Science of the Total Environment 809 (2022) 152226

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv

Excess emergency department visits for cardiovascular and respiratory

diseases during the 2019–20 bushfire period in Australia: A two-stage
interrupted time-series analysis
Bo Wen, Yao Wu, Rongbin Xu, Yuming Guo , Shanshan Li
⁎ ⁎
School of Public Health and Preventive Medicine, Monash University, Level 2, 553 St Kilda Road, Melbourne, VIC 3004, Australia

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• The first study using ITS model to quantify

the 2019 bushfire effect in Australia
• The 2019–20 bushfires in Australia led to
considerable excess ED visits.
• Bushfires could increase the risk for car-
diovascular diseases.
• Higher risks related to bushfires were
found in regions with lower SES.

A R T I C L E I N F O A B S T R A C T

Article history: The health effects of the unprecedented bushfires in Australia in 2019–20 have not been fully examined. We aimed to
Received 29 September 2021 examine the excess emergency department (ED) visits related to the 2019–20 bushfires in New South Wales (NSW).
Received in revised form 14 November 2021 We obtained weekly data of ED visits for cardiovascular and respiratory diseases in all the 28 Statistical Area Level
Accepted 3 December 2021
4 (SA4) regions in NSW during the bushfire seasons from 2017 to 2020. A two-stage interrupted time-series analysis
Available online 7 December 2021
was applied to quantify the excess risk for ED visits in 2019–20. The total number of excess ED visits, excess percent-
Editor: Hai Guo ages, and their empirical confidence intervals (eCIs) were calculated to estimate the impacts of the bushfire season. A
total of 416,057 records of cardiorespiratory ED visits were included in our analysis. The bushfire season in 2019–20
Keywords: was significantly associated with a 6.0% increase (95% eCI: 1.9, 10.3) in ED visits for respiratory diseases and a 10.0%
Bushfire increase (95% eCI: 5.0, 15.2) for cardiovascular diseases, corresponding to 6177 (95% eCI: 1989, 10,166) and 3120
Respiratory (95% eCI: 1628, 4544) excess ED visits, respectively. The percentage of excess ED visits was higher in regions with
Cardiovascular lower SES and high fire density. In the context of climate change, more targeted strategies should be developed to pre-
Emergency department visit vent adverse bushfire effects and recover from such extreme environmental events.

1. Introduction Australia were more intense and started much earlier than previous years
(Jalaludin et al., 2020). The unprecedented bushfires (also known as the
In the context of climate change, climate-related extreme events, such as Black Summer bushfires) in Australia burned at least 19.4 million hectares
heatwaves, monsoon flooding, and wildfires, are likely to increase in fre- of land, destroyed 5900 residential and public structures, and directly killed
quency and severity (Watts et al., 2019). Promoted by severe drought and 34 people (van Oldenborgh et al., 2020; Yu et al., 2020). Considering the in-
an increase in extremely hot temperatures, the 2019–20 bushfires in direct impacts and costs, the bushfires can be more catastrophic (Johnston

⁎ Corresponding authors.
E-mail addresses: [email protected] (Y. Guo), [email protected] (S. Li).

http://dx.doi.org/10.1016/j.scitotenv.2021.152226
0048-9697/© 2021 Elsevier B.V. All rights reserved.
B. Wen et al. Science of the Total Environment 809 (2022) 152226

et al., 2021). However, the health impacts of the 2019–20 bushfires in measurement is missing or insufficient (Reid et al., 2016a). Second, other
Australia have not been well studied. components in the bushfire smoke, such as volatile organic compounds
The health risks associated with wildfires include direct risks from expo- (VOCs), CO, and ozone, have often been overlooked due to difficulties mon-
sure to fires, risks from exposure to wildfire smoke (Chen et al., 2021; Ye itoring the components (Chen et al., 2021). An interrupted time-series anal-
et al., 2021), and other health effects such as delayed access to healthcare ysis could address these limitations, as it takes the fire event itself rather
facilities for patients with specific conditions (Finlay et al., 2012; Xu than any specific air pollutant level as the exposure variable, thus it can
et al., 2020). The primary compounds of the wildfire smoke known to be evaluate the overall health impacts of the fire event, accounting for both di-
harmful to human health include fine and coarse particulate matter rect and indirect impacts (Wilkinson et al., 2019). In this study, we aimed to
(PM2.5 and PM10), acrolein, benzene, carbon monoxide (CO), and polycy- examine the excess ED visits related to the 2019–20 bushfire season in New
clic aromatic hydrocarbons (PAHs) (Cascio, 2018; Doubleday et al., South Wales, Australia. A novel two-stage interrupted time-series design is
2020). During the bushfire period in 2019–20, the population-weighted applied in 28 sub-State geographical units (Statistical Area Level 4, SA4 re-
PM2.5 increased sharply to 98.5 μg/m3 on 14 January 2020 in Australia, gions) to provide precise estimations.
which was fourteen times more than the historical level (6.8 μg/m3)
(Borchers Arriagada et al., 2020). 2. Material and methods
A growing body of research shows that wildfire smoke exposure has ad-
verse health effects (Cascio, 2018; Reid et al., 2016a). In Australia, previous 2.1. Study design and area
studies have demonstrated that PM10, PM2.5, and black carbon concentra-
tion from bushfires may increase the risk of emergency department (ED) An interrupted time-series study was carried out to examine the effects
presentations for respiratory disease, chronic obstructive pulmonary dis- of 2019–20 bushfire season on weekly emergency department (ED) visits in
ease (COPD), and asthma (Crabbe, 2012; Johnston et al., 2002; Morgan New South Wales (NSW), Australia (Fig. 1). In this study, NSW was divided
et al., 2010). However, most of the previous studies used the concentration into a total of 28 SA4 regions to show the geographical variation of the ef-
of particulate matter as the exposure assessment method, which may under- fects. SA4 regions are the largest sub-State regions in the Main Structure of
estimate the effect of bushfires mainly for two reasons: First, wildfire fre- the Australian Statistical Geography Standard (ASGS) and are specifically
quently happens in rural areas where reliable and sustainable air quality designed to represent labor markets or groups of labor markets within

Fig. 1. Weekly ED visits for cardiorespiratory diseases and annual fire density in 2017–20 in NSW, Australia.

2
B. Wen et al. Science of the Total Environment 809 (2022) 152226

each State and Territory, according to the Australian Bureau of Statistics each SA4 region was constructed using population-weighted averages of
(www.abs.gov.au). Each of the SA4 regions was set for a minimum popula- the constituent SA1s with the following formula:
tion of 100,000 persons, providing the best sub-State socio-economic
breakdown in the ASGS. Weekly ED visits from 1 September 2019 to 29 n

∑ IRSADSA1j  POPSA1j
February 2020, which was known as the bushfire season, were compared j¼1
IRSADSA4i ¼
with those in the same periods of 2017–18 and 2018–19. POPSA4i

where i is the specific SA4 region in NSW, j is the specific SA1 region in the
2.2. Data sources SA4 region i, POP is the population for each SA1 or SA4 region, and n is the
total number of SA1 regions in the SA4 region i. SA4 region with an IRSAD
2.2.1. ED data score lower than the median value (986) in NSW was defined as low SES,
Weekly ED visits of each SA4 region were obtained from the National and with an IRSAD score higher than the median value was defined as
Non-admitted Patient Emergency Department Care Database high SES (Table S3).
(NNAPEDCD), which was provided by the Australian Institute of Health
and Welfare (AIHW) (Australian Institute of Health and Welfare, 2020). Di- 2.3. Statistical analysis
agnosis of ED visits for 2017–19 and 2019–20 was coded with the Interna-
tional Classification of Diseases, 10th edition, Australian modification (ICD- 2.3.1. Interrupted time-series analysis
10-AM), while the Systematized Nomenclature of Medicine - Clinical Terms A two-stage interrupted time-series (ITS) analysis was applied to quan-
- Australian version, Emergency Department Reference Set (SNOMED-CT- tify the excess risk for ED visits during the bushfire period in 2019–20. In
AU EDRS) was also used in 2017–18 to code ED presentations. The princi- the first stage, a quasi-Poisson time-series regression was applied in each
pal diagnosis data coded in SNOMED-CT-AU EDRS were mapped to ICD-10- SA4 region (Bhaskaran et al., 2013; Scortichini et al., 2020). We used a
AM codes using a mapping file. Two main disease groups were presented in dummy variable to indicate the 2019–20 bushfire period (coded 1) and
this study, including respiratory diseases (ICD-10-AM codes: J00–J99), and the 2017–19 bushfire period (coded 0) in the model (Lopez Bernal et al.,
cardiovascular diseases (I10–I15, I20–I25, I26–I28, I30–I52). 2016). We included a linear term for the number of weeks to model long-
term trends, and an indicator for the month to control for seasonality.
Weekly mean temperature and weekly relative humidity were also con-
2.2.2. Meteorological data trolled in the model. A natural spline function with 3 degrees of freedom
The daily meteorological data were obtained from the Scientific Infor- was respectively applied to weekly mean temperature and weekly relative
mation for Land Owners (SILO) database (https://www.longpaddock.qld. humidity.
gov.au/silo/), which were constructed using weather station observations In the second stage, we pooled the region-specific estimates for the ex-
from the Australian Bureau of Meteorology (Jeffrey et al., 2001). The cess risk using a random-effect meta-analysis with maximum likelihood es-
gridded data had a national coverage with 0.05° × 0.05° spatial resolution timation (Gasparrini et al., 2012; Sera et al., 2019). Predictors at the region
(approximately 5 km × 5 km). Daily maximum temperature, daily mini- level were further included in the meta-regression models to account for the
mum temperature, and daily mean relative humidity were extracted from residual heterogeneity. We assessed the heterogeneity and the performance
the database and linked to NSW and each SA4 region by calculating the av- for each meta-regression model by I2 statistic (%), Cochran's Q test, and
erage value of all grids overlaying the area. Daily mean temperature was ap- Akaike information criterion (AIC). The model with a lower I2 statistic as
proximately calculated as the mean of daily maximum and minimum well as a lower AIC was chosen as the final model (Table S4). Our initial
temperature (Xu et al., 2019b). Weekly mean temperature and relative hu- analyses show that, for respiratory diseases, the best model is the model
midity were then calculated for each SA4 region. that included the fire density as the only predictor. For cardiovascular dis-
eases, the intercept-only meta-regression model produced the best model
performance. Region-specific estimates of the excess risk were then calcu-
2.2.3. Fire density data lated by using the best linear unbiased prediction (BLUP), which could pro-
Fire density was represented by active fire counts during the study pe- vide precise estimations (Chen et al., 2018).
riod. Data of daily active fire counts were obtained from the NASA's Fire In-
formation for Resource Management System (FIRMS) using the Fire Data 2.3.2. Quantification of excess ED visits
Product (Collection 6) collected by the Moderate Resolution Imaging Relative risks (RRs) were used to represent the excess risk of the
Spectroradiometer (MODIS), aboard NASA's Terra and Aqua satellites 2019–20 bushfire season. Region-specific number of excess ED visits was
(Justice et al., 2002). An active fire algorithm was used to detect thermal estimated by equation:
anomalies on a per-pixel basis, which was validated to produce the final ac-
tive fire counts in each grid (Giglio et al., 2003). The gridded daily data had EV i ¼ V i  ðRRi −1Þ=RRi
a global coverage with 1 km × 1 km spatial resolution (https://earthdata.
nasa.gov/firms). The weekly active fire counts were calculated for each where i is the specific region, Vi is the number of ED visits for region i, RRi is
SA4 region by summing daily counts within the area (Earl and the relative risk for region i, which is predicted by the BLUP. The total num-
Simmonds, 2017). We classified the fire density as low and high according ber of excess ED visits (EV) was calculated by summing the EVi. The excess
to the difference in the fire counts between 2017–19 and 2019–20 percentages were then calculated for each region and for the whole of NSW
(Table S2). SA4 regions were defined as high density if the active fire counts through dividing the excess ED visits by the difference between the total ED
in 2019–20 were more than ten times greater than the average fire counts visits and the excess ED visits. Empirical 95% confidence intervals (eCIs)
in 2017–19. were calculated by Monte Carlo simulation of the coefficients with the as-
sumption of a multivariate normal distribution of the point estimates
(Gasparrini and Leone, 2014).
2.2.4. Socio-economic indexes for areas
The Index of Relative Socio-economic Advantage and Disadvantage 2.3.3. Stratification analyses
(IRSAD) was obtained from the Australian Bureau of Statistics to character- First, stratified analysis was performed to detect the risks in regions
ize the socioeconomic status (SES) for each SA4 region (https://www.abs. with different fire densities. Secondly, stratified analysis was performed
gov.au). IRSAD was constructed based on Census 2016 of Population and to detect the risks in regions with different SES. The excess number and
Housing and calculated at the SA1 (Statistical Area Level 1) level. SES of fraction of ED visits were computed for high or low fire density regions,

3
B. Wen et al. Science of the Total Environment 809 (2022) 152226

and high or low SES regions. We used random effect meta-regression fitted
by the maximum likelihood method to compare the effects estimated in dif-
ferent groups (Xu et al., 2019a).

2.3.4. Sensitivity analyses

Sensitivity analyses were conducted to test the robustness of our results.
First, we tested whether the results were reliable by changing the df (2, 3,
4) of natural splines for temperature and relative humidity or using linear
terms for temperature and relative humidity in the first stage. Secondly,
we also tested whether the results were robust by removing temperature
or relative humidity from the model. As the ambient temperature is related
to bushfires (Yu et al., 2020), it may underestimate the impacts of bushfire
season when adjusting for temperature. We used fixed effect meta-
regression to compare the effects estimated in the sensitivity analyses
with our primary models (Xu et al., 2019a). Besides, we also applied the
autoregressive integrated moving average (ARIMA) model as an alternative
analysis (Schaffer et al., 2021). The methods and results were shown in Sup-
Fig. 3. Association between the 2019–20 bushfire season and ED visits for
plementary materials.
cardiovascular and respiratory diseases in total and stratified by socioeconomic
We used R software (version 4.0.1) to perform all analyses. The pack- status (SES) and fire density. The exact values can be found in Table S4.
ages “dlnm” and “mixmeta” were used to perform the DLNM model and
meta-regression, respectively. Two tailed p-values less than 0.05 were con-
sidered statistically significant. we estimated the 2019–20 bushfire season was associated with the increase
of ED visits (RR = 1.05, 95% eCI: 1.02–1.09, P < 0.001). The RR for respi-
3. Results ratory diseases in low SES regions (1.12, 95% eCI: 1.06–1.18, P < 0.001)
was significantly higher than that in high SES regions (1.00, 95% eCI:
Weekly ED visits in bushfire periods were collected from 2017 to 2020. 0.96–1.04, P = 0.904). Similarly, the RR for respiratory diseases in high
During the study period, a total of 416,057 records were included in our fire density regions was significantly higher than that in low fire density re-
analysis. Fig. 1 shows the mean weekly ED visits and average fire counts gions. For ED visits of cardiovascular diseases, the RR of 2019–20 bushfire
of each SA4 region in 2017–19 and 2019–20, respectively. Overall, the season was 1.10 (95% eCI: 1.07–1.13, P < 0.001). However, there was no
mean weekly ED visits for respiratory diseases were 4204 in 2019–20, significant difference when stratified by SES and fire density. Region-
which was higher than that (3857) in 2017–19 (Table S1). By contrast, specific RRs were shown in Fig. S3 (respiratory diseases) and Fig. S4 (car-
the mean weekly ED visits for cardiovascular diseases (1323) in 2019–20 diovascular diseases).
was lower than that in 2017–19 (1380). The fire density shows a dramatic The estimated excess ED visits and the excess percentage for each SA4
increase in 2019–20 for most of SA4 regions, especially for those located in region during the bushfire period in 2019–20 are shown in Table 1. The
coastal areas (Fig. 1). bushfire season in 2019–20 was significantly associated with a 6.0% in-
Trends of weekly ED visits with and without the bushfire period in crease (95% eCI: 1.9, 10.3) in ED visits for respiratory diseases and a
2019–20 are shown in Fig. 2. It can be found that the observed weekly 10.0% increase (95% eCI: 5.0, 15.2) in ED visits for cardiovascular diseases,
ED visits for respiratory diseases and cardiovascular diseases during the corresponding to 6177 (95% eCI: 1989, 10,166) and 3120 (95% eCI: 1628,
bushfire period in 2019–20 were higher than the predicted weekly ED visits 4544) excess ED visits, respectively. Table 2 shows excess percentages of ED
if the bushfire period in 2019–20 kept the same density as the periods in visits stratified by fire density and SES. When stratified by fire density, the
2017–19. percentage excess of ED visits for respiratory diseases was higher in high
The relative risks (RR) during the bushfire period in 2019–20 are shown density regions. When stratified by SES, the percentage excess of ED visits
in Fig. 3 and exact values are shown in Table S5. For respiratory diseases, was higher in low SES regions, especially for respiratory diseases.

Fig. 2. Modeling of ED visits during the study periods in NSW, Australia. Points show observed weekly ED visits in NSW during the included months from 2017 to 2020. Black
line represents modeled weekly ED visits aggregated at the state level. Orange line represents predicted weekly ED visits if the 2019–20 bushfires were the same level as those
in 2017–19. The shadowed area represents the 95% confidence interval. Methods of the modeling were described in Supplementary eMethods.

4
B. Wen et al. Science of the Total Environment 809 (2022) 152226

Table 1
Number of ED visits, estimated excess visits, and excess percentage (%) for different SA4 regions in NSW in 2019–20.
SA4 region Fire density in SES Respiratory diseases Cardiovascular diseases
2019–20 group
Total Excess visits Excess percentage Total Excess visits Excess percentage
(%) (%)

Capital Region High High 5111 1088 (773, 1381) 27.0 (17.8, 37.0) 1369 120 (61, 177) 9.6 (4.6, 14.9)
Central Coast High Low 5885 −2 (−197, 186) −0.0 (−3.2, 3.3) 2014 193 (107, 275) 10.6 (5.6, 15.8)
Central West High Low 5031 762 (535, 978) 17.9 (11.9, 24.1) 1545 158 (92, 221) 11.4 (6.4, 16.7)
Coffs Harbour - Grafton High Low 3431 476 (332, 614) 16.1 (10.7, 21.8) 933 78 (38, 117) 9.2 (4.2, 14.4)
Far West and Orana Low Low 3614 42 (−70, 150) 1.2 (−1.9, 4.3) 960 95 (54, 135) 11.0 (6.0, 16.4)
Hunter Valley exc Newcastle High Low 5119 573 (390, 750) 12.6 (8.2, 17.2) 2136 195 (104, 283) 10.1 (5.1, 15.3)
Illawarra Low High 4008 −10 (−144, 120) −0.2 (−3.5, 3.1) 1415 127 (65, 185) 9.8 (4.8, 15.1)
Mid North Coast High Low 4941 783 (551, 1003) 18.8 (12.5, 25.5) 1639 143 (72, 210) 9.5 (4.6, 14.7)
Murray Low Low 880 7 (−21, 34) 0.8 (−2.3, 4.0) 269 24 (12, 35) 9.7 (4.6, 15.0)
New England and North West High Low 5394 1685 (1207, 2108) 45.4 (28.8, 64.1) 1642 144 (72, 212) 9.6 (4.6, 14.9)
Newcastle and Lake Macquarie Low Low 4318 −19 (−166, 122) −0.4 (−3.7, 2.9) 1788 157 (80, 231) 9.6 (4.7, 14.8)
Richmond – Tweed High Low 5435 197 (44, 346) 3.8 (0.8, 6.8) 1529 134 (67, 197) 9.6 (4.6, 14.8)
Riverina High Low 4131 234 (118, 347) 6.0 (2.9, 9.2) 998 90 (47, 132) 9.9 (4.9, 15.2)
Southern Highlands and Shoalhaven High Low 2465 180 (108, 250) 7.9 (4.6, 11.3) 953 86 (44, 125) 9.9 (4.8, 15.2)
Sydney - Baulkham Hills and Hawkesbury High High 1931 68 (14, 121) 3.7 (0.7, 6.7) 684 66 (36, 95) 10.7 (5.6, 16.0)
Sydney - Blacktown Low High 4104 −18 (−158, 116) −0.4 (−3.7, 2.9) 1271 119 (64, 172) 10.4 (5.3, 15.6)
Sydney - City and Inner South Low High 2771 −13 (−107, 78) −0.5 (−3.7, 2.9) 830 74 (37, 109) 9.7 (4.7, 15.0)
Sydney - Eastern Suburbs Low High 2766 −13 (−107, 78) −0.5 (−3.7, 2.9) 908 83 (44, 121) 10.1 (5.1, 15.4)
Sydney - Inner South West Low High 7662 −36 (−296, 216) −0.5 (−3.7, 2.9) 2318 196 (96, 292) 9.3 (4.3, 14.4)
Sydney - Inner West Low High 2747 −13 (−106, 77) −0.5 (−3.7, 2.9) 985 86 (43, 127) 9.5 (4.5, 14.8)
Sydney - North Sydney and Hornsby Low High 3397 −16 (−131, 96) −0.5 (−3.7, 2.9) 914 87 (47, 125) 10.5 (5.4, 15.8)
Sydney - Northern Beaches Low High 2416 −11 (−93, 68) −0.5 (−3.7, 2.9) 862 79 (42, 115) 10.1 (5.1, 15.4)
Sydney - Outer South West High Low 3786 3 (−121, 123) 0.1 (−3.1, 3.4) 983 87 (44, 128) 9.7 (4.7, 15.0)
Sydney - Outer West and Blue Mountains High High 3945 294 (178, 407) 8.1 (4.7, 11.5) 1293 110 (54, 164) 9.3 (4.3, 14.5)
Sydney - Parramatta Low High 5108 −24 (−197, 144) −0.5 (−3.7, 2.9) 1470 132 (69, 193) 9.9 (4.9, 15.1)
Sydney - Ryde Low High 1657 −8 (−64, 47) −0.5 (−3.7, 2.9) 450 40 (20, 58) 9.6 (4.6, 14.9)
Sydney - South West Low Low 4927 −22 (−189, 140) −0.4 (−3.7, 2.9) 1388 133 (73, 190) 10.6 (5.6, 15.9)
Sydney - Sutherland Low High 2333 −11 (−90, 66) −0.5 (−3.7, 2.9) 858 83 (46, 118) 10.7 (5.6, 16.0)
New South Wales 109,313 6177 (1989, 10,166) 6.0 (1.9, 10.3) 34,404 3120 (1628, 4544) 10.0 (5.0, 15.2)

Fig. S2 illustrates the geographical variation of percentage excess ED cardiovascular diseases was still inconclusive (Cascio, 2018; Reid et al.,
visits associated with the 2019–20 bushfire season for two diseases at the 2016a). Some studies documented that the association was non-
SA4 region level. Results of sensitivity analyses showed that the results significant between wildfire and total cardiovascular diseases (Johnston
were robust when changing the df (2–7) of spline function or using a linear et al., 2014; Rappold et al., 2011). Other studies showed there were signif-
function for temperature and relative humidity (Table S6). When removing icant associations between wildfire smoke and higher morbidity risks for
temperature and relative humidity from the main model, the RR showed specific cardiovascular outcomes (e.g., ischemic heart diseases, out-of-
significant increases compared with our primary models (Table S7). hospital cardiac arrests) (Haikerwal et al., 2015; Karanasiou et al., 2021).
The findings in this work provided new evidence that bushfires may lead
4. Discussion to an increase of ED visits for cardiovascular diseases. Nevertheless, further
research is still warranted.
To the best of our knowledge, this is the first study using a two-stage ITS In this study, ED visits associated with respiratory diseases were signif-
method to quantify the effects of the 2019–20 bushfire season in Australia. icantly increased during the bushfire period. The results were consistent
By use of weekly time-series data of ED visits in NSW, we found evidence of with the studies carried out in Sydney and North Carolina, which reported
an association between the 2019–20 bushfire season and elevated ED visits. that ED attendances for respiratory diseases and asthma increased com-
The percentage excess in ED visits associated with the bushfire season was pared with non-smoky periods and counties (Johnston et al., 2014;
6.0% (95% eCI: 1.9, 10.3) for cardiovascular diseases and 10.0% (95% eCI: Rappold et al., 2011). Besides, the associations were found to be varying
5.0, 15.2) for cardiovascular diseases in NSW. The excess risks seemed to be in different sub-state regions, which could be possibly owing to the spatial
varying in different sub-state regions. The risk of ED visits for respiratory and temporal distribution of the bushfires. We found that the excess ED
diseases was much higher in high fire density and low SES regions. visits for respiratory diseases were higher in high fire density regions. It is
In our study, we found a higher risk of ED visits for cardiovascular dis- thus critical to identify regions where wildfires are more active, especially
eases, which was in line with some of the previous studies (Tinling et al., for more densely populated areas (Sadasivuni et al., 2013).
2016; Wettstein et al., 2018). General knowledge has been well established In this study, we found there were 6177 excess ED visits for respiratory
that ambient air pollution has cardiovascular clinical effects, including in- diseases and 3120 ED visits for cardiovascular diseases attributed to bush-
creased mortality and hospital admissions (Mannucci et al., 2019; fire exposure. Two previous studies have examined the health burden of
Thurston et al., 2017). However, the association between wildfire and the 2019–20 bushfire season in Australia using modeling methods
(Borchers Arriagada et al., 2020; Ryan et al., 2021). A total of 2345.6 and
18.5 hospital admissions could be attributed to PM2.5 and ozone exposure
Table 2 of 2019–20 bushfires in NSW, respectively (Ryan et al., 2021). For cause-
Excess percentages of ED visits (%) in 2019–20 stratified by fire density and SES. specific hospital admissions, the bushfire PM2.5 in 2019–20 was estimated
Stratification Subgroup Respiratory diseases Cardiovascular diseases in another study to be associated with 577 (95% CI: 81, 357) and 1050
Total 6.0 (1.9, 10.3) 10.0 (5.0, 15.2) (95% CI: 0, 2204) hospital admissions for cardiovascular diseases and respi-
Fire density High density 12.6 (7.5, 18.0) 10.0 (5.0, 15.2) ratory diseases in NSW (Borchers Arriagada et al., 2020). The figures, how-
Low density −0.3 (−3.6, 3.0) 10.0 (5.0, 15.2) ever, were much lower than the estimations in our study, which may imply
SES Low SES 9.0 (4.4, 13.7) 10.1 (5.1, 15.3) that it could significantly underestimate the health burden related to the
High SES 2.6 (−1.1, 6.4) 9.9 (4.9, 15.1)
bushfires based only on PM2.5. By using a two-stage ITS model, we could

5
B. Wen et al. Science of the Total Environment 809 (2022) 152226

model the pre-event (the 2019–20 bushfire season) trends of the outcome and more bushfires. This study could provide evidence on the impacts of
and construct the counterfactual scenario to quantify the excess risks bushfires and help to develop more targeted policies and strategies to pre-
(Leibel et al., 2020). As a result, this design allowed us to consider the vent adverse effects and recover from the disaster, especially in the context
health impacts of the bushfires from a more comprehensive perspective. of climate change and the COVID-19 pandemic.
In this study, we also found that SES could potentially modify the influ-
ences of bushfires, which is that regions with lower SES had higher excess CRediT authorship contribution statement
ED visits for both respiratory diseases and cardiovascular diseases. This
finding was consistent with previous studies (Jones et al., 2020; Rappold BW: Data curation, Formal analysis, Methodology, Visualization,
et al., 2012; Reid et al., 2016b). For populations with lower SES, they Writing – original draft. YW: Validation, Visualization, Data curation,
may have a greater prevalence of preexisting health conditions, which Writing – original draft. RX: Methodology, Writing – review & editing. SL
lead to a higher risk during the bushfire period. By contrast, populations and YG: Conceptualization, Supervision, Data curation, Methodology,
with higher SES may have more protective methods, including moving to Funding acquisition, Writing – review & editing.
areas unaffected by bushfires and using air filters. Moreover, policies for re-
covery have halted due to the COVID-19 pandemic following the 2019–20 Funding
bushfire season, which may exacerbate the effects of bushfire events in
those rural communities (Usher et al., 2021). This study was supported by Australian Research Council
The health effects of bushfire season can be mainly explained by expo- (DP210102076), and Australian National Health and Medical Research
sure to air pollution in the process. Small particles in wildfire smoke have Council (APP2000581). BW, YW, RX was supported by China Scholarship
been shown to result in increased oxidative stress at the cellular level Council [grant number 202006010043, 202006010044, 201806010405]
(Stowell et al., 2019). Wildfire smoke is likely to disrupt lung endothelial (https://www.csc.edu.cn/chuguo/s/1844, https://www.csc.edu.cn/chuguo/
cell barrier integrity, cause damage to the respiratory epithelium, and s/1267). SL was supported by the Early Career Fellowship of the Australian
therefore lead to lung inflammation, increased airway reactivity, and lung National Health and Medical Research Council [grant number
dysfunction (De Sario et al., 2013; Olivieri and Scoditti, 2005; Wang APP1109193] (https://www.nhmrc.gov.au/). YG was supported by the Ca-
et al., 2010). In this study, we also observed that the bushfire effect on car- reer Development Fellowship of the Australian National Health and Medical
diovascular diseases is plausible. There are three principle underlying path- Research Council [grant number APP1163693] (https://www.nhmrc.gov.
ways (Cascio, 2018; Chen et al., 2021). First, air pollutants can react with au/). The funding bodies did not play any role in the study design, data collec-
neural receptors in the lung and directly activate the autonomic nervous tion, data analyses, results interpretation and writing of this manuscript.
system. Second, oxidative stress and systemic inflammation led by air pol-
lutants could result in increased platelet activation and thrombosis, and in- Declaration of competing interest
flammation of vascular endothelia. Third, ultrafine fraction and gases in the
air pollutants could translocate through alveolar membranes and contrib- The authors declare that they have no known competing financial inter-
ute to cardiovascular effects, including endothelial activation and injury. ests or personal relationships that could have appeared to influence the
Except for the smoke exposure, mental stresses, lack of medication, and work reported in this paper.
other psychosocial factors following bushfires may increase the risk for car-
diovascular diseases (Babaie et al., 2021; Hayman et al., 2015). Studies Appendix A. Supplementary data
have observed that the prevalence and incidence of diabetes, substance
misuse, hypertension, and cardiovascular diseases increased following nat- Supplementary data to this article can be found online at https://doi.
ural disasters and humanitarian crises (Greene et al., 2018; Keasley et al., org/10.1016/j.scitotenv.2021.152226.
2020; Kehlenbrink et al., 2019; Nozaki et al., 2013). For example, it was
found that cardiovascular events significantly increased in three weeks fol-
References
lowing the Great East Japan Earthquake in 2011 (Nozaki et al., 2013). Men-
tal stresses were regarded as the main driving factors because natural
Australian Institute of Health and Welfare, 2020. Australian Bushfires 2019–20: Exploring the
disasters could commonly lead to the damage of homes and properties, fi- Short-term Health Impacts. AIHW, Canberra.
nancial pressure, separation from friends and family, and even loss of fam- Babaie, J., Pashaei Asl, Y., Naghipour, B., Faridaalaee, G., 2021. Cardiovascular diseases in
natural disasters; a systematic review. Arch. Acad. Emerg. Med. 9, e36.
ily members (Babaie et al., 2021). Moreover, natural disasters could lead to
Baum, A., Barnett, M.L., Wisnivesky, J., Schwartz, M.D., 2019. Association between a tempo-
the decrease of health care access and further exacerbate the condition of rary reduction in access to health care and long-term changes in hypertension control
those with underlying diseases (Babaie et al., 2021; Baum et al., 2019). among veterans after a natural disaster. JAMA Netw. Open 2, e1915111.
As a result, attention should also be paid to the non-smoke health impacts Bhaskaran, K., Gasparrini, A., Hajat, S., Smeeth, L., Armstrong, B., 2013. Time series regres-
sion studies in environmental epidemiology. Int. J. Epidemiol. 42, 1187–1195.
of bushfires and other natural disasters. Borchers Arriagada, N., Palmer, A.J., Bowman, D.M., Morgan, G.G., Jalaludin, B.B., Johnston,
The main strength of this study is the use of ITS model, which enabled F.H., 2020. Unprecedented smoke-related health burden associated with the 2019–20
us to quantify the impacts of the bushfire season in a more comprehensive bushfires in eastern Australia. Med. J. Aust. 213, 282–283.
Cascio, W.E., 2018. Wildland fire smoke and human health. Sci. Total Environ. 624,
way rather than only focusing on the impacts of bushfire PM2.5. However, 586–595.
several limitations should be acknowledged. The analysis is based on the Chen, R., Yin, P., Wang, L., Liu, C., Niu, Y., Wang, W., et al., 2018. Association between am-
dataset released by AIHW and we were unable to apply further stratifica- bient temperature and mortality risk and burden: time series study in 272 main chinese
cities. BMJ 363, k4306.
tion analyses in different age groups due to the absence of relevant vari- Chen, G., Guo, Y., Yue, X., Tong, S., Gasparrini, A., Bell, ML., et al., 2021. Mortality risk attrib-
ables. This may be addressed in the future when more detailed data were utable to wildfire-related PM2.5 pollution: a global time series study in 749 locations. The
released. The data in this study were entirely collected from emergency de- Lancet Planet. Health 5 (9), e579–e587.
Chen, H., Samet, J.M., Bromberg, P.A., Tong, H., 2021. Cardiovascular health impacts of wild-
partments and the diagnoses are not coded by qualified clinical coders,
fire smoke exposure. Part. Fibre Toxicol. 18, 2.
which may result in outcome misclassification. Crabbe, H., 2012. Risk of respiratory and cardiovascular hospitalisation with exposure to
bushfire particulates: new evidence from Darwin, Australia. Environ. Geochem. Health
34, 697–709.
5. Conclusions
De Sario, M., Katsouyanni, K., Michelozzi, P., 2013. Climate change, extreme weather events,
air pollution and respiratory health in Europe. Eur. Respir. J. 42, 826–843.
In conclusion, our findings suggest that considerable excess ED visits for Doubleday, A., Schulte, J., Sheppard, L., Kadlec, M., Dhammapala, R., Fox, J., et al., 2020.
cardio-respiratory outcomes were associated with the 2019–20 bushfire Mortality associated with wildfire smoke exposure in Washington state, 2006–2017: a
case-crossover study. Environ. Health 19, 4.
season in NSW. The results indicate that the unprecedented bushfires led Earl, N., Simmonds, I., 2017. Variability, trends, and drivers of regional fluctuations in austra-
to a huge health burden, showing a higher risk in regions with lower SES lian fire activity. J. Geophys. Res. Atmos. 122, 7445–7460.

6
B. Wen et al. Science of the Total Environment 809 (2022) 152226

Finlay, S.E., Moffat, A., Gazzard, R., Baker, D., Murray, V., 2012. Health impacts of wildfires. Rappold, A.G., Cascio, W.E., Kilaru, V.J., Stone, S.L., Neas, L.M., Devlin, R.B., et al., 2012.
PLoS Curr. 4, e4f959951cce2c. Cardio-respiratory outcomes associated with exposure to wildfire smoke are modified
Gasparrini, A., Leone, M., 2014. Attributable risk from distributed lag models. BMC Med. Res. by measures of community health. Environ. Health 11, 71.
Methodol. 14, 55. Reid, C.E., Brauer, M., Johnston, F.H., Jerrett, M., Balmes, J.R., Elliott, C.T., 2016a. Critical
Gasparrini, A., Armstrong, B., Kenward, M.G., 2012. Multivariate meta-analysis for non-linear review of health impacts of wildfire smoke exposure. Environ. Health Perspect. 124,
and other multi-parameter associations. Stat. Med. 31, 3821–3839. 1334–1343.
Giglio, L., Descloitres, J., Justice, C.O., Kaufman, Y.J., 2003. An enhanced contextual fire de- Reid, C.E., Jerrett, M., Tager, I.B., Petersen, M.L., Mann, J.K., Balmes, J.R., 2016b. Differential
tection algorithm for MODIS. Remote Sens. Environ. 87, 273–282. respiratory health effects from the 2008 northern California wildfires: a spatiotemporal
Greene, M.C., Kane, J.C., Khoshnood, K., Ventevogel, P., Tol, W.A., 2018. Challenges and op- approach. Environ. Res. 150, 227–235.
portunities for implementation of substance misuse interventions in conflict-affected pop- Ryan, R.G., Silver, J.D., Schofield, R., 2021. Air quality and health impact of 2019–20 Black
ulations. Harm Reduct. J. 15. Summer megafires and COVID-19 lockdown in Melbourne and Sydney, Australia. Envi-
Haikerwal, A., Akram, M., Del Monaco, A., Smith, K., Sim, M.R., Meyer, M., et al., 2015. Im- ron. Pollut. 274, 116498.
pact of fine particulate matter (PM 2.5) exposure during wildfires on cardiovascular Sadasivuni, R., Cooke, W., Bhushan, S.J.E.M., 2013. Wildfire Risk Prediction in Southeastern
health outcomes. J. Am. Heart Assoc. 4, e001653. Mississippi Using Population Interaction. 251, pp. 297–306.
Hayman, K.G., Sharma, D., Wardlow, R.D., Singh, S., 2015. Burden of cardiovascular morbid- Schaffer, A.L., Dobbins, T.A., Pearson, S.-A., 2021. Interrupted time series analysis using
ity and mortality following humanitarian emergencies: a systematic literature review. autoregressive integrated moving average (ARIMA) models: a guide for evaluating
Prehosp. Disaster Med. 30, 80–88. large-scale health interventions. BMC Med. Res. Methodol. 21.
Jalaludin, B., Johnston, F., Vardoulakis, S., Morgan, G., 2020. Reflections on the catastrophic Scortichini, M., Schneider Dos Santos, R., De’ Donato, F., De Sario, M., Michelozzi, P., Davoli,
2019-2020 Australian bushfires. Innovation 1 (1), 100010. M., et al., 2020. Excess mortality during the COVID-19 outbreak in Italy: a two-stage
Jeffrey, S.J., Carter, J.O., Moodie, K.B., Beswick, A.R., 2001. Using spatial interpolation to interrupted time-series analysis. Int. J. Epidemiol. 49 (6), 1909–1917.
construct a comprehensive archive of australian climate data. Environ. Model Softw. Sera, F., Armstrong, B., Blangiardo, M., Gasparrini, A., 2019. An extended mixed-effects
16, 309–330. framework for meta-analysis. Stat. Med. 38, 5429–5444.
Johnston, F.H., Kavanagh, A.M., Bowman, D.M., Scott, R.K., 2002. Exposure to bushfire Stowell, J.D., Geng, G., Saikawa, E., Chang, H.H., Fu, J., Yang, C.E., et al., 2019. Associations
smoke and asthma: an ecological study. Med. J. Aust. 176, 535–538. of wildfire smoke PM2.5 exposure with cardiorespiratory events in Colorado 2011–2014.
Johnston, F.H., Purdie, S., Jalaludin, B., Martin, K.L., Henderson, S.B., Morgan, G.G., 2014. Environ. Int. 133, 105151.
Air pollution events from forest fires and emergency department attendances in Sydney, Thurston, G.D., Kipen, H., Annesi-Maesano, I., Balmes, J., Brook, R.D., Cromar, K., et al., 2017.
Australia 1996–2007: a case-crossover analysis. Environ. Health 13, 105. A joint ERS/ATS policy statement: what constitutes an adverse health effect of air pollu-
Johnston, F.H., Borchers-Arriagada, N., Morgan, G.G., Jalaludin, B., Palmer, A.J., Williamson, tion? An analytical framework. Eur. Respir. J. 49.
G.J., et al., 2021. Unprecedented health costs of smoke-related PM2.5 from the 2019–20 Tinling, M.A., West, J.J., Cascio, W.E., Kilaru, V., Rappold, A.G., 2016. Repeating cardiopul-
australian megafires. Nat. Sustain. 4, 42–47. monary health effects in rural North Carolina population during a second large peat wild-
Jones, C.G., Rappold, A.G., Vargo, J., Cascio, W.E., Kharrazi, M., McNally, B., et al., 2020. Out- fire. Environ. Health 15, 12.
of-hospital cardiac arrests and wildfire-related particulate matter during 2015–2017 Cal- Usher, K., Ranmuthugala, G., Maple, M., Durkin, J., Douglas, L., Coffey, Y., et al., 2021. The
ifornia wildfires. J. Am. Heart Assoc. 9, e014125. 2019–2020 bushfires and COVID-19: the ongoing impact on the mental health of people
Justice, C.O., Giglio, L., Korontzi, S., Owens, J., Morisette, J.T., Roy, D., et al., 2002. The living in rural and farming communities. Int. J. Ment. Health Nurs. 30, 3–5.
MODIS fire products. Remote Sens. Environ. 83, 244–262. van Oldenborgh, G.J., Krikken, F., Lewis, S., Leach, N.J., Lehner, F., Saunders, K.R., 2020. At-
Karanasiou, A., Alastuey, A., Amato, F., Renzi, M., Stafoggia, M., Tobias, A., et al., 2021. tribution of the Australian bushfire risk to anthropogenic climate change. Nat. Hazards
Short-term health effects from outdoor exposure to biomass burning emissions: a review. Earth Syst. Sci. 21, 941–960.
Sci. Total Environ. 781, 146739. Wang, T., Chiang, E.T., Moreno-Vinasco, L., Lang, G.D., Pendyala, S., Samet, J.M., et al., 2010.
Keasley, J., Oyebode, O., Shantikumar, S., Proto, W., McGranahan, M., Sabouni, A., et al., Particulate matter disrupts human lung endothelial barrier integrity via ROS- and p38
2020. A systematic review of the burden of hypertension, access to services and patient MAPK-dependent pathways. Am. J. Respir. Cell Mol. Biol. 42, 442–449.
views of hypertension in humanitarian crisis settings. BMJ Glob. Health 5, e002440. Watts, N., Amann, M., Arnell, N., Ayeb-Karlsson, S., Belesova, K., Boykoff, M., et al., 2019. The
Kehlenbrink, S., Smith, J., Ansbro, É., Fuhr, D.C., Cheung, A., Ratnayake, R., et al., 2019. The 2019 report of the lancet countdown on health and climate change: ensuring that the
burden of diabetes and use of diabetes care in humanitarian crises in low-income and health of a child born today is not defined by a changing climate. Lancet 394,
middle-income countries. Lancet Diabetes Endocrinol. 7, 638–647. 1836–1878.
Leibel, S., Nguyen, M., Brick, W., Parker, J., Ilango, S., Aguilera, R., et al., 2020. Increase in Wettstein, Z.S., Hoshiko, S., Fahimi, J., Harrison, R.J., Cascio, W.E., Rappold, A.G., 2018. Car-
pediatric respiratory visits associated with Santa Ana wind-driven wildfire smoke and diovascular and cerebrovascular emergency department visits associated with wildfire
PM2.5 levels in San Diego County. Ann. Am. Thorac. Soc. 17, 313–320. smoke exposure in California in 2015. J. Am. Heart Assoc. 7.
Lopez Bernal, J., Cummins, S., Gasparrini, A., 2016. Interrupted time series regression for the Wilkinson, A.L., Scollo, M.M., Wakefield, M.A., Spittal, M.J., Chaloupka, F.J., Durkin, S.J.,
evaluation of public health interventions: a tutorial. Int. J. Epidemiol. 46 (1), 348–355. 2019. Smoking prevalence following tobacco tax increases in Australia between 2001
Mannucci, P.M., Harari, S., Franchini, M., 2019. Novel evidence for a greater burden of ambi- and 2017: an interrupted time-series analysis. Lancet Public Health 4, e618–e627.
ent air pollution on cardiovascular disease. Haematologica 104, 2349–2357. Xu, R., Zhao, Q., Coelho, M., Saldiva, P.H.N., Abramson, M.J., Li, S., et al., 2019a. The associ-
Morgan, G., Sheppeard, V., Khalaj, B., Ayyar, A., Lincoln, D., Jalaludin, B., et al., 2010. Effects ation between heat exposure and hospitalization for undernutrition in Brazil during
of bushfire smoke on daily mortality and hospital admissions in Sydney, Australia. Epide- 2000–2015: a nationwide case-crossover study. PLoS Med. 16, e1002950.
miology 21, 47–55. Xu, R., Zhao, Q., Coelho, M., Saldiva, P.H.N., Zoungas, S., Huxley, R.R., et al., 2019b. Associ-
Nozaki, E., Nakamura, A., Abe, A., Kagaya, Y., Kohzu, K., Sato, K., et al., 2013. Occurrence of ation between heat exposure and hospitalization for diabetes in Brazil during 2000–2015:
cardiovascular events after the 2011 great East Japan earthquake and tsunami disaster. a Nationwide case-crossover study. Environ. Health Perspect. 127, 117005.
Int. Heart J. 54, 247–253. Xu, R., Yu, P., Abramson, M.J., Johnston, F.H., Samet, J.M., Bell, M.L., et al., 2020. Wildfires,
Olivieri, D., Scoditti, E.J.E.R.R., 2005. Impact of Environmental Factors on Lung Defences. 14, global climate change, and human health. N. Engl. J. Med. 383, 2173–2181.
pp. 51–56. Ye, T., Guo, Y., Chen, G., Yue, X., Xu, R., Coêlho, MSZS., et al., 2021. Risk and burden of hos-
Rappold, A.G., Stone, S.L., Cascio, W.E., Neas, L.M., Kilaru, V.J., Carraway, M.S., 2011. Peat pital admissions associated with wildfire-related PM2.5 in Brazil, 2000-15: a nationwide
Bog Wildfire Smoke Exposure in Rural North Carolina is Associated With Cardiopulmo- time-series study. Lancet Planet. Health 5 (9), e599–e607.
nary Emergency Department Visits assessed Through Syndromic Surveillance. 119, Yu, P., Xu, R., Abramson, M.J., Li, S., Guo, Y., 2020. Bushfires in Australia: a serious health
pp. 1415–1420. emergency under climate change. Lancet Planet. Health 4, e7–e8.