Industrial Wastewater Discharge and Disease Incidence in China: A Spatial Analysis of Public Health and Sustainable Development Implications

Abstract

With the continuous advancement of industrialization in China, industrial wastewater discharge has become a critical factor influencing water environmental quality, public health, and the long-term sustainability of regional development. This study systematically examines both the direct and spatial spillover effects of industrial wastewater on disease incidence. Based on panel data from 30 provincial-level regions in China over the period 2011–2020, a composite incidence index of four waterborne infectious diseases is constructed using the entropy weight method, and the Spatial Durbin Model (SDM) is employed to capture both local and cross-regional effects. The results show that industrial wastewater discharge significantly increases disease incidence and exhibits clear spatial spillover effects, suggesting that the associated health risks may extend beyond local boundaries. Moreover, the analysis suggests that the “Water Ten Plan” reduced both local effects and regional spillovers, highlighting the value of stricter discharge control and coordinated basin-level governance for sustainable regional development. Overall, this study uncovers the spatial health externalities of industrial pollution and provides empirical support for integrated policy approaches linking environmental governance with public health protection.

1. Introduction

Industrial wastewater discharge has become one of the major sources of global water pollution [1]. According to a report by the United Nations Educational, Scientific, and Cultural Organization (UNESCO), a large proportion of wastewater worldwide is discharged directly into water bodies without adequate treatment, posing severe threats to ecosystem security and human health [2,3]. Industrial wastewater often contains harmful substances such as heavy metals, organic pollutants, and pathogenic microorganisms, which can enter the human body through drinking water, irrigation systems, and the food chain, inducing a series of waterborne diseases [4,5]. In the context of accelerated urbanization and industrialization, developing countries are facing particularly severe water pollution challenges [6,7]. In China, for example, industrial wastewater has become a critical barrier to improving water quality and public health [8,9,10].

Existing studies have revealed the impacts of water pollution on human health from multiple perspectives. A growing body of empirical evidence shows that heavy metals, organic pollutants, and pathogens in industrial wastewater are associated with elevated risks of specific diseases once they enter the human body via water environments. Some studies have focused on severe chronic outcomes. Polluted water environments have been linked to higher risks of liver cancer, pancreatic cancer, brain cancer, and other long-term health problems [11,12,13]. To broaden the health perspective beyond single infectious outcomes, recent research has also shown that the health effects of polluted water may extend to digestive cancer mortality and other chronic consequences in China [9]. Prior literature indicates that water pollution is associated with a wide range of health outcomes, including waterborne infectious diseases, chronic illnesses, and cancer, with these effects exhibiting clear regional and population heterogeneity [14]. This broader pattern indicates that the health burden of water pollution is heterogeneous rather than limited to a single disease category. Against this background, evidence from Bangladesh shows that the incidence of skin diseases, diarrhea, and dysentery was substantially higher in severely polluted areas along the Turag River [15]. Similar findings have been reported in India, where communities near sewage overflow sites experienced a higher incidence of acute gastroenteritis, typhoid fever, dysentery, and hepatitis A [16]. Collectively, these studies indicate that water pollution is closely associated with adverse health outcomes [9,14].

Within this broader literature, industrial wastewater merits particular attention because it is a major pollution source whose impacts often extend beyond the place of discharge. As a typical watershed-based pollutant, its generation and diffusion are shaped by industrial location, regional production linkages, and cross-regional pollutant transfer [17,18]. As a result, the health and environmental effects of industrial wastewater are often not confined to the area where the pollution originates [17,18]. This understanding is further supported by provincial panel evidence from China, showing that industrial wastewater discharge remained at a relatively high level during the study period and exhibited clear spatial dependence [19]. Accordingly, industrial wastewater should not be treated simply as a local explanatory variable in analyses of its health consequences.

The health consequences of pollution may also exhibit a spatial dimension because environmental risks can extend beyond administrative boundaries through hydrological and related pathways [14,17,18]. Recent studies suggest that pollution and its associated spatial spillovers may affect health-related outcomes in neighboring regions [20]. Evidence from related environmental health research also indicates that conventional non-spatial panel models may be inadequate when spatial dependence is present [21].

To examine the relationship between pollution and health, existing studies have employed diverse approaches in both indicator construction and empirical identification. Lloyd and Wheeler proposed the Disease Product Index (DPI), which multiplies the incidence rates of diarrhea, typhoid fever, and hepatitis to capture the combined effects of water quality on multiple diseases [22]. A significant positive correlation was reported between the DPI and water pollution levels. Field survey evidence from Pakistan further showed that drinking groundwater contaminated by industrial wastewater was associated with a markedly higher hepatitis A infection rate, reaching 32–38% [23].

These studies provide a solid foundation for understanding the association between water pollution and health. However, several important limitations remain. First, most of the current studies focus on specific watersheds or individual diseases, lacking a comprehensive assessment of the health risks posed by multiple waterborne diseases, which limits the full depiction of complex exposure scenarios [14]. Second, spatial spillover effects in the pollution–health relationship have been insufficiently addressed in current studies. There is a general lack of attention to the dynamic evolution of spatial and temporal heterogeneity. Given these spatial diffusion characteristics, pollutants can be transported across regions through surface runoff and hydrological cycles, posing health risks not only locally but also to adjacent areas. This calls for simultaneous consideration of both localized impacts and spatial spillovers.

To address these gaps, this study selects four infectious diseases with clear waterborne pathways—bacillary dysentery, typhoid fever, hepatitis A, and hepatitis E. We construct a composite incidence index using the entropy weight method to comprehensively measure the regional conditions of waterborne diseases. Based on panel data from 30 provincial-level regions in China between 2011 and 2020, we employ the Spatial Durbin Model (SDM) to systematically assess the direct and indirect effects of industrial wastewater discharge on public health in both local and neighboring areas. Furthermore, we examine changes in the spatial relationship between pollution and health outcomes before and after the implementation of China’s national water pollution control policy—the “Water Ten Plan.” This design allows us to assess both localized effects and spatial spillovers and to examine whether the pattern of association changed after the implementation of the “Water Ten Plan.”

This study contributes to the literature in three main respects. First, it develops a multi-disease indicator system, thereby broadening the assessment of water-related public health risks. Second, it brings localized effects and spatial spillovers into a unified analytical framework, helping to identify how industrial wastewater discharge influences health outcomes across regions. Third, it evaluates whether national pollution control policies were accompanied by changes in pollution-related health risks, providing evidence from China on the role of public intervention in mitigating environmental health risks.

2. Materials and Methods

2.1. Spatial Weight Matrix

Spatial weight matrices are used to quantify the correlation between different spatial units. Spatial weight matrices are generally divided into adjacency, geographical distance, and economic distance weight matrices [24,25]. Combined with the context of this study, potential cross-regional dependence in industrial wastewater discharge and disease incidence may exist in geographic space. At the same time, environmental governance policies such as the “River Chief System” and the “Water Ten Plan” mostly use administrative boundaries as the unit of responsibility division, making an adjacency-based matrix suitable for capturing spatial dependence among neighboring provinces. Therefore, this study constructed the adjacency matrix to examine the spatial dependence of industrial wastewater discharge and its association with health risks in neighboring regions. Specifically, a rook contiguity matrix was adopted, in which a weight of 1 was assigned when two provincial-level regions shared a common boundary and 0 otherwise [24,25,26]. The formula used is as follows:

$$W=\left[\begin{array}{ccc}\begin{array}{cc}{W}_{11}& W_{12}\\ W_{21}& W_{22}\end{array}& \begin{array}{c}\dots \\ \dots \end{array}& \begin{array}{c}{W}_{1n}\\ W_{2n}\end{array}\\ \begin{array}{cc}⋮& ⋮\end{array}& ⋮& ⋮\\ \begin{array}{cc}{W}_{n1}& W_{n2}\end{array}& \dots & W_{nn}\end{array}\right]$$

(1)

$$W_{ij}=\left\{\begin{array}{l}1, Region i is geographically adjacent to region j.\\ 0, Region i is not geographically adjacent to region j.\end{array}\right.$$

(2)

2.2. Spatial Econometric Model

To examine the spatial association between industrial wastewater discharge and disease incidence, this study constructed a spatial econometric model based on the geographic adjacency matrix. According to Anselin, the spatial econometric model is specified as follows:

$$Y=\rho WY+\beta X+\theta WX+\gamma X_c+\mu$$

(3)

$$\mu =\lambda W\mu +\epsilon ,\epsilon \sim N\left(0,{\sigma }^{2}I\right)$$

(4)

where $Y$ represents the explained variable; $X$ is the core explanatory variable; $X_c$ represents a set of control variables, and $W$ represents the spatial weight matrix and is expressed in Equation (1); $\rho$ and $\theta$ represent the spatial autoregressive coefficients; $\beta$ is the parameter vector associated with $X$, and $\gamma$ is the coefficient vector for control variables; $\mu$ and $\epsilon$ represent the error terms, where $\epsilon$ is distributed with a normal distribution. When $\rho \ne 0$, $\theta =0$ and $\lambda =0$, the model corresponds to the Spatial Autoregressive Model (SAR); when $\rho =0$, $\theta =0$ and $\lambda \ne 0$, the model corresponds to the Spatial Error Model (SEM); when $\rho \ne 0$,$\theta \ne 0$ and $\lambda =0$, the model conforms to the Spatial Durbin Model (SDM) [27].

2.3. Variable Description

2.3.1. Explained Variable

The explained variable is the composite incidence index. This study selected four types of waterborne infectious diseases—bacillary dysentery, typhoid fever, hepatitis A, and hepatitis E—as core indicators, given their close association with water quality conditions and water-related transmission risks documented in prior studies.

The Entropy Weight Method (EWM) is an objective weighting method that determines the weights based on the degree of variability of the indicators and is widely used in multi-indicator evaluation systems. Li et al. found that the incidence rate of infectious diseases was given the highest weight due to its high volatility in constructing the index of the health level of the population, which indicated its importance in the assessment of public health [28]. Drawing on this approach, this study applied the entropy weight method to integrate the incidence rates of the four diseases and constructed a composite incidence index. The specific steps are as follows:

1.

Data normalization

To eliminate the magnitude difference, the incidence rate data of the four types of diseases in each provincial-level region were normalized using min–max scaling:

$$z_{ij}={\displaystyle \frac{x_{\mathit{ij}}-\min \left(x_j\right)}{\max \left(x_j\right)-\min \left(x_j\right)}}$$

(5)

where $x_{ij}$ is the incidence rate of disease $j$ in region $i$, and $z_{ij}$ is the standardized value.

2.

Entropy and Weight Calculation

First, the entropy value for each indicator was computed as follows:

$$e_j=-{\displaystyle \frac{1}{\ln n}}\sum _{i=1}^n{p}_{ij}\ln p_{ij}$$

(6)

where

$$p_{ij}={\displaystyle \frac{z_{ij}}{\sum z_{ij}}}$$

(7)

Then, the divergence degree and entropy weight were calculated as follows:

$$d_j=1-e_j$$

(8)

$$w_j={\displaystyle \frac{d_j}{\sum d_j}}$$

(9)

A larger $w_j$ indicates greater spatial variability in the incidence rate of disease $j$, suggesting that it contributes more significantly to the overall health risk profile.

3.

Index Aggregation

Finally, the composite incidence index for each provincial-level region was obtained through the weighted sum method:

$$Incidenc{e}_{ij}=\sum _{j=1}^4{w}_j\cdot z_{ij}$$

(10)

This composite indicator reflects its importance in health risk assessment while preserving the relative differences among diseases.

2.3.2. Explanatory Variables

The explanatory variable is industrial wastewater discharge, measured by the total volume of industrial wastewater discharged in each provincial-level region. The unit of the original data of industrial wastewater discharge is 10,000 tons. To eliminate the magnitude difference, a logarithmic transformation was applied to this variable in this study.

2.3.3. Control Variables

This study includes industrial added value, urbanization rate, per capita water consumption, health expenditure, and population density as control variables. Industrial added value is measured by the annual industrial added value of each provincial-level region at constant 2000 prices. It is included to capture differences in industrial scale and potential pollution pressure across regions, as regional industrial conditions are closely related to industrial wastewater pollution and its governance outcomes [29]. The urbanization rate is measured by the proportion of the urban resident population to the total resident population. It reflects differences in population concentration, infrastructure, and environmental exposure, and related studies have treated urbanization as an important socioeconomic factor in analyses of pollution and public health [30]. Per capita water consumption is calculated as total water consumption divided by the total population. It is included to reflect differences in water-use intensity and possible exposure conditions. Health expenditure is measured by the share of total health expenditure in GDP and is included to reflect regional differences in health resource input [31]. Population density is defined as the number of permanent residents per square kilometer. It is included to capture demographic concentration and has commonly been controlled for in analyses of pollution and public health [32]. To improve dimensional comparability and reduce potential heteroscedasticity, industrial added value, per capita water consumption, and population density are expressed in natural logarithmic form.

2.3.4. Data Sources

This study used a balanced panel dataset covering 30 provincial-level regions in China from 2011 to 2020. Tibet, Hong Kong, Macao, and Taiwan were excluded because consistent data for the full study period could not be obtained, and the final sample therefore consisted of 300 observations. Industrial wastewater discharge data were sourced from the China Environmental Statistical Yearbook. Data on disease incidence were obtained from the China Health Statistical Yearbook. Data for control variables were collected from the China Statistical Yearbook and the China Health Statistical Yearbook. To ensure consistency in measurement units and mitigate the impact of heteroscedasticity, all non-ratio variables were log-transformed. Definitions and descriptive statistics of the variables are provided in

Table 1. Variables description and descriptive statistics.

Variables	Mean	SD	Min	Median	Max
Bacillary dysentery	12.78	13.41	0.12	8.78	94.45
Typhoid fever	0.76	1.35	0.00	0.34	10.94
Hepatitis A	2.00	2.80	0.09	1.14	26.17
Hepatitis E	1.88	1.16	0.19	1.56	5.45
Incidence	0.27	0.12	0.05	0.26	0.73
Industrial wastewater discharge (lniww)	64,456.07	57,111.65	5573.00	42,689.50	246,298.00
Industrial added value (lniav)	8756.66	8011.03	415.40	6464.95	39,353.90
Urbanization rate (ur)	0.59	0.12	0.35	0.57	0.90
Per capita water consumption (lnpcwc)	500.53	414.07	161.20	433.40	2657.40
Health expenditure (hep)	6.53	1.81	3.14	6.34	12.97
Population density (lnpd)	469.74	707.29	7.87	286.90	3949.56

SD: standard deviation; Min: minimum; Max: maximum. Industrial added value is calculated at constant prices in 2000.

. All statistical analyses were conducted using Stata 18.0.

To ensure the robustness of the regression estimates, the main explanatory variables were tested for multicollinearity using the Variance Inflation Factor (VIF). As shown in

Table 2. Multicollinearity test.

Variables	VIF	1/VIF
Industrial wastewater discharge	5.96	0.168
Industrial added value	5.28	0.189
Urbanization rate	2.87	0.349
Per capita water consumption	1.93	0.518
Health expenditure	1.93	0.519
Population density	1.65	0.606
Mean VIF	3.27

, all VIF values are below the commonly accepted threshold of 10, indicating that multicollinearity is not a serious concern in the model [33].

3. Results

3.1. Spatial Autocorrelation

To investigate the spatial correlation characteristics of environmental health risks, this study applied Global Moran’s I and Local Moran’s I indices [27,34,35]. The global Moran’s I index is used to determine whether global spatial agglomeration is significant. The local Moran’s I index is used to identify local spatial clustering characteristics and regional heterogeneity. These two indices enable a comprehensive analysis of spatial agglomeration patterns and their evolution over time.

3.1.1. Global Spatial Correlation

Table 3. Global Moran’s I index of industrial wastewater discharge and composite incidence index.

Variables	Industrial Wastewater Discharge			Composite Incidence Index
	Moran’s I	Z	p-Value	Moran’s I	Z	p-Value
2011	0.266 **	2.463	0.014	0.414 ***	3.806	0.000
2012	0.223 **	2.121	0.034	0.401 ***	3.740	0.000
2013	0.257 **	2.406	0.016	0.332 ***	3.100	0.002
2014	0.232 **	2.193	0.028	0.243 **	2.348	0.019
2015	0.275 ***	2.561	0.010	0.352 ***	3.264	0.001
2016	0.292 ***	2.800	0.005	0.350 ***	3.169	0.002
2017	0.252 **	2.395	0.017	0.408 ***	3.666	0.000
2018	0.246 **	2.346	0.019	0.473 ***	4.201	0.000
2019	0.252 **	2.387	0.017	0.429 ***	3.807	0.000
2020	0.223 **	2.179	0.029	0.415 ***	3.705	0.000

Note: *** and ** indicate statistical significance at the 1% and 5% levels, respectively.

reports the global Moran’s I values of industrial wastewater discharge and composite incidence index across 30 provincial-level regions in China from 2011 to 2020. Moran’s I values of industrial wastewater discharge remained positive throughout the period, indicating persistent spatial clustering. Among them, eight years are significant at the 5% level and two years at the 1% level. Spatial dependence peaked in 2016, with a Moran’s I of 0.292. The ten-year average Moran’s I was 0.252, suggesting a stable and significant positive spatial autocorrelation in the spatial distribution of pollution.

In contrast, the composite incidence index exhibits even stronger spatial dependence. It reaches the 1% significance level in nine out of ten years and the 5% level in the remaining year. Moran’s I value peaked in 2018 at 0.473, indicating the strongest spatial clustering of disease transmission during that year. The average Moran’s I over the decade was 0.382, significantly higher than that of wastewater discharge, suggesting a more pronounced spatial aggregation of infectious diseases. This pattern is consistent with the possibility that environmental exposure and regional interaction may jointly shape the spatial distribution of disease incidence.

In summary, Moran’s I values of both indicators deviate significantly from a random spatial pattern, indicating clear spatial dependence in environmental pollution and public health risks.

3.1.2. Spatial Distribution Patterns

Figure 1 presents the choropleth maps of the composite incidence index and industrial wastewater discharge in 2011 and 2020. In the regression analysis, industrial wastewater discharge was log-transformed, whereas Figure 1 presents the original values to better display the spatial distribution. As shown, both variables exhibit clear regional variation rather than a random spatial pattern.

For the composite incidence index, relatively higher values were mainly observed in western and southwestern provinces in 2011, while by 2020, the high-value areas were more concentrated in southwestern China, especially Guangxi and Yunnan. By contrast, many northern and eastern provinces remained at relatively low levels in both years.

For industrial wastewater discharge, relatively high levels were concentrated in eastern and central China in 2011. By 2020, high-value areas were concentrated in a smaller number of provinces, particularly along the eastern coast. Most western provinces, by contrast, continue to exhibit comparatively low levels.

Overall, the maps provide descriptive evidence of non-random spatial distribution for both variables and support the subsequent spatial autocorrelation tests and spatial econometric analysis.

3.1.3. Local Spatial Correlation

To facilitate temporal comparison, this study focuses on two cross-sectional years, 2011 and 2020, covering 30 provincial-level administrative units. The regional codes are reported in

Table A1. Provincial-level region codes.

N	Regions
1	Beijing
2	Tianjin
3	Hebei
4	Shanxi
5	Inner Mongolia
6	Liaoning
7	Jilin
8	Heilongjiang
9	Shanghai
10	Jiangsu
11	Zhejiang
12	Anhui
13	Fujian
14	Jiangxi
15	Shandong
16	Henan
17	Hubei
18	Hunan
19	Guangdong
20	Guangxi
21	Hainan
22	Chongqing
23	Sichuan
24	Guizhou
25	Yunnan
26	Shaanxi
27	Gansu
28	Qinghai
29	Ningxia
30	Xinjiang

in Appendix A. Figure 2 visually presents the spatial correlation features between industrial wastewater discharge and the composite incidence index through local Moran’s I scatter plots.

The local Moran’s I scatter plots reveal significant spatial clustering characteristics of the composite incidence index and industrial wastewater discharge across China’s 30 provincial-level regions in both 2011 and 2020. Data points in the first and third quadrants of the local Moran scatter plots indicate positive spatial correlations, representing “high–high” (HH) and “low–low” (LL) clusters, respectively [36]. Over 65% of the regions are distributed in the HH and LL quadrants, suggesting that spatial homogeneity dominates. The HH clusters of the composite incidence index have remained stable in southwestern regions (e.g., Yunnan, Guizhou, Guangxi), while LL clusters are concentrated in areas with lower population density (e.g., Heilongjiang, Inner Mongolia).

In contrast, the HH clusters of industrial wastewater discharge have persistently been concentrated in manufacturing hubs such as the Yangtze River Delta and the Shandong Peninsula. Meanwhile, southeastern coastal regions (e.g., Guangdong, Fujian) transitioned from HH to LL clusters during the observation period, suggesting that the spatial distribution of industrial wastewater discharge may have changed over time. The proportion of outlier regions (low–high and high–low clusters) is relatively low, further confirming weak spatial heterogeneity. Moreover, the differences in evolving spatial patterns suggest that regional health risks are relatively persistent over time, possibly because they are shaped by the uneven allocation and limited short-term adjustability of healthcare resources. In contrast, industrial pollution emissions seem to be more sensitive to policy intervention, and their spatial changes may be related to industrial relocation during the 13th Five-Year Plan period.

3.2. Empirical Analysis of Spatial Econometric Model

3.2.1. Model Selection and Specification Tests

Previous analyses have confirmed significant spatial correlations between industrial wastewater discharge and the composite incidence index. However, while Moran’s I test can detect spatial autocorrelation, it cannot discern the specific types of spatial effects. To determine the optimal spatial econometric model specification, this study sequentially conducted LM, Hausman, and LR tests.

First, the Lagrange Multiplier (LM) test was employed to identify spatial effect types. A classical linear regression (OLS) was performed on panel data to obtain a model without spatial terms, followed by constructing LM test statistics using the residuals. Both standard and robust LM tests were implemented. As shown in

Table 4. LM tests.

Test	Statistic	p Value
LM spatial lag	56.123 ***	0.000
Robust LM spatial lag	60.315 ***	0.000
LM spatial error	30.745 ***	0.000
Robust LM spatial error	34.938 ***	0.000

Note: *** indicate statistical significance at the 1% level.

, the LM-lag and LM-error statistics are significant at the 1% level, indicating coexisting spatial lag effects and spatial error dependence. Following Anselin et al., when both dependencies are significant, relying solely on SAR or SEM may yield inconsistent estimates [37]. Thus, the Spatial Durbin Model (SDM), which simultaneously incorporates spatial lag and error terms, was considered.

To address unobserved heterogeneity under this complex specification, all three models (SAR, SEM, and SDM) were estimated, and Hausman tests were subsequently employed to assess fixed effects [27,38,39]. The results revealed a monotonic increase in χ² values with model complexity, ranging from 14.29 (SAR) to 25.99 (SDM) (SAR: ${\chi }^2$ = 14.29, $p$ = 0.046; SEM: ${\chi }^2$ = 16.90, $p$ = 0.018; SDM: ${\chi }^2$ = 25.99, $p$ = 0.017), with all p-values below the 5% threshold. This trend aligns with Mundlak and Wooldridge’s theoretical expectation that when unobserved individual effects are correlated with the regressors, the “zero-correlation” assumption underlying the random-effects model is systematically violated, leading to biased estimates [40,41]. Consequently, fixed-effects specifications with year fixed effects were uniformly adopted to mitigate omitted-variable bias.

After establishing the fixed-effects specification, this study further conducted a systematic comparison of the SDM, SAR, and SEM spatial econometric models using likelihood ratio (LR) tests. The SDM’s log-likelihood (324.07) substantially exceeded those of the SAR (271.49) and SEM (269.39) models, indicating a clear improvement in model fit. This finding is consistent with the finite-sample properties highlighted by Lee and Yu (2010) [42], who demonstrate that fixed-effects spatial panel estimators yield more reliable inference under limited time dimensions. The LR statistics rejected SDM’s degeneration into SAR (${\chi }^2$ = 105.14, $p$ < 0.01) or SEM (${\chi }^2$ = 109.35, $p$ < 0.01), consistent with Debarsy and Ertur’s Monte Carlo findings that when the true data-generating process is an SDM, estimating a simplified SAR or SEM model leads to systematic underestimation of spatial spillover effects [43]. The LM, Hausman, and LR test results support the use of the SDM with year fixed effects as the baseline specification. Relative to the SAR and SEM models, the SDM provides a better fit to the observed spatial dependence structure in the data.

In conclusion, the Spatial Durbin Model (SDM) was selected as the primary estimation framework. Unlike SAR or SEM, SDM accounts for spatial lags in both dependent and independent variables, enabling the identification of industrial pollution’s direct local effects and regional spillover mechanisms on public health. Detailed estimates are presented in

Table 5. Model regression results.

Variables	SAR	SEM	SDM
lniww	0.0345 (2.19) **	0.0415 (2.7) ***	0.0656 (4.43) ***
lniav	−0.0081 (−0.55)	−0.0069 (−0.52)	−0.0791 (−4.95) ***
ur	−0.1085 (−1.51)	0.0604 (0.72)	0.0742 (0.94)
lnpcwc	−0.0029 (−0.23)	0.0044 (0.36)	0.0244 (2.04) **
hep	0.0133 (2.44) **	0.017 (3.04) ***	0.0133 (2.32) **
lnpd	−0.0049 (−0.66)	−0.0184 (−2.24) **	0.0064 (0.64)
W×lniww	/	/	0.0652 (2.1) **
Log-likelihood	271.4942	269.3911	324.0662
σ2	0.0089 (11.87)	0.0088 (11.74)	0.0066 (12.11)
R2	0.1497	0.0355	0.2520
Hausman test	14.29 **	16.90 **	25.99 **
LR test	105.14 ***	109.35 ***	/

Note: *** and ** indicate statistical significance at the 1% and 5% levels, respectively. Parentheses enclose t-statistics. The symbol “/” indicates that the term is not applicable in the corresponding model specification.

.

3.2.2. Regression Results of the Spatial Durbin Model

The regression results indicate that industrial wastewater discharge (lniww) is significantly and positively associated with the composite incidence index both locally and in neighboring regions. In the SDM model, the regression coefficient is 0.0656 ($p<0.01$), indicating that after controlling for other variables, an increase in industrial wastewater discharge is associated with a higher composite incidence index in the local region. Meanwhile, the regression coefficient of the spatial lag term is 0.0652 ($p<0.05$), suggesting that pollution discharge in neighboring regions also exerts a significant spillover effect on local health risks. Overall, the results suggest that pollution-related health risks exhibit cross-boundary spatial association, which implies that pollution control may require greater interregional coordination rather than relying solely on isolated local responses.

Regarding control variables, industrial added value (lniav) is not significant in both the SAR and SEM models but shows a significant negative impact in the SDM model ($\gamma$ = −0.0791, $t$ = −4.95, $p$ < 0.01). This change suggests that traditional models may overlook the positive externality of economic structure in spatial environmental governance. Under the premise of pollution control, increased industrial output may be associated with higher levels of technological investment and pollution control, thereby mitigating public health risks [44]. Additionally, per capita water usage (lnpcwc) becomes significant for the first time in the SDM model ($\gamma$ = 0.0244, $t$ = 2.04, $p$ < 0.05), indicating a certain degree of spatial heterogeneity and a nonlinear relationship between water resource usage and health. In densely populated areas with water scarcity, the health impact per unit of water usage may be further amplified, suggesting that the ecological implications of water usage behavior warrant further in-depth exploration.

The proportion of total health expenditure to GDP (hep), a key indicator of public health financial capacity, shows a significant positive impact across all three models, with an estimated value of 0.0133 ($t$ = 2.32, $p$ < 0.05) in the SDM model. While this result is statistically significant, its direction deviates somewhat from conventional theoretical expectations. Typically, increased healthcare investment is expected to effectively improve regional health levels and reduce disease burden. However, in this study, a higher health expenditure ratio is associated with a higher composite incidence index of waterborne diseases. This can be interpreted from multiple perspectives. First, rising medical expenditures may reflect a reactive response to elevated disease prevalence rather than proactive investment, implying a reverse causation. In regions with severe epidemic or disease burdens, governments and households are compelled to increase health spending, producing an observed positive correlation without implying beneficial causal effects. Second, greater health expenditure may strengthen diagnostic and surveillance systems, thereby raising the apparent incidence rate. Higher health expenditures enhance detection capabilities, thereby increasing disease reporting rates [45].

3.2.3. Spatial Effect Decomposition

Building on the effect decomposition framework proposed by LeSage and Pace (2009) [39], this study further delineates the spatial action pathways of various variables through the Spatial Durbin Model. The results show that the estimated spatial indirect effect of industrial wastewater discharge accounts for a substantial share of the total effect. Under the baseline specification, the indirect effect is larger than the direct effect, indicating that cross-regional association may represent an important component of the overall relationship. This finding is consistent with the presence of spatial spillovers in environmental health risks and suggests that policies confined to a single jurisdiction may be insufficient to address cross-regional externalities.

Furthermore, the indirect effect of industrial added value is −0.2483 ($t$ = −4.51, $p$ < 0.01), which is higher than its direct effect (−0.0956, $t$ = −5.27, $p$ < 0.01). This suggests that regional industrial structure restructuring or green transformation generates significant positive spatial externalities, thereby reducing the public health burden in adjacent areas. In contrast, population density in neighboring areas imposes notable adverse spillovers, with a significant indirect effect of (0.1278, $t$ = 5.43, $p$ < 0.01), highlighting the need to establish cross-border epidemic monitoring points and high-density population health intervention mechanisms in population convergence zones. Additionally, health risks associated with urbanization development warrant sufficient attention. Although the local urbanization rate does not exhibit a significant direct impact, its indirect effect is highly significant (−1.7099, $t$ = −9.44, $p$ < 0.01). Consequently, integrated regional strategies within urban clusters are required, focusing on transportation infrastructure, cross-regional health service provision, and joint control of environmental pollution sources.

Overall, the SDM results suggest that the relationship between environmental pollution and public health has an important spatial dimension. This indicates that pollution control should not rely solely on isolated responses by individual jurisdictions but may require greater coordination among neighboring administrative regions. Strengthening regional coordination mechanisms is therefore crucial for curbing the spatial spillover of pollution and achieving synergistic improvements in both ecological quality and public health. The results of the spatial effect decomposition are presented in

Table 6. Spatial effect decomposition results.

Variables	Direct Effect	Spatial Spillover Effect	Total Effect
lniww	0.074 (4.38) ***	0.1236 (2.47) **	0.1977 (3.25) ***
lniav	−0.0956 (−5.27) ***	−0.2483 (−4.51) ***	−0.3439 (−5.01) ***
ur	−0.0264 (−0.38)	−1.7099 (−9.44) ***	−1.7362 (−9.18) ***
lnpcwc	0.0209 (1.6)	−0.0488 (−0.99)	−0.0279 (−0.49)
hep	0.011 (1.92)	−0.0361 (−2.73) ***	−0.0251 (−1.7) *
lnpd	0.0146 (1.48)	0.1278 (5.43) ***	0.1424 (5.07) ***

Note: ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. Parentheses enclose t-statistics.

.

3.3. Robustness Tests

To assess the robustness of the estimation results, this study re-estimates the models by replacing the spatial adjacency matrix with an inverse-squared distance matrix and substituting total wastewater discharge (lntww) for industrial wastewater discharge as the key explanatory variable. Owing to data availability constraints, total wastewater discharge data for different regions are available only for 2011–2017, as China stopped publishing such data after 2017. The data are obtained from the China Environmental Statistical Yearbook and are reported in units of 10,000 tons. To ensure comparability and improve model stability, they are transformed using natural logarithms prior to estimation.

In 2015, the State Council of the People’s Republic of China introduced the Water Pollution Prevention and Control Action Plan (“Water Ten Plan”), which imposed stricter standards on industrial wastewater discharges and established cross-boundary water quality assessment mechanisms. As this policy may have reshaped both the intensity of industrial wastewater pollution and the spatial dynamics of its diffusion, this study applies a segmented regression design to further test the robustness of the model. Specifically, the policy’s effective year (2015) is taken as the breakpoint, and the full sample is divided into two sub-periods: the pre-policy period (2011–2015) and the post-policy period (2016–2020). The spatial econometric models are then estimated separately for each sub-period. A comparative summary of the robustness test results, together with the baseline SDM estimates, is reported in

Table 7. Baseline estimation and robustness tests.

Variables	Baseline	Alt. Matrix & Alt. Variable	Subsample: 2010–2015	Subsample: 2016–2020
lniww	0.0656 (4.43) ***		0.1444 (4.69) ***	0.0574 (3.45) ***
lntww		0.0718 (1.88) *
lniav	−0.0791 (−4.95) ***	−0.0191 (−0.65)	−0.1321 (−4.84) ***	−0.0764 (−3.92) ***
ur	0.0742 (0.94)	0.0898 (1.05)	0.2414 (1.98) **	0.1118 (1.00)
lnpcwc	0.0244 (2.04) **	0.0121 (0.75)	0.0097 (0.56)	0.0164 (1.02)
hep	0.0133 (2.32) **	0.0467 (5.88) ***	0.045 (4.11) ***	0.0000 (0.00)
lnpd	0.0064 (0.64)	−0.026 (−2.36) **	−0.0099 (−0.67)	0.0117 (0.89)
W×lniww	0.0652 (2.1) **	0.298 (3.18) ***	0.1642 (2.08) **	0.0324 (0.95)
Log-likelihood	324.0662	211.9186	164.8437	175.1545
σ2	0.0066 (12.11)	0.0078 (10.23)	0.0064 (8.6)	0.0054 (8.45)
R2	0.2520	0.4593	0.4660	0.3753

Note: ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. Parentheses enclose t-statistics.

.

Across all robustness tests, the signs, significance levels, and directions of spatial effects of the core variables remain consistent with those in the baseline model, indicating that the research conclusions are robust to alternative model specifications, variable definitions, and sample periods. Furthermore, the segmented regression results provide additional evidence of the structural influence of the “Water Ten Plan” on the spatial dynamics of pollution.

In the post-policy period, the coefficients of both local effects and spatial spillovers showed a clear decline, and the spillover effect was no longer statistically significant. This indicates that the “Water Ten Plan” policy not only effectively reduced local pollution emissions but also weakened the cross-regional transmission of pollution. These findings suggest that the policy’s emphasis on cross-boundary coordination and water quality assessment may have been associated with improvements in water environment management.

4. Discussion

Industrial wastewater discharge is positively associated with waterborne disease incidence in China, and this relationship has a clear spatial dimension. Pollution-related health risks are not confined to the province where wastewater is discharged but are also associated with conditions in neighboring provinces. Water pollution, therefore, appears to be not only a local environmental problem but also a cross-regional public health issue.

This pattern is broadly consistent with previous research showing that wastewater pollution may affect health through multiple exposure channels and that both environmental risks and health outcomes often exhibit spatial clustering [11,15,17]. The analysis further shows that, at the provincial level in China, the association between industrial wastewater discharge and disease incidence extends across adjacent regions. This suggests that local environmental conditions and health outcomes are embedded in a broader regional context.

The post-policy analysis suggests that this relationship became weaker after the implementation of the “Water Ten Plan”. This pattern is broadly consistent with the policy emphasis on stricter discharge control and stronger cross-regional coordination in water governance.

From a policy perspective, the findings suggest that environmental health governance may benefit from greater coordination across regions. When pollution-related health risks are spatially associated, isolated local responses may be insufficient to address the full range of environmental health impacts. Strengthening coordination in pollution control, monitoring, and public health response may therefore improve the effectiveness of environmental governance.

Several limitations should also be noted. The health indicators cover only four waterborne infectious diseases and therefore do not capture the full health burden associated with water pollution. The spatial weight matrix used in the baseline analysis may not fully reflect hydrological linkages, economic interaction, or distance decay. In addition, the provincial scale of the data may conceal important within-province variation. These limitations point to clear directions for future research.

5. Conclusions

Using panel data for 30 provincial-level regions in China from 2011 to 2020, this study examines the spatial relationship between industrial wastewater discharge and waterborne disease incidence. The results show that both variables exhibit significant spatial dependence and that industrial wastewater discharge is positively associated with disease incidence at both the local and neighboring-regional levels.

The spatial effect decomposition further shows that the indirect effect accounts for a substantial share of the total effect, suggesting that cross-regional association is an important feature of pollution-related health risks. The post-policy analysis also indicates that this relationship became weaker after the implementation of the “Water Ten Plan” and that the spillover term was no longer statistically significant in the later period.

Overall, pollution control and health protection should not rely solely on isolated local responses. More effective governance may require stronger coordination across adjacent regions in order to address the cross-boundary nature of environmental health risks. Therefore, improving industrial wastewater management is important not only for reducing pollution and disease but also for promoting healthier and more sustainable regional development.