Wildfire particulate matter in Shasta County, California and respiratory and circulatory disease-related emergency department visits and mortality, 2013-2018.

Wildfire smoke harms health. We add to this literature by evaluating the health effects of California's 2018 Carr Fire and preceding wildfire seasons in Shasta County.
METHODS: With data from the Shasta County Health and Human Services Agency, we examined the link between weekly wildfire fine particulate matter (PM2.5) exposure estimated using a spatiotemporal multiple imputation approach and emergency department (ED) visits and mortality using time-series models that controlled for temporal trends and temperature.
RESULTS: Between 2013 and 2018, Shasta County experienced 19 weeks with average wildfire PM2.5 ≥5.5 μg/m3 (hereafter, "high wildfire PM2.5 concentration"). Among all Shasta County Zip Code Tabulation Areas (ZCTAs; n = 36), we detected no association between high wildfire PM2.5 concentrations and respiratory or circulatory disease-related ED visits or mortality. Subsequent analyses were confined to valley ZCTAs (n = 11, lower elevation, majority of population, worse air quality in general). In valley ZCTAs, high wildfire PM2.5 was associated with a 14.6% (95% confidence interval [CI] = 4.2, 24.9) increase in same-week respiratory disease-related ED visits but no increase in the subsequent 2 weeks nor on circulatory disease-related mortality or ED visits or all-cause mortality. Two weeks after high wildfire PM2.5 weeks, respiratory disease-related deaths decreased (-31.5%, 95% CI = -64.4, 1.5). The 2018 Carr Fire appeared to increase respiratory disease-related ED visits by 27.0% (95% CI = 4.0, 50.0) over expectation and possibly reduce circulatory disease-related deaths (-18.2%, 95% CI = -39.4, 2.9).
CONCLUSIONS: As climate change fuels wildfire seasons, studies must continue to evaluate their health effects, particularly in highly exposed populations.

What this study adds

California experienced one of its worst wildfire seasons ever in 2018. The Carr Fire in Shasta County burned nearly 230,000 acres, destroyed over 1,500 homes, and cost $162 million to contain. This study provides the first evidence of an association between smoke from the 2018 wildfires and morbidity and mortality in human populations. In the most populated, low-elevation areas, we found that increased weekly wildfire particulate matter (PM2.5) was associated with respiratory disease-related emergency department (ED) visits, but not circulatory disease-related ED visits or all-cause or respiratory disease- or circulatory disease-related deaths. Similarly, the Carr Fire itself appeared to result in increased respiratory disease-related ED visits.

Introduction

In California, wildfires account for 71% of total fine particulate matter (PM2.5) on days that exceed US Environmental Protection Agency (USEPA) standards.[1] This proportion will likely increase in the future. In addition to PM2.5, wildfires generate volatile organic compounds (VOCs), polycyclic aromatic hydrocarbons (PAHs), and trace metals and contribute to elevated ozone levels, all of which potentially harm health.[2-4] Global research suggests a relationship between wildfire smoke exposure and respiratory disease-related emergency department (ED) visits, but evidence is mixed regarding circulatory disease-related ED visits and cause-specific mortality.[5-14] We know of no studies that evaluate health implications of California’s 2018 wildfire season, one of its most destructive to date.

Methods

We used data from the Shasta County Health and Human Services Agency from January 2013 to December 2018 to assess associations between Zip Code Tabulation Areas (ZCTAs) weekly average wildfire PM2.5 and ED visits and mortality. We specifically evaluated the Carr Fire, among California’s most destructive fires, which ignited on 23 July 2018 and burned nearly 230,000 acres in Shasta and Trinity Counties before it was contained on 30 August 2018 and extinguished in January 2019. Located in Northern California, Shasta County is mountainous (minimum elevation = 43 m; maximum elevation = 4,247 m) and relatively rural with an overall population density of ~19 individuals per km2. It contains 36 ZCTAs, which can be grouped into 11 valley (lower elevation, majority of population [~88%], worse air quality in general) and 25 mountain ZCTAs (eTable 1; http://links.lww.com/EE/A113 and eFigure 1; http://links.lww.com/EE/A113).

We identified weekly counts of respiratory and circulatory disease-related ED visits and all-cause and respiratory and circulatory disease-related mortality using International Classification of Diseases codes (460–519 [ICD-9] and J00–J99 [ICD-10] and 390–459 [ICD-9] and I00–I99 [ICD-10], respectively) for Shasta County residents.

To estimate daily ZIP code-level wildfire and nonwildfire PM2.5 concentrations, we used a multistage approach (see eMethods; http://links.lww.com/EE/A113). Briefly, we first identified smoke-plume exposed ZIP codes/days with the National Oceanic and Atmospheric Administration’s (NOAA) Hazard Mapping System (HMS) and overall PM2.5 concentrations with USEPA monitoring data from the Redding and Lassen stations and then used a spatiotemporal multiple imputation approach to estimate daily ZIP code-level wildfire PM2.5 concentrations, which we aggregated to the weekly-level for analyses. We estimated the intraclass correlation coefficient for the ZIP code-specific wildfire PM2.5 concentrations—the ratio of the between ZIP code over the sum of the within-ZIP code contributions to total variability—as 0.01. This indicated the majority of variation in wildfire PM2.5 occurred over time, rather than across ZIP codes (eFigure 2; http://links.lww.com/EE/A113). This fact and because Shasta County has low population density resulting in low ZCTA-level counts of health outcomes, we conducted weekly analyses across all ZCTAs combined and then restricted to ZCTAs in the valley. We defined weeks in which wildfire PM2.5 equaled or exceeded the study period mean concentration of 5.5 μg/m3 as high wildfire PM2.5 weeks. We selected this level, which represented the mean nonzero wildfire PM2.5 concentration a priori. We also linked weekly average maximum temperature from NOAA’s Local Climatological Data Daily Summary from the Redding, CA station. Redding is the largest city in Shasta County: over 50% of residents overall and 68% of residents in valley Zip Code Tabulation Areas (ZCTAs) live there.

For analyses, we used Box-Jenkins transfer function models[15] to estimate the association between time-series measuring weekly average wildfire PM2.5 concentrations and the logarithm of respiratory and circulatory disease-related ED visits and all-cause and respiratory and circulatory disease-related mortality. This approach detects and controls autocorrelation—including trends, cycles, and persistence of high and low observations—that can lead to spurious associations arising from shared autocorrelation (e.g., seasonality) and make estimates of true associations less precise. Because Shasta County has low population density resulting in low ZCTA-level counts of health outcomes, we conducted weekly analyses across all ZCTAs combined and then restricted to ZCTAs in the valley.

We estimated the relationship between wildfire PM2.5 and health outcomes in two ways. First, we estimated the association between weekly PM2.5 ≥5.5 μm/m3 (hereafter, “high wildfire PM2.5”) and outcomes. Second, we specified a binary “Carr Fire” variable equal to 1 during the Carr Fire and 0 otherwise. Because our outcome was log(Y), model coefficients approximate the percent change in each outcome when wildfire PM2.5 was high. All tests controlled for temperature and adjusted outcomes for autocorrelation.

Results

This study included 36 ZCTAs containing approximately 180,000 residents of Shasta County, California, between 2013 and 2018, when 51 major wildfires (>5,000 acres burned) occurred within 100 km of the County. Between 2013 and 2018, Shasta County experienced 19 weeks with average wildfire PM2.5 ≥5.5 μg/m3. Wildfire PM2.5 concentrations reached their weekly average maximum (48 μg/m3) during the 2018 Carr Fire (Figure A). Maximum average weekly temperature followed a seasonal pattern, with an annual average of 25.6 °C. Across the 314-week study period, there were a weekly average (SD) of 142 (64) respiratory disease-related ED visits, 31 (7) circulatory disease-related ED visits, 35 (7) all-cause deaths, 5 (2) respiratory disease-related deaths, and 11 (3) circulatory disease-related deaths. Respiratory disease-related ED visits exhibited strong seasonality (Figure B), whereas other outcomes did not.

Weekly data in Shasta County, California valley Zip Code Tabulation Areas from January 2013 to December 2018. A, Weekly mean wildfire PM2.5 concentration and (B) observed (points) and expected (lines) respiratory disease-related ED visits. The orange rectangle denotes the weeks of the 2018 Carr Fire.

We observed no associations between weekly high wildfire PM2.5 or the Carr Fire and ED visits or deaths overall. However, in models restricted to valley ZCTAs (n = 11), weekly high wildfire PM2.5 was associated with a 14.6% (95% confidence interval [CI] = 4.2, 24.9) increase in respiratory disease-related ED visits in the same week but no increase the subsequent 2 weeks (eTable 2; http://links.lww.com/EE/A113). We observed no association between high wildfire PM2.5 and circulatory disease-related ED visits or all-cause or circulatory disease-related mortality. High wildfire PM2.5 was potentially associated with decreased respiratory disease-related deaths 2 weeks later (–31.5%, 95% CI = –64.4, 1.5).

The Carr Fire appeared to increase respiratory disease-related ED visits by 27.0% (95% CI = 4.0, 50.0) over expectation, with no association for circulatory disease-related ED visits and all-cause deaths or respiratory disease-related deaths, and a potentially protective association with circulatory disease-related deaths (–18.2%, 95% CI = –39.4, 2.9).

Further adjustment of models for weekly average nonwildfire PM2.5 did not materially change results. For example, we observed a 14.7% (95% CI = 1.6, 27.7) increase in respiratory disease-related ED visits during weeks where wildfire PM2.5 was ≥5.5 μg/m3.

Discussion

We observed a 14.6% increase in respiratory disease-related ED visits in weeks where wildfire PM2.5 was ≥5.5 μg/m3 in valley ZCTAs of Shasta County, CA, between 2013 and 2018, with a 27.0% increase during the 2018 Carr Fire. No associations were seen for circulatory disease-related ED visits or all-cause or circulatory or respiratory disease-related deaths.

Our sample size precluded examination of specific respiratory and circulatory diseases, but results were consistent with three prior studies that reported associations between wildfire smoke and combined respiratory visits in California.[6,16,17] Prior research has consistently demonstrated increased risk of asthma- and chronic obstructive pulmonary disease-related ED visits associated with increased wildfire PM2.5.[5,6,16,18,19] For example, prior studies have reported associations between 3-day average wildfire PM2.5 and respiratory disease-related ED visits[18] and subdaily overall PM2.5 during wildfire season and ambulance dispatches for adverse respiratory outcomes.[13] We were unable to assess relevant daily or subdaily lagged results due to a relatively small population (approximately 190,000 people) and sparse daily counts. This is a limitation that may have contributed to our null results for all outcomes other than respiratory ED visits. Further, average weekly versus daily nonwildfire PM2.5 and maximum temperature may not adequately control confounding by these factors. Other studies that have taken a similar approach to ours, but at the daily level, have identified increases in daily respiratory ED visits[19,20] and adverse respiratory effects of specific wildfires, including the 2017 California Lilac Fire[12] and the Fall 2007 California firestorm,[6] for example.

Importantly, our analyses controlled for weekly average maximum temperature. The average maximum temperature exceeded 41 °C (106 °F) during the first week of the Carr Fire. Higher temperatures could have exacerbated the effect of the fire, as observed during the 2010 Moscow heatwave and wildfire,[21] or could have encouraged residents to stay indoors, thereby reducing exposure. We were unable to consider the role of other potentially important wildfire-related air pollutant exposures beyond PM2.5, including VOCs, PAHs, or secondarily formed ozone. In mutually adjusted models, Reid et al[16] found PM2.5, but not ozone, was associated with increased asthma-related ED visits during a large 2008 California wildfire. However, ozone levels do increase during wildfires and may impact health in some cases. In addition, by examining total wildfire PM2.5, we capture an overall wildfire mixture, but could not know which specific components of PM2.5 were related to the outcomes of interest.

As climate change fuels more frequent and destructive wildfires,[1,22] studies must continue to evaluate their health effects, particularly in highly exposed populations. Future studies may wish to consider compounding environmental and social effects of wildfire PM2.5 with other exposures such as extreme heat, co-occurring air pollution, or low socioeconomic status. In addition, characterizing population mobility such as evacuation and other responses to wildfire warning systems has the potential to inform wildfire policy and preparedness to reduce adverse health effects.

Conflicts of interest statement

The authors declare that they have no conflicts of interest with regard to the content of this report.

Supported, in part, by R00 ES027023, P30 ES009089, and R01 ES030616 from the National Institute of Environmental Health Sciences.