89 days, forming the analysis interval. Valid dates of the first positive specimen collec- tion were referenced if a patient was asymp- tomatic or if the date of symptom onset was missing or invalid. Patients without a valid date of symptom onset or date of first positive specimen collection were excluded from the analysis. Air pollution data were merged onto the individual patient-level COVID-19 dataset through an inner join by Federal Information Processing Standard (FIPS) county codes. Average air pollutant concentrations and the median air tempera- ture during the defined 90-day interval were calculated for each COVID-19 case accord- ing to the patient’s county of residence. To account for intra-county variation, the study was limited to MSAs, which have a greater number of air pollution reporting sites within each county. The cases for this study were defined as anyone with a lab-confirmed positive PCR or antigen test result for SARS-CoV-2 reported in Indiana who died as a result of COVID-19 ( n = 2,564). Controls were defined as those who had a lab-confirmed positive PCR or antigen test for SARS-CoV-2 reported in Indi- ana who did not die as a result of COVID-19 ( n = 45,368). Two backward stepwise logistic regression models were performed in R-4.2.1 via RStudio Desktop 2022.07.2+576 to calculate OR s and 95% confidence intervals (CIs) with adjust- ment for race, ethnicity, and the presence of preexisting conditions as confounders. These models included the potential exposures— mean sulfur dioxide (SO 2 ), mean Air Quality Index (AQI), mean nitrogen dioxide (NO 2 ), and particulate matter (PM 2.5 and PM 10 ) con- centrations—that were calculated during a 90-day interval from each patient’s first sign of infection. Mean ozone (O 3 ) and mean carbon monoxide (CO) concentrations were excluded from the model due to a lack of variability in the data. The binary outcome of interest for these models was a reported COVID-19 death. An Akaike information criterion (AIC) for both models is reported in the results section. Model 1 examined the risk of death among individuals with a positive test for SARS- CoV-2 reported to the Indiana Department of Health ( N = 53,459). This analysis assessed the entire study population. The model accounted for median air temperature during the 90-day exposure interval, race, ethnicity,
age, and the presence of at least one preexist- ing condition before the positive test result. The final regression equation for Model 1 is: ln ( p ) = 5.5667 – 0.4007 x NO 2 + 1 – p 0.8272 x PM 2.5 – 0.1940 x AQI + 0.0052 x Age – 0.0193 x Race – 0.0163 x ° F represents the concentration of NO 2 (ppb), x PM 2.5 represents the concentration of PM 2.5 (μg/m 3 ), and x AQI represents the con- where x NO 2 centration calculated mean AQI during the exposure interval. Additionally, the model retained age ( x Age ), race ( x Race ), and medial air temperature ( x °F ) during the interval. Model 2 examined the risk of death among reported COVID-19 cases who had a previ- ously diagnosed chronic lung disease ( n = 3,672) at the time of COVID-19 illness onset. This model accounted for median air temper- ature during the 90-day exposure interval, as well as race, ethnicity, and age. The final regression equation for Model 2 is: ln ( p ) = 13.8044 – 1.7762 x NO 2 + 1 – p 0.6070 x PM 2.5 + 0.9709 x PM 10 + 0.0038 x Age – 0.4616 x ° F where x NO 2 represents the concentration of NO 2 (ppb), x PM 2.5 represents the concentra- tion of PM 2.5 (μg/m 3 ), and x PM10 represents the concentration of PM 10 (μg/m 3 ) during the exposure interval. Additionally, the model retained age ( x Age ) and medial air temperature ( x °F ) during the interval as covariates. Results Study Population The study population consisted of all reported confirmed and probable COVID-19 cases in Indiana between March 31, 2020, and December 31, 2020. There were 53,459 patients with positive test results for SARS- CoV-2 reported who met our study inclu- sion criteria. Of the 53,459 patients, 2,564 patients died (4.7%) as a result of COVID- 19 infection; these cases are the ones whose data we used for our analysis. The remaining 45,368 patients in the sample who did not have a reported COVID-19 death were the cases we used as the control group for our analysis (Table 1).
Model 1 had an AIC of -19.71 and produced a statistically significant association between the 90-day average concentration of PM 2.5 and the risk of mortality from COVID-19. For every 1 μg/m 3 increase in PM 2.5 , there was a 4.5990 greater risk of death from COVID-19. No other statistically significant association between air pollution and mortality risk was observed from this model (Table 2). Model 2 had an AIC of 5.68. This model examined the approximate risk of mortality for individuals with previously diagnosed chronic lung disease. Two statistically signifi- cant associations were observed. The concen- tration of NO 2 appears to have a protective association with mortality from COVID-19 ( OR = 0.1609). For every 1-unit increase in the 90-day average ppb concentration of NO 2 , mortality from COVID-19 for individu- als with chronic lung conditions decreased by a factor of approximately 0.16. Additionally, PM 10 was found to have a statistically signifi- cant association with increased COVID-19 mortality. For every 1 μg/m 3 increase in the 90-day average concentration of PM 10 , there was a 2.7235 increased risk of mortality from COVID-19 (Table 3).
Discussion
Findings In both of the models, there was at least one statistically significant exposure variable. In Model 1, which examined risk of mortality from COVID-19 in relation to air pollution exposure, a statistically significant associa- tion between 90-day average PM 2.5 concen- tration and mortality from COVID-19 was observed. In Model 2, which examined the risk of air pollution exposure among COVID- 19 patients who had a previously diagnosed chronic lung disease, there was a statistically significant association between 90-day aver- age PM 10 concentration and mortality from COVID-19. Model 2 did not show any sig- nificance between the relationship of average 90-day PM 2.5 concentration and COVID-19 mortality, though. Interpretation We hypothesize that an average 90-day PM 2.5 concentration can increase the risk of mortal- ity from COVID-19 in the general population. Other studies have also observed this rela- tionship, and previous researchers have sug-
9
September 2024 • Journal of Environmental Health
Powered by FlippingBook