Data

In this paper, we use three evaluation indices—the ecological environment, economic environment and social environment—as the explanatory variables of the model. The data are obtained from the CSMAR and EPS databases, and some missing data are filled in according to the data from the statistical bureaus of each province in China. According to the availability of data, comprehensive evaluation indices of the environmental quality of 30 provinces excluding Tibet in 2014, 2016, 2018 and 2020 were selected as the research objects.

The explanatory variables used medical expenditures from the CFPS data, and the covariates included 24 variables in three categories: living habits, household situation, and individual characteristics. The data come from the China Family Panel Studies (CFPS), which is a tracking survey of representative samples of villages, households, and family members across the country to reflect China’s economic development and social changes. The microdata are selected from the 2014, 2016, 2018, and 2020 survey data of people older than 60 years. For handling partially missing data and outliers, we employed methods such as moving averages, ARIMA forecasting, and interpolation. These methods ensured the reliability and accuracy of our research results. Consequently, we confirmed a usable sample size of 18,862. Table 1 shows the descriptions of the microvariables and descriptive statistics.

Table 1 Description of microvariables and descriptive statistics (years: 2014, 2016, 2018, 2020; place: 30 provinces in China)VariableDependent variable

Medical Expenditure: The primary dependent variable of the study, the medical expenditure of the elderly, is measured through patients’ medical insurance records and direct medical cost records, including the total of outpatient and hospitalization expenses.

Independent variable

Environmental Quality Index: The main explanatory variable in this study is the composite environmental quality index, assessed using various environmental monitoring parameters such as PM2.5, water quality levels, and noise levels. This index is developed through a comprehensive evaluation system of environmental quality, utilizing the entropy weight-TOPSIS method to calculate a composite index for each province as a proxy indicator.

Control variables

To ensure the accuracy of our study results, we have included the following control variables, covering 24 variables listed in Table 1 aside from medical expenses and environmental factors:

Individual Characteristics: This includes demographic factors such as age, gender, education level, and residential status, as well as health-related variables like chronic disease status, past hospitalizations, and overall health condition. Psychological factors such as satisfaction with life, medical services, and trust in doctors are also included, as these have been shown to significantly correlate with health status and medical needs.

Family Circumstances: Economic status indicators such as per capita family income and pension receipt, along with family dynamics such as the number of family members and frequency of meals shared with family, affect an individual’s access to and affordability of medical services.

Lifestyle Habits: Behaviors including smoking, drinking, frequency of physical activities, and sleep duration, which can impact an individual’s health status.

Measurement of the comprehensive environmental quality assessment indexComprehensive evaluation index system for environmental quality

Environmental problems urgently need to be solved. Domestic and foreign research organizations mostly use index evaluation methods to measure environmental quality, and the determination of an evaluation index is the basis of this method. In the selection of an evaluation index, it is necessary to consider whether the index can include all the core contents of environmental quality. Therefore, this paper reviews relevant literature and documents such as the National Environmental Protection Standards of the People’s Republic of China and refers to the studies of Bernardo et al. [25], Mourhir et al. [26] and Li and Wang [27] and combines the constraints of benefit analysis, scientificity, practicability, and data availability to construct a comprehensive evaluation index system of environmental quality, which contains three criteria layers for ecological, economic, and social environments and 10 indicator layers, as shown in Fig. 2.

Fig. 2

Comprehensive environmental quality evaluation system

Next, the indicators are further explained. For the ecological environment, forest coverage is an important indicator of ecological balance and a reflection of abundant resources, representing the condition of the original ecological environment, while an air quality of Grade 2 or above indicates a low pollution index, with less PM2.5, PM10, etc., and the environmental noise in key urban areas reflects the liveability of the area; for the economic environment, the green space per capita reflects the green space in a certain period of time; for the economic environment, the green space per capita reflects the green space in a certain period of time. In terms of the economic environment, the green space per capita reflects the development of the urban economy in a certain period of time, and the greening difference between regions is gradually decreasing, while the growth rate of energy consumption per unit of GDP and the investment in environmental infrastructure reflect the level of energy consumption and the economic cost invested in environmental pollution. In terms of the social environment, the amount of water resources per capita, the amount of nitrogen oxides per unit of nitrogen oxides, the amount of domestic oxygen demand and the amount of sulfur dioxide emissions are chosen to measure the social governance capacity.

Entropy weight-TOPSIS method

Entropy weight method is an interpretation based on the basic principles of information theory, and it is also an objective assignment method to determine the weights according to the degree of change of each indicator, which fully demonstrates the information value of the data itself and avoids the possibility of human judgment and manipulation. The basic idea of entropy weight method is that the smaller the entropy of the evaluation index, the greater the degree of its discrete, the more information it contains, the greater the role in the comprehensive judgment system, so the corresponding weight is greater, and the TOPSIS method is a sorting method close to the ideal solution to evaluate the advantages and disadvantages of each index according to the size of the order. The entropy weight-TOPSIS method combines the two, weighting by entropy weighting method and sorting by TOPSIS method, so that the evaluation results have stronger objectivity.

Compared to other environmental quality assessment methods such as Principal Component Analysis (PCA), Analytic Hierarchy Process (AHP), and Data Envelopment Analysis (DEA), the entropy weight-TOPSIS method determines weights based on the amount of information inherent in the data itself, thus minimizing the influence of subjective factors. It comprehensively considers the impact of various indicators to produce relatively reasonable comprehensive evaluation results. Additionally, its calculation process is relatively simple, making it suitable for processing and analyzing large-scale data. Zou et al. [28] utilized the entropy weight-TOPSIS method to assess the progress of China’s green energy consumption revolution. The study indicates that the entropy method can objectively determine the weights of various indicators, while the TOPSIS method is used to comprehensively evaluate the progress [28]. The specific steps are as follows:

Step 1: Construct the original decision matrix. Assuming that there are \(m\) provinces, and each province has \(n\) evaluation indicators, the value of evaluation indicator \(j\) for province \(i\) is \(}\).

$$A})_},(i = 1,2,...,m,j = 1,2,...,n)$$

(1)

The positive indicator is calculated by the formula:

$$ = \frac - }}}} - }}}$$

(2)

The negative indicator is calculated as:

$$ = \frac} - }}} - }}}$$

(3)

Then through the normalization process, all the indicators are converted to dimensionless, and the standardized decision matrix is obtained as \(B(})\). In the formula:

$$} = }/\sum\limits_^m ^n }} }$$

(4)

\(}\) denotes the quantitative value of evaluation indicator \(j\) for province \(i\).

Step 3: Calculate the entropy value of evaluation indicators. \(m\) is the number of provinces in China, in this paper, \(m\) = 30. \(n\) is the number of evaluation indicators, in this paper, \(n\) = 10.

$$ = - \frac\sum\limits_^m ^n }} } \ln }$$

(5)

When \(} = 0\), \(\ln }\) is meaningless, then \(}\ln } = 0\) can be considered.

The distance from the infuence indicator of each province to the ideal solution \(\) is:

$$S_^+ = \sqrt ^m } - x_i^+ )}^2}} }$$

(12)

The distance from the infuence indicator of each province to the virtual worst solution \(\) is:

$$S_^* = \sqrt ^m } - x_i^*)}^2}} }$$

(13)

$$} = S_^*/(S_^+ + S_^*)$$

(14)

where \(} \in [0,1]\), denotes the comprehensive evaluation index of environmental quality. According to the calculated value of \(}\), the evaluation objects are ranked in order of superiority or inferiority from the largest to the smallest.

Weighting analysis

After completing the construction of the comprehensive environmental quality evaluation index system, the entropy weight-TOPSIS method was used to calculate the weights of each index, as shown in Table 2.

Table 2 Weights of the evaluation indicators (years: 2014, 2016, 2018, 2020; place: 30 provinces in China)

As shown in Table 2, at the indicator level, there are five positive and five negative indicators. Among the positive indicators, water resources per capita have the greatest weight, followed by investment in environmental infrastructure, forest coverage, green space per capita in parks, and the number of days when air quality reaches level 2 or higher. This indicates that when more water is available per capita, there is less pollution and better ecological quality. The weighting of environmental noise in key urban areas in the negative indicators is too large compared to that in the other indicators, and the lower the environmental noise is, the better the sense of residents’ living experience.

For the criterion layer, the weights of the ecological environment, economic environment and social environment are 26.84%, 30.55% and 42.61%, respectively; i.e., the entropy weight of the social environment is the highest, and that of the ecological environment is the lowest. According to the principle of the entropy weight method combined with the distribution results, the social environment data contain the least information entropy, and the degree of dispersion and the difference between provinces are the greatest; therefore, environmental quality problems are difficult to solve by focusing on the social environment. The social environment is a large group, from the state down to enterprises and individuals. The rapid development of industry is accompanied by extremely serious pollution. In 2021, the National Development and Reform Commission issued a policy system to improve the dual-control policy of energy consumption, and some regions have begun to limit electricity to reduce the emissions of nitrogen oxides (NOx), sulfur dioxide (SO2), and other pollutants from enterprises and factories and to achieve the target tasks of carbon peaking and carbon neutrality. For the ecological environment, which carries the least weight, there is less variation among provinces, and combined with the influence of some common factors such as laws, policies and culture, people have a stronger sense of protecting the ecological environment.

Environmental quality analysis

The data of 30 provinces in 2014, 2016, 2018 and 2020 are then put into the entropy weight-TOPSIS model to obtain the comprehensive evaluation index of environmental quality, which is shown in Table 3. The distribution map of environmental quality is drawn according to the relevant data, as shown in Fig. 3.

Table 3 Measurement results of the comprehensive evaluation indices of the environment quality in each province (years: 2014, 2016, 2018, 2020; place: 30 provinces in China)Fig. 3

Distribution of environmental quality in each province (years: 2014, 2016, 2018, 2020; place: 30 provinces in China)

As shown in Table 3, the top three provinces are Jiangxi, Guangxi and Sichuan. With the continuous increase in energy savings and emission reduction, environmental quality improved annually in 2018 compared with 2014, which is attributed to the cost invested in China’s current environmental governance. Among them, Shanghai has the lowest overall environmental quality evaluation index compared with the other provinces. As an economically developed region with a population of nearly 30 million people, Shanghai has a large amount of automobile exhaust emissions that cause serious air pollution, and at the same time, the surrounding areas are densely populated with industries, which pollutes industries and results in poor environmental quality. Figure 3 shows that the environmental quality in China is generally greater in the western region than in the eastern region and greater in the southern region than in the northern region. Due to the low population density in the western region, the environmental pressure is low, and the emission of pollutants is low. The northern region is colder than the southern region, uses more coal, and has a dry climate with low vegetation cover and low precipitation, leading to an increase in airborne pollutants and unfavourable precipitation.

Model construction

Traditional linear regressions are first created to analyse the effects of environmental quality and medical expenditures, including least squares estimation and fixed effects models. Compared with least squares regression, fixed effects are able to control for the effects of unobservable factors caused by individuals and time. The fixed effects model was developed as follows:

$$} = + } + } + + + }$$

(15)

where \(}\) denotes the medical expenditure of individual i in year t, \(W\) denotes the composite air quality index of the time and area where individual i is located, \(}\) denotes the control variables, \(\) denotes the fixed effect of area i, \(\) is the fixed effect of year t, and \(}\) is the random disturbance term.