key: cord-0723130-5fnk6fgy
authors: Peng, Wenjia; Dong, Yilin; Tian, Meihui; Yuan, Jiacan; Kan, Haidong; Jia, Xianjie; Wang, Weibing
title: City-level greenness exposure is associated with COVID-19 incidence in China
date: 2022-02-03
journal: Environ Res
DOI: 10.1016/j.envres.2022.112871
sha: 8a45ee22a9696ccb525fade90c181b7f2b469d9d
doc_id: 723130
cord_uid: 5fnk6fgy

Accumulating studies have suggested an important role of environmental factors (e.g. air pollutants) on the occurrence and development of coronavirus disease 2019 (COVID-19). Evidence concerning the relationship of greenness on COVID-19 is still limited. This study aimed to assess the association between greenness and COVID-19 incidence in 266 Chinese cities. A total of 12,377 confirmed COVID-19 cases were identified through February 29th, 2020. We used the average normalized difference vegetation index (NDVI) during January and February 2020 from MOD13A2 product, to represent the city-level greenness exposure. A generalized linear mixed-effects model was used to estimate the association between NDVI exposure and COVID-19 incidence using COVID-19 cases as the outcome. We evaluated whether the association was modified by population density, GDP per capita, and urbanization rate, and was mediated by air pollutants. We also performed a series of sensitivity analyses to discuss the robustness of our results. Per 0.1 unit increment in NDVI was negatively associated with COVID-19 incidence (IRR: 0.921, 95% CI: 0.898, 0.944) after adjustment for confounders. Associations with COVID-19 incidence were stronger in cities with lower population density, lower GDP per capita, and lower urbanization rate. We failed to detect any mediation effect of air pollutants on the association between NDVI and COVID-19 incidence. Sensitivity analyses also indicated consistent estimates. In conclusion, our study suggested a beneficial association between city-level greenness and COVID-19 incidence. We could not establish which mechanisms may explain this relationship.

Environmental factors play a large role in the occurrence and development of a 1 variety of health outcomes. Built environment refers to the artificially constructed 2 structures and infrastructure to provide for human activities, including land use, 3 transport network, and green space. Green space is defined as open pieces of land 4 which are covered by trees, grasslands, or other vegetation (de Keijzer et al., 2019b) . 5 In recent years, an accumulating body of evidence suggested that the presence of 6 green space is beneficial for a broad range of health outcomes including diabetes 7 (Yang et al., 2019b) , metabolic syndrome (de Keijzer et al., 2019a) , hypertension 8 (Yang et al., 2019a) , obesity Sarkar, 2017) , cardiovascular disease 9 (Jia et al., 2018) , depression (Gascon et al., 2018; Pun et al., 2018) and mortality 10 (James et al., 2016; Kasdagli et al., 2021) . However, previous studies mainly focused 11 on chronic non-infectious diseases and mental disorders. Evidence concerning the 12 relationship of greenness with infectious disease is still limited. Although, the 13 underlying mechanisms by which greenness benefits health are not fully elucidated, 14 several pathways have been investigated. Firstly, access to more green could reduce 15 certain harms, such as exposure to air pollution and heat (Gunawardena et al., 2017;  16 Hankey and Marshall, 2017) . Secondly, living in greener areas might promote 17 individuals to participate in physical activity (James et al., 2017; Sadeh et al., 2019) , 18 which is critical to help the immune system against virus (Shahrbanian et al., 2020) . 19 Thirdly, exposure to more greenness is beneficial for relieving mental and physiologic 20 stress (Markevych et al., 2017) .

To the best of our current knowledge, there has been no study to examine the 1 association between greenness exposure and COVID-19 incidence in China.

2 Therefore, we utilized an ecological study based on 266 Chinese cities to explore: 1) 3 whether greenness exposure was negatively associated with COVID-19 incidence; 2) 4 whether the association was modified by potential covariates; 3) whether the 5 association was mediated by air pollutants. 

All data in the study were obtained from publicly available sources. Table S1 8 presented the descriptive and source of the outcome, exposure, potential covariates, 9 air pollutants, and meteorological factors. 

As of February 29 th , 2020, the total number of COVID-19 confirmed cases in 12 China was 79824, with 66907 cases in Hubei province, accounting for more than 80% 13 of total cases. Therefore, cities in Hubei province were not included to avoid bias. 14 City-level COVID-19 confirmed cases were collected from the National Health 

We quantitatively assessed the city-level greenness using normalized difference vegetation index (NDVI) from vegetation product of Moderate-resolution imaging spectroradiometer (MODIS) sensor aboard the National Aeronautics and Space Administration (NASA) Terra satellite, namely MOD13A2, available at https://ladsweb.modaps.eosdis.nasa.gov/search/. MOD13A2 product provides vegetation index at one kilometer (km) spatial resolution. The algorithm for this product was based on the best available pixel value from all the acquisitions. NDVI values ranged from -0.2 to 1, with higher values indicating more greenness. We downloaded NDVI data from January 1 st to February 29 th ,2020. The mean value of NDVI was assigned to each city, representing the greenness exposure. The extraction of vegetation index was done using R statistical software, Version 4.0.3 (University of Auckland, New Zealand). The negative value was set to NA (not available) which represented water features.

According to prior knowledge, we collected city-level variables from numerous public sources to adjust for confounding bias. From the 7 th national population census, we obtained the following city-level covariates: total population, the proportion of older adults aged over 65 years old (older people), male to female ratio (gender ratio), average education years of the population aged over 15 years old (education years), urbanization rate. Population density (person/sq. km.) was calculated using total population divided by land area. From the China city statistical yearbook of 2019, we J o u r n a l P r e -p r o o f collected GDP per capita, hospital beds (hospital beds), and doctors per 1,000 people (doctors).

A government response index (GRI) indicator from publicly-accessible data was collected by the Coronavirus Government Response Tracker at https://www.bsg.ox.ac.uk/research/research-projects/covid-19-government-response-tr acker . The GRI was calculated using a series of indicators about governments' policy response to the COVID-19 pandemic, mainly including containment and closure, economic response, and health systems from 1 January 2020. The GRI ranges from 0 to 100, with higher values indicating a higher overall response level. We collected the average provincial level GRI within the time span of 60 days from January 1 st to February 29 th , 2020.

We also considered the influence of population movement on COVID-19

incidence. An intra-city movement intensity, defined as the proportion of people traveling within cities, was derived from Baidu map migration big data (http://qianxi.baidu.com/) from January 1 st to February 29 th , 2020. Population movements are anonymously collected at the city level based on individual geographic location.

We collected air pollutants from the Bureau of Ecological Environment of each city, including fine particulate matter (PM2.5), nitrogen dioxide (NO2), and carbon monoxide (CO). The daily concentrations of PM2.5, NO2, and CO were available from January 1 st to February 29 th , 2020. We averaged daily concentrations of these J o u r n a l P r e -p r o o f 8 pollutants and then aggregated them into each city. Meteorological data on daily mean temperatures (°C) and daily mean relative humidity (%) during the same period were also collected from the China Meteorological Data Sharing Service System

A generalized linear mixed-effects model was used to evaluate the association of NDVI and COVID-19 incidence with a random intercept for provinces to account for a potential correlation in cities within the same province. We included a log population offset in the model. Hence, the model could be expressed as follows:

where E (•) is an expected value of incident cases, β 0 is the intercept, is the regression coefficient of explanatory variables .

The incidence rate ratio (IRR) with its 95% confidence interval (CI) for per 0.1 unit increment in greenness was reported. All the covariates were scaled. GDP per capita and population density were also log-transformed due to severe skewness. The

Spearman correlation coefficient (rs) and variance inflation factor (VIF) were calculated to diagnose the potential multicollinearity among multiple variables (Table   S2 and S3). The main model was performed by removing highly correlated variables (rs > 0.6). We used a threshold value of 5 for VIF to identify the presence of multicollinearity.

Stratified analyses by population density, GDP per capita, and urbanization rate were performed to discuss the effect modification. We categorized these covariates into two groups based on the median. The effect modification was tested by adding J o u r n a l P r e -p r o o f interaction as a product term into the model, which was adjusted for all the covariates, except for the interaction covariates. Mediation analysis was conducted to investigate the potential mediators (air pollutants) on the role of NDVI and COVID-19 incidence.

The significance of the mediation effect was tested through 500 bootstraps resamples of the estimated indirect effect. We reported the average direct effect (ADE) and

average causal mediation effect (ACME).

To test the robustness of our results, we performed a series of sensitivity analyses.

First, to assess the sensitivity of the main model to confounders, we removed the confounders in turn. Second, we excluded several cities with less than 10 and 5 confirmed COVID-19 cases. Third, NDVI in the winter season (January and February) might not completely represent the level of greenness in area with snow cover. We also obtained the NDVI data from July to September 2019, the greenest months in one

year. The association of NDVI in this period with COVID-19 incidence was also assessed.

All the statistical analyses were conducted in R statistical software, Version 4.0.3 (University of Auckland, New Zealand) using lme4 package (for main model analysis)

or mediation package (for mediation analysis). Research ethics board approval was not required because all data were publicly available and aggregated at the city level.

A total of 266 cities were enrolled in our study. 12377 confirmed COVID-19 cases were identified through February 29 th , 2020. The COVID-19 incidence per J o u r n a l P r e -p r o o f 100,000 individuals was 0.98. Table 1 presented the descriptive statistics of the characteristics of 266 cities. 70 cities had 10 or fewer and 28 had 5 or fewer confirmed cases. The median value of NDVI during January and February 2020 was 0.387. Figure 1 displayed the spatial distribution of COVID-19 confirmed cases in China. Except for Hubei province as the center of the outbreak, COVID-19 incident cases were more prevalent in municipalities (e.g. Chongqing, Beijing, and Shanghai) and the capital cities of provinces. Figure 2 visualized the spatial variation of NDVI in

China. Generally, greenness density is more prominent in the eastern and southern cities of the country. Figure 3 showed the results from the main analysis and stratified analysis. In the main adjusted model by population density, older people, GDP per capita, hospital beds, doctors, GRI, intra-city movement intensity, and temperature, we found an inverse association between city-level NDVI and COVID-19 incidence, with an IRR 

J o u r n a l P r e -p r o o f Table 2 showed the mediation analysis by PM2.5, NO2, and CO. The ADEs were statistically significant. We failed to detect any significant mediation effect of air pollutants on the association between NDVI and COVID-19 incidence. 

The results from a series of sensitivity analyses were provided in 

In this ecological study with 266 Chinese cities, we found that city-level greenness exposure was negatively associated with COVID-19 incidence. Per 0.1 unit increment of NDVI was associated with a 7.9% decrease in COVID-19 incidence.

The results provided scientific evidence of the protective effect of greenness on infectious disease. Coincidentally, consistent with our results, studies from the United States and Canada also suggested a negative association of greenness with COVID-19

incidence (Ciupa and Suligowski, 2021; Klompmaker et al., 2021; Stieb et al., 2020) .

In addition, several studies found an inverse association between greenness and COVID-19 mortality (Klompmaker et al., 2021; Russette et al., 2021) . Existing evidence was mainly conducted in developed countries. This is the first study to investigate the association between greenness exposure and COVID-19 incidence in the China region.

In stratified analyses, we found a stronger protective relation of NDVI on COVID-19 incidence in cities with lower population density. The result was not consistent with an ecological study based on approximately 3000 counties that found stronger associations in densely populated counties (Klompmaker et al., 2021) , and they explained a positive association of greenness and COVID-19 mortality in the lowest population density that an increase in greenness limited the access to health care. Differences in the association might be due to the heterogeneity of greenness variation and other confounders. The population is relatively stable in cities with lower population density in China. Individuals living in these areas have fewer opportunities to contact the nonnative population. Studies (Frank and Engelke, 2005; Shen et al., 2017) suggested that the increase in population density had a positive relationship with the accumulation of air pollutants, which increased the risk of COVID-19 infection. When stratified by GDP per capita, we found a stronger association in cities with lower GDP per capita. The GDP per capita is often considered a reflection of regional living standards. Similar to previous studies (Maas et al., 2006; McEachan et al., 2016; Sarkar, 2017) , health benefits of greenness were more prominent in populations with low income and low social-economic status. A possible explanation for this finding is that individuals with a lower socioeconomic status tend to be less mobile than those with higher counterparts, and spend more time in the vicinity of their homes, which promoted them contact with green space frequently (Maas et al., 2009) .

Numerous studies have investigated the association of air pollutants with various COVID-19 outcomes in different regions. In general, exposure to the high concentration of air pollutants was associated with increased incidence (Adhikari and Yin, 2020; Vasquez-Apestegui et al., 2020; Zhu et al., 2020) , mortality (Konstantinoudis et al., 2021; Liang et al., 2020; Wu et al., 2020) and case fatality (Liang et al., 2020; Tian et al., 2021; Yao et al., 2020) . In contrast, exposure to more greenness has been suggested to reduce air pollutants (Dadvand et al., 2012; Nowak et al., 2006) . One possible mechanism of greenness associated with decreased COVID-19 incidence was attributed to absorb air pollutants. Air pollution exposure could trigger oxidative stress and induce inflammation reactions, which may eventually deteriorate the immune system (Qin et al., 2020) . Air pollution exposure has also been associated with increased infection of SARS-CoV-2 (the virus that causes COVID-19). In our mediation analysis, we failed to detect any significant J o u r n a l P r e -p r o o f mediation effects of air pollutants on NDVI and COVID-19 incidence association.

Due to the unavailable of individuals' residential addresses, the air pollutants were collected and averaged at the scale of city-level, which was inevitable to bring into potential misclassification bias. In addition, China adopted strict measures to control the spread of COVID-19 during the initial phase of COVID-19 outbreak. The closure of factories and the restriction of vehicles lead to the reduction of air pollutants. This might be one of the explanations for the non-significant mediation effect by air pollutants in our study. More mechanistic studies on greenness -air pollutants -COVID-19 association are therefore needed to address the confusion.

There are two main strengths in our study. First, to our knowledge, this is the first study to quantitatively assess the association of greenness exposure with COVID-19 incidence in the China region. Second, we have performed a series of sensitivity analyses through the exclusion of cities with less than 10 and 5 confirmed COVID-19 cases and using greenness data from other months. The results remained robust and consistent.

However, several limitations should be acknowledged. First, this study is an ecological study with aggregated data at the city level. We could not obtain the individual-level confounding factors. Thus, the results might be affected by the ecological fallacy, which might bias the results to derive the causality. Second, we used NDVI as a proxy measurement of greenness, which could not provide information about the composition of green spaces, such as parks, forests, and J o u r n a l P r e -p r o o f agricultural land, and also the structure of vegetation, such as trees, grass, and shrubs.

Third, epidemic control strategies across cities did not take into account, such as the closure of community, wearing masks, banning orders, and stay-at-home order.

However, we think that there are few differences in the control strategies with the strict intervention measures implemented by the Chinese government. In addition, we obtained two indices, namely GRI and intra-city movement intensity, as a proxy of keeping social distance and population movement. Last but not the least, the first wave outbreak of COVID-19 was occurred during the period of the Chinese New Year, experiencing the largest population mobility. Assignment of greenness exposure was inevitable to bring potential misclassification bias. Thus, based on the above limitation, the results from our study should be interpreted with caution, and need to validate in other studies with less bias.

In conclusion, our study suggested a beneficial association between city-level greenness and COVID-19 incidence. This association was stronger in cities with lower population density, lower GDP per capita, and lower urbanization rate.

Although causality was unable to be derived from this study, the results might be important toward increasing more greenness and decreasing air pollution to protect human health. It is important for public health implications to research how modifiable factors are associated with COVID-19 outcomes to guide policymaking.

Further studies on the individual level that can access plausible biologic mechanisms are needed to uncover the protective effect of greenness.

J o u r n a l P r e -p r o o f

Term Effects of Ambient Ozone, PM2.5, and Meteorological Factors on COVID-19 Confirmed Cases and Deaths in Queens

Epidemiology, causes, clinical manifestation and diagnosis, prevention and control of coronavirus disease (COVID-19) during the early outbreak period: a scoping review

The coronavirus pandemic in five powerful charts

Green-Blue Spaces and Population Density versus COVID-19 Cases and Deaths in Poland

Surrounding greenness and exposure to air pollution during pregnancy: an analysis of personal monitoring data

Long-term exposure to greenspace and metabolic syndrome: A Whitehall II study

Green and blue spaces and physical functioning in older adults: Longitudinal analyses of the Whitehall II study

The global community needs to swiftly ramp up the response to contain COVID-19

Multiple Impacts of the Built Environment on Public Health: Walkable Places and the Exposure to Air Pollution

Long-term exposure to residential green and blue spaces and anxiety and depression in adults: A cross-sectional study

Utilising green and bluespace to mitigate urban heat island intensity

Urban Form, Air Pollution, and Health

Association between community greenness and obesity in urban-dwelling Chinese adults

Exposure to Greenness and Mortality in a Nationwide Prospective Cohort Study of Women

GPS-Based Exposure to Greenness and Walkability and Accelerometry-Based Physical Activity

Cardiovascular diseases in middle aged and older adults in China: the joint effects and mediation of different types of physical exercise and neighborhood greenness and walkability

Associations of air pollution and greenness with mortality in Greece: An ecological study

County-level exposures to J o u r n a l P r e -p r o o f greenness and associations with COVID-19 incidence and mortality in the United States

Long-term exposure to air-pollution and COVID-19 mortality in England: A hierarchical spatial analysis

Urban Air Pollution

Case-Fatality and Mortality Rates in the United States

Genomic characterisation and epidemiology of 2019 novel coronavirus: implications for virus origins and receptor binding

Social contacts as a possible mechanism behind the relation between green space and health

Green space, urbanity, and health: how strong is the relation?

Covid-19: WHO declares pandemic because of "alarming levels" of spread, severity, and inaction

Exploring pathways linking greenspace to health: Theoretical and methodological guidance

The association between green space and depressive symptoms in pregnant women: moderating roles of socioeconomic status and physical activity

Applications of laboratory findings in the prevention, diagnosis, treatment, and monitoring of COVID-19

Air Pollution Removal by Urban Trees and Shrubs in the United States

Association of neighborhood greenness with self-perceived stress, depression and anxiety symptoms in older U.S adults

Dysregulation of Immune Response in Patients With Coronavirus 2019 (COVID-19) in Wuhan, China

Greenspace exposure and COVID-19 mortality in the United States

Residential greenness and increased physical activity in patients after coronary artery bypass graft surgery

Residential greenness and adiposity: Findings from the UK Biobank

Physical Activity for Improving the Immune System of Older Adults During the COVID-19 Pandemic

Urbanization-induced population migration has reduced ambient PM2.5 concentrations in China

An ecological analysis of long-term exposure to PM2.5 and incidence of COVID-19 in Canadian health regions

Ambient air pollution and low temperature associated with case fatality of COVID-19: A nationwide retrospective cohort study in China

Association Between Air Pollution in Lima and the High Incidence of COVID-19: Findings from a Post Hoc Analysis

Air pollution and COVID-19 mortality in the United States: Strengths and limitations of an ecological regression analysis

Community greenness, blood pressure, and hypertension in urban dwellers: The 33 Communities Chinese Health Study

Associations of greenness with diabetes mellitus and glucose-homeostasis markers: The 33 Communities Chinese Health Study

Association of particulate matter pollution and case fatality rate of COVID-19 in 49 Chinese cities

Association between short-term exposure to air pollution and COVID-19 infection: Evidence from China

We express our sincere gratitude to editors and anonymous reviewers for providing valuable comments on this paper. 

The data that support the findings of this study are available from publicly available sources.