Environmental Research 216 (2023) 114089

Contents lists available at ScienceDirect

Environmental Research
journal homepage: www.elsevier.com/locate/envres

COVID-19 epidemic spread and green areas Italy and Spain between 2020
and 2021: An observational multi-country retrospective study
Andrea Falco a, b, *, Prisco Piscitelli b, Domenico Vito b, Federico Pacella b, Cristina Franco a,
Manuela Pulimeno b, Paolo Ambrosino a, Javier Arias c, Alessandro Miani b, d
a
University of Sannio, Benevento, Italy
b
Italian Society of Environmental Medicine, Milan, Italy
c
Dean of Medical School, Complutense University, Madrid, Spain
d
Department of Environmental Sciences and Policy, University of Milan, Italy

A R T I C L E I N F O A B S T R A C T

Keywords: Several studies have proposed that environmental factors influencing human wellbeing, such as chronic expo­
COVID-19 sures to high levels of particulate matter, could indirectly or even directly affect also the severity of COVID-19
Urban green areas disease in case of infection by novel coronavirus SARS-COV2. This study has investigated the association between
Hospitalizations
COVID-19 infections, hospitalizations or deaths and the extension of public green areas (km2 per 100,000 based
Mortality
Contagions
on OECD data of 2014), an indicator that has been chosen as independent endpoint variable to test the research
Environment hypothesis in 10 Italian and 8 Spanish Provinces with more than 500.000 inhabitants, including capitals (Rome
and Madrid) and bigger cities (Bologna, Catania, Florence, Genoa, Milan, Naples, Palermo, Turin and Venice for
Italy; Barcelona, Valencia, Seville, Zaragoza, Malaga, Las Palmas and Bilbao for Spain). Two different method­
ologies have been applied: a bottom-up approach was applied to Spanish institutional data concerning conta­
gions/hospitalizations/deaths and the extent of public green areas for each responder to an official questionnaire
in the frame of a nationwide survey (with detailed data granularity per province) containing specific georefer­
enced information; a top-down approach was used for Italy, starting from the official figures of contagions/
hospitalizations/deaths of each province and linking them to the OECD statistics about the extension of public
green areas in the different areas. Linear and generalized models were used for statistical analyses including also
PM2.5 in a multivariate approach (with annual average concentrations from official air quality monitoring
stations) and were able to adjust for the different number inhabitants living in each province, in order to take
into account the difference in contagion dynamics related to the different density of population. The results
obtained for Spain are consistent with those observed for Italy, as for both countries, it has clearly emerged a
statistically significant association between COVID-19 clinical features (contagions, hospitalizations, and deaths)
and the extension of public green areas, as well as the annual average concentrations of PM2.5 (with this latter
variable loosing statistical significance in some province). Therefore, the extension of public green areas and air
pollution seem to have a high correlation with COVID-19 severity.

1. Introduction of local communities and they have been questioned to worsen the dy­
namics of contagion or clinical outcomes of COVID-19 pandemic
COVID-19 is an infectious disease caused by the SARS-CoV-2 coro­ (Bakadia et al., 2021). Several studies have shown that the population
navirus. Most people infected with the virus experience mild to mod­ previously subjected to additional stressors, including chronic environ­
erate respiratory illness and recover without requiring special mental pressures are more susceptible to severe forms of COVID-19
treatments (Rothan and Byrareddy, 2020). Current lifestyles, especially (World Food Program, 2020; Valderrama et al., 2020; Xianyu et al.,
in cities, are characterized by heavy chronic stress and exposures to 2018). Environmental stressors refer to environmental factors causing
environmental hazards. These elements can influence the health status pressure on wellbeing status (Guski, 2001): such factors include

* Corresponding author. University of Sannio, 84100, Benevento, Italy.

E-mail address: [email protected] (A. Falco).

https://doi.org/10.1016/j.envres.2022.114089
Received 8 April 2022; Received in revised form 6 August 2022; Accepted 6 August 2022
Available online 22 August 2022
0013-9351/© 2022 Published by Elsevier Inc.
A. Falco et al. Environmental Research 216 (2023) 114089

available food, the presence of parasites or interactions with the envi­ 2. Materials and methods
ronment that can affect the physical and mental health condition of the
population in a certain area such as air quality. For the proposed multicentric analysis, open datasets of confirmed
Actually, such environmental stressors can represent a source of cases, hospital admissions and deaths published by the Spanish Institute
oxidative stress (Patel et al., 2018), that can be defined as an imbalance of Public Health (Istituto de Salud Carlos III, ISCIII) and by the Italian
favoring the oxidative systems compared to the antioxidants, resulting Civil Protection Department were used. For Italy, the number of COVID-
in impaired functioning of certain enzymes, transduction of cellular 19 infections from January 2020 to December 2021 were assessed with
signals and lowering of immune defense against pathogens, as well as regard to the average per-capita availability of urban green areas per
altered cell cycle, cell differentiation, and regulation of capillary dila­ 100,000 inhabitants as resulting from the database of the Organization
tion. Oxidative stress is involved as a trigger or associated with several for Economic Cooperation and Development (OECD, 2014). For Spain,
disease complications and especially regarding SARS-COV2 syndrome the number of infections from January 2020 to December 2021, were
(Cecchini and Cecchini, 2020). The exogenous and endogenous sources then merged at municipal level with the data of household expenditure
of oxidative stress could potentially favor the system oxidants versus of the INE (Istituto Nacional de Estadistica), taking into account the
antioxidants; thereby, the increased production of reactive oxygen degree of the area of residence, in a progressive manner from a value of
species (ROS) during the oxidative stress could promote viral multipli­ higher urbanization (1) to a value of higher rural density (9). In addi­
cation (Reshi et al., 2014). The COVID-19 exacerbation, hypothetically tion, in order to evaluate the incidence of air pollution, data from PM2.5
connected to effects of oxidative stress damages induced by exposure to monitoring stations at provincial level were used. The source of PM2.5
environmental stressors, showcase the promising way to be investigated data for Italy was the National Program for Air Quality Assessment
further for sustainable control of the pandemic (Rothan and Byrareddy, (Viias). The source of pollution data for Spain was the annual report
2020; Bakadia et al., 2021; World Food Program, 2020; Valderrama edited by the Spanish Ministry of Ecological Transaction. To assess the
et al., 2020; Xianyu et al., 2018; Guski, 2001; Patel et al., 2018; Cecchini presence of a statistically significant association between a reduced
and Cecchini, 2020). In this study, we want to explore the potential severity in the burden of COVID-19 and the higher availability of urban
benefits of green areas on the exposure to environmental stressors, green areas, we analyzed 11 Italian provinces (Rome, Milan, Naples,
exploring the possible association between the decrease in COVID-19 Turin, Palermo, Genoa, Florence, Bari, Bologna, Catania and Venice),
mortality or hospitalizations and the high availability of urban green and 8 Spanish provinces (Madrid, Barcelona, Valencia, Seville, Zar­
areas (UGA) in cities and provinces with more than 500.000 inhabitants. agoza, Malaga, Las Palmas, Bilbao). In addition, the value of average
Such relationship can let to better understand on how the different annual concentrations of PM 2.5 (category of particulate air pollutant
pre-existing environmental stressors and COVID-19-related stressors are that is 2.5 μm or smaller in size) was evaluated as a corrective factor to
further worsening the effects of the viral disease by inducing the gen­ assess - on the primary analysis - how air pollution impacted the rates of
eration of oxidative stress (Bakadia et al., 2021). Standing to EU infections, hospitalizations and death due to COVID-19.
COPERNICUS classification, urban green areas are defined as any area
with vegetation within or partly embraced by urban fabric (Copernicus). 2.1. Italy
This class is assigned for urban greenery, which usually has recreational
or ornamental character and is usually accessible for the public; it is The sample for the year 2020 consisted of 3432 timepoints recorded
applicable to parks inside settlements, with or without public access, from 1st of January 2020 to 31st December 2020. The sample for the
ornamental gardens, mansions’ green grounds, botanical and zoological year 2021 was composed of 4015 time points recorded from 1st of
gardens situated inside settlements or in contact-peripheral zone of January 2021 to 31st of December 2021. Each time point contained the
settlement, city squares with greenery, inner spaces of city blocks or any observations concerning the number of infected cases, the number of
vegetated areas that can potentially be used for recreational purpose. hospital admissions and the total deaths due to COVID-19. As already
Standing to “Urban green spaces and Health Report” edited by WHO in mentioned, the selected provinces were Bologna, Catania, Florence,
2016) (Europe, 2016), urban green spaces, can promote mental and Genoa, Milan, Naples, Palermo, Rome, Turin and Venice. The endpoint
physical health, and reduce morbidity and mortality in urban residents variable taken into account for the analysis was the average per-capita
by providing psychological relaxation and stress alleviation, stimulating availability (expressed in km2) of green urban areas per 100,000 for
social cohesion, supporting physical activity, and reducing exposure to green areas in urban areas (OECD, 2014) per each province. Then, the
air pollutants, noise and excessive heat. Several epidemiological studies variable of the incidence of green areas was then divided into 3 cate­
evaluated the effects of urban green space on the health outcomes of gories (Low, Medium, High presence of green areas) with a cut-off to the
study participants (Europe, 2016). However, the mechanisms underly­ first quartile (Q1) and second quartile (Q2), as described below:
ing links between green space access and health are complex and The derived variable is composed as follows:
interconnected and synergic. Hartig et al. (World Health Organization, The number of COVID-19 infections, hospitalizations, and deaths is
2016) suggested that there are four main interactive pathways through distributed as follows:
which nature or green space may contribute to health: improved air
quality, enhanced physical activity, stress reduction and greater social 2.1.1. Spain
cohesion. Regarding the current pandemic, all these co-benefits can The sample consisted of 1472 household units, with a total popula­
alleviate the pressure of environmental stressor and finally determine a tion of 3680, registered by INE between the year 1998 to the year 2019.
rebalancing of oxidative stress and a consequential reduction of the The selected provinces are Madrid, Barcelona, Valencia, Seville, Zar­
impact of COVID-19 (De Petris et al., 2021). Furthermore, Urban green agoza, Malaga, Las Palmas and Bilbao. The statistical robustness of the
spaces, as part of a wider environmental context, have the potential to in sample was significantly demonstrated by INE (Spanish National Insti­
a preventative way that can be more efficient than just intervening on tute of Statistics). No households were registered in the province of
the consequences of illnesses (Hartig et al., 2014). This study has Barcelona, which was therefore excluded from the analysis. The
investigated the association between COVID-19 infections, hospitaliza­ endpoint variable involved for the analysis is the type of residence in
tions or deaths and the extension of public green areas (km 2 per 100, which the household unit is located. The original variable of residence
000 based on OECD data of 2014), an indicator that has been chosen as type was so subdivided: 1 Urban luxury, 2 High Urban, 3 Urban, 4 Low
independent endpoint variable. The research hypothesis was tested in Urban, 5 Rural industrialized, 6 Fishing Rural, 7 Agriculture Rural. The
Italy and Spain, two Mediterranean countries with similar data were then grouped on the basis of high exposure to urbanization (1,
characteristics. 2) medium exposure to urbanization (3, 4, 5) and low exposure to ur­
banization (6, 7), thus resulting distributed as follows:

2
A. Falco et al. Environmental Research 216 (2023) 114089

The average per-capita availability of green areas in Spain for the Table 2
selected provinces, estimated in 2014 (OECD) was around 84.79 for Derived categorized variable for Green Areas in Italy (2020).
100.000 inhabitants, as described below: Items N Percentage Cumulative N Cumulative Percentage
The ISCIII database on the number of infections, hospitalizations and
1 (Low) 936 27.27 936 27.27
deaths due to COVID-19 from January 2020 to December 2021 for the 2 (Medium) 1872 54.55 2808 81.82
selected provinces was merged to the household expenditure question­ 3 (High) 624 18.18 3432 100
naire (INE database).
The sample is distributed as follows (Table 8):
Table 3
2.2. Data analysis Derived categorized variable for Green Areas in Italy,2021.
Items N Percentage Cumulative N Cumulative Percentage
Descriptive data are presented for numeric variables in mean, stan­
1 (Low) 1095 27.27 1095 27.27
dard deviation (SD), quartiles (Q1 and Q2). For categorical variables,
2 (Medium) 2190 54.55 3285 81.82
data are presented in percent (%), 95% confidence interval (95% CI), 3 (High) 730 18.18 4015 100
cumulative absolute frequency, and cumulative relative frequency.
Confidence intervals are presented calculated using the Clopper- Pear­
son method. Tests applied in this paper are examined at an alpha sig­
Table 4
nificance less than/equal to 0.05 (5%). Continuous variables are
COVID-19 demographic table (Contagions, Hospital Admissions, Deaths), Italy
compared through the t-student (t-value) test statistic. The Multiple 2020
Linear Regression Analysis Method was applied to analyze the correla­
Variable: Contagions
tion between the COVID-19 indicators (contagion, hospital admissions
and deaths) confirmed in each province and the green zone indicators. Mean SD Q1 Q3
Based on the data from the selected observation variable, multiple linear 14885.45 26866.71 1503.50 16312.50
regression models can be constructed as shown in the equation: Variable: Hospital admissions
Mean SD Q1 Q3
yi = β0 + β1*xi + εi. 1175.71 1906.83 101.00 1488.00
Variable: Deaths
The model has been developed in three different steps: Mean SD Q1 Q3
1412.49 3795.66 570.50 8292.00
- The first step considers y as the dependent variable, representing the
number of confirmed cases over time, measured in days. While the
independent variables represent the parameter of green areas per Table 5
capita per 100,000 inhabitants in the provinces, the level of PM2.5 Demographic table COVID-19 (Contagions, Hospital Admissions, Deaths), Italy
per capita per 100,000 inhabitants in the provinces and the total 2021
population to ensure homogeneity of the population in the clusters. Variable: Contagions
- The second step considers y as the dependent variable, representing Mean SD Q1 Q3
the number of confirmed hospital admissions over time, measured in
127858.80 85335.55 62479.00 197897.00
days. While the independent variables represent the parameter of
Variable: Hospital admissions
green areas per capita per 100,000 inhabitants in the provinces, the Mean SD Q1 Q3
level of PM2.5 per capita per 100,000 inhabitants in the provinces 1037.29 1205.99 243.00 1494.00
and the total population to ensure homogeneity of the population in Variable: Deaths
the clusters. Mean SD Q1 Q3
1388.10 1737.24 71.74 2466.48
- The third step considers y as the dependent variable, which repre­
sents the number of confirmed deaths over time, measured in days.
While the independent variables represent the parameter of green
areas per capita per 100,000 inhabitants in the provinces, the level of Table 6
Derived categorized variable for Type of Residence in Spain (1998) to 2019.
PM10 per capita per 100,000 inhabitants in the provinces and the
total population to ensure homogeneity of the population in the Items N Percentage Cumulative N Cumulative Percentage
clusters. 1 (Low) 300 20.38 300 20.38
2 (Median) 515 34.99 815 55.37
β i are the unknown parameters of the corresponding independent 3 (High) 657 44.63 1472 100.00
variables; ε is called the error term. The above linear regression model
can be used to predict daily virus trends and determine the correlation
between each independent variable and the dependent variable. Table 7
The models are then compared between/inside countries taking into Descriptive values for green area in Spain (2014) (OECD) for the selected
consideration R squared adjusted of to demonstrate the speed of the provinces.
event at each step. Variable: Green Areas per-capita per 100.000
In addition, derived the green areas in three categories (high ur­ Mean Median SD Q1 Q3
banization, medium urbanization and low urbanization) as reference
84.79 46.59 77.141 23.18 171.53

Table 1
Descriptive values for green areas in Italy (OECD, 2014) for the selected point the first and third quartile for Italy and the high urban, medium
provinces. urban and rural items for Spain, the logistic regression is applied as
Variable: Green Area pro capita 100.000 shown in the equation
Mean SD Q1 Q3
Log (P/1-P) = a+bx
209.65 268.02 22.11 366.64

3
A. Falco et al. Environmental Research 216 (2023) 114089

Table 8 The model is presented in 3 steps:

COVID-19 demographic table (Contagions, Hospital Admissions, Deaths) by
provinces, Spain 2020 - The first step considers y as the dependent variable, representing the
Variable: Contagions province (see Table 1). While the independent variables represent
Mean SD Q1 Q3
the derived green area variable (Table 2, Table 3, Table 6), the level
of PM2.5 per capita per 100,000 population in the provinces and the
119536.80 132616.66 57259.00 96574.00
total cases (see Table 5) (see Table 4).
Variable: Hospital admissions
Mean SD Q1 Q3 - The second step considers y as the dependent variable, representing
17087.48 23526.43 5547.00 10453.00 the province. While the independent variables represent the derived
Variable: Deaths green area variable (Table 2, Table 3, Table 6), the level of PM2.5 per
Mean SD Q1 Q3 capita per 100,000 population in the provinces and number of hos­
3144.01 4050.18 1170.00 1911.00
pital admissions.
- The third step considers y as the dependent variable, representing the
province. While the independent variables represent the derived
Table 9 green area variable (Table 2, Table 3, Table 6), the level of PM2.5 per
COVID-19 demographic table (Contagions, Hospital Admissions, Deaths) by capita per 100,000 population in the provinces and the number of
provinces, Spain 2021
deaths.
Variable: Contagions

Mean SD Q1 Q3 The fitting of goodness model then are compared in terms of Akaike
258749.30 226463.90 133017.00 257646.00
Information Criterion (AIC).
Variable: Hospital admissions
Mean SD Q1 Q3 3. Results
13337.95 12896.57 5967.00 12917.00
Variable: Deaths
In a preliminary analysis, we evaluated the difference in terms of
Mean SD Q1 Q3
1911.65 1273.60 1183.00 2381.00 available urban green areas in the selected provinces for Italy and Spain.
The results showed a significant difference of the two areas, in line with
the literature. In fact, in Spain there are about 7 million trees planted in a
population of 47 million people, while in Italy there are 22 million trees
Table 10
Results of the preliminary statistical analysis.
planted in a population of 59 million people. The results are shown
below:
Mean Difference 95 %CI Wilcoxon-test p-value
The test was conducted with a two-tailed t-test and was highly sta­
124.85 113.98–135.72 <0.001 tistically significant. Linear regression shows high significance for green
areas and pollution with a negative trend for green areas and a positive
trend for pollution. This shows that significant increases in covid inci­
Where P represents the probability of the event.
dence are lower for green areas. The models were corrected for pop­
The logistic model is implemented to identify the protective or risk
ulations between provinces. It can be seen that the corrections tend to
value of derived areas by the calculation of Odds Ratio (OR):
balance the population weights between very dense and not so dense
OR = exp(log(P/1-P)) area. The results are shown below:
The trends are similar for Spain and Italy (see Table 14) (see
Interpretated as the exponential of the regression parameter. Table 15). We can see that the R squared of the regressions in Spain are
The ORs have a breakpoint at value 1. This means that value 1 is a lower than those in Italy. This trend is justified because the exposure to
value of zero risk of exposure to green areas for the spread of the virus. If urban green areas in Italy is higher than in Spain, as shown in the pre­
the value is greater than 1 then it means that green areas are a risk factor liminary analysis. It can also be noted that the effect of covid mortality is
for the spread of the virus. If the Odds Ratio value is less than 1 then it higher in the year 2021 than in 2020. In 2020 Spain, the effect of
means that green areas are a protective value for the spread of the virus. pollution on Covid is found to be non-significant and very low. In
The category of high urbanization areas is adopted as the reference addition, ORs were evaluated for green areas categorized by a derived
value for the analysis. ORs are presented with their respective 95% variable. The results are shown below:.
confidence intervals. From the results it can be seen that the green areas are significantly

Table 11
Linear Multivariable Regression – Italy (2020) with dependent variable Covid endpoints.
Endpoint Parameter Standard t-Value p-value 95% CI Adjusted R-squared
Error

Contagions 0.15
Population 0.01 0.00 17.58 <.0001 0.01 0.01
Green Area − 11.23 1.98 − 5.67 <.0001 − 15.11 − 7.34
Air pollution 99.57 11.91 8.36 <.0001 76.21 122.93

Hospitalized 0.11
Population 0.00 0.00 14.25 <.0001 0.00 0.00
Green Area − 1.16 0.15 − 7.94 <.0001 − 1.45 − 0.87
Air pollution 8.38 1.01 8.29 <.0001 6.40 10.37

Deaths 0.06
Population 0.01 0.00 8.70 <.0001 0.00 0.01
Green Area − 20.83 2.99 − 6.96 <.0001 − 26.69 − 14.96
Air pollution 161.01 20.69 7.78 <.0001 120.44 201.58

4
A. Falco et al. Environmental Research 216 (2023) 114089

Table 12
Linear Multivariable Regression – Italy (2021) with dependent variable Covid endpoints.
Endpoint Parameter Standard t-Value p-value 95% CI Adjusted R-squared
Error

0.83
Contagions Population 0.07 0.00 111.42 <.0001 0.07 0.07
Green Area − 68.22 2.58 − 26.40 <.0001 − 73.28 − 63.15
Air pollution 367.33 15.55 23.63 <.0001 336.85 397.81

Hospitalized 0.08
Population 0.00 0.00 14.50 <.0001 0.00 0.00
Green Area − 0.45 0.09 − 5.12 <.0001 − 0.62 − 0.27
Air pollution 2.35 0.60 3.91 <.0001 1.17 3.53

Deaths 0.08
Population 0.04 0.00 12.22 <.0001 0.03 0.05
Green Area − 115.52 13.40 − 8.62 <.0001 − 141.78 − 89.25
Air pollution 796.50 92.64 8.60 <.0001 614.86 978.14

Table 13
Linear Multivariable Regression – Spain (2020) with dependent variable Covid endpoints.
Endpoint Parameter Standard t-Value p-value 95% CI Adjusted R-squared
Error

Contagions 0.05
Population 0.09 0.04 2.52 0.01 0.02 0.16
Green Area − 10.86 4.84 − 2.25 0.03 − 20.39 − 1.34
Air pollution 5.95 4.46 1.33 0.18 − 2.83 14.72

Hospitalized 0.04
Population 0.02 0.01 2.52 0.01 0.00 0.03
Green Area − 1.97 0.87 − 2.26 0.02 − 3.69 − 0.26
Air pollution 0.55 0.80 0.69 0.49 − 1.03 2.13

Deaths 0.03
Population 0.00 0.00 2.49 0.01 0.00 0.01
Green Area − 0.31 0.15 − 2.09 0.04 − 0.61 − 0.02
Air pollution 0.01 0.14 0.07 0.94 − 0.26 0.28

Table 14
Linear Multivariable Regression – Spain (2021) with dependent variable Covid endpoints.
Endpoint Parameter Standard t-Value p-value 95% CI Adjusted R-squared
Error

Contagions 0.69
Population 0.15 0.02 6.94 <.0001 0.11 0.20
Green Area − 7.04 3.31 − 2.13 0.03 − 13.53 − 0.55
Air pollution 22.01 3.36 6.55 <.0001 15.42 28.61

Hospitalized 0.06
Population 0.01 0.00 6.93 <.0001 0.01 0.01
Green Area − 0.40 0.19 − 2.12 0.03 − 0.77 − 0.03
Air pollution 1.14 0.19 5.92 <.0001 0.76 1.52

Deaths 0.18
Population 0.00 0.00 6.92 <.0001 0.00 0.00
Green Area − 34.99 16.70 − 2.10 0.04 − 67.75 − 2.23
Air pollution 243.08 16.97 14.32 <.0001 209.78 276.37

protective against the spread of the SARS-COV2 virus. In Italy, the ORs 4. Discussion
are much more remarkable than in Spain. This is because in Spain (7
million trees in all the country per 47 million inhabitants) the avail­ This study has investigated the association between COVID-19 clin­
ability of green areas for urban areas is lower than Italy (22 million trees ical outcomes and the extension of public green areas (with three
per 59 million inhabitants), as demonstrated in the preliminary analysis different categories: high urban areas, medium urban and rural areas) in
by difference t-test on green area per-capita. Otherwise, the value of the provinces with more than 500.000 inhabitants of two different Medi­
OR indicates a risk in Spain for the value “Medium” vs. “High”. This is terranean countries (8 provinces in Spain and 10 in Italy) for years 2020
because morphologically the territory of Spain on green maps is more and 2021. Two different methodologies were applied: a bottom-up
desertic than that of Italy. Therefore, the green areas on average are approach was applied to Spanish institutional data both concerning
“higher” than the highly green areas which also have a lower greenery contagions/hospitalizations/deaths and the extent of public green areas
compared to the high Italian green areas, as shown in Table 7. for each responder to an official questionnaire in the frame of a
nationwide survey (with detailed data granularity per province)

5
A. Falco et al. Environmental Research 216 (2023) 114089

Table 15
Logistic Multivariable models for Contagion, Hospitalized, Deaths - Italy, 2020.
Without Adjusted for PM2.5 With Adjusted for PM2.5

Endpoint Categories OR 95% CI p-value AIC OR 95% CI p-value AIC

Contagions Low vs Median 0.17 0.14 0.20 <.0001 15752.10 0.13 0.11 0.15 <.0001 14957.23
Low vs High 0.10 0.08 0.12 0.12 0.10 0.14
Median vs High 0.61 0.53 0.70 0.92 0.80 1.07
Hospitalized Low vs Median 0.18 0.16 0.22 <.0001 14249.77 0.13 0.11 0.15 <.0001 13751.33
Low vs High 0.12 0.10 0.15 0.13 0.11 0.16
Median vs High 0.65 0.56 0.75 1.04 0.88 1.22
Deaths Low vs Median 0.17 0.15 0.21 <.0001 14237.24 0.12 0.10 0.15 <.0001 13729.83
Low vs High 0.11 0.09 0.14 0.13 0.11 0.16
Median vs High 0.65 0.56 0.76 1.06 0.91 1.25

Table 16
Logistic Multivariable models for Contagion, Hospitalized, Deaths – Italy, 2021.
Without Adjusted for PM2.5 With Adjusted for PM2.5

Endpoint Categories OR 95% CI p-value AIC OR 95% CI p-value AIC

Contagions Low vs Median 0.18 0.15 0.21 <.0001 19275.10 0.07 0.06 0.08 <.0001 17294.50
Low vs High 0.10 0.08 0.13 0.04 0.03 0.05
Median vs High 0.58 0.51 0.67 0.65 0.56 0.74
Hospitalized Low vs Median 0.18 0.16 0.22 0.284 16757.90 0.12 0.11 0.15 <.0001 16096.00
Low vs High 0.11 0.09 0.13 0.12 0.10 0.15
Median vs High 0.58 0.51 0.67 0.97 0.84 1.13
Deaths Low vs Median 0.18 0.16 0.22 <.0001 16744.50 0.13 0.11 0.15 0.191 16110.50
Low vs High 0.11 0.09 0.13 0.13 0.11 0.15
Median vs High 0.59 0.52 0.68 0.99 0.86 1.15

Table 17
Logistic Multivariable models for Contagion, Hospitalized, Deaths - Spain, 2020.
Without Adjusted for PM2.5 With Adjusted for PM2.5

Endpoint Categories OR 95% CI p-value AIC OR 95% CI p-value AIC

Contagions Low vs Median 0.22 0.11 0.44 0.04 1026.61 0.20 0.10 0.41 0.04 860.40
Low vs High 0.27 0.14 0.54 0.29 0.14 0.59
Median vs High 1.25 0.80 1.95 1.44 0.89 2.32
Hospitalized Low vs Median 0.24 0.12 0.48 <.0001 1016.23 0.22 0.11 0.45 0.04 850.17
Low vs High 0.29 0.15 0.59 0.31 0.15 0.64
Median vs High 1.23 0.78 1.93 1.42 0.88 2.31
Deaths Low vs Median 0.22 0.11 0.44 0.02 1025.36 0.20 0.10 0.42 0.05 859.36
Low vs High 0.27 0.14 0.55 0.29 0.14 0.60
Median vs High 1.25 0.80 1.95 1.43 0.89 2.32

Table 18
Logistic Multivariable models for Contagion, Hospitalized, Deaths - Spain, 2021.
Without Adjusted for PM2.5 With Adjusted for PM2.5

Endpoint Categories OR 95% CI p-value AIC OR 95% CI p-value AIC

Contagions Low vs Median 0.78 0.59 1.04 <.0001 5282.52 0.67 0.50 0.91 <.0001 4366.90
Low vs High 0.81 0.62 1.06 0.70 0.53 0.94
Median vs High 1.03 0.84 1.26 1.05 0.85 1.30
Hospitalized Low vs Median 0.78 0.59 1.03 <.0001 5275.66 0.66 0.49 0.90 <.0001 4358.78
Low vs High 0.80 0.61 1.05 0.70 0.52 0.93
Median vs High 1.03 0.84 1.26 1.05 0.85 1.30
Deaths Low vs Median 0.80 0.61 1.06 <.0001 5304.81 0.69 0.51 0.93 0.191 4387.80
Low vs High 0.82 0.63 1.08 0.72 0.54 0.96
Median vs High 1.03 0.84 1.26 1.05 0.84 1.29

containing specific georeferenced information; a top-down approach in each province, in order to take into account the difference in conta­
was used for Italy, starting from the official numbers of contagions/ gion dynamics related to the higher density of population. The results
hospitalizations/deaths of each province and linking them to the OECD obtained for Spain are consistent with those observed for Italy, as for
statistics (year 2014) about the extension of public green areas in the both countries, it has clearly emerged a statistically significant associ­
different provinces. The linear and generalized models used for statis­ ation between COVID-19 clinical features (contagions, hospitalizations,
tical analyses included also PM2.5 in a multivariate approach (with and deaths) and the extension of public green areas in the provinces, as
annual average concentrations from official air quality monitoring sta­ well as the annual average concentrations of PM2.5 (with this latter
tions) and were able to adjust for the different number inhabitants living variable loosing somewhere statistical significance). Therefore, the

6
A. Falco et al. Environmental Research 216 (2023) 114089

extension of public green areas and air pollution seem to have a high 26, protective public health measures in the view of pandemic pre­
correlation with SARS-CoV-2 activity. It should be noticed that there is a paredness should include a strong governmental commitment if refor­
substantial difference between Spain and Italy concerning the extension estation and air pollution reduction.
of green areas belonging to the intermediate cut-off (medium urban
areas) vs. those included in highest class (rural areas with lower ur­ CRediT authorship contribution statement
banization rate). Actually, living in a medium urban area in Spain does
not represents a protective factor towards COVID-19 severe clinical Andrea Falco: Writing,Statistician. Prisco Piscitelli: Writing,
outcomes, because the extension of public green areas in medium urban Project administration. Domenico Vito: Writing, Study Design. Feder­
and high urban areas in Spain does not significantly differ (due to the ico Pacella: Validation. Cristina Franco: Quality Check. Manuela
inclusion of desert zones with less dense vegetation in the “green” Pulimeno: Investigation. Paolo Ambrosino: Supervision. Javier Arias:
areas”). This study is based only on official data provided by the national Supervision. Alessandro Miani: Project administration.
and European agencies, with a high level of granularity. Our results are
consistent with studies that show how the incidence of green spaces, Declaration of competing interest
especially forests, can influence the transmission of COVID-19 (Bin and
Yi, 2021). These researches have proposed that the development of The authors declare that they have no known competing financial
COVID-19 epidemic in urban spaces can be correlated and eventually interests or personal relationships that could have appeared to influence
predicted based on changes in temperature and deforestation. This as­ the work reported in this paper.
sociation can potentially trigger indeed corresponding prevention and
control measures that can be promptly adopted to protect public health. References
It has been suggested that it is necessary to grasp the influence of climate
factors and that active measures can be taken to control COVID-19 Bakadia, B.M., Boni, B.O.O., Ahmed, A.A.Q., Yang, G., 2021. The impact of oxidative
outbreaks when virus activity is at a low level (Lin et al., 2020). stress damage induced by the environmental stressors on COVID-19. Life Sci. 264,
118653.
Currently, the specific mechanism of the interaction between green Berdejo-Espinola, V., Suárez-Castro, A.F., Amano, T., Fielding, K.S., Oh, R.R.Y., Fuller, R.
environment and virus activity is unknown. But COVID-19 has been A., 2021. Urban green space use during a time of stress: a case study during the
proved to produce less dramatic effects in green areas (Berdejo-Espinola COVID-19 pandemic in Brisbane, Australia. People and Nature 3 (3), 597–609.
https://doi.org/10.1002/pan3.10218.
et al., 2021). For these reasons, Governments should pay particular Bin, J., Yi, L., 2021. Green Spaces, Especially Forest, Linked to Lower SARS-CoV-2
attention to the environment when formulating measures to prevent and Infection Rates: A One-Year Nationwide Study. https://doi.org/10.1101/
control the epidemic situation, which can cause SARS-CoV-2 to have 2021.08.04.21261420.
Cecchini, R., Cecchini, A.L., 2020. SARS-CoV-2 infection pathogenesis is related to
higher activity in presence of scarce green areas. All the countries should oxidative stress as a response to aggression. Med. Hypotheses 143, 110102. https://
strengthen decisions in favor of environmental policies to protect air, doi.org/10.1016/j.mehy.2020.110102.
water and land across Europe. The preventive and control measures Copernicus. Green urban areas definition. available at: https://land.copernicus.eu/user
-corner/technical-library/corine-land-cover-nomenclature-guidelines/html/inde
related to COVID-19 epidemic should be strengthened when the activity x-clc-141.html.
of the virus is low, so as to suppress the trend of subsequent epidemic De Petris, S., Squillacioti, G., Bono, R., Borgogno-Mondino, E., 2021. Geomatics and
outbreaks or waves and achieve the purpose of epidemic prevention and epidemiology: associating oxidative stress and greenness in urban areas. Environ.
Res. 197, 110999.
control. At the same time, attention should be paid to the prevention and
Europe, W.H.O., 2016. In: Urban Green Spaces and Health. A Review of Evidence. WHO
control of environmental factors such as increasing the urban green Europe, Copenhagen.
areas and reducing air pollution. Our results are also significantly worse Guski, R., 2001. Environmental stress and health. In: Int. Encycl. Soc. Behav. Sci.
for hospitalized and mortality in areas with lower availability of green Elsevier, pp. 4667–4671. https://doi.org/10.1016/B0-08-043076-7/03832-8.
Hartig, T., Mitchell, R., De Vries, S., Frumkin, H., 2014. Nature and health. Annu. Rev.
spaces and highly polluted environments. So, it could be hypothesized Publ. Health 35, 207–228.
that fewer green public areas and higher air pollution also lead to an Lin, Shaofu, Fu, Yu, Jia, Xiaofeng, Ding, Shimin, Wu, Yongxing, Huang, Zhou, 2020.
increase in suspended solids in the atmosphere (as proposed by some Discovering correlations between the COVID-19 epidemic spread and climate. Int. J.
Environ. Res. Publ. Health 17 (21), 7958. https://doi.org/10.3390/ijerph17217958.
authors) which provides ideal conditions for attachment, replication and Patel, R., Rinker, L., Peng, J., Chilian, W.M., 2018. Reactive oxygen species: the good and
transmission of the virus 19. Finally, the combination of high pollution the bad. In: React. Oxyg. Species Living Cells. InTech.
and lower availability of green areas generally reduces human resistance Reshi, M.L., Su, Y.-C., Hong, J.-R., 2014. RNA viruses: ROS-mediated cell death. Int J Cell
Biol 2014, 467452.
to viruses (in terms of hospital admissions and mortality) and improves Rothan, H.A., Byrareddy, S.N., 2020. The epidemiology and pathogenesis of coronavirus
viral transmission rate. disease (COVID-19) outbreak. J. Autoimmun. 109, 102433.
Valderrama, J., Hansen, S.K., Pato, C., Phillips, K., Knowles, J., Pato, M.T., 2020. Greater
history of traumatic event exposure and PTSD associated with comorbid body
5. Conclusions
dysmorphic disorder in a large OCD cohort. Psychiatr. Res. 289, 112962.
World Food Program, 2020. COVID-19 Will Double Number of People Facing Food
The results of our study show that huger extension of public green Crisesunless Swift Action Is Taken. https://www.wfp.org/news/covid-19-wi
ll-double-number-people-facing-food-crises-unless-swift-action-taken. (Accessed 30
areas in province is significantly associated with less severe COVID-19
August 2020). Accessed.
clinical outcomes in terms of contagions, hospitalizations, and espe­ World Health Organization, 2016. Urban Green Spaces and Health (No. WHO/EURO:
cially deaths. The adjusted additional effect of air pollution (namely 2016-3352-43111-60341). World Health Organization. Regional Office for Europe.
PM2.5 annual average concentrations) was shown to turn in similar Xianyu, J., Feng, J., Yang, Y., Tang, J., Xie, G., Fan, L., 2018. Correlation of oxidative
stress in patients with HBV-induced liver disease with HBV genotypes and drug
results, with lower levels of PM2.5 further reducing the burden of resistance mutations. Clin. Biochem. 55, 21–27. https://doi.org/10.1016/j.
negative clinical outcomes. In line with the resolutions adopted at COP- clinbiochem.2018.03.014.