Seasonal extreme temperatures and short-term fine particulate matter increases child respiratory hospitalizations in a sparsely populated region of the intermountain western United States

Background Western Montana, USA, experiences complex air pollution patterns with predominant exposure sources from summer wildfire smoke and winter wood smoke. In addition, climate change related temperatures events are becoming more extreme and expected to contribute to increases in hospital admissions for a range of health outcomes. Few studies have evaluated these exposures (air pollution and temperature) that often occur simultaneously and may act synergistically on health. Methods We explored short-term exposure to air pollution on childhood respiratory health outcomes and how extreme temperature or seasonal period modify the risk of air pollution-associated hospitalizations. The main outcome measure included all respiratory-related hospital admissions for three categories: asthma, lower respiratory tract infections (LRTI), and upper respiratory tract infections (URTI) across western Montana for all individuals aged 0–17 from 2017–2020. We used a time-stratified, case-crossover analysis and distributed lag models to identify sensitive exposure windows of fine particulate matter (PM2.5) lagged from 0 (same-day) to 15 prior-days modified by temperature or season. Results Short-term exposure increases of 1 μg/m3 in PM2.5 were associated with elevated odds of all three respiratory hospital admission categories. PM2.5 was associated with the largest increased odds of hospitalizations for asthma at lag 7–13 days [1.87(1.17–2.97)], for LRTI at lag 6–12 days [2.18(1.20–3.97)], and for URTI at a cumulative lag of 13 days [1.29(1.07–1.57)]. The impact of PM2.5 varied by temperature and season for each respiratory outcome scenario. For asthma, PM2.5 was associated most strongly during colder temperatures [3.11(1.40–6.89)] and the winter season [3.26(1.07–9.95)]. Also in colder temperatures, PM2.5 was associated with increased odds of LRTI hospitalization [2.61(1.15–5.94)], but no seasonal effect was observed. Finally, 13 days of cumulative PM2.5 prior to admissions date was associated with the greatest increased odds of URTI hospitalization during summer days [3.35(1.85–6.04)] and hotter temperatures [1.71(1.31–2.22)]. Conclusions Children’s respiratory-related hospital admissions were associated with short-term exposure to PM2.5. PM2.5 associations with asthma and LRTI hospitalizations were strongest during cold periods, whereas associations with URTI were largest during hot periods. Classification: environmental public health, fine particulate matter air pollution, respiratory infections


Classi cation
: environmental public health, ne particulate matter air pollution, respiratory infections BACKGROUND Less than 1% of the world experiences daily concentrations of ne particulate matter air pollution (< 2.5 µm in aerodynamic diameter; PM 2.5 ) that is less than the recommended daily safe levels (Yu et al. 2023).
The daily safe thresholds and related policies have been set based on rigorously designed epidemiological cohort and time series studies (e.g., Pope et al. 1995, Pope et al. 2009), con rmed through rigorous re-analysis and subsequent studies over the last several decades (Krewski et Fajersztajn et al. 2013).Associative impact studies overwhelmingly corroborate a correlative link between respiratory health outcomes and exposure to air pollutants (e.g., US EPA 2009, Kim and Kabir 2015, Liu et al. 2017), as well as delayed exposures through both short-term (i.e., 1 day-1 month; e.g., Galan et al. 2003, Wong et al. 2008, Xie et al. 2019) or long-term timeframes (1 month-1 year; e.g., Orr et al. 2020, Landguth et al. 2020).Inhaling PM 2.5 can produce in ammation and oxidation stress, triggering cellular damage and increasing the risk of respiratory disease (Black et al. 2017).Ambient PM 2.5 air pollution, particularly in urban and higher-income country settings, has been signi cantly reduced over the last 40 years (McClure and Jaffe 2018, Ford et al. 2018).However, in some areas of the world, and speci cally for our rural and intermountain study setting of Montana, USA, exposure to PM 2.5 continues to increase due to residential wood combustion for heat in the winter season and wild re smoke events during the summer (or wild re) season.In the 2022 State of the Air report (American Lung Association 2022), Montana received failing grades for eight counties based on the number of unhealthy and hazardous air-quality days due to severe wild res and use of residential wood stoves.Regarding wood stoves, Montana ranks second in the USA in the proportion of households that heat with wood fuel (7.4% compared to 1.7% in the USA; ACS 2020).Chemical Mass Balance source apportionment studies have shown that residential wood stoves are the largest source of ambient PM 2.5 during the winter months (55.5-82%; Ward 2009, Ward and Lange 2010, Ward 2016).Studies evaluating the health impacts associated with residential sources of PM 2.5 are limited and often suffer from challenges related to sparse populations and uncertain generalizability (Noonan and Balmes 2010, Sigsgaard et al. 2014).
The second air quality threat in the mountain west region is smoke from nearby and distant wild res, a PM 2.5 source that is projected to worsen with climate change (Ford et al. 2018, O'Dell et al. 2019).A growing body of literature is focused on the health effects of PM 2.5 speci cally derived from wild re smoke.Health impacts from wild re smoke exposures range from irritation of the eyes and respiratory tract to respiratory morbidity, with growing evidence supporting an association with all-cause mortality (Reid et al. 2016).In particular, hospitalizations and emergency department visits related to respiratory infections and preexisting conditions, such as asthma and COPD, are consistently elevated during and shortly following wild re events (Reid et al. 2016, Cascio 2018).Several factors complicate the evaluation of wild re exposures and healthcare usage on health outcomes.These include uncertainty in lag effects and potential non-linear response curves that may indicate lower healthcare utilization during extremely high wild re smoke events, perhaps mediated through behavior changes that are not at play in urban settings where the moderately elevated PM exposures are less recognizable or notable by community members (Gould et al. 2023).
In parallel, global exposure to extreme temperatures has grown and is expected to worsen with climate change.Extreme temperature events, both cold and hot, are known to be associated with excess mortality and increased hospital admissions for a range of health outcomes (USGCRP 2016, Gasparrini et al. 2017).Hotter days in the summer will cause increased levels of illness and death by compromising the body's ability to regulate its temperature, or by exacerbating health problems.Cold temperatures in the winter can cause blood vessels to constrict, heightening cardiovascular issues, and irritating the airways triggering respiratory problems and lower immunity.Speci cally focusing on the respiratory-related health categories in this study, literature for temperature-associated respiratory health effects are mixed with respect to hot versus cold temperature extremes and depending on the respiratory categories studied.For example, it is well known that cold (and dry) conditions can increase the survival rate of in uenza viruses and enhance viral spread (e.g., Lowen et al. 2007).A recent review concluded that both extreme heat and cold could signi cantly increase the risk of asthma (Han et al. 2022).Sheerens et al. (2022) showed that higher temperatures may worsen dyspnea, while colder temperature may trigger cough and phlegm symptoms among COPD patients.
Multiple rigorous studies of have observed impacts on health of PM 2.5 and temperature, but have considered increases in these exposures separately; however, these exposures often occur simultaneously and may act synergistically on health.The potential for interactive effects based on these two climaterelevant factors is important as current population risk estimates and corresponding policy recommendations are based largely on epidemiological studies quantifying the effects of PM 2.5 and temperature considered in isolation.A systematic review of several studies, almost entirely in urban populations, indicate su cient ndings of moderate quality to support synergistic effects for temperature and air pollution (Anenberg et al. 2020), although such evidence for pediatric respiratory outcomes is extremely limited (Winquist et al. 2014).Assessment of these questions in rural communities also is limited, but a recent case crossover study in California for all age cardiorespiratory hospitalization showed strong evidence for a synergistic effect between wild re speci c PM 2.5 and extreme heat (Chen et al. 2023).Although the authors did not speci cally evaluate rurality, they found weaker interactions between temperature and air pollution in counties with higher education attainment, health insurance coverage, income, and automobile ownership, possibly due to greater capacity to reduce harmful exposures in individuals of higher socioeconomic status.
For the study presented here, we evaluated associations between short-term or delayed ne particulate matter (PM 2.5 ) on three respiratory health outcomes assessed at the individual level.We additionally assessed modi cation of these associations by temperature and season.We focused on a rural and sparsely populated service area in western Montana, USA, from 2017-2021.This area of the inter-Rocky Mountains is experiencing more frequent exceedance of daily air quality standards in the summer due to increases in wild re smoke events with the largest source of ambient PM 2.5 in the winter due to residential wood stoves.This area of the northern hemisphere also has more winter cold months than summer warm months, though average annual temperatures are rising.

Study Area, Population, and Respiratory Health Outcomes
The study protocol was approved by the Institutional Review Board (IRB) at the University of Montana.
Initial study approval was obtained by the University of Montana-Missoula Institutional Review Board on 6 July 2021 (#97-21).Health data were previously collected administrative data; thus informed consent requirements did not apply.

Study area
For our study, we are focused on western Montana, USA (Fig. 1).The study area covers 45 out of 361 Montana Zip Code Tabulation Areas (ZCTA) across 8 of the 56 counties (Deer Lodge, Granite, Lake, Mineral, Missoula, Powell, Ravalli, and Sanders).The total population within this area was approximately 233,657 in 2020 that includes one small city (Missoula, population total = 73,948) surrounded by sparsely populated areas (US Census Bureau 2020).According to the US Census Bureau's de nition of rurality, this study area is de ned as having 72.3% of the population living in rural areas.For context, the US has 19.3% of the population living in rural areas (Ratcliffe et al. 2016).This region of the inter-Rocky Mountains is experiencing more frequent exceedance of daily air quality standards in the summer months (particularly in July-September; Landguth et al. 2020) due to increases in wild re smoke events.
The largest source of ambient PM 2.5 in the winter is due to residential wood stoves (Ward 2009).At this northern hemisphere latitude (45-49 o N), Montana experiences more winter cold months (3.4  Hospital admissions data: Hospital data were collected from December 2017-September 2021 for one hospital that predominantly serves Missoula County in western Montana, United States, with 14,814 respiratory-related records.These data included nine sources of admissions type: clinic, inpatient, emergency, observation, outreach clinic, preadmission outpatient clinic, professional services, provider clinic, and telemedicine clinic.We removed records with residential addresses that fell outside western Montana, resulting in a reduced spatial sample of 14,071 records.We further right-censored the date range beginning on 13 March 2020 to remove the in uence of the COVID-19 pandemic on result ndings ( nal n = 10,133).Data included individuals aged 0-17 with a respiratory-coded infection (see Case de nitions for health outcomes below and Table 1).A strictly protected health protocol was implemented through data use agreements between the records provider and the University of Montana, where personal identi ers were removed, and residential addresses were geocoded and geomasked.The individual-level data and corresponding spatiotemporal daily PM 2.5 exposure values were used in casecrossover analyses (see Case-crossover design and analysis).

Case de nitions for health outcomes
For this study, cases related to upper respiratory tract infections (URTI), lower respiratory tract infections (LRTI), and asthma were rst identi ed using the International Classi cation of Diseases,10th Revision, Clinical Modi cation diagnosis codes (ICD-10-CM) and sorted following the case de nitions of the Armed Forces Health Surveillance Center (AFHSC 2015) (Table 1).We further identi ed and split case de nitions based on each infection's upper and lower airway occurrences.Records were classi ed by condition when a related diagnosis code of interest was found in the primary diagnosis eld ( rst-listed) or any secondary diagnosis eld (1-8).Records were selected once for each associated category.For example, records with more than one URTI code were only counted once for the URTI category.If a record had codes for URTI, LRTI, and asthma, the record was counted once in each of the three categories.Hospital data in the study period and area are shown as total counts for each respiratory category in Fig. 2, along with average weekly PM 2.5 ..83,J11.89, J21.0, J21.8, J21.9, H65, H66, H66.9 LRTI J20, J20.0, J20.1, J20.2, J20.3, J20.4,J20.5, J20.6, J20.7, J20.8, J20.9, J21, J21.0, J21.1, J21.8, J21.9, J09.X1, J09.X2, A37, A37.00, A22.concentration estimates to explore health outcome impacts of PM 2.5 across spatiotemporal domains speci c to the rural and intermountain western USA.We extracted daily PM 2.5 measurements for each case event address location on date of hospital visit for the case-crossover modeling, along with PM 2.5 at the address location for reference days (see Case-crossover design and analysis).
Delayed PM 2.5 exposure effects: To identify sensitive windows of PM 2.5 exposure and test the effects of short-term and delayed PM 2.5 effects on respiratory health, we investigated the effect of PM 2.5 from 0 to 15 days before the admission date with 3 variations of PM 2.5 lags: (i) single day -single day PM 2.5 lags include just the day previous to the case event, (ii) cumulative days -cumulative days of PM 2.5 include cumulative PM 2.5 for all k days prior to the case event day, for k = 1, …, 15, and (iii) weekly averageweekly average PM 2.5 lags include rolling averages of PM 2.5 over 1-week periods prior to the case event, including 0-6, 1-7, …, 9-15 days prior.

Temperature
We included daily maximum temperature modeled by gridMET (Abatzoglou 2013) extracted to each individual location, date of admission and corresponding reference days for the case-crossover model.
We used the 15th, 50th, and 85th percentiles of temperature (cutoffs for colder = -0.7 0 C, median = 6.2 0 C, and hotter = 20.7 0 C) to estimate the effect of the interaction of PM 2.5 and temperature on the three respiratory outcomes.

Season
In Montana, PM 2.5 levels spike during summer season due to the primary source of wild re smoke and during the winter season due to the primary source of wood smoke (Ward 2009).We therefore included a categorical season predictor that is assumed to be associated with the exposure of interest and potentially also associated with the respiratory health outcomes of interest.We included a northern hemisphere season as a categorical variable that included summer (June-August), fall (September-November), winter (December-February), and spring (March-May).The model also employed an interaction between season and PM 2.5 to allow for differential effects by season of PM 2.5 on the health outcomes.In addition, we included a three-way interaction between season, PM 2.5 , and temperature to assess potential different interactive effects of PM 2.5 and temperature by season.

Statistical modeling
Case-crossover design and analysis Introduced in environmental health studies by Maclure (1991), case-crossover designs compare an individual's (case) exposure immediately prior to or during the de ning case event with that same individual's exposure at different reference times.This method is attractive because it compares individuals with themselves and controls for time invariant confounders and secular trends by design.
Since the seminal Maclure (1991) study, several variations on choosing control days to minimize biases have emerged, and convergence to a time-strati ed case-crossover design has evolved as the recommended approach for minimizing sources of bias (see Wu et al. 2021 for review).
Thus, we evaluated the synergistic effect of temperature extremes speci c to a season and the 3 types of PM 2.5 lag predictors on the risk of each respiratory infection outcome (asthma, LRTI, or URTI) using a time-strati ed case-crossover design widely used in studies of short-term environmental health exposures (e.g., Talbot et al. 2014; Wei et al. 2019).We created case-crossover datasets for each respiratory outcome, with paired case events and either 3 or 4 controls.The case event day was de ned as the date of hospital admission.We then identi ed matched control days as the same weekdays from other weeks of the same month and year in the same geocoded location of residence (i.e., of the same person).
Matching by day of the week controlled for potential confounding by factors that vary within a week (e.g., weekend/weekday differences in hospital admission rates).We selected control days both before and after the case day to minimize bias from long-term time trends in PM 2.5 (Levy et al. 2001).Time-invariant factors between the case event day and each control day, such as age, sex, race, socioeconomic status, and other short timeframe changing health behaviors, were assumed unlikely to change (Lu and Zeger 2007).
For each respiratory binary response (asthma, LRTI, or URTI) and corresponding case-crossover dataset, we applied a conditional logistic model to estimate the odds ratio of the effect of PM 2.5 at 3 temperature levels (colder, median, hotter) and within 4 seasons (Chen et al. 2023).The conditional logistic model accounts for the case/control matching by comparing the exposures of each matched case and control, and calculating a weighted average across all matched sets (Di et al. 2017).We estimated odds ratios and corresponding 95% con dence intervals for relative odds of hospitalization for each 1 µg/m 3 increase in PM 2.5 , within each temperature/season combination, and for each of the 3 respiratory outcomes.An odds ratio larger than one indicates the exposure variable increases odds of hospitalization.Modi cation of the effects of PM 2.5 on each respiratory health outcome by temperature was assessed by including multiplicative interaction terms for PM 2.5 and temperature (colder, median, hotter).We employed a similar approach to assess interactions between PM 2.5 and season.The threeway interaction terms for PM 2.5 xTemperaturexSeason were not included due to model instability.To account for the effects of temperature or season, appropriate linear combinations of coe cients were utilized using the 'biostat3' R package (Tillander et al. 2022).

Distributed lag modeling
Distributed lag modeling is a exible approach that allows one to simultaneously model exposureresponse and lag-response relationships (Gasparrini 2014).Using the case-crossover modeling framework described above, distributed lag models were used to determine sensitive windows of PM 2.5 exposure on risk of respiratory hospital admission.These models used lagged PM 2.5 concentration as the exposure variable along with PM 2.5 xTemperature, and PM 2.5 xSeason interactions.

RESULTS
In summary, we analyzed respiratory hospital admission data for a sparsely populated region in western Montana, USA.During the study period ( Table 3 Modi cation of the effect of PM 2.5 exposure on respiratory health by temperature or season.The largest relative odds ratios (OR) and corresponding 95% con dence intervals (CI) with P-values (Pval) are displayed here to estimate the increase in odds of hospitalization for each 1 µg/m 3 increase in PM 2.5 at the given lagged cumulative days or weekly average for each temperature/season combination, and for each of the 3 respiratory outcomes: (A) asthma, (B) lower respiratory tract infections (LRTI), and (C) upper respiratory tract infections (URTI).For each respiratory health outcome condition, the three models are presented for PM 2.5 , PM

DISCUSSION
We found short-term increases in PM 2.5 air pollution were positively associated with children's respiratory related hospital visits for a patient population in western Montana, USA.These effects were found for categories of respiratory related visits of asthma (peak odds at lag of 7-13 days), LRTI (peak odds at a lag of 6-12 days), and URTI (peak odds after 13 accumulated days).These results are consistent with ndings from past studies.While, in general, consistency of ndings implies an association between increased respiratory risk and increased PM 2.5 , the length of the observed lag effect does vary (as highlighted below).These links between increased respiratory risk and increased short-term PM 2.5 are well established.That PM 2.5 impacts vary by temperature and season is a novel contribution of our study.
We showed that PM 2.5 associations with asthma and LRTI hospitalization are strongest during colder temperatures.By contrast PM 2.5 effects on URTI hospitalization were elevated during hotter temperatures and during the summer season.
. Asthma and PM 2.5 exposure: Numerous studies link air pollution to asthma.Several reviews have highlighted this connection, speci cally for exacerbating existing asthma, but also with an increase of new-onset asthma (Guarnieri et 2004) found a lag of 7-10 days.Our study, using the DLM, places the PM 2.5 associated increased risk in children's asthma events at a weekly average lag of 7-13 days.
Asthma, extreme temperatures, and seasonal effects: Our study indicated the highest risk for asthma hospitalization at PM Limitations: Air pollution case-crossover studies are not without limitations.First, a note on sample size.These data cover 821 days with an average of 12.8 events per day among all three outcomes (1.06, 0.84, and 10.9 events for asthma, LRTI, and URTI, respectively).It has been suggested in simulation studies of pollution effects, that thousands of observation days with an average of tens of events per day are needed (Winquist et al. 2012).We observed instability in some of our estimates, particularly for those models with low sample sizes (LRTI and PM 2.5 xSeason).Second, unmeasured time-variant factors might have provided additional confounding in uence and could possibly impact estimates (Bateson andSchwartz, 1999, 2001).To the degree that these factors occur at the individual level, e.g., immunity or vaccination status, the impact is likely to be negligible given the case-crossover design.Third, error in diagnostic coding is possible.Some cases may not be accurately categorized, and it is possible that such coding errors could be differential with respect to season.Finally, the assessment of exposure could be subject to measurement (and modeling) error, especially in a rural, sparsely populated study area with only a limited number of xed air quality monitors contributing to the estimates of PM 2.5 (Swanson et al. 2022).However, we expect this error would have results in attenuated effect estimates (Wu et al. 2019).

CONCLUSIONS
Western Montana, USA, is a sparsely populated region of the inter-Rocky Mountains with complex air pollution patterns.This region is experiencing more frequent exceedance of daily air quality standards due to increases in wild re smoke events during their summer/wild re season months.However, the region also experiences elevated levels of PM 2.5 during winter months from wood stove use with complex mountain meteorology and inversion effects.Here, we explored short-term PM 2.5 effects on three childhood respiratory health outcomes (asthma, LRTI, and URTI) and how other factors, such as extreme temperature or seasonal period, modify the risk of air pollution-associated hospitalizations.We found associations between elevated PM 2.5 exposures and hospital visits for all respiratory categories.We found interaction effects with extreme temperatures and during high impacted PM 2.5 seasons.We found increased risk for asthma and LRTI associated with elevated levels of PM 2.5 in colder temperatures, while increased risk for URTI associated with elevated levels of PM 2.5 in hotter temperatures or the summer season.Communities in the western US will experience increases in morbidity and mortality related to higher frequency of extreme temperature and wild re events (Whitlock et al. 2017).At present, policy and public health messaging related to air pollution and extreme temperatures ow through different agency pathways.For example, extreme cold and heat advisories often occur in advance based on local National Weather Service forecasting, and air quality advisories often occur in real-time according to EPA-based Air Quality Index measures.Communities at risk of wild re smoke exposures and extreme temperature events need locally-informed guidance, integrated strategies that address these compound risks, and communication approaches that include local knowledge and trusted sources.Local communities will be increasingly burdened with developing and sustaining strategies for adaptation and resilience to climate change, but we lack rigorous and reproducible models for such strategies, particularly as applicable to rural communities in the mountain west that additionally suffer from limited infrastructure that can be leveraged for mitigation.
al. 2014, Tiotiu et al. 2020).A recent meta-analysis of 84 studies including children, adults, or both found that outdoor air pollutants were associated with an increased risk of asthma exacerbations at lag 0-1 days (Huang et al 2021).The study also conducted age-based subgroup analyses of children (0-14) and adults (>14) and found children with asthma were more susceptible to outdoor air pollution (Huang et al. 2021).However, various other time-series studies using air pollutants have observed a lag effect with varying results from 0-5 days (Ostro et al. 2001, Lee et al. 2006, Halonen et al. 2008, Khalili et al. 2018, Lu et al. 2020, Puvvula et al. 2022) to 6-7 days (Chien et al 2018, Dabrowiecki et al. 2022).In general, lag effects past 7 days are typically not observed, though Gu et al. ( Abbreviations LRTI -lower respiratory tract infections

Figure 1 Study
Figure 1

Table 2 Explanatory
Variable Descriptions.A summary of variables used in modeling is provided.Interactions between these variables were also considered.
(Swanson et al. (2022)9,J45, J45.2, J45.20, J45.21, J45.22,J45.3,J45.30,J45.31,J45.32, J45.4,J45.40,J45.41,J45.42,J45.5, J45.50, J45.51, J45.52, J45.9, J45.90, J45.901, J45.902, J45.909, J45.99, J45.990, J45.991 Exposure, Outcome, and Other Explanatory Variables of Interest All environmental exposure data used in modeling are summarized in Table 2. -PM 2.5 exposure assessmentThe daily time-series dataset of PM 2.5 surface concentrations was previously developed, and details are reported elsewhere(Swanson et al. (2022).Brie y, these data were produced from air quality station observations, satellite data, and meteorological data to produce daily 1-km resolution surface PM 2.5 1 December 2017-13 March 2020), we observed 10,133 respiratory admissions among 8,128 unique patients, including 794 asthma, 638 LRTI, and 8,392 URTI.Figure2illustrates the weekly case counts and seasonal patterns across the time period studied.Daily mean PM 2.5 exceeded the United States Environmental Protection Agency 24-hr standard 35 µg / m 3 on 6 of 821 days (0.7%).Notably, all 6 of these days that exceeded the daily standard occurred during August 2018 when a prolonged air pollution event was experienced in the area due to smoke transport from extensive wild re activity in the western US and Canada.In what follows, we report results for each respiratory outcome (asthma, LRTI, URTI).Consistent results occurred for the distributed lag models that used cumulative days of PM 2.5 and weekly average of PM 2.5 with the strongest relationships between PM2.5 and the respiratory outcomes occurring around 7-13 day lag.Single day PM 2.5 values are shown only in supplementary material.temperaturesoranyseasonwere seen, although case frequencies for some of these model groups were small (e.g., n = 34 in the summer season).URTI: The increase in the risk of hospitalization for URTI associated with each 1 µg / m 3 increase in PM 2.5 modi ed by temperature or season can be found in Fig.3, Figures A.7-A.9, and Table3C.For the main effect PM 2.5 only model, highest odds estimates for URTI hospitalizations were associated with 1 µg / m 3 increase in cumulative PM 2.5 in the 13 days prior to a healthcare visit [OR = 2.43 95% CI: (1.05-5.64)].In general, for cumulative lag models the odds estimates remained elevated beyond 0-5 days.The higher frequency of URTI outcomes, relative to asthma and LRTI outcomes, allowed for consistent ndings of interactive effects by temperature and season and indicated that PM 2.5 effects were present in hotter rather than colder conditions.At cumulative lag (0-13 days), summer days yielded the highest PM 2.5 association with 3.35-fold increased odds of URTI hospitalization [95% CI: (1.85-6.04)].Similar patterns were observed for hotter days and summer/spring seasons across multiple lag periods.
Asthma: The increase in the risk of hospitalization for asthma associated with each 1 µg / m 3 increase in PM 2.5 modi ed by temperature or season can be found in Figures A.1-A.3 and Table3A.For the main effect of PM 2.5 , we observed associations with asthma hospitalizations at weekly average lag 7-13 [OR = 1.87, 95% CI: (1.17-2.97);Fig. A.1C].In colder temperatures and during the same lag of 7-13 days prior to a healthcare visit, a 1 µg / m 3 increase in PM 2.5 was associated with 3.11-fold greater odds of asthma [95% CI: (1.40-6.89);Fig. A.2C].Accumulated PM 2.5 (0-13 days) resulted in consistent ndings with patterns of a stronger odds for PM 2.5 at colder versus median temperatures [OR = 2.43 95% CI: (1.05-5.64)and OR = 1.97 95% CI: (1.03-3.76),respectively; Fig. A.2B].Finally, in the winter season and during the same lag of 7-13 days prior to an admission, a 1 µg / m 3 increase in PM only model was associated with LRTI hospitalizations at weekly average lag 6-12 days [OR = 2.18 95% CI: (1.20-3.97);Fig. A.4C].In colder temperatures and during the same lag of 6-12 days prior to a healthcare visit, a 1 µg / m 3 increase in PM 2.5 was associated with 2.61-fold greater odds of LRTI hospitalization [95% CI: (1.15-5.94);Fig. S5C].No signi cant associations for PM 2.5 effects on LRTI hospitalization during hotter 2.5 xTemperature (grouped by Colder, Median, and Hotter), and PM 2.5 xSeason (grouped by Fall, Winter, Spring, and Summer).Bolded values indicate an OR with 95% CI that occurred above 1.0.UN indicates unstable estimates.n indicates the sample size for model and group where available, noting that temperature was interacted as a continuous variable and doesn't de ne groups.
(Lam et al. 201919Fanget al. 2021, Lam et al. 2016ratures or during the winter season.Of course, above the 450N parallel, these two factors for colder temperatures and winter season are no doubt, con ated.However, very cold and dry or very hot and humid climate conditions have been shown to exacerbate asthma conditions(Cong et al. 2017, Fang et al. 2021, Lam et al. 2016).An animal model demonstrated that high and low temperatures can aggravate airway in ammation in mice suggesting that asthmatics are more at-risk during exposures to high and low temperature extremes(Deng et al. 2020).A recent review found that extreme cold exposures were associated with an increased risk of asthma by 19.77%(Han et al. 2023).Seasonal effects on asthma are inconclusive most likely because a range of temperature conditions have been shown to affect asthma risk.However, increased asthma risk has been observed in fall and winter seasons(Teach et al. 2015).LRTI and PM 2.5 exposure: In this study, LRTI encounters for children increased with elevated PM 2.5 and peaked at a weekly average lag of 6-12 days Studies on this category of respiratory infections or speci c infections within this category (e.g., bronchitis or pneumonia) vary in their ndings.To discuss a few, numbers of acute lower respiratory infections for young children in Utah, USA, were found to increase after 1 week of increased PM 2.5 and peak after 3 weeks of an increased exposure(Horne et al. 2018), while a similar study and results from Korea found acute lower respiratory infection hospitalizations to be associated with an increase in the 7-day running average of PM 2.5(Oh et al. 2021).Zhu et al. (2017)did not nd a signi cant effect of short-term PM 2.5 on childhood lower respiratory diseases in China, but did observe the effect with other air pollutants (PM 10 , NO 2 , and SO 2 ).In New York, USA, increases in PM 2.5 from the previous 7 days were found to be associated with hospital admissions for culture-negative pneumonia and bacterial pneumonia(Croft et al. 2018).To further illustrate variability in results, a metaanalysis review of short-term exposure to PM 2.5 and pneumonia-related hospitalizations found variable results across study populations, where elderly subgroups showed an increased risk ratio with unclear lag effects and younger patients did not have a signi cant increase in visits(Kim et al. 2020).LRTI, extreme temperatures, and seasonal effects: Our study showed the highest risk estimates for LRTI as a function of PM 2.5 during colder temperatures but insu cient sample size to assess any seasonal PM 2.5 mediated effects.These results are in line with past studies showing LRTI to be a signi cant cause URTI and PM 2.5 exposure: URTI have also been extensively studied and linked to air pollutants.Here, we found a positive association between PM 2.5 and children's URTI hospitalizations with a peak response at 13 days of accumulated PM 2.5 prior to an admission.As with asthma and LRTI, past research has shown that study population, study region, methodology, and type of upper respiratory tract infection can produce variations in the length of the delayed short-term effects.For example, in Beijing, China, a positive association between PM 2.5 and increased in uenza cases suggested a 1-2 month delayed response(Liang et al. 2014).In Hefei, China, increasing concentrates of most all pollutants at lag days 3-6 were associated with increased URTI in children aged 0-14 years (Li et al. 2018), while in Suzhou City, China, PM 2.5 showed a signi cant association with these infections in children under 3 years old with a lag of 3 weeks(Zhang et al. 2019).In Kenya, a 2-week delayed response in children's URTI from PM 2.5 exposure was observed(Larson et al. 2022).In Poland, moderate exposure to air pollution over 12 weeks was associated with an increased risk of URTI in children aged 3-12 years (Rataiczak et al. 2021).URTI, extreme temperatures, and seasonal effects: Our study showed relationships between increased risk of URTI hospitalization and delayed and elevated PM 2.5 , temperature, and season.Elevated levels of PM 2.5 accumulated across 13 days during hotter temperatures or during the summer season yielded the highest risk of children's URTI hospitalization.In general, URTI are thought to be more common in colder temperatures because colder exposure impairs nasal antiviral immunity(Eccles 2002, Huang et al. 2023).Viral infectious diseases affecting the upper tract way, such as in uenza, have strong seasonal effects in winter temperate regions and are associated with colder temperatures(Price et al. 2019).However, not all URTI spike in winter months in northern temperate sites, and others, like enterovirus and parain uenza virus, can occur in summer months and respiratory syncytial virus can occur earlier than in uenza in fall months(Lam et al. 2019).Rhinoviruses and adenoviruses can circulate throughout the year with occasional peaks in autumn and winter for rhinoviruses and early spring for adenoviruses(Jacobs et al. (Ahmed et al. 2006al.2019).In summary, most respiratory viruses follow a seasonal pattern but not all URTI are viruses, and some factors can increase the incidence of URTI, like mass crowding(Ahmed et al. 2006), and, as was observed in this study, air pollution.