Resources, n.d.-a). We used the daily average of PM 2.5 in our analyses as both a continuous variable and categorical variable using Air Quality Index (AQI) categories (Wisconsin Department of Natural Resources, n.d.-b). To handle the small number of random missing PM 2.5 values, we imputed missing values with the mean of nearby daily values (1 day before and 1 day after), as done by Chi et al. (2019). Gaps of more than 1 day were not imputed. Meteorological Exposures We downloaded weather data from the National Oceanic and Atmospheric Adminis- tration (NOAA) for the dates December 18, 2018–July 14, 2022, for the seven rural coun- ties included in this analysis. We included an additional 2 weeks before and after the case date range as potential control dates. We used only the maximum daily temperature from this data set. To handle the small number of missing values, we imputed missing values with the mean of nearby daily values (1 day before and 1 day after). As temperature data for Forest County were unavailable, we used temperatures from the adjacent Vilas County for Forest County measurements. Statistical Analysis We used a retrospective, time-stratified, case- crossover study design to test for associations between PM 2.5 exposures and asthma exac- erbations. We applied conditional logistic regression for multivariate analyses to produce hazard ratios (HR) and 95% confidence inter- vals (CI) (SAS PHREG with TIES=BRESLOW; Wang et al., 2011). We combined clinical encounter data with daily PM 2.5 measure- ments and daily maximum temperature from each county. Control exposure periods were selected by matching the day of the week for 1–2 weeks prior to and 1–2 weeks after the exposure date for each exacerbation ( n = 3–4). We created lag days for PM 2.5 exposures on the day of the exacerbation (lag0) and 1–6 days prior to the exacerbation (lag1–6). We also included a 7-day PM 2.5 average to represent the total exposure to PM 2.5 in the 7-day period prior to the exacerbation. All models were adjusted for daily maxi- mum temperature (°F). In secondary analy- ses, we assessed potential e¥ect modification by sex, race and ethnicity, age, season, and county of residence. Race and ethnicity were evaluated as potential e¥ect modifiers given
TABLE 1
Main Characteristics of Asthma Exacerbations Among Residents From Ashland, Dodge, Forest, Grant, Sauk, Taylor, and Vilas Counties in Wisconsin
Characteristic
# (%) or Mean ± SD
Exacerbations
1,424 (100)
Type of exacerbation
Emergency department visit
1,267 (89.0)
Hospitalization
80 (5.6) 77 (5.4)
Observation stay
PM 2.5 (µg/m
3 ) Mean: All possible days Mean: Exacerbation days
6.7 ± 4.7 7.5 ± 4.8 35 ± 21.7
Mean age at time of exacerbation (years)
Age group (years) 0–4
103 (7.2)
5–17
231 (16.2) 406 (28.5) 525 (36.9) 159 (11.2)
18–34 35–64
≥65
Race and ethnicity Asian
5 (0.4)
Black
57 (4.1) 77 (5.5) 74 (5.3)
Hispanic
Native American
White
1,179 (84.7)
Not provided
32
continued on page 10
Methods
sin Department of Natural Resources. Rural county classifications were based on the Fed- eral Oce of Rural Health Policy data files (Health Resources and Services Adminis- tration, 2024). Second encounters for cases occurring within 7 days of a prior exacerba- tion and exacerbations with missing PM 2.5 data were excluded from the analysis. PM 2.5 Exposure We used PM 2.5 data collected by the Wis- consin Department of Natural Resources from all rural counties with air monitors ( n = 7). Throughout Wisconsin, the Wisconsin Department of Natural Resources operates a network of air monitors that use feder- ally approved methods for measuring air quality (Wisconsin Department of Natural
Case Definition of Asthma Exacerbation
To identify asthma exacerbations, we used hospital discharge data collected by the Wis- consin Hospital Association that had been shared with the Wisconsin Department of Health Services. We defined an asthma exac- erbation as any emergency department visit, hospitalization, or observation stay with an ICD-10 (International Classification of Dis- ease, 10th revision) code of J45 that occurred during January 1, 2019–June 30, 2022. We restricted our analysis to Wisconsin residents of seven rural counties (Ashland, Dodge, Forest, Grant, Sauk, Taylor, and Vilas) with PM 2.5 air monitors operated by the Wiscon-
9
January/February 2025 • Journal of Environmental Health
Powered by FlippingBook