Wildfires are associated with increased emergency department visits for anxiety disorders in the western United States

Abstract

As wildfires increasingly impact the global economy and public health, understanding their effects is crucial. Particularly, the relationship between wildfires and anxiety disorders remains unclear. In this study, we explore this association by analyzing 1,897,865 emergency department visits for anxiety disorders in the western United States. We examined records from 2007 to 2018, using a case-crossover design and conditional logistic regression to assess the impact of wildfire-related exposures on these visits. Here we show that exposure to wildfire smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ is positively linked with emergency department visits for anxiety disorders. This effect is more pronounced in women and girls and in older adults, highlighting their vulnerability. Notably, major smoke events (smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ contributed ≥75% of the total ${\mathrm{P}\mathrm{M}}_{2.5}$) significantly amplify this risk. These findings underscore the psychological impacts of wildfires and their smoke, suggesting a need for targeted disaster risk reduction and climate risk management strategies, especially for vulnerable groups such as older adults and women. Our results call for increased climate awareness and tailored risk communication to mitigate these emerging health challenges.

Attributable to anthropogenic climate change repercussions, wildfires have become a growing threat to the global economy, public health^1,2, and ecosystem³. Over the past 20 years, rising temperatures and shifting drying patterns have substantially escalated global fire potential¹, resulting in expanded burned areas and prolonged fire seasons^4,5. The impact of wildfires, however, is not uniformly distributed across the globe, with variations stemming from differences in climatology and geographical conditions. The western United States has long been identified as a major fire-prone region, due predominantly to the recurrent presence of fire weather conditions, increasing droughts, and the abundance of fuel resources^2,6,7. The widespread occurrence of wildfires in the western United States has caused a surge in smoke emissions. In recent years, wildfire smoke has been estimated to account for around 40% of the total particulate matter with an aerodynamic diameter of 2.5 μm or less (${\mathrm{P}\mathrm{M}}_{2.5}$) across the whole western United States and even 50% in the northwestern region, exacerbated substantially from the 15–20% contribution in the early 2000s⁸. Previous epidemiological studies have linked wildfire smoke exposure to a variety of adverse health outcomes, such as headache⁹ and respiratory diseases¹⁰. However, the psychological consequences of wildfires remain largely unclear to date¹¹.

Mental health disorders, including anxiety disorders, have been another expanding threat to global public health in the past decades. According to the Global Burden of Disease study, mental health disorders contributed a total of 125.3 million (95% uncertainty interval (UI) = (93.0–163.2)) disability-adjusted life years in 2019, increased substantially from 80.8 million (95% UI = (59.5–105.9)) in 1990¹². Anxiety disorders, in particular, were the most prevalent mental health condition, affecting 970.1 (95% UI = (990.9–1,044.4)) per million individuals in 2019¹². The Global Burden of Disease 2019 revealed that anxiety disorders ranked among the top 25 leading contributors to the global disability-adjusted life years for all populations and were among the top 10 contributors for women and girls and all adolescents¹³. Regions with high sociodemographic indices, including western Europe, Australia, and high-income North America, exhibited the highest age-standardized incidence rates of anxiety disorders¹⁴.

In recent years, the psychological impacts of climate change, termed ‘eco-anxiety’ or ‘climate anxiety’, have become increasingly recognized¹¹. With anxiety as the most common symptom, climate change has been linked to a variety of psychological disorders such as depression, post-traumatic stress disorder, and eating disorders¹⁵. Nevertheless, the extent to which natural disasters, such as wildfires, contribute to climate anxiety remains largely unclear. Despite the parallel trends of escalating wildfire activities and a rising prevalence of anxiety disorders, the incidence of anxiety disorders following wildfires remains under-researched¹⁶. Although a handful of studies have established the link between ambient ${\mathrm{P}\mathrm{M}}_{2.5}$ and anxiety disorders in the United States^17,18, China^19,20, and the United Kingdom²¹, evidence concerning wildfire smoke exposure remains inconclusive²². For example, ref. 23 reported that the prevalence of being unable to stop or control worrying more than half the time during the past 14 days elevated by 30% (prevalence ratio = 1.30, 95% confidence interval (CI) = (1.03–1.65)) among adults exposed to medium or heavy smoke for six weeks or more in the past year in Oregon, USA²³. Similarly, ref. 24 found that wildfire smoke events were associated with heightened anxiety and depression in the rural Washington State community²⁴. However, the absence of high-performance quantitative measurements of wildfire smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ has limited the potential for large-scale population studies that feature higher statistical power and validity.

In addition to the inhalation of smoke ${\mathrm{P}\mathrm{M}}_{2.5}$, wildfire itself may have adverse psychological effects that ultimately result in anxiety disorders. For example, a case study in Canada showed a positive association between prevalence of generalized anxiety disorders and witnessing homes being destroyed by the wildfire or living in a different home among survivors of the 2016 Fort McMurray wildfire²⁵. As reviewed by ref. 16, trauma-related factors, such as property loss or fearing the loss of loved ones, could increase the risk of anxiety disorders following wildfire events¹⁶. However, those studies were limited in scope and did not quantify the magnitude of risk associated with wildfire exposure. Furthermore, it is also imperative to investigate whether certain population groups are more susceptible to post-wildfire anxiety disorders since some individuals, such as women, are considerably more vulnerable to anxiety and stress-related disorders^14,26. In addition, research has consistently shown sex differences related to behavioral responses to crisis information and psychological impacts before, during, and after the end of a disastrous event²⁷. However, the extent of these sex-based differences in psychological responses to wildfire-related exposures has not been thoroughly measured.

Therefore, the present study aimed to investigate the impact of wildfires on emergency department (ED) visits for anxiety disorders in the western United States using a satellite-driven exposure dataset. The exposures of interest were selected on the basis of the two primary pathways that potentially underlie the association between wildfires and anxiety disorders: (1) the inhalation of wildfire smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ and (2) the psychological effects of wildfire smoke events and active fire points. We also evaluated whether the association between wildfire and anxiety disorders differed among distinct sex, age, race, and ethnic groups. The findings of this study will enhance our mutual understanding of the mental health impact of wildfires and promote environmental justice by benefiting vulnerable populations.

Results

Study population

This study analyzed 1,897,865 ED visits for anxiety disorders (Table 1). Of the 1,296,468 patients who visited the ED during the International Classification of Diseases, Ninth Revision (ICD-9) period (before 1 October 2015), 88.5% (1,147,099) were diagnosed with anxiety, dissociated, and somatoform disorders while the remaining 11.5% (149,369) patients were diagnosed as reaction to severe stress and adjustment disorders. During the ICD-10 period, of the 601,397 patients, the majority (512,468, 85.2%) had a diagnosis code of F41, denoting Other (non-phobic) anxiety disorders. An additional 61,486 (10.2%) patients were diagnosed as reaction to severe stress and adjustment disorders (F43). Phobic anxiety disorders (F40), obsessive-compulsive disorder (F42), associative and conversion disorders (F44), somatoform disorders (F45), and other nonpsychotic mental disorders (F48) accounted for 901 (0.1%), 1,476 (0.2%), 8,683 (1.4%), 14,567 (2.4%), and 1,816 (0.3%) participants, respectively (Supplementary Table 1). The majority of cases (1,513,553, 79.8%) originated from California, while Arizona, Nevada, Oregon, and Utah contributed 184,772 (9.7%), 63,643 (3.4%), 67,953 (3.6%), and 67,944 (3.6%) cases, respectively. Among all participants, 60.5% (1,148,133) were women and girls, with the percentage varying between 59.0% and 61.8% across different states. Adults accounted for 83.0% (1,575,972) of the total ED visits, while children and adolescents made up 8.0% (152,405), and older adults accounted for 8.9% (169,448). Out of all the participants, 1,181,866 (62.3%) were non-Hispanic. The highest proportion of Hispanic participants was observed in California (580,410, 38.3%), while the lowest proportion was found in Oregon (6,371, 9.4%). A total of 1,228,284 (64.7%) participants were white, while Black, Asian, and other racial groups accounted for 8.1% (154,493), 4.0% (76,312), and 23.1% (438,776), respectively.

Table 1.

Study population

	Arizona	California	Nevada	Oregon	Utah	Overall
N	184,772	1,513,553	63,643	67,953	67,944	1,897,865
Sex
Men and boys	75,710 (41.0%)	594,981 (39.3%)	25,697 (40.4%)	27,372 (40.3%)	25,972 (38.2%)	749,732 (39.5%)
Women and girls	109,062 (59.0%)	918,572 (60.7%)	37,946 (59.6%)	40,581 (59.7%)	41,972 (61.8%)	1,148,133 (60.5%)
Age group
5–17	14,797 (8.0%)	120,235 (7.9%)	4,718 (7.4%)	6,232 (9.2%)	6,423 (9.5%)	152,405 (8.0%)
18–64	152,477 (82.5%)	1,258,049 (83.1%)	53,879 (84.7%)	54,913 (80.8%)	56,654 (83.4%)	1,575,972 (83.0%)
65+	17,498 (9.5%)	135,269 (8.9%)	5,046 (7.9%)	6,808 (10.0%)	4,867 (7.2%)	169,488 (8.9%)
Ethnicity
Non-Hispanic	132,933 (71.9%)	893,532 (59.0%)	51,800 (81.4%)	60,348 (88.8%)	43,253 (63.7%)	1,181,866 (62.3%)
Hispanic	50,398 (27.3%)	580,410 (38.3%)	10,744 (16.9%)	6,371 (9.4%)	8,264 (12.2%)	656,187 (34.6%)
Missing	1,441 (0.8%)	39,611 (2.6%)	1,099 (1.7%)	1,234 (1.8%)	16,427 (24.2%)	59,812 (3.2%)
Race
White	114,490 (62.0%)	959,218 (63.4%)	38,497 (60.5%)	56,605 (83.3%)	59,474 (87.5%)	1,228,284 (64.7%)
Black	10,623 (5.7%)	132,832 (8.8%)	7,905 (12.4%)	2,089 (3.1%)	1,044 (1.5%)	154,493 (8.1%)
Asian	2,110 (1.1%)	70,372 (4.6%)	1,823 (2.9%)	1,004 (1.5%)	1,003 (1.5%)	76,312 (4.0%)
Other	57,549 (31.1%)	351,131 (23.2%)	15,418 (24.2%)	8,255 (12.1%)	6,423 (9.5%)	438,776 (23.1%)

As illustrated in Table 2, the averaged 48 hour exposures to smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ and background ${\mathrm{P}\mathrm{M}}_{2.5}$ were 0.93 ± 3.18 μg m⁻³ and 9.91 ± 4.42 μg m⁻³, respectively. A total of 629,622 smoke events occurred during the study period, including 142,436 that occurred on case days and 486,186 that occurred on control days. We also identified 11,931 extreme wildfire smoke events, with 2,809 taking place on case days and 9,122 on the control days.

Table 2.

Wildfire-related exposures

	Control days (N = 6,452,523)	Case days (N = 1,897,865)	Overall (N = 8,350,546)
Smoke ${\mathrm{P}\mathrm{M}}_{2.5}$
Mean (s.d.)	0.930 (3.17)	0.935 (3.20)	0.931 (3.18)
Background ${\mathrm{P}\mathrm{M}}_{2.5}$
Mean (s.d.)	9.90 (4.42)	9.91 (4.41)	9.91 (4.42)
Smoke events	486,186	142,436	629,622
Major smoke events	9,122	2,809	11,931
	Control days (N = 14,574)	Case days (N = 4,264)	Overall (N = 18,838)
Active fire points a
Mean (s.d.)	2.92 (7.48)	3.16 (8.34)	2.97 (7.68)
Cumulative FRPE a
Mean (s.d.)	211 (814)	415 (821)	362 (816)

^aWe included only days with active fire points for these statistics. Nevada was excluded from the analyses for fire points due to insufficient zip code information.

A total of 4,264 cases occurred on days with active wildfire points in the same zip code (Nevada was excluded). On case days with active fire points, the mean fire count was 3.16 ± 8.34, with an average cumulative fire radiative power (FRP) of 415 ± 821. We also identified 14,574 control days with active fire points. The average fire count and cumulative FRP were 2.92 ± 7.48 and 211 ± 814, respectively (Table 2).

The effects of wildfire smoke ${\mathrm{P}\mathrm{M}}_{2.5}$

We found that a 10 μg m⁻³ increase in 48 hour exposure to smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ was associated with a 0.6% increase in the risk of ED visits for anxiety disorders (odds ratio (OR) = 1.006 (1.000, 1.013), P = 0.040) in model 1 (Table 3). The OR for smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ in the model adjusted for background ${\mathrm{P}\mathrm{M}}_{2.5}$ was 1.004 (95% CI = (0.998, 1.010), P = 0.169).

Table 3.

The associations between wildfire-related exposures and ED visits for anxiety disorders

Exposures	Model 1		Model 2
	OR (95% CI)	P value	OR (95% CI)	P value
Smoke ${\mathrm{P}\mathrm{M}}_{2.5}$a	1.006 (1.000, 1.012)	0.040*	1.004 (0.999, 1.010)	0.169
Smoke events	1.005 (0.998, 1.013)	0.183	1.005 (0.997, 1.012)	0.230
Major smoke events	1.063 (1.015, 1.113)	0.009*	1.056 (1.009, 1.106)	0.019*
Fire points	1.002 (0.999, 1.005)	0.066	1.002 (0.999, 1.005)	0.068
Cumulative FRPb	1.000 (0.999, 1.001)	0.200	1.000 (0.999, 1.001)	0.205

Model 1 included only wildfire-related exposures and the covariates, while model 2 also adjusted for background ${\mathrm{P}\mathrm{M}}_{2.5}$. The two-sided P values were derived from conditional logistic regression models using a z test without adjustments for multiple comparisons.

^aThe OR was for a 10 μg m^–3 increase in 48 hour smoke ${\mathrm{P}\mathrm{M}}_{2.5}$.

^bThe OR was for an interquartile range (67.75 W) increase in the cumulative FRP.

^*P < 0.05.

As illustrated in Fig. 1, a 10 μg m⁻³ increase in smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ was associated with a 1.1% increase in the risk of ED visits for anxiety disorders among women and girls in model 1 (OR = 1.012, 95% CI = (1.002, 1.021), P = 0.020). Smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ was also a significant risk factor for ED visits for anxiety disorders in older men (OR per 10 μg m⁻³ = 1.033 (1.011, 1.056), P = 0.003), older women (OR per 10 μg m⁻³ = 1.040 (1.006, 1.074), P = 0.027), and adult women (OR per 10 μg m⁻³ = 1.012 (1.001, 1.023), P = 0.027). No significant association between smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ and ED visits for anxiety disorders was observed for children and adolescents, regardless of sex.

Fig. 1 |. The age- and sex-stratified analysis for 48 hour smoke ${\mathrm{P}\mathrm{M}}_{2.5}$.

Model 1 included only 48 hour smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ and the covariates. Model 2 also adjusted for background ${\mathrm{P}\mathrm{M}}_{2.5}$. The ORs are for a 10 μg m⁻³ increase in smoke ${\mathrm{P}\mathrm{M}}_{2.5}$. The statistics were derived from a total of 1,897,865 participants, including 749,732 men and boys and 1,148,133 women and girls. The age group 5–17 consisted of 152,405 participants (56,767 boys and 95,638 girls). The age group 18–64 consisted of 1,575,972 participants (644,424 men and 931,548 women), and the age group 65+ consisted of 169,488 participants (48,541 men and 120,947 women, respectively). Data are presented as ORs and their 95% CI.

The race- and ethnicity-stratified analysis (Supplementary Fig. 1) demonstrated that 48 hour exposure to smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ was associated with the risk of ED visits for anxiety disorders in white women and girls (OR = 1.012 (1.000, 1.024), P = 0.045). We also observed marginally significant associations in other women and girls (OR = 1.021 (0.998, 1.044), P = 0.068) and Hispanic women and girls (OR = 1.017 (0.997, 1.036), P = 0.073).

The effects of smoke events

We found that a smoke event was not significantly associated with the risk of ED visits for anxiety disorders in the general population, regardless of the adjustment for background ${\mathrm{P}\mathrm{M}}_{2.5}$ (Table 3). Nevertheless, as demonstrated in Fig. 2 (upper panel), the age- and sex-stratified analysis revealed the association between smoke events in the past 48 hours and an elevated risk of ED visits for anxiety disorders among all women and girls (OR = 1.013 (1.001, 1.025), P = 0.035) and adult women (OR = 1.015 (1.002, 1.029), P = 0.021). Both associations remained significant after adjusting for background ${\mathrm{P}\mathrm{M}}_{2.5}$ (OR = 1.012 (1.000, 1.025), P = 0.044 for all women and girls and OR = 1.015 (1.002, 1.029), P = 0.026 for adult women). We found no significant association between smoke events and ED visits for anxiety disorders in men and boys among all age groups.

Fig. 2 |. The age- and sex-stratified analysis for smoke events and major smoke events.

Model 1 included only smoke events and other covariates. Model 2 also adjusted for background ${\mathrm{P}\mathrm{M}}_{2.5}$. The upper panels are for smoke events in which wildfire smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ contributed ≥25% of the total ${\mathrm{P}\mathrm{M}}_{2.5}$. The lower panels are for major smoke events in which wildfire smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ contributed ≥75% of the total ${\mathrm{P}\mathrm{M}}_{2.5}$. The statistics were derived from a total of 1,897,865 participants, including 749,732 men and boys and 1,148,133 women and girls. The age group 5–17 consisted of 152,405 participants (56,767 boys and 95,638 girls). The age group 18–64 consisted of 1,575,972 participants (644,424 men and 931,548 women), and the age group 65+ consisted of 169,488 participants (48,541 men and 120,947 women, respectively). Data are presented as ORs and their 95% CI.

Furthermore, our results (Table 3) showed that a major smoke event was associated with an increased risk of ED visits for anxiety disorders in the general population (OR = 1.063 (1.015, 1.113), P = 0.009), even adjusting for background ${\mathrm{P}\mathrm{M}}_{2.5}$ (OR = 1.056 (1.009, 1.106), P = 0.019). While this association was not observed in women and girls (Fig. 2, lower panel), we found significant impacts of major smoke events on both adult men (OR = 1.082 (1.013, 1.156), P = 0.018) and older men (OR = 1.236 (1.052, 1.452), P = 0.010). However, we also found that major smoke events were negatively associated with ED visits for anxiety disorders in all and in girls and adolescents (OR = 0.821 (0.691, 0.974), P = 0.008 and OR = 0.636 (0.474, 0.852), P = 0.002, respectively).

The effects of active fire points and FRP

As shown in Table 3, we found a marginally significant association between the number of active fire points and ED visits for anxiety disorders in the general population (OR per fire point = 1.002 (0.999, 1.005), P = 0.066). No association was observed for the cumulative FRP (OR per 67.75 W = 1.000 (0.999, 1.001), P = 0.200). However, the age- and sex-stratified analysis (Fig. 3) showed that an active fire point is associated with a 1.1% increase in the risk of ED visits for anxiety disorders among all older adults (OR = 1.011 (1.000, 1.022), P = 0.042). In addition, per interquartile range increase (67.75) in the cumulative FRP is also associated with an increased risk of ED visits for anxiety disorders in all older adults (OR = 1.003 (1.001, 1.006), P = 0.008) and older men (OR = 1.003 (1.001, 1.006), P = 0.020) (Supplementary Fig. 2). Adjusting for background ${\mathrm{P}\mathrm{M}}_{2.5}$ did not change the associations for both fire counts and FRP.

Fig. 3 |. The age- and sex-stratified analysis for active fire points.

Model 1 included only the total number of active fire points in the past 48 hours and other covariates. Model 2 also adjusted for background ${\mathrm{P}\mathrm{M}}_{2.5}$. The statistics were derived from a total of 1,897,865 participants, including 749,732 men and boys and 1,148,133 women and girls. The age group 5–17 consisted of 152,405 participants (56,767 boys and 95,638 girls). The age group 18–64 consisted of 1,575,972 participants (644,424 men and 931,548 women), and the age group 65+ consisted of 169,488 participants (48,541 men and 120,947 women, respectively). Data are presented as ORs and their 95% CI.

Sensitivity analyses

As shown in Supplementary Fig. 3, we found that the same-day and 48 hour average exposure to wildfire smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ were positively associated with ED visits for anxiety disorders. The ORs for same-day and 48 hour average smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ were 1.006 (95% CI = (1.001, 1.012), P = 0.030) and 1.006 (95% CI = (1.000, 1.012), P = 0.040), respectively. However, exposures with an extended temporal coverage showed no association with anxiety disorders. The different degrees of freedom for temperature and relative humidity (RH) did not significantly alter the association between smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ and ED visits for anxiety disorders (Supplementary Figs. 4 and 5).

Discussion

The present study investigated the association between wildfire-related exposures and ED visits for anxiety disorders in the western United States from 2007 to 2018. Our findings suggest that wildfire is associated with an increased risk of anxiety disorders through multiple pathways. First, we identified a positive link between wildfire smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ and ED visits for anxiety disorders. While previous epidemiological evidence is inconclusive on this specific association, the mental health impact of ambient ${\mathrm{P}\mathrm{M}}_{2.5}$ has been widely documented. For example, ref. 21 found that per 10 μg m⁻³ increase in ${\mathrm{P}\mathrm{M}}_{2.5}$ was associated with an OR of 2.31 (95% CI = (2.15, 2.50)) for symptoms of nerves, anxiety, tension or depression from UK Biobank data²¹. Ref. 19 further revealed that certain ${\mathrm{P}\mathrm{M}}_{2.5}$ constituents, including organic carbon and elemental carbon, were associated with anxiety and depression symptoms¹⁹. In addition, our results indicate that the link between wildfire smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ and ED visits for anxiety disorders occurs within 48 hours post-exposure. This finding aligned with previous studies that showed exposure to ambient ${\mathrm{P}\mathrm{M}}_{2.5}$ was associated with ED visits for anxiety disorders in the same day²⁸ and with a 1 day lag²⁹. Given that the effects of wildfire smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ cannot be completely separated from those of ambient ${\mathrm{P}\mathrm{M}}_{2.5}$ in our study (as illustrated in Fig. 1), it is plausible that wildfire smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ may induce neuroinflammation and exhibit neurotoxic effects through mechanisms similar to those of ambient ${\mathrm{P}\mathrm{M}}_{2.5}$, which have been supported by multiple toxicological studies^30,31.

In addition to the inhalation of smoke ${\mathrm{P}\mathrm{M}}_{2.5}$, our findings suggest that wildfire has a direct psychological impact that increases the risk of anxiety disorders. Specifically, we found that a major smoke event (in which smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ contributed ≥75% of the total ${\mathrm{P}\mathrm{M}}_{2.5}$) was associated with a 6.3% increase in the risk of ED visits for anxiety disorders in the general population (OR = 1.063 (1.015, 1.113), P = 0.009). During such major smoke events, the average wildfire smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ concentration was 53.56 ± 30.14 μg m⁻³. This concentration corresponds to a cumulative OR of only 1.030 (1.000, 1.061) if we assume a linear concentration–response relationship for smoke ${\mathrm{P}\mathrm{M}}_{2.5}$. The discrepancy between these two ORs suggests that major smoke events may be associated with ED visits not solely through the neurotoxicity of ${\mathrm{P}\mathrm{M}}_{2.5}$. In addition, adjusting for background ${\mathrm{P}\mathrm{M}}_{2.5}$ slightly modified but did not change the significance of major smoke events, highlighting that the effect of major smoke events is partially independent of background ${\mathrm{P}\mathrm{M}}_{2.5}$. Furthermore, we also identified that the number of active fire spots and FRP were associated with ED visits for anxiety disorders among older adults, and these effects were independent of background ${\mathrm{P}\mathrm{M}}_{2.5}$. Taken together, our findings suggest that the direct psychological impact of wildfires also contributed to their association with anxiety disorders, which possibly explains why previous studies reported a surge in the prevalence of anxiety disorders after megafire events^25,32.

Our stratified analyses further suggest that women and girls are more susceptible to wildfire-related exposures with regard to anxiety disorders. Specifically, significant associations between smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ and anxiety disorders were observed only among adult women and all older adults. Besides, a smoke event is positively associated with anxiety disorders in all women and girls, while only major smoke events are linked to anxiety disorders among men and boys. The sex disparity is possibly attributable to sex hormones as previous literature has reported that estradiol and progesterone are associated with increased vulnerability to anxiety, trauma, and stress-related disorders in women and girls^26,33. Such vulnerability not only results in an increased prevalence of anxiety disorders but also exacerbates the severity of anxiety symptoms and the overall burden of diseases in women and girls^12,14. The comprehensive impact of wildfires on women and girls needs to be emphasized since they are also more susceptible to the respiratory outcomes associated with wildfires³⁴.

The older adult population is also disproportionately affected by wildfires, especially in the western United States. Ref. 35 reported that regions in California that experience significant wildfire effects tend to have a higher concentration of older adult residents³⁵. Factors such as household wealth, insurance coverage, and health conditions may hinder the ability of older individuals to adapt and recover from the adverse effects of wildfires³⁶. Our findings indicate that, compared with other age groups, older adults are more vulnerable to the anxiety disorders associated with wildfire-related exposures. Living with anxiety is associated with higher outpatient and inpatient costs among the older adult population. It is estimated that the excess annual adjusted health-care costs of anxiety and of comorbid anxiety and depression reached $80.0 and $119.8 million per 1 million older adults in Canada³⁷. The total burden of disease attributable to wildfires is expected to grow among the older adult population, given the population aging and the continuous exacerbation of climate change in the United States.

Our study has several advantages. First, our study comprises 1,897,865 ED visits from five western US states from 2007 to 2018. To the best of our knowledge, this is the largest and most comprehensive research on the association between wildfire-related exposures and anxiety disorders; previous studies typically focused on a small region with several hundred subjects. Second, we used high-performance satellite-driven exposure datasets to quantify wildfire-related exposures. With this advance, we were able to identify multiple pathways through which wildfires are associated with an increased risk of anxiety disorders.

The present study also has some limitations. To begin with, all the ED visits data were at the zip code level. That might increase the risk of exposure misclassification due to the inability to ascertain precise residential addresses for all participants. In addition, while the active-fire-point data sourced from the Fire Information for Resource Management System (FIRMS) stands as the foremost available dataset currently, it is not without its limitations as a proxy for fire origins because we are not able to determine the actual range of the impacted of area. This might result in an underestimation of the health impact of active wildfire points, especially for those who suffer from the impact of the fire points but do not reside in the corresponding zip codes. Moreover, our study used ED visits for anxiety disorders as the outcome of interest. With this dataset, we are not able to obtain the previous anxiety diagnosis for the participants, which might be an effect modifier for the association between wildfires and anxiety. In addition, only a small portion of patients with anxiety symptoms will visit the emergency department. Therefore, we were not able to identify the association between wildfire-related exposures and milder anxiety symptoms. This may ultimately result in an underestimation of the mental health impact associated with wildfires.

Furthermore, the correlations between wildfires and a spectrum of other psychological conditions beyond anxiety disorders, including mood disorders, post-traumatic stress disorder, eating disorders, and sleep disturbances, remain incompletely understood. Further investigations are warranted to understand the extent to which wildfire events contribute to climate-related anxiety and the associated burden of disease across the diverse populations around the world.

Conclusions

In this study, we used satellite-driven exposure datasets to examine the association between wildfire-related exposures and ED visits for anxiety disorders in the western United States from 2007 to 2018. Our results suggest that wildfires are associated with an increased risk of anxiety disorders through two different pathways: the inhalation of smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ and the direct psychological impact of smoke events and active fire points. The age- and sex-stratified analysis further revealed that women and girls and the older adult population are more susceptible to anxiety disorders associated with wildfire-related exposures.

In light of the escalating wildfire activity amid the global climate crisis, our study underscores the urgent need for immediate and effective climate actions. The evident disparities in exposure, susceptibility, and resiliency to climate-related disasters call for enhanced disaster risk reduction and climate risk management strategies, including climate awareness and risk communication tailored for vulnerable subpopulations such as older adults and women and girls. These steps are crucial to decelerate the pace of climate change, protect the vulnerable population from the climate crisis, and consequently, promote mental health—a fundamental human right—in the western United States and globally.

Methods

ED visits for anxiety disorders

We obtained data on ED visits from five western US states: Arizona (2010–2018), California (2007–2018), Nevada (2009–2016), Oregon (2014–2018), and Utah (2007–2016). Anxiety disorders, including anxiety, dissociative, stress-related, somatoform, and other nonpsychotic mental disorders, were defined as the ICD-10 codes F40–F48. Their corresponding ICD-9 codes (300, 308, 309.0, 309.1, 309.24, 309.28, 309.29, 309.3, 309.4, 309.8, and 309.9) were applied for records before 1 October 2015 (Supplementary Table 1). Our study included only patients whose primary reason for the ED visit was anxiety disorders. This selection criterion was employed to mitigate potential confounding biases arising from acute panic responses associated with other medical conditions, such as asthma and various respiratory diseases. This study was approved before being conducted by the Emory University Institutional Review Board (STUDY00004823).

Exposure assessments

Satellite remote sensing is a powerful tool for measuring air pollution^38–40 and dynamic environmental events such as wildfires⁴¹, cyclones⁴², and lightning strikes⁴³ due to its unparalleled advantage in spatiotemporal coverage. Our study utilized a satellite-driven dataset to determine levels of exposure related to wildfires. Given the potential pathways through which wildfires may be associated with anxiety disorders, our exposures of interest included wildfire smoke ${\mathrm{P}\mathrm{M}}_{2.5}$, smoke events, major smoke events, active wildfire points, and the cumulative FRP of the active wildfire points. Specifically, exposure to wildfire smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ was assessed with a 1 km satellite-driven model. In brief, we considered two different scenarios of ${\mathrm{P}\mathrm{M}}_{2.5}$ pollution: total ${\mathrm{P}\mathrm{M}}_{2.5}$ and background ${\mathrm{P}\mathrm{M}}_{2.5}$ (non-smoke ${\mathrm{P}\mathrm{M}}_{2.5}$). The total ${\mathrm{P}\mathrm{M}}_{2.5}$ model was trained with the data from regions that were affected by wildfire smoke. The background ${\mathrm{P}\mathrm{M}}_{2.5}$ model was trained with data from the smoke-free regions (regions that were not affected by wildfire smoke). We then used both models to predict ${\mathrm{P}\mathrm{M}}_{2.5}$ concentrations under different scenarios and subtract the background ${\mathrm{P}\mathrm{M}}_{2.5}$ from the total ${\mathrm{P}\mathrm{M}}_{2.5}$ to get wildfire smoke ${\mathrm{P}\mathrm{M}}_{2.5}$. The detailed description of this model has been documented elsewhere⁴⁴. We further aggregated wildfire smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ and background ${\mathrm{P}\mathrm{M}}_{2.5}$ to the zip code level to match the ED visits data. We did a vaguer aggregation (aggregate by the first four digits of the zip codes) for all patients in Nevada, given that the full zip codes were not available for this state. We define smoke events and major smoke events as smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ contributed ≥25% and ≥75% of the total ${\mathrm{P}\mathrm{M}}_{2.5}$ in the past 48 hours, respectively.

The moderate-resolution imaging spectroradiometer 1 km daily active wildfire points and their corresponding FRP from 2007 to 2018 were obtained from the FIRMS data archive⁴⁵. The fire-spot location is reported as the centroid of the 1 km² pixel. Although the actual burning area can be smaller than the pixel size, these fire spots must be sufficiently large to emit strong thermal Infrared radiation detectable from space. We further excluded the low-quality active fire points with a confidence score less than or equal to 33%. This study included two proxies of exposure to wildfire: the number of active wildfire points within a given zip code and the cumulative FRP for all those fire points. The state of Nevada was excluded from this analysis given that the complete zip code information was unavailable.

Meteorological data

The Daymet daily maximum and minimum temperatures in degrees Celsius and vapor pressure in pascals at 1 km resolution from 2007 to 2018 were obtained from the Oak Ridge National Laboratory data archive⁴⁶. We further calculated the daily mean temperature by averaging the daily maximum and minimum temperatures. Relative humidity was calculated using the Magnus formula⁴⁷. Similar to wildfire smoke ${\mathrm{P}\mathrm{M}}_{2.5}$, we also aggregated the Daymet meteorological data to the zip code level to align with ED visits records.

Statistical analysis

We used a case-crossover design to investigate the association between wildfire exposure and ED visits for anxiety disorders. Specifically, if an ED visit occurred on day $X$, we used all the other dates within the same month that fell on the same day of the week as day $X$ to serve as self-control days. The exposures of interest included (1) 2 day average wildfire smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ (the day of the ED visits and the previous day); (2) the occurrence of a wildfire smoke event (wildfire smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ contributed ≥25% of the total ${\mathrm{P}\mathrm{M}}_{2.5}$ in the past two days); (3) the occurrence of a major wildfire smoke event (wildfire smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ contributed ≥75% of the total ${\mathrm{P}\mathrm{M}}_{2.5}$ in the past two days); (4) the total number of active wildfire points within a given zip code in the past two days; and (5) the cumulative FRP of all the wildfire points within a given zip code in the past two days.

We employed a two-stage conditional logistic regression model to examine the association between the exposures of interest and ED visits for anxiety disorders. Stage one (model 1) considered only the effects of the exposures of interest, while stage two (model 2) controlled for background ${\mathrm{P}\mathrm{M}}_{2.5}$. To control the influences of temporal trends, holidays, and meteorological conditions, common covariates for both models included federal holiday indicators and natural splines for the day of year (DOY), mean temperature, and RH. The degrees of freedom were four for DOY and six for temperature and RH. The detailed model specification is shown as equation (1):

$$\begin{gathered} \text{Logit}P\left(Y\right)={\beta }_0+{\beta }_1\text{exposures}\text{of}\text{interest}+{\beta }_2\text{background}{\mathrm{P}\mathrm{M}}_{2.5} \\ +{\beta }_3\text{holidays}+\mathrm{n}\mathrm{s}\left(\text{DOY}\right)+\mathrm{n}\mathrm{s}\left(\text{temp}\right)+\mathrm{n}\mathrm{s}\left(\mathrm{R}\mathrm{H}\right) \end{gathered}$$ (1)

where $Y$ denotes an ED visit for anxiety disorders and $P\left(Y\right)$ is the probability of the occurrence of that visit; $\mathrm{n}\mathrm{s}\left(\right)$ denotes the natural splines; exposures of interest included 2 day average smoke ${\mathrm{P}\mathrm{M}}_{2.5}$, smoke events, major smoke events, the total number of active fire points within a 2 day period, and their cumulative FRP; each exposure of interest was analyzed independently. Model 1 incorporated the exposures of interest and other covariates, while model 2 further adjusted for background ${\mathrm{P}\mathrm{M}}_{2.5}$ levels.

Although an advantage of a case-crossover study is that the individual-level time-invariant variables (for example, age, sex, and race) are automatically controlled for in the design stage, we conducted multiple stratified analyses to determine whether the association between wildfire-related exposures and ED visits for anxiety disorders varied across different population groups. Specifically, we evaluated whether the association varied across different sexes (men and boys/women and girls), age groups (children and adolescents aged 5–17, adults aged 18–64, and older adults aged 65 and above), races (white/Black/Asian/other), and ethnicities (Hispanic/non-Hispanic).

We also conducted some sensitivity analyses to see whether the results were stable under different model settings. First, we explored various lagged effects of smoke ${\mathrm{P}\mathrm{M}}_{2.5}$, ranging from the same-day exposure of the ED visits (day 0) to the five-day moving average (day 0 plus four previous days). Second, we investigated the effects of different degrees of freedom for DOY (4–8), temperature (4–6), and RH (4–6).

All analyses were performed with R (version 4.0.5). The package survival (version 3.2–10) was used to run the conditional logistic regression models.

Data availability

As part of the data use agreement with individual states, we are prohibited from sharing the ED visits data to protect identifiable health information. However, similar data used in this analysis can be directly requested from the corresponding health departments or hospital association. The Daymet daily 1 km meteorological data are publicly available in the Oak Ridge National Laboratory data archive (https://doi.org/10.3334/ORNLDAAC/2129). The active-fire-point data are publicly available in the FIRMS database (https://firms.modaps.eosdis.nasa.gov/active_fire/). The zip code level wildfire smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ data used for this study can be found on Figshare (https://doi.org/10.6084/m9.figshare.25016510). The original 1 km wildfire smoke ${\mathrm{P}\mathrm{M}}_{2.5}$ data are available on reasonable request from the corresponding author. Source data are provided with this paper.