Effects of air pollution on emergency room visits for mental disorders: risks and effect modification by comorbid physical disorders and personal characteristics

Emerging evidence suggests that exposure to ambient air pollution negatively impacts mental health. However, little is known about potential moderators of such effects, including pre-existing physical disorders and personal characteristics. This study investigated if the effects of daily changes in levels of air pollutants among individuals with mental disorders (MDs) vary depending on comorbid physical disorders, age, sex, or race/ethnicity. We used a time-stratified case-crossover design that compared the levels of PM2.5 and NO2 on days leading up to MD-related emergency room (ER) visits to levels on control days. The sample consisted of individuals with MDs for their primary diagnosis, including both patients with a second diagnosis and those without. We conducted a stratified analysis to examine potential effect modifications by individuals’ demographic characteristics (sex, age, and race/ethnicity) or a diagnosis of a physical disorder (cardiovascular disease, diabetes mellitus, and respiratory disease). We found that both PM2.5 and NO2 were significantly associated with ER visits for MDs. Per 10 µg m−3 increase in daily PM2.5 and per 10 ppb increase in NO2 concentration were associated with 1.07% (95% CI: 0.81%, 1.34%) and 0.56% (95% CI: 0.42%, 0.69%) increase in ER visits for MDs, respectively. We also found significantly greater susceptibility among younger persons (below 18 years old), Black, and individuals with respiratory disease. Exposures to both PM2.5 and NO2 were significantly associated with ER visits for MDs, and these adverse effects were more pronounced among youth, Black and individuals with respiratory disease as a comorbid physical disorder.


Introduction
Mental disorders (MDs) are among the top ten leading causes of disease burden worldwide (Patel et al 2018, GBD 2019 Mental Disorders Collaborators and others 2022) with a projected annual economic toll rising to $6 trillion by 2030 (Trautmann et al 2016, The Lancet Global Health 2020).Furthermore, MDs are frequently intertwined with physical disorders, potentially complicating the treatment of both conditions (Fleet et al 1996, Ortega et al 2006, Cohen et al 2007, Scott et al 2007).Given the prevalence and burden of mental illness, improving our understanding of relevant risk factors, especially modifiable environmental factors, is crucial in order to take steps towards disease control and prevention (Wei et al 2022).
Emerging studies (Basu et al 2018, Engemann et al 2019, Yoo et al 2021, 2022b, Nori-Sarma et al 2022, Wang and Li 2023) have demonstrated adverse effects of both ambient temperature and a lack of greenspace on emergency room (ER) visits for MDs.Likewise, there is some evidence that ambient air pollution also affects ER visits for MDs, including depression, schizophrenia, substance abuse, and anxiety disorders (Szyszkowicz 2007, Cho et al 2014, Szyszkowicz et al 2016, 2018, Pun et al 2017, Thilakaratne et al 2020, Nguyen et al 2021, Yoo et al 2022a).In a recent meta analysis on air pollution exposures and depression,

Environmental data
We used 1 × 1 km 2 gridded daily PM 2.5 and NO 2 data for NYS during 2005-2016 from an ensemble-based model, which integrated three machine learning algorithms (specifically, neural network, random forest, and gradient boosting algorithms) and inputs from satellite data, meteorological data, chemical transport model outputs, and land use data (Di et al 2019a(Di et al , 2019b)).The predication estimates had good performance with a 10-fold cross-validated R 2 of 0.86 and 0.79 for daily PM 2.5 and NO 2 , respectively Daily temperature for NYS during the study period (2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016) was acquired from Daymet (Thornton et al 2020).Daymet model outputs 1 km gridded daily estimates of weather conditions, including minimum temperature and maximum temperature, using inputs from ground based metrological stations for all North America.We calculated daily mean temperature by taking an average of daily maximum and minimum temperatures.The daily mean relative humidity was also calculated based on both daily minimum and maximum temperature of each 1 km grid (Thornton et al 1997, Spangler et al 2019).Detailed equations used to calculate the daily relative humidity can be found in appendix B.

Statistical analysis
Case-crossover design is widely used to estimate acute associations of time-varying exposures, including ambient air pollutant concentrations and extreme temperature, with daily mortality or morbidity in a particular geographic region (Janes et al 2005, Armstrong et al 2014).Given a sample of subjects who experienced the event, the level of environmental exposure associated with event ('case') is compared with exposure at 'control' (or 'referent') times.
In the present study, we used a time-stratified case-crossover design with a conditional logistic regression model to estimate the association between air pollution exposure and ER visits for MDs.Specifically, the model formulation for the conditional logistic model with a single pollutant model (i.e.PM 2.5 and NO 2 ) can be written in equation ( 1) as: Y ij indicates whether the event that the i-th individual' ER visit for mental disorder occurs in time interval j in the pre-specified reference window k(j) (Y ij = 1, event; 0, not).The set of reference periods is denoted as k(j) whose size may vary per month.The spatially and temporally resolved exposures at the location s i (i.e. the residence of i-th individual) and time j (day) for an air pollutant (PM 2.5 or NO 2 ) and temperature, and relative humidity are denoted as X s i jl , Temp s i i , RH s i j , respectively.β represents the log-relative risk of ER visits for MDs associated with the l-th air pollutant X s i jl .The smoothing function is denoted as ns with the corresponding degree of freedom df 1 , df 2 for temperature and relative humidity, respectively.For each individual, the levels of PM 2.5 and NO 2 exposures on the day when the individual visited the ER for MDs ('case' day) were compared with the exposure estimates on the other days in the same calendar month, referred to as 'control' days hereafter.By selecting control days from the same day of week within the same month of the case day, we assessed the associations between changes in individuals' exposure to air pollutants (i.e.PM 2.5 and NO 2 ) and their mental health outcomes while effectively controlling for long-term temporal trends, seasonality, day of the week, and potential time-invariant confounders, such as the individual's age, sex, socioeconomic conditions, and other fixed participant characteristics (Janes et al 2005, Carracedo-Martnez et al 2010).
In the conditional logistic regression model, we investigated the effects of PM 2.5 and NO 2 exposures on ER visits for MDs at both single day lags (from the day of ER visit to 3 days prior to ER visit; i.e. lag 0 -lag 3) and moving average lags for the period up to 3 lagged days (i.e.lag 01 -lag 03).A set of measurements estimated for control days (three to four days depending on the month of year) were also included.We further adjusted for potential time-varying confounders by using natural cubic splines with 5 degrees of freedom (df) for ambient daily temperature and 3 df for daily mean relative humidity.To determine the main lag, we conducted a sensitivity analysis across different lag structures by fitting separate models with the same parameter settings as in the main model.We also developed a two-pollutant model where both PM 2.5 and NO 2 are simultaneously included to ensure the robustness of our results.
Stratified analyses were conducted for PM 2.5 and NO 2 exposure effects on ER visits for MDs to examine effect modification by sex (female and male), age groups (below 18 years old, 19-34 years old, 35-49 years old, 50-64 years old, and above 65 years old), and race/ethnicity (non-Hispanic Black (Black), non-Hispanic White (White), non-Hispanic Asian (Asian), and Hispanic), and comorbid physical disorders that potentially increase sensitivity to air pollutants.In terms of the latter, we selected three secondary diagnoses of physical disorders that are frequently studied at the individual level (Denollet et  For the stratified analyses by comorbid physical disorders, we calculated the relative effect modification (REM) index, which is the ratio between the odds ratio (OR) when comorbidity is present and when the comorbid health condition is absent (Altman and Bland 2003, Lavigne et al 2014, Oudin Aström et al 2020).In the case of comorbid physical disorders, the REM would be interpreted as the relative increase in MD-related ER visit for persons with a comorbid physical health condition compared to those without (i.e.ED visits primarily for MD without a specific comorbidity condition).The statistical significance of the REM was tested with the calculation of its 95% confidence interval.
Results are presented as the percent change (i.e.[OR-1] × 100) in daily ER visits for MDs associated per 10 µgm 3 increase in ambient PM 2.5 and 10 ppb increase in ambient NO 2 concentrations at the main lag.Estimates were considered statistically significant if the 95% posterior interval did not overlap with zero.The statistical significance of effect modification was tested by calculating the 95% confidence interval (95% CI) of the difference between the effect estimate within each pair of subgroups (Zeka et al 2006, Wang et al 2018, Chen et al 2019): where Q1 , Q2 denote the model coefficients in the conditional logistic regression model for sex, age group, race/ethnicity, and comorbid physical conditions.In equation ( 2), the corresponding standard errors are denoted as σ2 1 , σ2 2 , respectively.All statistical analyses were conducted in R software (version 4.2.2) using survival and spline packages to analyze conditional logistic regression models.

Results
A total of 3655 497 ER visits were made in NYS between 1 January 2005 and 31 December 2016 with mental health-related illnesses as the primary reason for the visit.Among them 1748 556 visits (≈48%) were made only for MD (i.e.absence of secondary diagnosis), and 1906 941 visits (≈52%) were made for MD and a comorbid disorder being present as a secondary diagnosis, which could include any additional physical or mental disorder.The characteristics of ER visits for a MD as the primary diagnosis (regardless of presence of secondary diagnostics) are summarized in table 1.Most patients were 19 to 34 years of age (1150 219; 31.47%),35 to 49 years old (1015 124; 27.77%) and 50 to 64 years old (707 839; 19.36%).Relatively smaller percentages of cases were aged 18 or younger (558 158; 15.27%) or 65 or above (224 157; 6.13%).There were more visits (2164 544; 59.21%) by males than females.We also found that White patients made up 43.03% (1572 877) of ER visits for MDs, while 24.19% (884 286) were by Black patients, 16.26% (594 384) by Hispanic patients, and 1.76% (64 401) by Asian patients.
The environmental conditions, including the multi-year average of the daily mean temperature and relative humidity, and PM 2.5 and NO 2 , are summarized in table 1.The study period average of the daily mean temperature was 10.03 • C with standard deviation 10.20 • C (range: −17.63 • C to 31.23 • C) and the mean daily relative humidity was 69.56% with standard deviation 10.80% (range: 28.23% to 95.80%) also shows a summary of each pollutant's levels.Pearson correlation coefficients showed that NO 2 was negatively correlated with mean temperature (r = −0.08;p-value < 0.001) while PM 2.5 was positively correlated with mean temperature (r = 0.19; p-value < 0.001).The muti-year average (2005-2016) of PM 2.5 and NO 2 are illustrated in figure 1, which were positively correlated (r = 0.35; p-value < 0.001).
The comorbid conditions with MDs are summarized in table 2. A total of 1906 941 individuals visited the ER for MDs as a primary diagnosis with underlying physical or mental disorders, which included CVD 136 608 (7.16%), diabetes mellitus 47 798 (2.51%), and respiratory disease 65 522 (3.44%).
The short-term effect of air pollution exposure for both PM 2.5 and NO 2 on daily ER visits for MDs are presented in table 3. The main lag for the subsequent analysis was determined as two days prior to ER visits (lag 2) by the sensitivity analysis whose details are summarized in appendix D. Results based on this two-day lag show the percent change in ER visits for MDs per 10 µg m −3 increase in daily PM 2.5 concentration (1.07%; 95% CI: 0.81%, 1.34%) and per 10 ppb increase in NO 2 concentration (0.56%; 95% CI: 0.42%, 0.69%).In other words, elevated levels of either PM 2.5 or NO 2 were associated with increased ER visits for MDs two days subsequent to the exposure in the sample as a whole.Similar results were found when both PM 2.5 and NO 2 were simultaneously considered, suggesting that the results were generally robust for the inclusion of a two-pollutant model (see appendix E).Note: N represents the total observations (i.e., a total number patients who visited ER for MDs during the study period and the total number of days over which environmental were collected, respectively).Mean refers to the average number of daily ER visits during the study period and the multi-year average of daily environmental condition.SD denotes standard deviation, and Q1, Q2, and Q3 denotes 25, 50, and 75% of the distribution, respectively.In terms of potential effect modification, analyses focused on comorbid physical disorders found that individuals with a secondary diagnosis of respiratory diseases were significantly more susceptible to PM 2.5 (but not NO 2 ) than those without secondary diagnosis.As presented in table 4, the REMs for CVD, diabetes mellitus, and a variable reflecting the presence of any of these three secondary physical disorders were not Table 3. Percent changes in daily ER visits for MDs and 95% confidence interval (CI) per 10 µgm −3 increase in PM25 and per 10 ppb increase in NO2.

PM2.5 NO2
Total cases 1.07 statistically significant indicating that the potency of air pollution exposure did not vary as a function of these conditions.In terms of demographic variables, there was no evidence of differential effects for sex for either PM 2.5 or NO 2 (see table 3).In contrast, there was evidence of differential effects depending on patients' age and race/ethnicity.In terms of the former, the youth (age 0-18 years old) were significantly more susceptible to both PM 2.5 and NO 2 exposures than each of the other adult population age groups (all p-values ⩽ 0.001), while none of the other age groups differed from each other.In the case of race/ethnicity, Black patients were significantly more susceptible to PM 2.5 (1%; 95% CI: 0.49%, 1.52%) and NO 2 (0.73%; 95% CI: 0.48%, 0.99%) compared to White patients (p-value ⩽ 0.044 for PM 2.5 and p-value ⩽ 0.018 for NO 2 ), whereas none of the other pairwise comparisons were statistically significant.

Discussion
We examined the effects of acute air pollution exposure on ER utilization for MDs and potential effect modifications by comorbid physical disorders and personal characteristics using 3.6 million ER visits that occurred in NYS, United States, between 2005 and 2016.To our knowledge, this is one of first studies to examine whether the effects of ambient air pollution exposure on mental health outcomes vary depending on pre-existing physical disorders, or other characteristics such as sex, age, and racial/ethnic identification, in the US at the population level.Unlike previous studies that were based on specific age groups (e.g., the young or elderly) or relatively small samples of the population, the present study used a representative sample that cut across all age groups and race/ethnicities within the state of New York.Using a case-crossover design, we examined (1) the effect of air pollution exposure on ER visits for MDs and (2) the differential susceptibility to the effects of air pollution exposure by subgroups defined by the presence of pre-existing physical disorders and demographic characteristics.
Consistent with previous studies (Cho et al 2014, Szyszkowicz et al 2016, Oudin et al 2018, Brokamp et al 2019), we found evidence that elevations in either PM 2.5 and NO 2 were associated with increased likelihood of visiting an ER for a mental disorder.Of particular interest, effects were strongest at a 2 day lag.In other words, increased ER visits for MDs were occurring two days subsequent to exposure to these pollutants.This result will potentially help inform our understanding pathways and mechanisms by which acute exposures to these pollutants leads to ER as it constrains the mechanism to something that unfolds over the matter of days.
We also examined whether the strength of the effect of air pollution exposure on ER visits for MDs varied as a function of pre-existing physical health conditions.Based on their high frequency as a secondary diagnosis among individuals who visit ERs for MDs in our sample, we focused on CVD, diabetes mellitus, and respiratory disease.The present study found the statistically significant association between ambient air pollutants and ER visits for MDs varied depending on pre-existing health conditions.Specifically, results indicated that patients with respiratory disease were significantly more susceptible to short-term exposure to PM 2.5 (but not NO 2 ) than in those without a specific comorbidity condition, including respiratory disease, in terms of ER visits for MDs.
Although our study did not find evidence of effect modification for diabetes mellitus or CVD, several previous studies reported that CVD is associated with increased sensitivity to air pollutants in terms of mental health outcomes.For example, Kim et al (2010), Cho et al (2014), and Kim et al (2016) reported that both long-term and short-term exposures to ambient particulate matter (PM 10 and PM 2.5 ) were more strongly associated with suicidality and depression among individuals with pre-existing CVD conditions compared to those who did not have these medical conditions.However, there are potentially critical differences in sample characteristics across these studies.For example, Pun et al (2017) found significant effect modification of associations between long-term PM 2.5 exposure and major depressive disorders by stroke and COPD in a racially and ethnically diverse population.In contrast, Cho et al (2014) did not observe any significant risk in individuals with underlying COPD within a racially homogeneous population.We suspect that the mixed results are due to the limited investigation of representative samples and differences in a study design (i.e. a diagnosis of pre-existing condition and the scope of physical illness, a long-term and short-term exposure, and the misclassification of exposure estimates).
Finally, the present study found that the youth (age 0-18 years old) were significantly more susceptible to exposure to both PM 2.5 and NO 2 than other age groups.In other words, air pollutants had stronger effects on the younger population compared to adults.These results are consistent with findings from existing studies.For example, Brokamp et al (2019) found that acute exposure to ambient PM 2.5 increased the psychiatric ER utilizations among children and adolescents, although their studies further demonstrated that the effect modification from community deprivation was inconsistent across different lags.Similarly, Thilakaratne et al (2020) reported elevated risk from exposure to NO 2 among the youth.Furthermore, our study revealed that both PM 2.5 and NO 2 significantly increased the percent change of ER visits for MDs to a greater extent among Black compared to White patients.Racial discrepancies in physical health outcomes, such as CVD and respiratory diseases, associated with air pollution exposure, are well documented (Ostro et al 2001, Nishimura et al 2013, Humphrey et al 2019, Clougherty et al 2021, Ward-Caviness et al 2021, Grant et al 2022).However, studies on the increased susceptibility to mental illness among racial/ethnic minorities, including Black persons, associated with air pollution exposures are still rare, and some studies have reported non-significant effects among Blacks (Thilakaratne et al 2020, Nguyen et al 2021).
Why would some individuals, such as those with respiratory disease, youth and Blacks, be more susceptible to air pollutants than other individuals?We speculate that among individuals with respiratory diseases, exposure to air pollutants such as PM 2.5 likely generates physical and psychological distress in the form of difficulty breathing, anxiety, fear and irritability.Such escalating distress may lead to seeking care at the ER, particularly among those who have an existing mental disorder, such as panic disorder or generalized anxiety disorder.In the case of children this escalation of distress is likely quite alarming to parents leading them to seek emergency care for their children.In contrast to comorbid respiratory diseases, it seems less likely that individuals with other types of comorbid physical disorders, such as CVD or diabetes, would experience this type of acute and escalating physical and psychological distress following exposure to air pollutants.If confirmed by future research, it may be important for mental health treatment providers to address this potential pattern of escalating distress among their patients with comorbid respiratory diseases and develop mitigation strategies, which might include reducing exposure.Finally, why would individuals identifying as Black be more susceptible to the effects of air pollutants on ER visits compared to Whites?Here we speculate that race serves as a marker of disparities in access to healthcare, such that White persons may have more healthcare options available beside visiting the ER compared to Blacks (Chow et al 2003, Garland et al 2005, Alexandre et al 2010, Ault-Brutus 2012).In other words, when experiencing distress resulting from air pollutant exposure, Whites may be more likely to schedule an office visit with their primary care doctor, whereas Black persons be less likely to have that option available and instead they would rely on the ER.If confirmed by future research, these results would further highlight race/ethnicity based inequities in health care and the need for reform.We hope that future studies at the individual level directly test these post-hoc explanations.
The strength of the present study includes the use of time-stratified case-crossover study design, which enabled us to effectively control confounding effects, such as socio-economic conditions and seasonal and monthly variation associated with air pollution.Furthermore, air pollution exposure associated with the event and control days were estimated from the local average, i.e., 1 × 1 km 2 surrounding individuals' home address, rather than a regional average, which has been used in most previous studies.Consequently, our estimates of air pollution are likely more precise than those used in prior studies.
It is worth noting that the demographic composition of the present study might have an effect on our findings.According to the U.S. Census Bureau (2020), the ethical/racial composition of NYS is 56.4% White, 15.0% Black, 19.1% Hispanic, and 9.2% Asian.In our sample, the demographic composition of ER visits for MDs showed a similar pattern nearly 43% of ER visits were made by White, 24% by Black, 16% by Hispanic, and 1.7% by Asian patients.Consequently, our findings may not generalize to other regions with different ethnic/racial compositions.Lastly, although our case-crossover design controlled for individuals' socio-economic conditions, we did not examine whether this variable itself modified effects of air pollutants.Other study designs, such as ecological analysis at small spatial scales, might enable researchers to investigate individuals' neighborhood level risk factors.

Conclusions
The present study suggests that short-term exposure to ambient PM 2.5 and NO 2 increase ED visits for MDs two days subsequent to exposure.Further, some individuals are more susceptible to these effects than others.Specifically, short-term exposure to these air pollutants have stronger effects on the youth (0-18 years old), individuals who identify as Black, and those with pre-existing respiratory disease.Further research is needed to confirm our findings at the person-level and to determine the underlying mechanisms by which air pollution can put individuals with MDs at risk for ER visits.

Appendix E. Percent change in daily ER visits for MDs from two-pollutant model
We examined the cumulative effect of air pollution exposure on ER visits for mental disorders from the two pollutant model, which includes both PM 2.5 and NO 2 simultaneously.The results in table E.4 were comparable to the estimates obtained from a single model where the effect of PM 2.5 and NO 2 was assessed separately (see table 3).
Table E.4.Percent changes in daily ER visits for MDs and 95% confidence interval (CI) per 10 µgm −3 increase in PM25 and per 10 ppb increase in NO2 from Two-pollutant model.

Table 1 .
Summary of daily ER visits for MDs and environmental variables(New York state, 2005-2016).

Table 2 .
Number and percentage of ER visits for MDs by pre-existing physical disorders.

Table D . 2 .
Percent changes in daily ER visits for MDs and 95% confidence interval (CI) per 10 µg m −3 increase in PM25 and per 10 ppb increase in NO2 across different lag specifications.

Table D .
3. Percent changes in daily ER visits for MDs and 95% confidence interval (CI) per 10 µgm −3 increase in PM25 and per 10 ppb increase in NO2 by comorbid physical disorders and relative effect modification (REM) across different lag specifications.