Time Trends of Greenspaces, Air Pollution, and Asthma Prevalence among Children and Adolescents in India

The prevalence of childhood asthma contributes to the global burden of the disease substantially. Air pollution in India has increased. In this study, we examine the associations among greenspaces, air pollution, and asthma prevalence in children and adolescents over a large, diverse population in India. We used state-wide global burden of disease data on asthma from age 0 to 19 years in 2005, 2011, and 2017. For greenspace, we used the normalized differential vegetation index (NDVI), which is the surface reflectance of light during photosynthetic activity. NDVI, air pollutants (PM2.5, PM10, SO2, NO2, and O3), weather, and socio-demographic factors were included in generalized estimating equation (GEE) models to estimate their associations with childhood asthma prevalence over time. Novel data visualization illustrated the complex spatial distributions. NDVI was associated with asthma prevalence (β = 0.144; 95% CI = 0.10, 0.186; p < 0.0001) for high PM2.5, along with high levels of both gaseous air pollutants, SO2, and NO2 ((β = 0.12; 95% CI = 0.08, 0.16; p < 0.0001) and (β = 0.09; 95% CI = 0.05, 0.13; p < 0.0001)). However, NDVI and high O3, had a strong negative association with asthma prevalence (β = −0.19; 95% CI = −0.26, −0.11; p < 0.0001). We observed additional effects of the interaction between the NDVI and high concentrations of PM2.5, PM10, NO2, and O3, assuming that these associations share a common pathway, and found interaction effects for asthma prevalence. Given the changing environmental conditions that interplay over geographical characteristics on the prevalence of asthma, further studies may elucidate a better understanding of these complex associations.


Introduction
Asthma is an important public health respiratory disease affecting 1-18% of the population in various countries [1]. Asthma ranks as the second most prevalent disease and the second leading cause of mortality among chronic respiratory diseases [2]. In children, the global prevalence of asthma varies from 5 to 20%, contributing substantially to the disease burden [3,4]. If uncontrolled, it results in persistent symptoms associated with asthma, increased hospitalization, reduced lung functions, and impaired quality of life [5,6].
Asthma is multifactorial, and we are yet to understand the causal mechanisms fully; however, genes, the surrounding environment, and behaviors play a part in asthma development [7][8][9]. Environmental factors may account for the substantive variations in asthma prevalence across different regions of the world [9][10][11][12]. Air pollutants, especially from traffic emissions (TRAP), lead to the worsening of existing respiratory symptoms [13][14][15] and increase the risk for new-onset asthma [16]. Globally, 5-10 million and 9-23 million Int. J. Environ. Res. Public Health 2022, 19, 15273 2 of 17 annual asthma emergency room visits (ERVs) were attributable to particulate matter of diameter 2.5 µm or less (PM 2.5 ) and ozone (O 3 ), respectively [17,18]. Each year, 16 new million pediatric asthma cases occur due to PM 2.5 , and about 4.5 million premature deaths were attributable to air pollution itself [19,20]. Ambient air pollution due to PM 2.5 contributes to 50% of the global total disease burden greatly in major developing countries (China and India) and 17.4 million DALYs in children younger than five years globally [21].
Most of the observational studies on ambient air pollution (AAP) and asthma focus on Australia, Europe, and the US [22][23][24] but very little in the South Asia region, especially in India. The estimated mortality rates from ambient air pollution (AAP) and asthma are the greatest in the South Asian subcontinent, and India alone accounted for more than 1.09 million deaths due to AAP in 2015 [21,25]. The average concentration in India of PM 2.5 [65.2 (78.61-96.56) µg/m 3 ] is 6 times more than the WHO permissible limits for PM 2.5 (10 µg/m 3 ) [26].
There is considerable interest in understanding the effect of the natural green environment on asthma, but the evidence is unclear [14,27,28]. A study in Spain observed a 60% higher prevalence of asthma in children living close to greenspaces such as parks [28], while another study from New York observed that an increased level of greenness was associated with a 17% higher prevalence of asthma in children aged 4-7 years [29], and another showed declines in prevalence among pre-school children in a New York study [27]. A recent review of studies found an association between urban greenspaces and asthma among children, although findings were inconsistent, and there was no conclusive interpretation [30]. When both greenspace and air pollution effects were studied together, the impacts of air pollution on asthma prevalence were lesser for children living in areas with higher levels of greenspaces and increased access to greenspaces [27], while this was contradicted in other studies showing an increase in asthma prevalence [31,32].
Studies in the US, Canada, and Europe have assessed both greenspace and air pollution with childhood asthma prevalence [27][28][29]32,33], but no studies have discussed South Asia. There are almost half a billion children in India, and about 6% of the children in India have asthma [10]. India comprises approximately 18% of the global population [34], and it is quite heterogeneous; diversity exists across states in environmental and topographical features. There are no studies on greenness and asthma prevalence in children and adolescents across all states and union territories in India. Studies on air pollution globally estimated the exposure rate is highest in India [21,35], and 14 of the top 20 world's most polluted cities in terms of particulate matter of PM 2.5 are in India [36,37]. India enormously contributes to global asthma emergency room visits due to air pollutants PM 2.5 (30%), O 3 (23%), and nitrogen dioxide (NO 2 ) (15%) [17]. Over the last two decades, air pollution has increased in India, and particulate matter contribution to AAP is estimated to increase in India over the next ten years drastically [38,39]. In this study, we sought to estimate the level of greenness in each state of India and its association with asthma prevalence in children and adolescents. We also assessed the role of air pollutants (PM 2.5 , PM 10 , SO 2 , NO 2 , and O 3 ) in these associations. We hypothesized that exposure to increased greenness over time shows lower asthma prevalence in children and adolescents. Increased levels of air pollutants in these areas will modify the association.

Study Design and Population
India is a diverse country spread across 15,200 km of land and 7517.24 km of coastline, and it is home to 1.3 billion residents. India comprises 29 states and 7 union territories. The health system in India operates at the state level [40], and so we present the results of an analysis of the 29 states and 1 union territory-Delhi, the national capital. The remaining six small union territories (Andaman and Nicobar Islands, Chandigarh, Dadra and Nagar Haveli, Daman and Diu, Lakshadweep, and Puducherry) were not included due to the non-availability of data for all the study variables.

Outcome: Asthma Prevalence Data
The prevalence of asthma data by location, age, and year we derived from the Global Burden of Disease (GBD) data visualization tool developed by the Institute for Health Metrics and Evaluation (IHME), USA [41]. The state-wise prevalence of asthma was estimated using the terms defined in the GBD studies [42][43][44]. We calculated asthma prevalence estimates for the 29 states and Delhi for children aged 0 to 19 years across three time periods-2005, 2011, and 2017. Currently, GBD is the most reliable source in India, as it collates the data from the national databases of the federal government and extensive multicenter studies in India, including the International Study of Asthma and Allergies in Childhood (ISSAC) and the India Study on Epidemiology of Asthma, Respiratory Symptoms and Chronic Bronchitis (INSEARCH) studies [39,43,45].

Exposures and Other Variables
Area Greenness: Greenness was obtained from Landsat Thematic Mapper Surface Reflectance images (https://earthexplorer.usgs.gov/, accessed on 5 May 2020), which we used to estimate the normalized differential vegetation index (NDVI). The Landsat images constituted the surface reflectance images generated every 16 days at a 30-m pixel spatial resolution and selected images for November. We obtained the next closest image if we missed or could not generate the surface reflectance images for that specific 16-day period. Similarly, the NDVI for each state was estimated across the three time points of 2005, 2011, and 2017. The selected cloud-free (<10 percent clouds) images avoided contamination with the NDVI values. An NDVI value of '0 means no vegetation, and values close to '1 represent the highest greenness [46]. The NDVI captured the density of greenness at a spatial resolution of 30 m and was calculated using the Quantum Geographical Information System (QGIS), Gnu General Public License, Version 2, 1991 Free Software Foundation, Inc. (Boston, MA, USA) tool v.3.12.3 9 [47]. The ratio of visible light (R) to the near-infrared (NIR) light reflected by the vegetative growth comprises the NDVI, that is, the surface reflectance of light during photosynthetic activity [46]. Greenness in a spatial area is defined as an average value of NDVI. The average NDVI values were calculated for the years 2005, 2011, and 2017, representing the annual greenspace exposure for the respective time periods in our study. The NDVI proportions were calculated using the following expression NDVI = (NIR − R)/(NIR + R)".
Air pollution: The exposure estimates of air pollutants were from open data sources from the statutory organization the Central Pollution Control Board (CPCB) and its counterparts, the State Pollution Control Boards (SPCBs) of India. The annual average concentrations of particulate matter up to diameters of 2.5 µm (PM 2.5 ) and 10 µm (PM 10 ), nitrogen dioxide (NO 2 ), sulfur dioxide (SO 2 ), and ozone (O 3 ) in µg/m 3 were from the routine air quality monitoring stations of India for each of the years 2005, 2011, and 2017. We considered a small number of air pollutant data from reports, web pages, and details from the environmental information system (ENVIS) of the central pollution control board and state government. The published literature across 30 states in India contributed to some of the data on the average annual air pollutants, and data with different units of measure, such as parts per billion (ppb), were converted into micrograms per meter cube. Northeast India presents only recent years' data, as the air quality monitoring stations across all seven states are recent. CPCB had first set the air quality standards in India in 1982, which were further revised by the year 2009, and the standards were higher than the air quality guidelines of WHO [48,49]. The National Air Quality Monitoring Program (NAMP), India, has set the National Ambient Air Quality Standards (NAAQS), with permissible limits for each of the air pollutants to protect public health, vegetation, and property [50].
Weather and other variables: The meteorological data for the states in India were recorded from online points of data collection-www.weatheronline.in, www.wunderground.com, and the Indian Meteorological Department (IMD); accessed on 15 December 2020-spanning 2005, 2011, and 2017. The meteorological parameters utilized were the annual average maximum temperature (degree Celsius), relative humidity (percentage), and average annual rainfall (millimeters). Population distribution in India is measured using the density of the population. Population density (PD) data across all states in 2005, 2011, and 2017 were included from the Census of India website, Refs. [51][52][53], representing the number of persons per square kilometer (km 2 ). The Social Progress Index (SPI) is a tool reproduced by the Institute for Competitiveness and Social Progress Imperative that comprises facets of social progress on basic human needs, the foundation of wellbeing, and opportunity, which includes various components [54,55]. The SPI supplements economic success measures by directly measuring social and environmental outcomes. The scores range on a scale of 0-100 for each state. A higher state score indicates a better understanding of the relationship between economic gain and social progress.

Statistical Analysis
We used data visualization methods to illustrate the distribution of asthma prevalence in the India map chart across 29 states and Delhi over the years 2005, 2011, and 2017. The estimated mean NDVI values of the 29 states and 1 union territory (SUTs) projected spatial distributions of average greenness over 2005, 2011, and 2017. Similar graphical methods described the differences in the population rates with PM 2.5 and O 3 levels and asthma prevalence. Two-dimensional multiple line graphs were plotted to understand the pattern of association between asthma prevalence and the mean NDVI (Y1 and Y2-axis) over the periods on the horizontal axis (X-axis), created using Stata IC16 software(StataCorp, College Station, TX, USA). Statistical software R, v.3.6.3, was used to plot the multiple variables, using gg-plot for bubble charts and contour plots. The bubble plot is a multi-variable chart similar to a scatterplot and describes the average concentrations of the pollutants and the mean NDVI across the X-axis and Y-axis.
Spearman's correlations were estimated to determine the levels of correlation among air pollutants (PM 2.5 , PM 10 , SO 2 , NO 2 , and O 3 ), and between air pollutants and meteorological factors (maximum temperature, relative humidity, and average annual rainfall).
We first separately analyzed the association between the NDVI and the prevalence of asthma in each year (2005, 2011, and 2017), controlling for each air pollutant (PM 2.5 , PM 10 , SO 2 , NO 2 , and O 3 ). For each year, the regression model approach that we took was modeling the outcome as a Poisson regression and adjusting it for meteorological parameters, PD, and SPI. We then used a generalized estimating equation (GEE) to estimate the association between the measure of greenness (NDVI) as the primary exposure variable and the prevalence of asthma as the outcome, using a Poisson link and combining the three time points in a single model. Additionally, we also examined the interaction between the NDVI and high air pollutant levels over the three time periods across all states in the GEE model. The air pollutant concentrations higher than the 75th centile were considered high. All statistical tests were 2-sided; effect estimates with a 95% confidence interval (CI) were reported and considered to have strong evidence of association if the p-value was less than 0.05. The statistical data analyses were performed using Stata IC 16 (StataCorp, College Station, TX, USA).

Results
The overall prevalence (min-max) of childhood and adolescent asthma for 0-19 years  (Table 1). Asthma in children and adolescents was more prevalent across the northeastern regions of India than in the rest ( Figure 1A). The prevalence rates of asthma across all the states in India show a progressive increase from 2005 (1275 cases per 100,000) to 2011 (1776 cases per 100,000), and then present a downward trend until 2017 (1419 cases per 100,000) ( cluding high O3 levels, the NDVI shows a strong negative association with asthma prevalence (β = 0.19; 95% CI = 0.26, 0.11; p < 0.0001).
The strata-specific effects of low vs. high air pollution on the relationship between the NDVI and asthma prevalence are presented in Figure 4. In all cases except SO2, there was an interaction between the air pollutants and the NDVI in the GEE models ( Table 3). The largest difference in the slopes can be observed in the PM2.5 and NO2 plots, where in the states with low concentrations of air pollutants, an increase in the mean NDVI is associated with an increase in asthma prevalence, with a greater rate of change. In Figure 1A, the rates of asthma prevalence in ages 0 to 19 years are shown on a spectrum of dark blue (0 prevalence) to dark red (highest prevalence) for the years 2005 (left panel), 2011 (middle panel), and 2017 (right panel). In Figure 1B   The lines on the graph represent the temporal trends of the series of data (broken blue lines represent the asthma prevalence rates and the green lines show the mean NDVI) over the three time periods of 2005, 2011, and 2017 along the horizontal axis (X-axis); the Y1-axis on the left is the asthma prevalence rate (blue), and the Y2-axis on the right is the mean NDVI (green).
Selected India states and union territories (SUTs) of the present study: (1) Andhra Pradesh, (2)   As expected, a strong correlation exists between PM 2.5 and PM 10 , while there were weak to moderate correlations among the other air pollutants in this study (Supplementary Materials, Table S1). The maximum temperature (temp) levels were positively but weakly correlated with PM 2.5 , PM 10 , and NO 2 . The distributions of the NDVI, asthma rates, and particulate matter (PM 2.5 and PM 10 ) show a moderate negative correlation between the percentage of exposure to particulate matter (PM 2.5 and PM 10 ) and the mean NDVI ( Figure 3). Compared to the distribution of particulate matter, many states with exposure to gaseous pollutants (SO 2    The NDVI was associated with increased asthma prevalence rates in all three years (2005, 2011, and 2017) even after adjusting for individual pollutants (Supplementary Materials, Table S2). However, in the models adjusted for O 3 , the NDVI had no association with the asthma prevalence rates in 2011 and 2017.
We also analyzed all three years in one model. The univariate GEE models of high particulate matter, SO 2 , NO 2 , and O 3 demonstrated a strong negative association with the prevalence of asthma, with only high PM 10 showing a positive association ( Table 2). The evidence of the association of the NDVI and air pollutants with asthma prevalence remained the same when adjusting for the weather parameters, socio-demographic factors such as the SPI, and population density, except for NO 2 in this study. The NDVI, after adjustment for high levels of PM 2.5 , was positively associated with asthma prevalence (β = 0.144; 95% CI = 0.10, 0.186; p < 0.0001). In contrast, there was no association between the NDVI and asthma prevalence after adjustment for PM 10 at high levels (β = 0.035; 95% CI = −0.006, 0.076; p = 0.096). The NDVI in the models, with high levels of both gaseous air pollutants, SO 2 , and NO 2 , was strongly positively associated with asthma prevalence ((β = 0.12; 95% CI = 0.08, 0.16; p < 0.0001) (β = 0.09; 95% CI = 0.05, 0.13; p < 0.0001)). In the model including high O 3 levels, the NDVI shows a strong negative association with asthma prevalence (β = 0.19; 95% CI = 0.26, 0.11; p < 0.0001).
The strata-specific effects of low vs. high air pollution on the relationship between the NDVI and asthma prevalence are presented in Figure 4. In all cases except SO 2 , there was an interaction between the air pollutants and the NDVI in the GEE models ( Table 3). The largest difference in the slopes can be observed in the PM 2.5 and NO 2 plots, where in the states with low concentrations of air pollutants, an increase in the mean NDVI is associated with an increase in asthma prevalence, with a greater rate of change.       Data presented for selected Indian states and union territories of the present study: In Figure 1A, the rates of asthma prevalence in ages 0 to 19 years are shown on a spectrum of dark blue (0 prevalence) to dark red (highest prevalence) for the years 2005 (left panel), 2011 (middle panel), and 2017 (right panel). In Figure 1B

Discussion
Our findings highlight that asthma prevalence was highest in the northeastern regions, followed by some parts of the eastern coastal plains, while semi-arid areas show lowprevalence estimates. As expected, the northeastern and east coast regions exhibited the highest levels of greenspaces, while the hot arid regions and the snow-capped regions exhibited the lowest levels of greenspaces. The present ecological study is the first statewide study to show that greenspace is associated with childhood asthma prevalence in India. Our results suggest that greenspaces are positively associated with asthma prevalence in children, though this is not uniform across the geographical regions of India. Although this study enabled us to estimate asthma prevalence trends across the three periods (2005, 2011, and 2017), the established time trend and effect estimates indicate shifts in these trends over those years. Although no studies have assessed greenspace and childhood asthma in India, several studies in other regions have evaluated the effect of greenspace and asthma but reported inconsistent associations. Studies in Spain, New York, and Lithuania have observed that children (9-12 years, 7 years, and 4-6 years) residing close to greenspaces had an increased prevalence of asthma [28,29,32]. In contrast, studies from Cincinnati, Texas, the European region, and Canada have found no such association [56][57][58], and others from Spain, Australia, and Mexico have observed protective effects [27,59,60]. South Asian countries lack extensive studies on greenness and asthma to compare with our findings.
There are several hypotheses to explain these differences in the observations between greenspaces and their association with asthma. The outcomes, whether adverse (promoting asthma) or protective against asthma, depend on many co-factors. The timing of exposure to greenspaces may be a factor, that is, early-life exposure may have different outcomes than late-life exposure [31,33]. The type of greenspace seems to be relevant; while residences near forest greenspaces are protective, when they are near parks, an increase in the risk of asthma was observed [28]. Several studies have observed that parks and urban spaces may use exotic and non-native trees, which may increase exposure to pollens that have higher allergenicity [29,61]. Additionally, unlike forest greenspaces, there is a higher exposure to pesticides and fertilizers in urban greenspaces [62], increasing the risk of asthma. In some greenspaces, greater exposure to fungal spores increases the risk for asthma [63,64]. Protection from asthma has been observed due to improved air quality [19,22] and improved biodiversity, both at the macrobiota and the microbiota levels [62,63]. Decreased biodiversity has been associated with increased immune system dysfunction [65][66][67]. Higher greenness was associated with decreased air pollution, especially particulate matter, which offered protection in situations of heavy traffic pollution [59] but not in moderate or low traffic pollution. Studies have observed that children residing near a greater density of greenness are engaged in greater physical activity, while children in urban homes with less greenness lead more sedentary lifestyles and are obese, both of which are associated with a greater risk of asthma [68]. Differences in host responses due to various characteristics of the urban built-up environment, especially grey surfaces (which include industries, transport services, and the urban fabric) [69], indoor and outdoor environments, and differences in climate and geography, are all known to influence the complex interaction between greenspaces and asthma. In addition, residents of areas with higher greenness have been observed to have less stress, which impacts the psycho-neuroimmune and hypersensitive reactions to allergens [70][71][72][73]. The differences in the observations may also be due to the variable definitions used to diagnose asthma across studies and variable tools used to quantify greenspaces.
Only a few studies have evaluated the impact of greenspaces and air pollution on asthma, and the evidence is mixed [31,57]. Wide variations have been observed regarding the burden of air pollution, asthma prevalence, and greenspaces over time in different states in diverse countries such as India. When analyzed, a strong positive association was shown between asthma prevalence and greenspaces at high levels of PM 2.5 , PM 10 , NO 2 , and SO 2 , and a strong negative association was shown with O 3 . By examining interactions over time, our study observed the influence of greenspaces and air pollutants (PM 2.5 , PM 10 , SO 2 , NO 2 , and O 3 ) on changes in childhood asthma prevalence at different concentration levels of air pollutants. In Australia, greater quantities of greenspace may buffer the impacts of heavy traffic exposure on childhood asthma [59]. Children living in areas with greenspace coverage of more than 40% had a much lower prevalence of asthma than children living in areas with less than 20% greenspace coverage [59]. An eight-year observational study in European (Spain, Germany, and Sweden) birth cohorts showed no associations among asthma, outdoor green environments, and exposure to NO 2 [58]. The direction of associations among greenspaces, air pollution, and asthma are not always uniform, especially at different time points. In the US (New York), an increase in greenspace was associated with a 29% lower prevalence of asthma in the areas of increased proximity to air pollution sources [27], but a 17% increase in asthma prevalence was observed in areas with high traffic volumes (>1000 vehicles daily) [29]. Similar observations were made by Dadvand et al. in Spain on asthma prevalence, air pollution, and the NDVI at two time points (2012, 2014), and contradictory results have been found [28,31,33]. In another European study (Lithuania), higher greenness levels increased the risk of asthma prevalence when adjusting for PM 2.5 and NO 2 [32]. These differences in the direction of association in the same cities (New York, Spain) in two different time periods could be related to differences in the age range of the participants included in the two studies, the diverse analytical sample, the type of study design, the definition of asthma, changes in the quality and quantity of the air pollutants, or the vegetation in the city, similar to other study observations [28,29,32,56,74].
Several potential limitations are important to discuss in our nationwide study. First, the environmental exposure data across the states over all the time points were incomplete due to various reasons, such as, currently, there is a total of 703 air pollutant monitoring stations across all the states compared to 630 stations in the year 2005, while 3000 stations are expected to be established across Indian cities and towns by 2022 [75]. Nevertheless, we managed to retrieve additional data from national and state websites, reports of CPCB, and other studies, leading to uncertainties in the data quality. Still, we used a robust and efficient GEE model to obtain unbiased estimations of the average population. Second, our study presents the findings for the annual average estimates rather than the monthly or daily data, and we might have lost the trends of the short-term effects of environmental exposures. Third, the focus of our study was exploratory, so graphically presenting the differences in the results overcomes this. In addition, using the NDVI tool for estimating greenspaces would have captured some differences in the rural and urban characteristics; however, we would have missed some relative differences in the suburban and urban areas. The NDVI does not account for the vegetation type (denser vs. scarce; natural vs. artificial), and hence, the association may not be detailed; comprehensive assessments can combine different greenspace estimation methods in future research. Studying sourcespecific roles of each air pollutant [76,77] and their composition was beyond our scope; still, they give a better understanding of the temporal variations along with their distributions across regions.

Conclusions
In India, greenspace varies over a wide range across states and is associated with increased asthma prevalence, but the association does depend on varying air pollutants in different states. Government initiatives are contributing to green growth, and the smart city initiatives developed in 2017 will continue to further boost the sustained development in urban cities. This is the first study in India, and our study findings contribute to the growing literature on greenspaces and their effects on health under variable environmental factors.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/ijerph192215273/s1, Table S1: Correlations and p-values between the air pollutants and meteorological variables; Table S2

Data Availability Statement:
The data presented in this study are available upon request from the corresponding author.