Does Urbanization Increase the Risk of Emerging Infectious Diseases in China? A Spatial Econometric Analysis

Abstract

The current COVID-19 pandemic has inspired more and more discussion about the relationship between urbanization and emerging infectious diseases (EID). This paper aims to examine the spatial effect of urbanization on EID incidence, such as hepatitis, syphilis and gonorrhea in China. Taking into account geographical and economic factors, the estimation results of the Spatial Error Model (SEM) show that urbanization has increased the risks of EID transmission from 2003 to 2019 in China. The spatial effects of urbanization are slightly different due to different types of infectious diseases, with a larger effect on syphilis comparing with hepatitis and gonorrhea. The regional heterogeneity test shows that the impact of urbanization on EID in eastern China is stronger than that in the Midwest especially when considering spatial correlation. Policy implications that include health must be integrated into urban planning, attaching more importance to ecological construction, improving regional joint prevention and control mechanisms, and more attention being paid to vulnerable groups. Considering the frequent occurrence of COVID-19 among cities in China, we emphasize the importance of enhancing the coordinated anti-crisis capacity of urban clusters and highlight the leading role of central cities.

1. Introduction

COVID-19 is now sweeping across countries and cities and emerging infectious diseases (EID) continue to be a significant impediment to public health in the 21st century [1,2]. EID refer to diseases that have newly appeared in the population, or have existed but are rapidly increasing in incidence or geographic range (e.g., tuberculosis, dengue fever, hepatitis B, C, COVID-19) [3]. Untangling the mechanisms underlying EID is one of the challenging scientific issues facing society, alongside demographic and social changes [4,5].

Factors influencing the emergence of EID are complex, with urbanization, driven by population growth and consumption, believed to be one of the main drivers [6,7,8]. Urbanization, with growing population size and density, leads to more frequent close contact between humans, increasing the risks of pathogen transmission [9,10]. In addition, land use associated with urbanization has effectively concentrated human, animal reservoir and vector at unprecedented densities [5,11], upsetting the balance of ecosystems, which may contribute to the emergence of EID [12,13].

As one of the largest developing countries, China’s urbanization has been a momentous event attracting worldwide attention [14,15]. Over the past three decades, the urban population in China has increased rapidly from 151.5 million (17.2% of total population) in 1973 to 842.9 million (60.3% of total population) in 2018 [16], and an increase of an additional 255 million will be reached by 2050 [17]. Meanwhile, infectious diseases remain a major public health issue facing the populous country [9]. Compared to 2018, the number of A and B infectious diseases increased by 0.3% and the number of deaths increased by 8.16% in 2019, with hepatitis being one of the major threats reported [18]. Does urbanization increase the transmission of EID in China? Urbanization and infectious diseases have spatial attributes. Taking into spatial factors, what is the impact of urbanization on EID? Moreover, are there regional differences? Answering these questions is helpful to deepen our understanding of the impact of urbanization on infectious diseases and thus better respond to the challenges of disease control in an increasingly urbanized country.

Numerous studies have been conducted to explore the role of urbanization in EID. Most research considered urbanization one of main factors contributing to the emergence of EID. They believed that urbanization with high density provided a host population large enough to support a continuous cycle of pathogens [19,20], favoring the transmission of diseases. Wilcox and Gubler [5] proposed a general model for the occurrence of zoonoses worldwide. The model linked demographic and social factors to land use and land cover changes to help explain the emergence of diseases. Wilcox and Gubler [21] elaborated on the risk parameters associated with urbanization, including population density, public infrastructure, and public services access to medical care, housing, water, sewage disposal, and waste management. Neiderud [22] concluded that megacities could become incubators for new epidemics, and zoonotic diseases could spread faster and become global threats. Connolly et al. [8] argued that the process of extended urbanization led to increased vulnerability to the spread of infectious diseases and highlighted three key factors, including demographic change, infrastructure, and governance.

As for empirical research, Keiser et al. [23] estimated that 200 million people (24.6% of the total African population) live at risk of contracting malaria in urban Africa based on United Nations figures on urbanization prospects. Saksena et al. [24] examined 10,820 communes and wards (classified as rural, peri-urban, and urban) of Vietnam and found that while urbanization alone was not a significantly independent risk factors of HPAI, it could lead to peri-urban places being the most likely HPAI “hot-spots” when other risk factors combined spatially. In a study of the COVID-19 Pandemic, Hamidi et al. [25] found metropolitan population (measured by density) to be one of the most significant predictors of infection rates, and larger metropolitan areas had higher infection and higher mortality rates.

Meanwhile, there are some other studies which argue that urbanization has decreased the risk of EID based on different urbanization indicators or methods. For example, Hamidi et al. [25] and Emilie et al. [26] concluded that destruction of vector habitats, improved housing conditions and access to curative measures could lead to a decline in disease incidence. Boussaa et al. [27] classified urbanization into intensely urban, urban areas near douars, urban douars, suburban douars, and waste-management sites. The result showed that urbanization had decreased sandfly populations, providing evidence that air and water pollution could hinder vector proliferation. Using population density as the indicator of urbanization, Wood et al. [28] argued that urbanization and wealth were associated with lower burdens for many diseases based on spatial and temporal analysis from 1990 to 2010. For the 24 infectious diseases selected, he found that most diseases had unique path models in the spatial analysis. Specifically, urbanization had inhibited the spread of diseases such as malaria, hepatitis, and tuberculosis, to varying degrees. Using the global database of the Global Rural–Urban Mapping Project (GRUMP), Qi et al. [29] detected that the transmission pattern of P. vivax was significantly lower in urban areas than in rural areas globally except for the Americas based on a set of 10,003 community-based P. vivax parasite rate surveys from America, Asia, and Africa. Hay et al. [30] defined Africa urbanization as locations with more than 1000 persons per km² and recalculated Africa’s malaria burden in 2000. The results showed that urban populations were on average subject to reduced levels of malaria transmission and severe disease than rural ones.

Although the existing research has provided a wealth of insights into the role of urbanization in EID, vastly broadened research perspectives and interdisciplinary methods are still needed to account for the underlying mechanism of the high incidence of EID [4]. First, current studies on China’s urbanization mainly focus on pollution, energy consumption, and economic development, while research on public health is insufficient. Second, most existing studies focus on the theoretical viewpoint of the role of urbanization in EID. Empirical studies are relatively few, especially in developing countries like China. Moreover, urbanization and EID have spatial attributes, which are often ignored in previous studies, leading to errors or deviations in estimating and analyzing process. Third, the existing empirical studies usually take the proportion or density of urban population as the measurement index of urbanization, ignoring the impact of comprehensive level of urbanization on EID. Fourth, the mechanisms of EID transmission are complex and vary not only from country to country but within countries. Existing studies are mostly concentrated at the national level, and there are relatively few studies on the differences within countries.

The contribution of this study is as follows. (1) This paper investigates the effects of urbanization on EID incidence, which enriches current research on public health in China and provides insights for understanding the transmission of COVID-19 from an economic perspective. (2) From the perspective of methodology, this study adopts the spatial panel model to measure the impact of urbanization on EID, hoping to make marginal contributions to the existing empirical research. (3) Considering the social environment effects, physical environment effects, and social services effects of urbanization, this paper uses the entropy method to construct the comprehensive index of urbanization, which can reflect the influence of comprehensive level of urbanization on EID. (4) From the perspective of regional differences, this paper examines the different impacts of urbanization on EID in the east and Midwest in China, providing implications for prevention and control of EID in different regions, especially during the COVID-19 pandemic.

The remainder of this paper is as follows. In Section 2, we introduce the method of spatial empirical analysis, and then define the relevant variables. Section 3 carries on the spatial empirical analysis. Section 4 is the discussion. Section 5 is the conclusion with some policy implications.

2. Methodology

2.1. Method

This paper adopts the Global Moran’s I index to measure the spatial autocorrelation of urbanization and the incidences. The calculation formula of Global Moran’s I is set as follows:

$$\mathrm{I}\left(\mathrm{d}\right)=\frac{{\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\sum }_{\mathrm{j}\ne 1}^{\mathrm{n}}{W}_{\mathrm{ij}}({\mathrm{x}}_{\mathrm{i}}-\bar{\mathrm{x}})({\mathrm{x}}_{\mathrm{j}}-\bar{\mathrm{x}})}{{\mathrm{S}}^2{\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\sum }_{\mathrm{j}\ne 1}^{\mathrm{n}}{W}_{\mathrm{ij}}}$$

(1)

where x_i denotes the variable in region i, $\bar{\mathrm{x}}$ represents the mean value of x_i, S² is the sample variance, and W_ij represents the spatial weight matrix. The value range of Moran’s I is [−1, 1]. There is a tendency of spatial clustering trend when Moran′s I is positive and a spatial dispersion trend when it is negative. The observed values show an independent distribution trend when the Global Moran’s I is zero [31].

2.2. Spatial Econometric Model

When the main source of spatial correlation is the lag term of EID incidence, the Spatial Lag Model (SLM) is established [32]. When the error term is spatially correlated, the main source of spatial correlation is the random error term [33]. Therefore, Spatial Error Model (SEM) is established. The spatial models are as follows:

SLM:

$${\ln \mathit{IID}}_{\mathit{it}}{=\beta }_0{+\rho W\ln \mathit{IID}}_{\mathit{it}}{+\beta }_1{\ln \mathit{UCI}}_{\mathit{it}}{+\beta }_2\ln X_{control}{+\epsilon }_{\mathit{it}}$$

(2)

SEM:

$$\begin{gathered} {\ln \mathit{IID}}_{\mathit{it}}{=\beta }_0{+\beta }_1{\ln \mathit{UCI}}_{\mathit{it}}{+\beta }_2\ln X_{control}{+\mu }_{\mathit{it}} \\ {\mu }_{\mathit{it}}{=\lambda W\mu }_{\mathit{it}}{+\epsilon }_{\mathit{it}} \end{gathered}$$

(3)

where x_i denotes the variable in region i, t represents the year in the sample period, lnIID_it represents the incidence of EID of each province in China, and lnUCI_it denotes the comprehensive index of urbanization in each province. X_control is the control variable, including international travelling, humidity, and political efficiency

The most commonly used spatial weight matrices include geographical weights matrix, inverse distance matrix, and economic distance weight matrix. This paper constructs geographical weights matrix (W₁) to investigate the effects of urbanizations on the incidences of EID. The geographical weights matrix is as follows:

$$W_1=\{\begin{array}{c}1,\mathrm{if}i\mathrm{and}j\mathrm{are}\mathrm{adjacent}\\ 0,\mathrm{otherwise}\end{array}$$

(4)

where i and j denotes different regions in China.

2.3. Variable and Data Source

Incidence of EID (IID). This paper adopts the incidence of hepatitis (IID_hep), the incidence of syphilis (IID_syp) and the incidence of gonorrhea (IID_gon) as proxy variables of EID, which is reported the main sources of the EID in China [18].

Urbanization comprehensive index (UCI). The effects of urbanization on EID mainly involve three themes: Social environmental effects, physical environment effects, and social service effects [34]. Social environment refers to the attributes of urban communities affecting individual behavior [34]. Physical environment includes land use, transportation, pollution, and housing. Social service mainly involves the service capacity of urban infrastructure such as education and medical care.

In order to reflect the social effect, physical effect, and social service effect of urbanization on EID, the paper selected 15 indicators from five aspects of population urbanization, economic urbanization, social urbanization, land urbanization, and ecological urbanization to evaluate the urbanization comprehensive index based on Zhang et al. [35] and Wei et al. [36] (

Table 1. Evaluation system of urbanization comprehensive index.

Classification	Index	Meaning
Population urbanization	Proportion of urban population (%)	Urban population/total population
	Population density (person/km2)	Urban population/urban area
Economic urbanization	GDP (100,000,000 yuan)	Gross Domestic Product
	Proportion of urban secondary and tertiary industry output value (%)	Added value of secondary and tertiary industry/GDP
	Income (yuan)	The disposable income of residents
Social urbanization	Medical Insurance Coverage (%)	Number of people participating in urban basic medical insurance/urban population
	Number of doctors owned per 10,000 person (person)	Number of urban doctors/urban population (10,000 person)
	Bed number owned per 10,000 person (beds)	Total number of beds in urban area/population in urban area (10,000 person)
	Public transport vehicles per 10,000 people	Public transport vehicles/urban population(10,000 person)
	Education level (%)	The ration of employed persons with bachelor degree
Land urbanization	Built-up area (km2)	Built-up area in urban
	Land area requisitioned(km2)	Land area requisitioned each year
	Urban road area per capita (m2)	Area of urban roads/urban population
Ecological urbanization	Green area of built-up area (%)	Green land area in urban built-up area/urban built-up area
	Cleaning area (10,000 km2)	Area of urban cleaning

). Among them, population urbanization mainly reflects the change of urban population proportion and density, and economic urbanization mainly reflects the change of urban economic structure, both of which have social environmental effects on EID. Land urbanization mainly reflects the expansion of urban scale and the change of landscape, while ecological urbanization mainly reflects the change of urban living environment, both of which have physical environment effect on EID. Social urbanization reflects the changes of a city’s public service capacity such as education and medical care, which can have social service effects on EID.

Entropy method was used to determine the sub-index weight and the urbanization comprehensive index. The larger the urbanization comprehensive index, the higher the degree of urbanization.

This paper uses the following variables as control variables:

International travelling (TRA): Connectivity is an important factor in disease transmission and EID can quickly spread around the world with international travel [37,38]. We used the number of inbound trips to China as a surrogate variable.

Humidity (HUM): Climate plays an important part to EID [9]. This paper used the humidity of Chinese provinces over the years as a substitute variable.

Political efficiency (POL): Government efficiency plays a key role in the control of the epidemic [39]. Studies have noted that the role of governments and organizations such as World Health Organization (WHO) is to take the necessary preventive measures quickly and effectively before the epidemic gets out of control [22]. Improving the transparency of government policies can promote public supervision and improve efficiency. The government’s response to administrative litigation cases represents its openness and transparency to a certain extent. This paper uses the product of administrative reconsideration case rate and administrative response case rate in each province of China to measure the efficiency of the government.

The sample used includes 30 provinces (municipalities) in China (excluding Tibet, Hong Kong, Macao and Taiwan) from 2003 to 2019. The data are derived from the China Statistical Yearbook (2004–2020), China City Statistical Yearbook (2004–2019), China Health Statistics Yearbook (2004–2020), China Environmental Statistics Yearbook (2005–2019), China Environment Yearbook (2004–2019), and China Law Yearbook (2004–2018).

3. Results

3.1. Spatial Autocorrelation Analysis

The Global Moran’s I of lnUCI, lnIID_hep, lnIID_syp, and lnIID_gon from 2003 to 2019 are significantly positive at the 5% significance level, indicating that there are spatial autocorrelations and it is necessary to consider the spatial effect.

At the same time, the regional distribution maps of urbanization comprehensive index and the EID incidence in 2003 and 2019 are shown in Figure 1, Figure 2, Figure 3 and Figure 4. As can be seen from Figure 1, the high urbanization index areas are mainly concentrated in the eastern region, especially in the coastal provinces like Shandong, Jiangsu, Zhejiang, Fujian, and Guangdong. In terms of the incidence, compared with 2003, hepatitis incidence in central provinces such as Hubei and Hunan, and in southeastern provinces such as Shandong, Fujian, Guangdong, and Hainan all increased in different proportions in 2019 (see Figure 2). The regional distribution of syphilis and gonorrhea rates show the similar trend to that of hepatitis incidence (see Figure 3 and Figure 4). Generally, the high-value urbanization comprehensive index area is the common high-value incidence area, while the low-value urbanization comprehensive index area is the common low-value incidence area. This shows that urbanization comprehensive index and EID incidence are basically consistent in regional distribution and spatial clustering in China. Based on spatial econometric models, the following section of the paper will further analyze the spatial effects and differences of the effects of urbanization on EID in China.

3.2. Empirical Analysis

Before the spatial regression, the LM test, LR test, and Hausman test were carried out to find an appropriate spatial model (

Table 2. Spatial models specification results.

Methods	W1
LM-lag test	29.016 ***
R-LM-lag test	3.075
LM-err test	395.534 ***
R-LM-err test	369.593 ***
LR-test joint significance spatial fixed effect	76.87 **
LR-test joint significance time fixed effect	625.97 ***
Hausman test	38.63 ***

Note: *** and ** denote statistical significance at the 1% and 5% significance levels.

). In the LM test, the model with higher statistical significance is better. If they pass the significance test at the same level, a robust LM test needs to be conducted, and then a suitable model according to the results of robust LM test can be selected [33]. The test results show that the results of SEM and SLM both pass the significant test at the 1% level at least, but in the robust LM test, the SLM fails to pass the significant test. Therefore, SEM is more suitable for the regression. The Hausman test shows the fixed model is feasible. Besides, both LR-test joint significance spatial and time fixed effect are significantly positive at least at the 5% level. Thus, SEM under the space-and-time fixed effect was selected to measure the impact of urbanization on EID incidence.

The regression results for the incidence of hepatitis, syphilis, and gonorrhea are listed in columns 1–3 of

Table 3. Estimation results for EID incidence.

	lnIIDhep	lnIIDsyp	lnIIDgon	lnIIDhep	lnIIDsyp	lnIIDgon
lnUCI	0.112 **	1.210 ***	0.774 ***	0.840 ***	3.504 ***	1.815 ***
	(0.1089)	(0.1285)	(0.1029)	(0.2375)	(0.2419)	(0.2392)
lnTRA	0.201 ***	0.213 ***	0.0319 *	0.176 ***	0.152 ***	0.044 **
	(0.0420)	(0.0477)	(0.0393)	(0.0417)	(0.0440)	(0.0399)
lnHUM	0.146	0.636 ***	−0.0773	0.132	0.450 **	0.0963
	(0.1711)	(0.1900)	(0.1622)	(0.1679)	(0.1742)	(0.1608)
lnPOL	0.0335	−0.0802	0.056	−0.011	0.0177	0.0231
	(0.0242)	(0.0266)	(0.0222)	(0.0235)	(0.0237)	(0.0221)
λ	0.0649 ***	0.0212 **	0.0108 **	0.0820 ***	0.0105 *	0.0428 *
	(0.0176)	(0.0182)	(0.0159)	(0.0182)	(0.0175)	(0.0181)
Adjust R2	0.47	0.49	0.56	0.43	0.76	0.52
N	510	510	510	510	510	510

Note: The t-values given in the parentheses. ***, **, and * denote statistical significance at the 1%, 5%, and 10% significance levels.

respectively. According to the regression results, spatial error regression coefficient λ are significant at least at 5% level under the time fixed effects and the space fixed effects. According to the regression results in Table 3, urbanization comprehensive index is positive correlated with the hepatitis incidence and is significant at least at the 5% level, illustrating that urbanization has led to an increase in the incidence of hepatitis during the sample period. The regression results of the incidence of syphilis and gonorrhea are basically consistent with that of hepatitis, with the values slightly larger than the latter.

Sustainable urban development is crucial to China’s sustainable development [40]. Over the past decade, China’s urbanization has been accompanied by rapid population growth, large-scale land development and dramatic changes in social structures. High-density living increases close contact between humans, which increases the risk of disease transmission [9]. Moreover, in the process of urbanization in China, overdevelopment of land has become a common phenomenon, challenging the bearing capacity of land. Massive new buildings, diverted rivers, and reduced vegetation affect the local ecological system, and break the balance between pathogen, host, and environment, leading to increased transmission of vector-borne pathogens, and thus increasing the risk of disease transmission to others, especially in crowded areas of poor sanitation [41,42,43]. Besides, the imbalance between urbanization and industrialization has become a significant feature in China [44]. According to the industrialization development theory, the urbanization rate in the world is about twice that of industrialization [45], while the ratio in China was about 1.46 in 2018, lower than the world average [44]. One of the direct effects of lagging urbanization is that public infrastructure construction and management in many Chinese cities have not kept pace with changes in social structures such as demographics [11]. This is detrimental to disease control as the occurrence of infectious diseases is not only the result of environmental disturbance, but also often the result of public health management failure [46].

As for the control variables, the number of international travelers is significantly positive with the incidence of hepatitis and syphilis at the 1% level, and is significantly positive at the 10% level with gonorrhea incidence. In an increasingly globalized world, connectivity plays an important part in the transmission of EID and the growth of international travel has contributed to the spread of pathogens between countries. This is also consistent with the research conclusion of Saksena et al. [24] and Wu et al. [11]. Humidity is significantly positive with syphilis incidence but not significant with hepatitis and gonorrhea incidence, which was not in line with our expectations. It is well known that changes in the natural environment are one of the important factors affecting EID. The insignificance may due to the fact that selection of the indicators is too simple to truly reflect the climate change. The coefficient of the government efficiency is also not significant. The reason may be that the government cannot learn and popularize public health knowledge fast enough to keep up with the pace of urbanization in China. Accordingly, the role of education and government effectiveness in EID control has not played out.

3.3. Robustness Test

To test the robustness of the impacts of urbanization on EID, this paper uses urban population ratio (POURB) instead of urbanization comprehensive index (UCI) as the explained variable for spatial regression. The spatial panel model and method chosen for robustness test are consistent with the above. It can be known from columns 4–6 of Table 3 that while the coefficients and significance of the variables have decreased or increased to a certain extent, the main results are basically consistent with the above, which shows that the impact of urbanization on EID is reliable and robust.

3.4. Regional Heterogeneity Analysis

Due to China’s vast territory and uneven population distribution, the process of urbanization varies greatly, especially in the eastern and western regions. Subject to the China’s Seventh Five-Year Plan of National Economy and Social Development, this paper divides the 30 provinces into the eastern regions and Midwestern regions. The eastern regions include Beijing, Tianjin, Hebei, Liaoning, Shanghai, Jiangsu, Zhejiang, Fujian, Shandong, Guangdong, and Hainan. The remaining provinces belong to the Midwestern regions. In order to better explain regional differences of EID, the data from 2003 to 2016 were taken as the training sample and the data from 2017 to 2019 were used as the test sample. The paper first performed regression of SEM based on the data from 2003 to 2016 (

Table 4. Regional heterogeneity of EID incidence.

		East			Midwest
	lnIIDhep	lnIIDsyp	lnIIDgon	lnIIDhep	lnIIDsyp	lnIIDgon
lnUCI	0.415 ***	0.197 **	0.524 ***	−0.0432	0.0323	0.0174
	(0.0564)	(0.0648)	(0.0490)	(0.0538)	(0.0620)	(0.0488)
lnTRA	0.616 ***	0.572 ***	0.532 ***	0.155 **	0.0469	−0.107 *
	(0.1114)	(0.1266)	(0.0958)	(0.0499)	(0.0655)	(0.0468)
lnPOL	0.0486	−0.0853 *	−0.0429	0.0263	0.063	0.0086
	(0.0366)	(0.0417)	(0.0313)	(0.0266)	(0.0322)	(0.0251)
lnHUM	0.0737	−0.58	−0.266	0.181	−0.537	−0.0327
	(0.3785)	(0.4354)	(0.3244)	(0.2303)	(0.2822)	(0.2163)
λ	0.296 ***	0.276 **	0.167 **	0.397 ***	0.128	−0.137
	(0.0759)	(0.0867)	(0.0766)	(0.0927)	(0.0938)	(0.0880)
Adjust R2	0.47	0.65	0.57	0.42	0.44	0.35
N	154	154	154	266	266	266

Note: The t-values given in the parentheses. ***, **, and * denote statistical significance at the 1%, 5%, and 10% level.

) and then predicted the regional incidence from 2017 to 2019. The actual incidence and the predicted incidence are shown in

Table 5. Comparison of the predicted and actual incidence.

	East			Midwest
IIDhep	Actual	Predicted	Error (%)	Actual	Predicted	Error (%)
2017	81.9303	77.9113	4.91%	103.5716	101.4126	2.08%
2018	78.5464	78.5821	−0.05%	101.4647	100.3163	1.13%
2019	80.7291	85.2425	−5.59%	98.1353	100.6672	−2.58%
MAPE(%)	3.4%			1.9%
IIDsyp	Actual	Predicted		Actual	Predicted
2017	39.6645	41.3718	−4.30%	37.0868	38.1885	−2.97%
2018	38.4109	40.2523	−4.79%	39.3105	39.1963	0.29%
2019	40.8127	41.4433	−1.55%	41.8447	39.3911	5.86%
MAPE(%)	3.5%			3%
IIDgon	Actual	Predicted		Actual	Predicted
2017	15.3427	14.8074	3.49%	6.5784	6.3319	3.75%
2018	13.8536	13.1165	5.32%	6.4158	6.2412	2.72%
2019	11.5645	12.1061	−4.68%	5.8532	6.2407	−6.76%
MAPE(%)	3.7%			4.4%

.

As can been in Table 4, the spatial error regression coefficient λ is significant in the east with the urbanization comprehensive index significantly positive at the 5% level at least. The spatial error regression coefficient λ in the Midwest is only significant when the explained variable is hepatitis incidence with the coefficient of lnUCI not significant. Moreover, the goodness of fit in the east are better those in the Midwest. These results indicate that urbanization in eastern China has a more significant effect on the transmission of EID than in the central and west regions. Generally speaking, the spatial regression results of the east are basically consistent with those of the whole sample while the Midwest does not show the similar trend with the whole country. These results are closely related to the imbalance of regional development in China. Compared with the central and western regions, urbanization in eastern China is developing at a faster pace. High-density population in the eastern cities increases the frequency of close contact and changes the way people socialize and live, which facilitates the spread of pathogens. In addition, with the rapid development of eastern cities, there has been an influx of migrant workers. Influenced by living conditions and economic income, they are more vulnerable to contracting the disease such as hepatitis, tuberculosis, and sexually transmitted diseases. They may delay seeking treatment due to lack of health insurance, thus increasing the risk of disease being transmitted to others.

The relative errors and MAPE (Mean Absolute Percentage Error) in Table 5 are used to evaluate the accuracy of predictions. The relative errors demonstrate the size of the predictive error to some extent, and MAPE estimates the mean accuracy of the entire forecast process. As is shown in Table 5, most of the relative errors are less than 5%, and MAPE ranges from 1.9% to 4.4%, which is below 5%, revealing the accuracy of the model. The comparison shows the regional incidence trends of EID in China to some extent. As for the hepatitis incidence, although the absolute number of the east is smaller than the Midwest, the incidence of the former is in a rising state, while the latter shows a certain downward trend. The syphilis incidence is basically in the same range in the east and Midwest, and both show a certain rising state. The absolute number of gonorrhea incidence in the east is more than twice as high as that in the Midwest although both regions have shown a certain downward trend.

4. Discussion

The prevalence of COVID-19 globally has received intensive attention and research on EID has become a hot topic. Given the complexity of EID, it is essential to consider its origins from an economic and social perspective, in addition to epidemiological efforts. Although many studies have speculated about the linkage between urbanization and EID, the empirical evidence is limited, especially in China. Using spatial panel data spanning nearly 20 years, this paper finds that urbanization has increased the transmission of EID in China, consistent with the conclusions of Tong et al. [9] and Wu et al. [11].

The construction of urbanization comprehensive index is helpful to measure the impact of urbanization on infectious diseases in China. However, the urbanization comprehensive index only covers 15 sub-indicators, which has certain limitations and cannot discriminate between the urbanized and the most urbanized areas, especially megacities. In addition, urbanization not only affects EID alone, but also may affect EID through other factors, such as industrialization. We fully recognize the shortcomings of our research and strive to address these problems in future research.

The comparison of the predicted and actual incidence shows that although urbanization in the east has a more significant impact on EID than that in the Midwest, its incidence has been controlled to some extent without rapid spread. A reasonable explanation is that although the higher population mobility in the eastern region promotes the transmission of EID, higher living standards, health and medical service, and even higher government efficiency can inhibit the epidemic of EID.

There are several limitations to our models. First, the sample only involves three kinds of EID, which is relatively small. Future studies could extend the sample to more types of EID. Secondly, the paper only uses one model in the prediction. Machine learning could be introduced in the future to make more accurate predictions. Thirdly, the pathogenesis of EID is complex, and each has its own unique pathogenesis. The particularity of EID is not considered in the paper. Fourth, external environment has a great impact on the spread of EID, and factors such as season, public cooperation, and government supervision should also be considered in the model.

5. Conclusions

This paper explores the spatial effect of urbanization on EID including hepatitis, syphilis and gonorrhea using panel data of 30 provinces in China from 2003 to 2019 by introducing geographical weights matrix, inverse distance matrix, and economic distance matrix. The main conclusions are as follows.

First, urbanization and the EID incidence are characterized by spatial differences and spatial autocorrelation. The spatial distribution of urbanization generally shows characteristics of aggregation in the east and southeast, which is basically consistent with that of EID.

Second, spatial regression results show that urbanization in China has led to an increase in the incidence of EID during the sample period. For different infectious diseases, the effect of urbanization varies to some extent. The effect of urbanization on syphilis is stronger than that on hepatitis and gonorrhea in sample period. High population densities, large-scale land development, and lagging urbanization can all be potential risks of EID transmission.

Third, the heterogeneity tests show urbanization in eastern China is more conducive to the spread of EID. Higher population density, changes in urban lifestyles, and the influx of migrant workers all bring higher risk of infection to eastern China.

From a policy perspective, we propose the following policy implications.

First, health must be integrated into urban planning. Urban health involves many sectors, including environment, medical, housing, and transport. Therefore, cross-sectoral cooperation is essential to epidemic control. Besides, the development of the health service system should keep pace with urbanization. The authorities should constantly promote the community-based public health services, improve the quality of professionals, and popularize public health education.

Second, more importance should be attached to the ecological construction in urban planning. The government can formulate strict biological protection laws, advocate for the application of green energy, and establish nature reserves and wetland parks to create a living environment where people and organisms can coexist harmoniously. In particular, policymakers must develop the political will to actively develop and implement these plans.

Third, regional joint prevention and control mechanisms should be constantly improved. The authorities should strengthen regional coordination on the release of EID warning information, establish open and transparent promulgate information system, and coordinate the administration of departments in the region. In particular, the health department should keep communication lines open, ensuring closer regional cooperation regarding medical information, diagnoses, and treatment plans.

Fourth, more attention should be paid to vulnerable groups. Urban floating population, as a vulnerable group, has become an accelerator of the transmission of EID due to the lack of adequate medical resources. Health policy makers have not taken fully into account the needs of migrants when formulating related policies and regulations. Access to insurance coverage, health care, and community support needs to be extended to the migrant population regardless of their residence and employment status.

Fifth, it is necessary to enhance the collaborative anti epidemic capacity of urban agglomerations and highlight the leading role of central cities. Urban agglomeration is a network system with frequent exchanges of people and logistics flow. As regional economic ties become ever closer, COVID-19 may spread rapidly among cities in China. The government should comprehensively evaluate the anti-risk ability of urban agglomeration, establish a systematic epidemic control scheme, intensify surveillance and case management of intercity floating population, and strengthen the function of community. In addition, as the center of regional economy, politics, and culture, central cities can help enhance regional resilience to COVID-19 through risk sharing, information sharing, material assistance, and experience sharing.

COVID-19 is continuing even as we complete this paper. It has gone beyond the realm of conventional epidemiology, which requires interdisciplinary efforts, including geography, epidemiology, sociology, and economics, to prevent and mitigate future outbreaks. Thus, while our analyses provide some insights into the impact of urbanization on EID from the perspective of spatial econometrics, more efforts are needed to broaden our understanding of the relationship between urbanization and EID.