Coupling coordination between ecological environment quality and public health of residents in the Yellow River Basin, China: A modified coupling coordination model approach

TL;DR

This study examines the relationship between environmental quality and public health in the Yellow River Basin, China, using a modified model to assess their coordination and identify influencing factors.

Contribution

The paper introduces a modified coupling coordination model to evaluate the interaction between ecological environment quality and public health.

Findings

Both ecological environment quality and public health levels in the Yellow River Basin showed an overall upward trend from 2011 to 2022.

The coupling coordination degree increased steadily across the entire sample and within all three major regions.

Regional disparities in per capita GDP significantly influence differences in coupling coordination levels.

Abstract

The continuous development of industrial technology has led to significant environmental pollution and climate change, both of which have severely impacted human health. Investigating the coupling coordination between ecological environment quality (EEQ) and public health of residents (PHR) is beneficial for enhancing public health and promoting sustainable development. This study uses panel data from 55 cities within urban agglomerations of the Yellow River Basin, China (YRBC) from 2011 to 2022 to construct evaluation index systems for both EEQ and PHR. The entropy method is first employed to quantify the development levels of these systems. Subsequently, a modified coupling coordination degree (CCD) model is applied to evaluate the coordination between the two systems. Furthermore, the study utilizes the Dagum Gini coefficient, Kernel density estimation, and Markov chains to analyze the spatiotemporal evolution of CCD. The Quadratic Assignment Procedure (QAP) is finally used to empirically test the factors influencing regional differences in CCD. The findings reveal that both EEQ and PHR levels in the YRBC exhibited an overall upward trend during the study period, although PHR showed declines in certain years. The CCD demonstrated a steady increase across the entire sample and within all three major regions. Analysis using the Dagum Gini coefficient indicates a narrowing disparity in CCD, with the Gini coefficient decreasing from 0.0617 in 2011 to 0.0536 in 2022. Kernel density estimation suggests that the CCD distribution curve has shifted rightward, becoming higher and steeper, indicative of reduced absolute differences in coupling coordination levels. QAP regression analysis reveals that factors such as regional disparities in per capita GDP significantly influence CCD regional disparities.

Figures

The flowchart of this paper.

The classifications of CCD.

The evaluation value of EEQ and PHR.

KDE estimation.

Equations

Full text

1. Introduction

Promoting global health and ensuring universal access to quality healthcare have been highlighted as key components of the United Nations Sustainable Development Goals (SDGs). Nevertheless, the intensification of anthropogenic activities has led to widespread ecological disruption, further aggravating numerous environmental challenges [1]. As these global environmental challenges intensify, the construction of an ecological civilization has become a focal point of international concern. The quality of the ecological environment is not only crucial for the Earth’s ecological balance and sustainable development but also directly impacts PHR. Environmental degradation, including air and water pollution and the loss of biodiversity, is closely linked to various public health issues. For instance, exposure to polluted air has been closely linked to higher risks of cardiovascular and respiratory illnesses. Similarly, contaminated water sources are major contributors to the spread of waterborne infections. Moreover, the deterioration of ecosystem functions—such as food provision and the natural purification of air and water—poses significant threats to both human health and overall quality of life. In addition to these direct effects, environmental degradation may also exert indirect influences on population health through alterations in living patterns, economic stability, and mental health conditions. Given the rising societal attention to health issues, it is increasingly important to investigate the complex interrelations between PHR and EEQ. Unraveling the dynamic interaction pathways between EEQ and PHR is crucial to promoting human well-being and achieving broader sustainability targets.

The YRBC encompasses four primary geomorphic regions: the Qinghai-Tibet Plateau, Inner Mongolia Plateau, Loess Plateau, and the North China Plain. Geographically, it spans eastern, central, and western parts of China, and is characterized by both abundant natural endowments and high population density [2]. Extending across nine provincial-level administrative units, the basin stretches approximately 5,464 kilometers and occupies a total area of about 795,000 square kilometers [3,4]. As a strategically important region, the sustainable development of the YRBC plays a pivotal role in ensuring China’s long-term economic vitality and social stability. In recognition of its significance, the Chinese government released the “Outline for Ecological Protection and High-Quality Development of the YRBC” in 2021, which serves as a comprehensive policy blueprint for present and future initiatives focused on ecological conservation and regional advancement. This document provides critical guidance and a foundation for the sustainable economic development of urban agglomerations with the basin. Therefore, this study focuses on these urban agglomerations, measuring the coupling coordination level between EEQ and PHR, and analyzing their spatiotemporal evolution patterns. This study endeavors to deepen insights into the coupling coordination status and its temporal evolution between EEQ and PHR, with the broader goal of supporting pathways toward green, low-carbon, and sustainable regional development. The results are expected to provide a theoretical framework and practical policy guidance for enhancing the synergy between EEQ and PHR in both present planning and future governance efforts.

Existing studies concerning the linkage between EEQ and PHR have largely concentrated on each domain independently rather than addressing their interplay. Regarding the ecological environment, Gu et al. (2020) systematically explored China’s ecological civilization strategy, summarizing its experiences and lessons, which provide valuable insights for other countries pursuing sustainable development [5]. Jiang et al. (2021) employed long-term NDVI datasets spanning from 1998 to 2018 to investigate the spatiotemporal evolution of ecological conditions across multiple administrative levels in China, including regions, provinces, and counties [6]. Fan et al. (2019), adhering to principles of scientific rigor, rationality, and representativeness, developed an evaluation index system for the ecological environment of 31 provincial capital cities in China and measured these using the entropy method [7]. Li et al. (2020) constructed an evaluation system that includes indicators such as greening coverage in urban areas to assess the ecological environment of 19 resource-based cities in Northeast China and studied the coordinated development between ecological and socioeconomic systems using a coupling coordination degree model [8]. Building upon China’s inaugural national policy on ecological civilization initiated in 2012, Wu et al. (2021) developed a comprehensive weighted index to assess the ecological civilization level across Chinese provinces. Their evaluation compared the periods 2007–2012 and 2012–2017, revealing that the national ecological civilization progress rate during 2012–2017 reached 14.94%, which was approximately 2.3 times higher than the rate observed in the preceding five years [9]. In the context of the Guangdong-Hong Kong-Macao Greater Bay Area—one of the most prominent bay areas globally—Yang et al. (2020) examined the ecological consequences of rapid urbanization. By incorporating remote sensing indicators such as vegetation coverage and the vegetation health index, they constructed an integrated ecological assessment framework to capture the spatiotemporal dynamics of environmental quality from 1987 to 2017, highlighting its transformation under increasing urban development pressure [10].

Regarding public health, numerous scholars have emphasized the detrimental effects of environmental degradation. Girard and Nocca (2020) underscored pollution as a major determinant negatively influencing human well-being [11]. Drawing on data from the 2017 China General Social Survey, Liu et al. (2023) employed a hierarchical linear modeling approach and confirmed that air pollution poses a significant threat to public health [12]. Similarly, Currie et al. (2009) investigated how three common air pollutants prevalent in the 1990s affected infant health in New Jersey. By integrating air quality monitoring data with information from birth records, their findings revealed a strong correlation between carbon monoxide exposure and adverse health outcomes during and shortly after birth [13]. He et al. (2016), through a time-series analysis using Ningbo data from 2009 to 2013, quantified the burden of air pollution on years of life lost (YLL), further underscoring the pressing need for effective pollution control measures [14]. In addition to pollution, socioeconomic disparities have also been examined in relation to health outcomes. Guo et al. (2022) utilized panel data from the China Family Panel Studies (CFPS) and applied a Probit model to analyze the health effects of income inequality in rural populations [15]. Pei et al. (2023), based on China General Social Survey data, explored the influence of environmental regulation on public health and discovered that such policies significantly enhance health outcomes, albeit with varying effects across population subgroups [16]. Furthermore, Yang et al. (2020), using data from the 2010 “Third Survey on the Social Status of Women in China,” identified urban-rural disparities in environmental awareness, revealing that urban residents are more sensitive to pollution issues such as air and noise, both of which were found to negatively affect urban health [17]. Most recently, Xie et al. (2024) matched 2018 CFPS data with the national forest city development list and used probit, logit, and instrumental variable methods to investigate how green urban initiatives influence public health. Their results showed significant improvements in both physical and mental health among residents, with the findings remaining robust under multiple empirical checks [18].

A review of the existing literature reveals that while comprehensive research has been conducted on both EEQ and PHR, these two interconnected systems mutually influence and constrain each other. Most current research predominantly focuses on the unidirectional impact of EEQ on PHR [19–21], with relatively few studies examining the coupling coordination between these systems in the YRBC. The YRBC, spanning the eastern, central, and western regions of China, is characterized by spatial proximity, frequent socioeconomic interactions, and integrated infrastructure, making it one of the most dynamic regions with significant growth potential in China’s economic development.

In response to the above context, this study centers on urban agglomerations within the YRBC region to explore the coupling coordination patterns between EEQ and PHR. Specifically, it seeks to address two central research questions: (1) How has the coupling relationship between EEQ and PHR evolved over time in the YRBC? and (2) What key determinants drive the level of coordination between these two systems? To tackle these issues, this study develops a comprehensive evaluation framework for both EEQ and PHR. After quantifying their respective development levels, a revised CCD model is adopted to examine the dynamic interactions between the two dimensions from 2011 to 2022. Furthermore, the research investigates the spatiotemporal trajectory of CCD across the region. The ultimate aim is to offer policy insights that can facilitate the synergistic development of ecological and public health systems, thereby contributing to the broader goal of sustainable and harmonious coexistence between humanity and nature. The overall research framework is illustrated in Fig 1.

The flowchart of this paper.

2. The index system and methodology

2.1. Index system and data sources

There is currently no consensus within the academic community on the assessment of EEQ and PHR. Different scholars have developed various evaluation index systems based on their specific research objectives; however, these systems lack uniformity.

There are two main approaches to evaluating EEQ. The first approach involves using a single indicator. For example, Dong et al. (2021) established a performance assessment framework for ecological civilization and applied Data Envelopment Analysis (DEA) to evaluate the ecological efficiency across 30 Chinese provinces [22]. In addition, some researchers have utilized quasi-natural experiments—such as the designation of national ecological civilization pilot zones [23,24] and the implementation of national forest city pilot policies [25]—as proxy variables for EEQ in empirical analyses. Another widely adopted approach involves the development of comprehensive indicator systems. Based on the conceptual framework and strategic objectives of ecological civilization, Mi et al. (2022) proposed an evaluation model encompassing 28 indicators, carefully selected to ensure both relevance and data accessibility [26]. At the national scale, Zhang et al. (2019) constructed an evaluation framework incorporating dimensions like green production, ecological environment, green living, and infrastructure [27]. Likewise, Meng et al. (2021) focused on the urban dimension by designing an index system that reflects green sustainability, coordinated green growth, and spatial development [28].

For the measurement and quantification of PHR, many existing studies rely primarily on singular indicators, such as morbidity or mortality rates [24]. However, several scholars have sought to establish more multidimensional assessment frameworks. For example, Jia et al. (2023) developed an index system for assessing physical health by incorporating six indicators—including perinatal mortality—while also considering data accessibility [29]. Similarly, Yang et al. (2022) constructed a self-reported health assessment tool that captures psychological well-being, physical condition, and overall life satisfaction to provide a more holistic measure of residents’ general health status [30]. Despite these advancements, research dedicated to comprehensive measurement approaches for PHR remains relatively sparse.

Building on existing studies [31–33], this study develops evaluation index systems for both EEQ and PHR, guided by official documents such as the “Evaluation Target System for Ecological Civilization Construction” released by the National Development and Reform Commission, and the “National Basic Public Health Service Standards (Third Edition)” issued by the National Health and Family Planning Commission. The EEQ assessment framework comprises six core indicators, encompassing dimensions such as environmental pressure and ecological status. Meanwhile, the PHR evaluation framework incorporates eight key metrics, including the availability of healthcare professionals and the provision of medical infrastructure. Detailed descriptions of these index systems are presented in Table 1.

Table 1: Evaluation index system for EEQ and PHR.

2.2. Data sources

The selection and classification of specific cities within the urban agglomerations of the YRCB in this study are based on the “Outline for Ecological Protection and High-Quality Development in the YRBC” issued by the Chinese government, as well as supplementary provincial-level planning documents of the YRBC. Additionally, relevant research was referenced [34]. A total of 66 prefecture-level cities were selected for analysis; however, 11 cities (or autonomous prefectures) were excluded due to severe data unavailability. These cities include Haidong, Yushu, Guoluo, Huangnan, Hainan, Haixi, and Haibei in Qinghai Province; Aba in Sichuan Province; Gannan and Linxia in Gansu Province; and Alxa League in Inner Mongolia Autonomous Region. The selected cities were further categorized into upstream, midstream, and downstream subregions. The specific cities and their regional classifications are detailed in Table 2.

Table 2: Classification of cities in the three regions.

The study period selected for this research is from 2011 to 2022, based on data availability. Information pertaining to EEQ indicators was collected from multiple sources, including the China Urban Statistical Yearbook, China Social Statistical Yearbook, and the China Research Data Service Platform (CNRDS). For PHR indicators, data were drawn from the China Health Statistical Yearbook, China Health and Hygiene Statistical Yearbook, China Health and Family Planning Statistical Yearbook, as well as the China Economic and Social Big Data Research Platform. For years and cities where data were missing, manual collection and compilation were conducted using sources such as regional statistical bureaus, government work reports, and other relevant documents. In cases where data gaps remained, interpolation methods were employed to fill them. All data are provided in the Supporting Information (S1 Dataset).

2.3. Research methods

2.3.1. Entropy method.

The entropy method is a widely adopted objective weighting technique, notable for its ability to minimize the impact of subjective judgment in the determination of indicator weights. In this study, it is applied to independently assess the development levels of EEQ and PHR. To construct the model, all indicators were initially standardized to eliminate dimensional inconsistencies across variables. Following this, the weight of each indicator was calculated based on its corresponding information entropy value. The detailed steps are as follows:

(1)Standardization of indicators to remove units and ensure comparability across variables.

Positive indicators:

$$[eqn]$$

Negative indicators:

$$[eqn]$$

(2)Calculation of indicator proportions and information entropy.

Indicator proportion:

$$[eqn]$$

Information entropy:

$$[eqn]$$

(3)Calculation of indicator weights.

$$[eqn]$$

(4)Calculation of the system score.

$$[eqn]$$

In the equations presented above, $[eqn]$ denotes the observed value of indicator $[eqn]$ for city $[eqn]$ in $[eqn]$ year; $[eqn]$ represents the entropy value corresponding to indicator $[eqn]$ ; $[eqn]$ refers to the weight assigned to indicator $[eqn]$ ; and $[eqn]$ are the comprehensive scores of the two systems, respectively.

2.3.2. Modified coupling coordination model.

The CCD model is an analytical method used to assess the level of interaction and coordinated development between two or more systems. It is commonly applied in fields such as environmental science and management science to measure the coordination and mutual influence of interconnected systems or subsystems. However, Jiang et al. (2017) found that the traditional CCD model has value calculation issues, preventing it from achieving optimal reliability and validity [35]. To address this issue, the present study adopts an improved version of the CCD model, as proposed by Xie et al. (2024) [36], in order to assess the interactive dynamics and coordination level between EEQ and PHR systems. The revised model is expressed through the following equations:

$$[eqn]$$ $$[eqn]$$

In this context, $[eqn]$ represent the composite indices for the EEQ and PHR systems, respectively. $[eqn]$ defined as $[eqn]$ , $[eqn]$ . Within the model framework, $[eqn]$ refers to the degree of coupling, which ranges between 0 and 1 and reflects the strength of the connection between the two subsystems. $[eqn]$ , on the other hand, captures the coupling coordination level, also on a scale from 0 to 1, where higher values suggest stronger integration and synergy between EEQ and PHR.

To enhance the interpretability of coupling coordination outcomes, this study builds upon the classification scheme proposed by Hou et al. (2022) [37], dividing the CCD into ten distinct categories, as illustrated in Fig 2.

The classifications of CCD.

2.3.3. Dagum Gini coefficient and its decomposition approach.

The Dagum Gini coefficient decomposition method, proposed by Dagum (1997) [38], effectively considers the distribution of subgroup samples and the overlap between sample data, enabling the decomposition of the sources of disparity. This approach overcomes the limitations of the traditional Gini coefficient and Theil index. In this study, the Dagum Gini coefficient and its decomposition method are used to characterize the regional disparities in the CCD between EEQ and PHR. The related computational expressions are provided as follows:

$$[eqn]$$ $$[eqn]$$ $$[eqn]$$ $$[eqn]$$

In these equations, $[eqn]$ refers to the total number of regional subgroups, while $[eqn]$ and $[eqn]$ denote the number of cities contained within each subgroup. $[eqn]$ and $[eqn]$ denotes the CCD between EEC and PHR, with $[eqn]$ being the average CCD. The Dagum Gini coefficient enables the total regional disparity (G) to be broken down into three components: intra-regional disparities ( $[eqn]$ ), inter-regional disparities ( $[eqn]$ ), and the intensity of transvariation or overlap between regions ( $[eqn]$ ), fulfilling the relationship $[eqn]$ .

2.3.4. Kernel density estimation (KDE).

KDE is a widely used non-parametric technique for estimating the probability density function of a continuous random variable. Unlike parametric methods, KDE does not require any prior assumptions regarding the distribution form of the data, making it highly suitable for applications such as data visualization, outlier detection, and distribution analysis. The general formulation of the kernel density function for a given variable is expressed as:

$$[eqn]$$

In this formula, $[eqn]$ refers to the selected kernel function, $[eqn]$ denotes the observed CCD between EEQ and PHR for each city in the dataset, $[eqn]$ represents the average value, $[eqn]$ is the total number of observations, and $[eqn]$ indicates the bandwidth parameter. This study adopts a Gaussian kernel function, known for its high precision, to estimate the dynamic distribution of the CCD between EEC and PHR across cities. The function expression is as follows:

$$[eqn]$$

2.3.5. Markov chain.

A Markov chain describes a sequence of random variables that transition between discrete states within a defined state space, where the future state depends solely on the current state, exhibiting the memoryless property. In this study, the Markov chain framework is employed to investigate the temporal dynamics of the coupling coordination development between EEQ and PHR in China. Initially, a conventional (non-spatial) Markov chain model is utilized to explore the internal transition patterns of CCD values. The quartile method is applied to classify CCD levels into four categories: low, medium-low, medium-high, and high. These categories serve as the basis for constructing the transition probability matrix. Subsequently, a spatial Markov chain model is introduced to further refine the analysis by accounting for spatial dependencies. This model extends the traditional Markov chain by integrating spatial lag effects through a spatial weight matrix, which captures the influence of neighboring regions via their weighted average CCD values. By incorporating spatial interaction mechanisms, this approach overcomes a key limitation of the traditional Markov chain, which typically neglects spatial interdependence among adjacent areas [39].

2.3.6. Quadratic Assignment Procedure (QAP).

The spatial disparities in CCD can be conceptualized as a form of inter-regional “relational structure”. Since variables representing such relational data are prone to autocorrelation and multicollinearity, conventional statistical approaches may not be appropriate for analyzing them. To address these limitations, this study adopts the QAP, a permutation-based technique, to explore the key determinants of regional CCD differences and uncover the underlying driving mechanisms.

QAP is widely applied in network analysis to assess the relationships between matrix-based data structures [40]. Unlike traditional regression models, it does not assume variable independence, thus effectively mitigating the effects of multicollinearity and spatial autocorrelation. In QAP-based econometric modeling, variables are represented as square matrices, rather than simple vectors. Drawing on the foundational logic and advantages of QAP, several researchers have developed difference matrices to investigate the sources of variation between regions. The QAP methodology comprises both correlation testing and regression analysis [41]. The correlation component measures the strength of association between matrix pairs, while the regression component evaluates how one dependent matrix is influenced by multiple independent matrices.

3. Regional disparities and temporal dynamics of CCD

3.1. Spatial variation in system CCD

3.1.1. Assessment of EEQ and PHR development levels.

To assess the development status of EEQ and PHR across the YRBC cities, this study applies the entropy-based weighting method to data from 55 cities over the period 2011–2022. The evaluation outcomes are illustrated in Fig 3.

The evaluation value of EEQ and PHR.

(1)EEQ development level. During the observation period, EEQ levels significantly improved. Across the entire sample, the overall level increased from 0.3130 in 2011 to 0.4002 in 2022, with an average value of 0.3656 over the 12 years, reflecting an overall growth of 27.83% during the sample period. Regionally, the midstream region experienced the largest increase, with a growth rate of 30.13%, followed by the upstream and downstream regions, which saw increases of 27.95% and 25.53%, respectively. The specific results for the upstream, midstream, and downstream regions are shown in Table 3.(2)PHR development level. The growth level of PHR during the observation period was relatively higher compared to that of EEQ. The overall national level increased from 0.3860 in 2011 to 0.4615 in 2022, with an average value of 0.4097 over the 12 years, reflecting an overall growth of 19.57% during the sample period. This growth rate is 8.26 percentage points lower than the overall growth rate of EEQ. Regionally, the upstream region exhibited the highest overall growth rate at 30.81%, slightly higher than that of the midstream and downstream regions, with the midstream region experiencing the lowest growth rate at 13.87%. The specific levels of PHR in the upstream, midstream, and downstream regions are detailed in Table 3.

Table 3: Level of EEQ and PHR in three subregions.

3.1.2. Results and analysis of system CCD.

Using the modified CCD model, this study evaluated the CCD between EEQ and PHR in 55 cities within the YRBC. Table 4 presents the CCD for the entire YRBC, as well as for the upstream, midstream, and downstream regions, from 2011 to 2022.

Table 4: Results of the CCD.

Across the entire sample, the CCD increased from “Near Coordination” in 2011 to “Primary Coordination” in 2022. By 2022, both the upstream and downstream regions had reached the “Primary Coordination” level, while the midstream region remained at the “Near Coordination” level. However, when considering the average levels throughout the observation period, the overall sample and each region maintained an average level of “Near Coordination”.

3.2. Spatial differences in CCD and decomposition of their sources

Table 5 presents the Dagum Gini coefficient and its decomposition results for the CCD. In terms of overall evolutionary patterns, although the CCD between EEQ and PHR in the YRBC exhibits an unbalanced state, the overall disparity shows a decreasing trend. The Gini coefficient declined significantly from 0.0617 in 2011 to 0.0536 in 2022. Regarding the sources of disparity and their contributions, intra-regional differences remained relatively stable throughout the observation period, initially decreasing and then slightly increasing from 2011 to 2022, with minimal overall variation. The contribution rate of inter-regional differences exhibited an irregular fluctuation trend, peaking at 25% in 2012 and reaching a low of 9.74% in 2020. The trans-variation density primarily explains the phenomenon of overlapping between regions. From 2011 to 2022, the trans-variation density remained relatively stable, with slight fluctuations in some years. Across all years, the contribution rate of trans-variation density consistently exceeded that of intra-regional and inter-regional differences, accounting for more than 45% of the total.

Table 5: Decomposition of regional differences in CCD.

3.3. Dynamic evolution of the distribution of CCD

Due to the heterogeneous resource endowments and varying stages of development among the upstream, midstream, and downstream regions—as well as across individual cities—the overall CCD trends observed at the YRBC-wide level may differ from those within each subregion. To capture these spatial and temporal variations in the CCD between EEQ and PHR, this study applies KDE as introduced by Parzen (1962) [42]. The analysis focuses on three key aspects of distributional change: central tendency, shape, and dispersion. KDE is conducted both for the full sample and separately for the three major regional groups. The results of this dynamic distributional analysis are visualized in Fig 4.

KDE estimation.

Fig 4(a) presents the temporal dynamics of the CCD distribution between EEQ and PHR across the full sample. Three key features are evident in the distributional evolution: First, with respect to the distribution’s central tendency, the peak of the density curve shifts progressively to the right, indicating a steady enhancement in the level of coordination between EEQ and PHR from 2011 to 2022. Second, in terms of curve shape, the density function becomes increasingly peaked and narrow over time, reflecting a marked decrease in the absolute disparities in CCD values. Third, regarding dispersion, the right-tail of the distribution becomes less prominent, suggesting a convergence in CCD levels among cities and a narrowing gap in coordination outcomes.

Fig 4(b), 4(c), and 4(d) illustrate the regional evolution patterns of CCD in the upstream, midstream, and downstream parts of the YRBC, respectively. From a distributional position perspective, all three subregions exhibit a rightward movement of the density curves, consistent with the overall trend, though the extent of shift varies among regions, signaling general progress in CCD. In the upstream region, the shape of the kernel density curve remains relatively stable, with minor fluctuations in peak height and width. In contrast, the midstream region experiences a decline in peak sharpness accompanied by a noticeable narrowing of the curve between 2016 and 2019. However, after 2018, the distribution widens significantly, reflecting increased variability. The downstream region displays a distributional trajectory largely aligned with that of the midstream area, showing similar transitions in shape and spread.

3.4. Markov chain analysis of CCD

3.4.1. Traditional Markov Transition Probability Matrix (MTPM).

To further investigate the internal dynamics and spatial transition characteristics of the CCD between EEQ and PHR, this study introduces the MTPM for analysis. First, the traditional Markov chain is utilized to examine the intrinsic trend characteristics of the CCD between the two systems. The quartile method is employed to categorize the CCD results of each city into four distinct levels: low, medium-low, medium-high, and high. The resulting transition probability matrix is presented in Table 6. An important observation is that the diagonal elements of the matrix are consistently higher than the off-diagonal entries. Specifically, the probabilities of maintaining the same CCD category over a one-year period are 75.93% for low, 73.68% for medium-low, 80.82% for medium-high, and 93.79% for high levels. These figures demonstrate a high level of temporal persistence in CCD status, indicating a tendency toward stability or path dependency—commonly referred to as “club convergence.” Moreover, the transition probabilities for both the low and high CCD groups are slightly higher than those for the two intermediate categories. This pattern implies the existence of a mild “Matthew Effect,” wherein regions already exhibiting either strong or weak coordination are more likely to maintain their relative positions over time, thereby reinforcing divergence at the extremes.

Table 6: MTPM for CCD (2011–2022).

3.4.2. Spatial Markov chain analysis.

To explore the spatial relationship between the CCD of EEQ and PHR among cities, this study calculated the global Moran’s I index for the CCD from 2011 to 2022 using Stata 17.0, based on the previously computed CCD data. The results are presented in Table 7.

Table 7: Global Moran’s I Index for CCD.

To further investigate the spatial interdependence in the CCD between EEQ and PHR across cities, this study employs the global Moran’s I index for spatial autocorrelation analysis. Based on previously calculated CCD values, the index was computed for the period 2011–2022 using Stata 17.0. The corresponding results are summarized in Table 7.

The findings reveal a statistically significant and positive Moran’s I across most years during the study period, implying that a city’s CCD is spatially correlated with those of its neighboring cities. This indicates the presence of spatial agglomeration patterns, where cities with similar levels of coordination tend to cluster together—manifested as high-high and low-low groupings. The Moran’s I values consistently fall within the range of 0.07 to 0.28, reflecting a moderately stable positive spatial correlation throughout the observed years.

The results indicate the necessity of incorporating spatial factors and establishing a spatial Markov transition probability matrix. The specific results are presented in Table 8. First, the four transition probability matrices under different types of spatial lags are not identical. This suggests that differences in the CCD of neighboring cities affect the probability of a city’s CCD shifting differently. Second, the diagonal elements in the transition probability matrices for all types of spatial lags are greater than the off-diagonal elements. This indicates that, under the influence of spatial spillover effects, the CCD between different levels remains relatively stable, demonstrating a “club convergence” phenomenon. Additionally, there are non-zero elements on both sides of the diagonal, implying that while there is a possibility for upward shifts in the CCD towards an ideal state, there is also a certain risk of downward shifts. Moreover, transitions tend to occur only between adjacent levels, making it difficult to achieve transitions across multiple levels. Furthermore, the impact of different lag types on the same level varies. For example, under the medium-low lag type, the probability of a low-level transitioning to a medium-low level is 18.64%, which is significantly higher than the transition probability under the low-level lag type. Finally, the impact of the same lag type differs across various levels. Under the medium-high lag condition, the probabilities of achieving an upward transition by one level for low, medium-low, and medium-high levels are 31.25%, 24.53%, and 8.06%, respectively, showing a decreasing trend. This suggests that the transition probability is influenced not only by the type of lag but also by the initial level of the CCD.

Table 8: Spatial Markov transition probability matrix for CCD.

4. QAP analysis

4.1. QAP correlation analysis

To examine the relationship between regional disparities in the CCD and a set of explanatory factors, this study employed the QAP correlation analysis. Utilizing the Ucinet software, a total of 2,000 random permutations were conducted to ensure the robustness of the statistical inference. The analysis treats the CCD difference matrix as the dependent variable, with several structural and socioeconomic variables as independent matrices. The resulting correlation outcomes are detailed in Table 9.

Table 9: QAP correlation analysis results.

The findings indicate that the regional differences in CCD are positively correlated with regional differences in per capita GDP (Var1), urbanization rate (Var2), industrial structure advancement (Var3), the number of software technology professionals (Var4), and market openness (Var5). Except for industrial structure advancement, the correlation coefficients for the other variables are significant at the 1% or 10% level. This suggests that disparities in the development of these factors can lead to changes in regional differences in CCD. The correlation coefficients between regional differences in per capita GDP, urbanization rate, the number of software technology professionals, and market openness with regional differences in the CCD are 0.695, 0.864, 0.232, and 0.406, respectively.

4.2. QAP regression analysis

To further investigate the magnitude and direction of influence exerted by the five explanatory variables, this study conducted a QAP regression based on 2,000 randomized permutations. The dependent variable in the analysis is the matrix representing regional disparities in the CCD, while the independent variables consist of five corresponding difference matrices reflecting key influencing factors. The regression results are reported in Table 10.

Table 10: QAP regression results on factors influencing regional differences in CCD.

The model yields an adjusted R-squared value of 0.802, suggesting that the selected variables collectively account for 80.2% of the observed variation in regional CCD differences. As presented in the table, standardized regression coefficients—used to remove unit-based discrepancies among variables—enable a direct comparison of each factor’s relative importance. Accordingly, these standardized coefficients are employed in this study to assess and interpret the specific contribution of each explanatory variable to regional CCD variation.

The results reveal the following:(1) The coefficient for regional differences in the number of software technology professionals is −0.004, but it is not statistically significant, suggesting that regional differences in the number of software technology professionals are not a primary factor contributing to regional differences in CCD. (2) The standardized regression coefficients for regional differences in per capita GDP, urbanization rate, industrial structure advancement, and market openness are all significantly positive, indicating that these are the main factors contributing to regional differences in CCD. Among them, regional differences in urbanization rate have the most significant impact, with a regression coefficient of 0.679, implying that widening regional disparities in urbanization levels will lead to greater regional differences in CCD. The impact of other variables follows in the order of regional differences in per capita GDP, industrial structure advancement, and market openness, with standardized regression coefficients of 0.235, 0.149, and 0.124, respectively.

5. Discussion

This research evaluated the development trajectories of EEQ and PHR across 55 cities in the YRBC over the period from 2011 to 2022, and subsequently measured the CCD between these two systems. In the context of increasing global emphasis on green transformation, sustainable development, and public health improvement, the outcomes of this study offer valuable insights into the coordination patterns and temporal evolution of EEQ and PHR. These findings contribute to building a theoretical framework and offer practical guidance for fostering integrated progress in ecological and health systems, both in current policy agendas and future planning. As such, the implications derived from the analysis merit further comprehensive discussion.

First, during the observation period, the CCD between EEQ and PHR has shown a stable upward trend. By 2022, the CCD level across the YRBC, as well as in the upstream and downstream regions, had reached the “primary coordination” level. The stable increase in CCD over time can likely be attributed to several factors. First, the continuous improvement of environmental policies and the implementation of stricter regulations may have played a key role in enhancing ecological conditions, which, in turn, positively impacts public health. As the government focuses on reducing pollution and promoting sustainable practices, environmental quality gradually improves, leading to better health outcomes for residents. Additionally, the development of healthcare infrastructure, particularly in areas with better environmental conditions, contributes to this positive trend.

When comparing the three major regions, the upstream area exhibits a higher CCD level than the midstream and downstream areas. This disparity can be attributed to several factors, including differences in economic development, environmental management strategies, and resource allocation. The upstream regions, typically less industrialized, may benefit from better environmental conditions, resulting in a stronger link between EEQ and PHR. In contrast, the more industrialized and urbanized midstream and downstream regions may face greater environmental challenges, which could hinder the improvement of their CCD.

Second, the distribution of the CCD is non-random and exhibits spatial autocorrelation. The spatial autocorrelation test results indicate that the autocorrelation coefficients for the CCD between EEQ and PHR consistently ranged between 0.07 and 0.28 from 2011 to 2022, demonstrating a certain degree of stability and showing a clear high-high and low-low clustering phenomenon. Possible reasons for this spatial clustering include the following: First, geographically proximate regions often share similar environmental and socio-economic conditions. For example, neighboring cities might benefit from similar natural resources, infrastructure, and economic activities, leading to comparable levels of EEQ and PHR. Additionally, policies and initiatives implemented at the regional level—such as those aimed at pollution control, healthcare improvements, or sustainable development—often have spillover effects on adjacent areas, further reinforcing the spatial clustering of the CCD.

Third, this study explored the factors influencing the CCD between EEQ and PHR. The QAP regression analysis revealed that the primary factors contributing to regional differences in CCD are disparities in per capita GDP, urbanization rate, industrial structure advancement, and market openness, while regional differences in the number of software technology professionals do not have a significant impact on these disparities.

Despite these findings and contributions, this study has certain limitations. First, although the study uses the most recent available data, the data for 2023 were not accessible, preventing the analysis and discussion of the latest conclusions for that year. Second, the study focuses on 55 cities within the YRBC. From a more micro-level perspective, conducting coupling coordination research on EEQ and PHR at the county level or in other regions would have significant practical relevance and application value. However, this remains an unexplored area and represents one of the limitations of this study. Naturally, it also points to an important direction for future research expansion.

6. Conclusions and policy recommendations

This study investigates the coupling coordination between EEQ and PHR across 55 cities in the YRBC over the period 2011–2022. Initially, comprehensive evaluation index systems were developed for both EEQ and PHR, with their respective levels assessed using the entropy weighting method. A modified CCD model was subsequently applied to quantify the coordination status between the two systems. The main conclusions drawn from the analysis are summarized as follows:

First, in terms of overall development levels, both EEQ and PHR in the YRBC showed a stable upward trend. However, the overall development level remains relatively low, indicating significant room for improvement. Over the 12-year period from 2011 to 2022, EEQ increased by 27.83%, outpacing the 19.57% growth rate of PHR. Second, the CCD calculated using the modified CCD model exhibited a steady upward trend across the YRBC as a whole and within the three major regions. Third, regarding the dynamic evolution trends, KDE reveals that the density curve of the CCD between EEQ and PHR in the YRBC has shifted to the right. Fourth, the QAP regression analysis shows that the main factors contributing to regional differences in the CCD are disparities in per capita GDP, urbanization rate, industrial structure advancement, and market openness, while regional differences in the number of software technology professionals do not significantly impact these regional disparities.

Based on the findings of this study, and drawing insights from international best practices, several policy recommendations are proposed to enhance the coupling coordination between EEQ and PHR in the YRBC:

First, strengthen environmental regulation and enforcement with smart monitoring. To improve EEQ, it is essential to reinforce environmental regulations. International experience, such as the European Union’s Industrial Emissions Directive (IED), demonstrates that integrated permitting and real-time emission monitoring can significantly reduce pollution. The YRBC can adopt and localize such smart monitoring systems, utilizing IoT sensors and satellite data to track industrial emissions in real-time, especially in heavily industrialized midstream and downstream regions. Such measures will help control pollution levels and contribute to a healthier environment, which in turn supports better public health outcomes.

Second, promote clean and renewable energy. The transition from traditional energy sources to clean and renewable energy is critical for reducing environmental degradation. Governments should provide incentives, such as subsidies and tax breaks, to encourage the development and adoption of renewable energy projects like solar, wind, and hydropower. Public-private partnerships should be fostered to accelerate technological innovation in the energy sector, further promoting the use of sustainable energy solutions that minimize environmental impact.

Third, increase investment in healthcare infrastructure. Given the large population base in China, particularly in the YRBC, it is crucial to expand healthcare infrastructure to ensure adequate access to medical services. Investments should be directed towards building new healthcare facilities, upgrading existing ones, and increasing the availability of medical personnel and resources. This will help to improve the PHR by ensuring that healthcare services are more accessible and of higher quality across the region.

Fourth, promote healthy lifestyles through evidence-informed public health initiatives and behavioral policy tools. Internationally, Finland’s North Karelia Project is a well-known example demonstrating that sustained, community-wide interventions can reduce major cardiovascular risk [43]. In the YRBC, governments should launch targeted public health campaigns that emphasize the benefits of physical activity, balanced nutrition, and preventive healthcare. Community-based programs can also be developed to encourage residents to engage in regular exercise and adopt healthier dietary practices, thereby improving the overall PHR.

Fifth, address regional disparities in development. The study highlights regional differences in the CCD, with upstream regions generally performing better than midstream and downstream areas. To address these disparities, policies should be tailored to the specific needs of each region. For instance, industrialized regions may require more stringent environmental controls and focused efforts to upgrade healthcare services, while less developed areas might benefit from targeted investments in infrastructure and capacity-building initiatives.

Supporting information

S1 DatasetCity-level panel data for 55 cities in the YRBC (2011–2022).(XLSX)