarding the function of intercity healthcare demand. Accounting for only 4% of Chinese cities, type 1 cities create 27% of the intercity healthcare demand, among which most is for the best diagnosis and treatment by their residents moving nationwide. Not rich in quality healthcare and accounting for 9% of Chinese cities, type 2 cities demand 20% of the intercity healthcare resources by their residents traveling regionwide for better diagnosis and treatment. The remaining 87% of cities are type 3 and 4 cities, which demand 53% of the intercity healthcare resources (Fig. 5).

The numbers in brackets refer to the percentage of cross-city healthcare utilization of the corresponding type.

Interdependent function and the benefit structure

The patient flows among different types of cities reflect the interdependency between the healthcare provision and demand cities, with flow size indicating intensity of interaction (Fig. 5). National hub cities serve type 4 cities the most (15%) rather than the most disadvantaged type 3 cities, indicating that accessibility to healthcare-rich cities offsets the local shortage of healthcare. Meanwhile, regional hub cities provide healthcare substantially to both type 3 and 4 cities. The advantaged type 1 cities depend substantially on occasional destination cities to provide healthcare to their residents (9%). One possible reason is that type 1 cities, mostly China’s first-tier megacities, have a concentration of jobs involving frequent traveling³². Therefore, they generate healthcare demands during business trips to a wide range of cities^33,34. It is notable that the unnoticed pseudo destination cities function to serve the advantaged type 1 and 2 cities, which host the majority of China’s migrant workers.

We discern the benefit structure of the IHS by examining the dual roles of cities. Combining the provision and demand roles yields nine compound modes involving more than three cities (Fig. 6a). For each city, we consider (1) the sufficiency of local healthcare provision for the local population and (2) the ability to obtain intercity healthcare provision for the local population. As a city both gains and loses healthcare resources within the IHS, we measure each city’s relative gains of intercity healthcare resources (Methods). The results indicate significant differences in the two attributes among different compound modes (Fig. 6b,c). Accordingly, we infer which compound modes benefit from the IHS regarding ensuring healthcare provision to the local population.

a, Nine compound modes combining the provision and demand roles. The numbers in the brackets refer to the number of cities at each junction. b, The sufficiency of local healthcare provision to the local population measured by the number of tertiary-A hospital beds per 1,000 persons (left), and the ability to obtain intercity healthcare provision to the local population measured by the relative gains of each mode under the IHS (right). A Shapiro–Wilk test is used for normality testing, and not all the variables for each mode of cities are normally distributed. The included P values in b correspond to two-sided Mann–Whitney tests for each comparison between different modes. c, The relationship of each mode between the sufficiency of local healthcare provision and the ability to obtain intercity healthcare provision.

Referring to Fig. 6a, cities at the junction of type A and type 1 are labeled as magnates, which means that they have rich healthcare resources while leveraging the nationwide IHS to further strengthen local healthcare provision. Accordingly, cities at the junction of type B and type 2 are local magnates. They benefit less from the IHS and depend mainly on a regional base. We describe cities at the junction of type D and type 2 as free-riders in that they do not contribute local resources to the IHS while obtaining intercity resources to compensate for the local shortage of quality healthcare.

We refer to cities at the junction of type C and type 4 as traders, which means that they offer local resources to remediate the megacities’ supply deficiency in meeting the healthcare needs of migrant workers, while obtaining intercity resources to serve local residents. They thus exchange benefits with other cities in the IHS. Similarly, cities at the junction of type D and type 4 are also traders that exchange their service function of meeting occasional healthcare demand from megacities for the supply to their local populations.

By contrast, refuges, which refers to cities at the junction of type C and type 3, can scarcely benefit from the IHS; they do not obtain intercity resources to offset their shortage of supply to the local population while providing substantially local resources to returning migrant workers. At the junction of type D and type 3, caretakers are also disadvantaged in the IHS. They do not obtain additional supply from other cities to serve local populations while providing the scarce local resources to meet the occasional healthcare demand from megacities. Givers are cities at the junction of type B and type 4, which share self-sufficient local resources with nearby cities but obtain few resources from the IHS.

In total, magnates, local magnates and free-riders constitute the winners of the IHS, respectively accounting for 4%, 4% and 5% of Chinese cities and accommodate 16%, 8% and 8% of the Chinese population. On the other side are refuges, caretakers and givers, which respectively account for 24%, 7% and 4% of Chinese cities and accommodate 13%, 6% and 6% of the Chinese population. Traders (27%) are in an ambiguous position in the benefit structure. Interestingly, 24% of the cities are at the junction of type O and type 3; they accommodate 9% of the population of China and are outsiders, as they demand little from the IHS and provide nothing in return. The results point toward a Matthew effect of accumulated advantage. The cities rich in local healthcare provision get richer through the IHS.