Trends in seasonal pollen and asthma-related morbidity among adults and children in a U.S. high-density urban center, 2001-2020

Abstract Objective To analyze the long-term trends in pollen counts and asthma-related emergency department visits (AREDV) in adult and pediatric populations in the Bronx. Methods Daily values of adult and pediatric AREDV were retrospectively obtained from three major Bronx hospitals using ICD-10 codes and pollen counts were obtained from the Armonk station from 2001–2020. Wilcoxon Ranked Sum was applied to compare median values, while Spearman correlation was employed to examine the association between these variables, for both decades and each season. Results The median value of pediatric AREDV increased by 200% from the 1st to 2nd decade (p < 0.001) and AREDV peak shifted from predominantly the spring season in the 1st decade to the fall and winter seasons in the 2nd decade. Seasonal patterns were consistent over 20 years with summer AREDV lower than all other seasons (9 vs. 17 per day) (p < 0.001). Spring tree pollen peaks were correlated with AREDV peaks (rho = 0.34) (p < 0.001). Tree pollen exceeding 100 grains/m3 corresponded to a median of 19.0 AREDVs while all other tree pollen (0 − 99 grains/m3) corresponded to a median of 15.0 AREDVs (p < 0.001). AREDVs sharply declined in 2020, coinciding with the emergence of COVID-19. Conclusions Pollen and AREDVs peak earlier in the spring and are more strongly interconnected, while asthma rates among children are rapidly rising, particularly in the fall and winter. These findings can advise targeted awareness campaigns for better management of asthma related morbidity.


Introduction
Asthma is a chronic respiratory disease that is increasing in prevalence and morbidity across the world (1,2). In 2020 alone, the National Hospital Ambulatory Medical Care Survey identified 1,600,000 emergency department visits with asthma as the primary diagnosis (3) and the Centers for Disease Control and Prevention estimated that the prevalence of asthma in the US was 8.4% for adults (≥18) and 5.8% for children (<18) (4). In particular, among all New York State (NYS) counties, the Bronx borough of New York City (NYC) has the highest asthma rate (5), nearly double the asthma related emergency department visits (AREDV) as that of the second highest county (6), and the highest age adjusted asthma death rate by far (7). In fact, the Bronx has one of the highest levels of asthma prevalence (13.0%) compared to the rest of the United States (7.7%) (6). Therefore, identifying asthma related longitudinal trends and risk factors in the Bronx is critical to minimizing its local burden.
Of note, prior studies have demonstrated that spring tree pollen levels are associated with AREDV counts in various urban centers (8)(9)(10)(11)(12)(13)(14)(15), including the Bronx (16)(17)(18)(19). This is especially important in light of the well-studied impact of asthma surveillance on AREDV, asthma-related hospitalizations (ARH), and asthma prevalence and morbidity across the general population (20)(21)(22)(23)(24)(25)(26). However, how both pollen levels and AREDVs have changed in high density urban centers, such as the Bronx, over an extended period of time is yet unknown (27). With the percentage of the global population living in urban areas projecting to increase from 54% in 2015 to 60% in 2030 and to 66% in 2050, it is necessary to broaden the knowledge of asthma trends in these high-density centers (28).
Our work expands on the existing research by collecting and analyzing tree, pollen, grass and AREDV data from the Bronx over a two-decade timeframe (2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019)(2020). The objectives of this paper are to (1) examine trends in pollen levels and AREDV between and among different age groups, (2) describe AREDV and pollen related trends that have changed between the two decades, and (3) statistically assess AREDV incidence and pollen variable associations between the two decades. The results of this study can be used to adjust for the timing and target population of government sponsored pollen peak warnings and, ultimately, minimize AREDV, and, ultimately, minimize AREDV and improve clinical outcomes in the Bronx and other high-density urban centers.

Participants and background
Study participants in this retrospective, population-based study included both adult and pediatric (i.e. individuals < 18 years old) males and females and amounted to a total of 119,302 adult and pediatric AREDVs. All patients presented for an AREDV at three major Bronx hospitals (Montefiore Medical Center's Moses and Weiler divisions and The Children's Hospital at Montefiore) during the specified study period from 1 January 2001 to 31 December 2020. Only the daily number of AREDVs, but no personally identifiable information, were collected and used for our statical analyses. Due to patient record de-identification, this study is considered exempt by the Einstein Institutional Review Board.

Asthma-related emergency department visits
Daily AREDV from 2001-2020 were determined through a retrospective search via our institution's Clinical Looking Glass (CLG) software, the analytical software tool used in conjunction with the electronic medical record at Montefiore Medical Center. Both adult and pediatric AREDVs were identified by the primary diagnosis of asthma (ICD-9 code 493, ICD-10 code J45) at time of discharge. For individual participants with more than one AREDV during a particular year, all visits were counted for that given year. Given the seasonal nature of AREDV trends, they were analyzed for each season separately. Date cutoffs for each season were determined according to meteorological, rather than astronomical, definitions, to better capture climate patterns. Winter was defined as 1st December to 28th (29th during leap years) February; spring: 1st March to 30th May; summer: 1st June to 31st October; fall: 1st September to 30th November. The AREDV numbers were aggregated into two decade-long time periods (2001-2010 and 2011-2020) in order to observe longitudinal trends between the two decades.

Tree, grass, and weed pollen counts
Daily tree, grass, and weed pollen counts were obtained from Fordham University's Louis Calder Center Aero-Allergen Monitoring station in Armonk, NY from 1 January 2001 to 31 December 2020. This site is 23 miles north of the Montefiore Medical Center and 24 miles north of Weiler Hospital. The Louis Calder station is the closest certified pollen counting site to the Bronx that has consistently reported daily pollen count values for the entire two-decade time frame. The center's certified technicians collect airborne pollen with a Burkard volumetric spore trap for microscopic analysis. At a minimum, technicians counted one horizontal traverse of the daily slide to determine the pollen particles per cubic meter. There were days when pollen levels were not collected or assumed to be zero; to avoid accuracy issues with arbitrary data interpolation and consistent with previous work (19), those days were excluded from our analysis. Also, to mitigate the effect of missing data on our study, we focused on peaks, seasonal trends, and upper quartile data. For our study, the pollen counts were further differentiated by tree, grass, and weed sub types.

Statistical methods
Total AREDV, adult AREDV, and pediatric AREDV as well as tree, grass, and weed pollen counts were calculated and compared for each season and decade. Wilcoxon Ranked Sum test was used to compare the median values for AREDV between decades for each season, except winter for which pollen counts are typically zero. Line graphs and box plots were used to visually present distribution of AREDVs and pollen counts by decade and age group. Spearman's correlation was used to examine the magnitude of the associations between individual tree, grass, and weed pollen counts and AREDV between decades and separately over the twenty years for each season. Findings with two-sided p values of 0.05 were statistically significant. Analyses were conducted using SPSS Version 28 (Armonk, NY) and SAS 9.4 (Cary, NC). MATLAB 2021a was used to determine yearly peak AREDV levels and generate graphs.

Comparing AREDV: 2001-2010 vs. 2011-2020
The median value of total AREDVs did not see a statistically significant change between the decades (15-16, p = 0.08) (Figure 1a), but when stratified by adult vs pediatric, pediatric AREDVs was shown to have increased from 3 to 6 (p < 0.001) (Figures 1b and 2a), while adult AREDV decreased from 10 to 9 (p < 0.001) (Table 1, Figure 1c). In the pediatric population, the maximum daily AREDV, when graphed annually as AREDV vs. day of year, occurred in the spring for 70% of the years and in the winter for 30% of the years during the first decade, but shifted to 30% in the springtime and 70% in the fall and winter during the second decade. The large increase in fall and winter pediatric AREDV during the second decade compared to the first can be seen in Figure 1b, which displays the weekly median of pediatric AREDV from 2001-2020. In the adult population, the maximum daily AREDV occurred 50% time in the spring and 50% in the winter during the first decade and 60% in the spring and 40% in the winter during the second decade. The values for adult AREDV were consistently lower during the winter, spring, and fall seasons in the second decade compared to the first (Figure 1c).

Seasonal AREDV: 2001-2020
Total AREDV values displayed cyclic behavior each year from 2001-2020 ( Figure 3). AREDVs were low and constant during the summer months, peaked in the early fall, peaked again in the early winter before peaking with the sharpest peak (greatest slope) in the spring. After the spring peak, AREDV values decreased steadily back to the low and constant values in the summer. The graph for the year 2009 was expanded to more clearly illustrate these three peaks ( Figure 4). This year was chosen as it had the clear visibility of all three peaks and sufficiently high number of total AREDVs and clear visibility of all three peaks. Notably, 2020 had consistently lower and 2008 had consistently higher total AREDV values, compared to all other years ( Figure 3). Finally, there was a decreased median value of total daily AREDV, adult daily AREDV, and pediatric daily AREDV in the summer compared to the Winter, Spring and Fall seasons (Summer: 9 total AREDV per day, Fall: 16 total AREDV per day, Spring: 17 total AREDV per day, Winter: 18 total AREDV per day; p < 0.0001) (Figure 2b). The tallest and most important AREDV peak in association with pollen counts is the Spring peak (18). While it Table 1. seasonal total, adult, and pediatric arEDV median values from 2001-2020 collected using the ClG software from montefiore medical Center's moses and Weiler hospitals in the Bronx, nY.   occurred at approximately the same time in both decades as decade-long medians (Figure 1a), it more often occurred earlier in the second decade as compared to the first on an annual basis (Supplemental Figures 1 and 2). In fact, during the first decade, there were 2 instances of AREDV peaks in April and 8 instances of peak AREDV in May, including two as late as the third and fourth weeks of May. In the second decade, there were three peaks in March, 1 in April, and 6 in May, all of which occurred during the first and second weeks (Supplemental Figures  1 and 2).

Spring tree pollen: 2001-2010 vs. 2011-2020
We recorded 1552 and 1406 tree pollen values from the first and second decades, respectively. The medians for both the upper quartile and overall spring tree pollen count decreased from 2001-2010 to 2011-2020 (458.00-155.0 p < 0.0001, 130.0-33.5 p < 0.0001, respectively). Tree pollen peaked every late-April to early-May for the first decade and peaked every mid-April to early-May for the second decade. The earliest incidence of peak pollen was at the tail end of the third week of April during the first decade, and 5 days earlier in the second. For the spring, summer, and fall seasons, there was a statistically significant decrease in the median upper quartile tree pollen counts from the first to second decade (Table 2). We previously reported that tree pollen counts >100 grains/m 3 corresponded with an increase in AREDV from the 2001-2008 timeframe (19). Throughout all seasons and both decades, tree pollen counts <100 grains/m 3 corresponded to a median of 15.0 AREDVs (± SD 8.43) while counts exceeding 100 grains/m 3 corresponded to a median of 19.0 AREDVs (± SD 12.08, n = 497), a 26.7% increase in AREDVs ( Figure 5).

Summer and fall grass and weed pollen: 2001-2010 vs. 2011-2020
Grass and weed pollen types both saw a decrease in its median upper quartile value for all reported seasons between the decades (p < 0.001) ( Table 2). Specifically, grass pollen decreased from 4 grains/cm 3 to 2 grains/cm 3 during the summer when it predominantly appears (p < 0.001). Weed pollen also decreased from 18 grains/cm 3 to 5 grains/cm 3 during the falls when it predominantly appears (p < 0.001).

Spearman analysis
Tree Pollen had a statistically significant positive association with AREDV over the two decades (Spearman rho = 0.22 p < 0.0001), while Grass Pollen did not have a statistically significant correlation, and Weed pollen had a statistically significant negative association with  AREDV (Spearman rho = −0.14 p < 0.0001). During the spring season over both decades, tree, grass, and weed pollen all had statistically significant positive correlations with AREDV (Supplemental Table 1). The association between spring tree pollen and AREDV strengthened from the first (Spearman rho = 0.11) to second decade (Spearman rho = 0.34) (p < 0.0001), as did weed pollen (Spearman rho = 0.17 p < 0.0001). During the summer of both decades, grass pollen had a significant positive correlation with AREDV, and during the fall time of both decades, weed pollen had a statistically significant positive association with AREDV (Supplemental Table 1).

Discussion
Our study aimed to assess intra and inter-decade asthma related trends in the Bronx, NYC for the first two decades of the twenty first century. As with previous studies, we focused on epidemiologic AREDV trends between age groups, seasons, and pollen types and levels. Unlike in previous studies, our report focuses on a high-density urban center for a prolonged, two-decade timeframe. Notably, while total daily AREDVs did not change significantly over the last two decades in the Bronx, pediatric daily AREDVs were shown to have doubled. The ever-increasing rates of asthma have been well documented before and are thought to be due to a combination of increased urbanization, increased rates of obesity, and the so-called 'hygiene theory' (29). Interestingly, however, we show that such is not the case for adult AREDV suggesting that current surveillance efforts within the adult population may be sufficient.
There is a well-documented annual cycle of AREDVs (19) that our findings are consistent with; they are lowest in summer and highest in the spring. The low summertime AREDVs can be explained by low tree pollen levels, warmer weathers causing decreased spread of respiratory viruses, and stronger medication adherence that is demonstrated by a medication sales spike in early May (11,30,31). However, pediatric AREDV peaks are now occurring in the fall and winter as well suggesting that this population is especially vulnerable in the context of symptomatic asthma. While the reason for this is unknown, it might be related to targeted awareness campaigns commonly occurring in the spring, such as National Asthma and Allergy Awareness Month related campaigns (32). If this were the case, it would suggest that an emphasis on fall and winter asthma surveillance and advisories could improve clinical outcomes among pediatric asthma patients in the Bronx. Several other factors might be contributing to the increased pediatric AREDV in the fall and wintertime, including increased spread of respiratory viruses on school opening, increased indoor aeroallergen exposure, and more extreme temperature changes (30,33). Indeed, Teach et. al demonstrated that pre-seasonal treatment with a corticosteroid boost or omalizumab was shown to significantly decrease fall asthma exacerbations (34). Similar medication adherence protocols could potentially guide clinical practice and minimize AREDV during the fall and wintertime for pediatric populations.
Interestingly, 2008 and 2020 stood out for their diverging AREDV counts. 2020 had remarkably lower AREDVs, possibly related to technical COVID related reasons, such as assuming COVID-19 as the primary diagnosis for all upper respiratory tract complaints in the early part of the pandemic (35)(36)(37). Alternatively, COVID related public health measures, such as mask wearing, social distancing, and quarantining, may have contributed to the decline of AREDV in the year 2020. This can be contrasted with reports of worsening allergic rhinitis symptoms (38) and increased positivity rates for common indoor inhalant antigens (39) during the COVID lockdowns. Indeed, decreased hospital admissions for asthma during the early COVID lockdowns in Scotland and Wales (40), the pre-pandemic voluntary COVID protective measures in Hong Kong (41), and the early NYC lockdowns at the Children's Hospital at Montefiore (CHAM) (42) have previously been reported. Further research is encouraged to determine if certain measures can be co-opted for the alleviation of the asthma related burden going forward. 2008 had remarkably higher AREDV, possibly related to the co-occurring financial crisis which impacted NYC significantly and led to gaps in health insurance coverage associated with lower asthma control (43).
Climate change has likely had a large impact on observed changes to asthma related variables. We found that spring tree pollen and AREDV peaks have been occurring slightly earlier in the year. This is in line with the oft cited phenomenon of 'season creep,' whereby spring has consistently begun earlier over the last number of decades in the Northern Hemisphere (44). Interestingly, unlike in an earlier study of multiple large North American cities (45), we found that tree, weed, and grass pollen counts have all been decreasing over the last two decades for all reported seasons in the Bronx. While it is unknown if our findings in the Bronx are paradigmatic of a broad climatic shift, asthma surveillance efforts may need to be fine-tuned to adequately adjust for the earlier AREDV spring peak incidence. Currently, the NYC Department of Health sends out city wide health advisories during the early spring to minimize the impact of the springtime pollen season. These advisories recommend that patients take appropriate prophylactic and therapeutic asthma medications and monitor daily pollen forecasts to regulate outdoor activity. If surveillance warnings are deployed earlier, the general population may be able to purchase asthma medications earlier, better respond to earlier AREDV springtime peaks, and ultimately manage their asthma conditions more appropriately.
The association between environmental pollen levels and AREDV, especially for spring tree pollen, has previously been reported (18). We report that tree pollen counts >100 grains/m 3 correspond to a 35.4% increase in AREDV. On Spearman analysis, we found mostly weak, but statistically significant associations between pollen levels and AREDV, with the strongest being for spring tree pollen. The strength of this association appears to have more than tripled between the decades on Spearman analysis.
Our findings are important to combat the rising rates of asthma, especially in urban centers where the global population is projected to increasingly reside (28) and in the context of rapid climate change. For example, our team developed an application for alerting asthma patients to rise in pollen levels early in asthma seasons. The changes that have occurred in our local environment over the last two decades are vitally important for targeted use of this or similar awareness campaigns.
Several limitations were present in this study. First was the method of pollen data collection. The Armonk station is one of the only daily pollen collecting sites in NYC and is close in proximity to the Bronx, but its findings may not be completely reflective of environmental circumstances throughout the borough. Also, there were many days when pollen levels were not collected, and those days were excluded from analysis. AREDVs were collected from our institution's electronic medical record, which may have missed some AREDV standardly assigned an alternative primary diagnosis at discharge. Additionally, we stratified our patient population by age alone, without further consideration of race, gender, or other social factors. Further, our choice to contrast two juxtaposed decades was statistically arbitrary and likely missed intra-decade trends. As well, the adoption of expanded Medicaid and primary care coverage for emergency department visits by NYS in 2014 may have served as an effect modifier for all reported trends. Interestingly, during the first year of Medicaid expansion, emergency department visits increased proportionally between states that did and that did not adopt the updated terms of coverage (46), possibly mitigating its impact on our findings. Finally, our findings concern a single high density urban center and so they may not be generalizable to other metropolitan or rural areas throughout the globe.

Conclusions/key findings
Ultimately, in high density urban centers, asthma surveillance is key to preventing increasing rates of asthma related morbidity. In this study, we show that several asthma related variables have changes over the last two decades in the Bronx, New York City. Pollen and AREDV occur earlier and are more strongly interconnected, and asthma rates among children are rapidly rising. These findings can advise targeted awareness campaigns for better management of asthma related morbidity.
Mani and Jai Shahani organized pollen count data from the Armonk Station, while Dr. Sunit Jariwala provided the abstracted asthma data from the electronic medical records. Kyle Mani, Raphael Miller, and Dr. Sunit Jariwala all drafted the manuscript. All authors read and approved the final manuscript.

Declaration of interest
The authors report no conflicts of interest. The authors alone are responsible for the content and writing of this article.

Funding source
None.