The Combined Effects of Hourly Multi-Pollutant on the Risk of Ambulance Emergency Calls: A Seven-Year Time Series Study

Background: Ambulance emergency calls (AECs) are seen as a more suitable metric for syndromic surveillance due to their heightened sensitivity in reflecting the health impacts of air pollutants. Limited evidence has emphasized the combined effect of hourly air pollutants on AECs. This study aims to investigate the combined effects of multipollutants (i.e., PM2.5, PM10, Ozone, NO2, and SO2) on all-cause and cause-specific AECs by using the quantile g-computation method. Methods: We used ambulance emergency dispatch data, air pollutant data, and meteorological data from between 1 January 2013 and 31 December 2019 in Shenzhen, China, to estimate the associations of hourly multipollutants with AECs. We followed a two-stage analytic protocol, including the distributed lag nonlinear model, to examine the predominant lag for each air pollutant, as well as the quantile g-computation model to determine the associations of air pollutant mixtures with all-cause and cause-specific AECs. Results: A total of 3,022,164 patients were identified during the study period in Shenzhen. We found that each interquartile range increment in the concentrations of PM2.5, PM10, Ozone, NO2, and SO2 in 0–8 h, 0–8 h, 0–48 h, 0–28 h, and 0–24 h was associated with the highest risk of AECs. Each interquartile range increase in the mixture of air pollutants was significantly associated with a 1.67% (95% CI, 0.12–3.12%) increase in the risk of all-cause AECs, a 1.81% (95% CI, 0.25–3.39%) increase in the risk of vascular AECs, a 1.77% (95% CI, 0.44–3.11%) increase in reproductive AECs, and a 2.12% (95% CI, 0.56–3.71%) increase in AECs due to injuries. Conclusions: We found combined effects of pollutant mixtures associated with an increased risk of AECs across various causes. These findings highlight the importance of targeted policies and interventions to reduce air pollution, particularly for PM, Ozone, and NO2 emissions.


Introduction
Air pollution is the leading cause of enormous public health burdens due to substantial excess morbidity and mortality [1].Based on the 2019 Global Burden of Disease (GBD), ambient air pollution is a major determinant of disease burden, particularly in relation to cardiovascular disease-related morbidity and death [2].Previous studies have established a correlation between short-term exposure to air pollutants and the incidence Toxics 2023, 11, 895 2 of 12 of cardiovascular disease, respiratory disease (i.e., respiratory tuberculosis; pneumonia; pulmonary embolism), malignant neoplasm of the trachea, bronchus and lung disease, and even death [3].For example, a time-stratified study demonstrated that exposure to different particular metrics was associated with nonfatal myocardial infarction in Germany, from 2005 to 2015 [4].Another 12-year study conducted in metropolitan France showed that long-term exposure to particular matter (PM) pollutants was associated with all-cause mortality [5].However, evidence examining the associations of multipollutants with disease onset is still limited and inconclusive.Moreover, although the association between disease onset and air pollutants was determined by conducting multipollutant models in such limited previous studies, most multipollutant models were used to examine the robustness of the pollutant of interest by including other pollutants as covariates in the regression model [6,7].Notably, previous studies have not yet assessed the predominant lag period for each air pollutant; therefore, an inaccurate lag period would bias the results in the multipollutant model.
Ambulance emergency calls (AECs) are a more suitable metric for syndromic surveillance due to their heightened sensitivity in reflecting the health impacts of air pollutants [8,9].They have also been regarded as a promising alternative to other types of administrative health data for determining the impacts of air pollution because the exact location (in latitude and longitude) and time (in minutes and seconds) of the dispatch call are recorded in the AECs dataset [10].Although some previous studies have examined the association between air pollutants and AECs, these findings were inconsistent because of inaccurate disease grouping criteria [8,11].For example, Chen et al. found that PM pollutants and SO 2 were positively associated with the risk of AECs in Shenzhen city, China [8], but not in Ichiki et al.'s study in Japan [11] and Straney et al.'s study in Perth, Australia, [12] due to the use of different lag periods and performing varied study designs.In addition, different patterns of AECs can be attributed to different seasons and daytime/nighttime periods.Identifying these potential patterns can be useful in helping people avoid exposure, but more evidence is needed to clarify whether these patterns will trigger air pollutant-related AECs.
Given the scarce evidence determining the association between muti-pollutant and well-grouped AECs with an accurate lag period, as well as the lack of the classification of potential patterns (i.e., seasons and daytime/nighttime) to trigger pollutant-related AECs, it is pivotal to concentrate on these research gaps in this study.Thus, we hypothesized that short-term air pollutants could jointly increase the number of cause-specific AECs.Therefore, we conducted a distributed lag nonlinear model (DLNM) to examine the predominant lag for each air pollutant (i.e., PM 2.5 , PM 10 , Ozone, NO 2 , and SO 2 ); then, we adopted the quantile g-computation model to estimate the combined effects of multipollutant on all-cause and cause-specific AECs.

Study Participants
The Shenzhen Center for Prehospital Care established a system consisting of 120 Emergency Medical Services (EMS) in 1994, comprising 73 emergency health institutions and 103 emergency stations.We used ambulance emergency dispatch data from 1 January 2013 to 31 December 2019 that were obtained from the Shenzhen Ambulance Emergency Center.We included patients aged 18 and older who utilized ambulance services for hospital transportation between 2012 and 2019 and had documented transfer time records.We excluded those with incomplete dispatch or delivery times, ambiguous prehospital diagnoses, or missing diagnostic data.Additionally, repeated calls from the same individuals were excluded.

Outcome Ascertainment
Our primary outcome was all-cause AECs, which we further classified into four causespecific categories, including AECs due to vascular, respiratory, reproductive diseases, and

Air Pollutant Exposure Ascertainment and Covariates
In this study, we obtained the hourly average city-level air pollutant concentration as the exposure for patients from the National Environmental Monitor Station during the study period.There are 23 city-level environmental monitoring stations in Shenzhen, which are located in Futian, Nanshan, Luohu, and other districts (Figure S1).We conducted a well-established distributed lag nonlinear model (DLNM) for nonlinear concentrationresponse and nonlinear lag-response functions to determine the predominant lag period, and we calculated the mean concentration of the predominant lag for each air pollutant as our exposure to be used in the quantile g-computation model [13][14][15].We defined each interquartile range (IQR) increment as the difference between the 25th and 75th centiles for each air pollutant.The primary model adjusted covariates as follows: an indicator day-of-week variable to account for short-term weekly variation; natural spline functions with 3 degrees of freedom (df) for temperature and relative humidity [13].
We collected the relevant demographic information such as age, sex, and admission date for each diagnosis.In our study, the cool season spans from 1 November to 31 March, while the warm season is from 1 April to 31 October.We segmented the time of day into daytime (07:00-19:00) and nighttime (20:00-6:00).New Year's Day, Chinese New Year's Day, Dragon Boat Festival, Tomb-Sweeping Day, Labour Day, and National Day were regarded as Chinese public holidays.We used an indicator variable for the day of the week to account for the short-term weekly variations.We generated a 'day of year' variable (0-365 days) to consider underlying temporal trends.In addition, we gathered meteorological data, including three-hour temperature and three-hour relative humidity, from 2013 to 2019 in Shenzhen from the World Meteorological Database.

Statistical Analysis
We followed a two-stage analytic protocol to estimate the associations between combined air pollutants (i.e., PM 2.5 , PM 10 , Ozone, NO 2 , and SO 2 ) and all-cause AECs as well as cause-specific AECs (i.e., vascular diseases, respiratory diseases, reproduction illnesses, and injuries), which included the distributed lag nonlinear model (DLNM) and the quantile g-computation model.
In the first stage, we determined the association of each air pollutant with AECs by using conditional logistic regression models to estimate odds ratios (ORs) with 95% confidence intervals (CIs) and examined the predominant lag for each air pollutant.We modeled the concentration-response function using a natural cubic B spline with 2 knots to account for potential nonlinear relationships and modeled the lag-response function using a linear function with 2 knots placed on the log scale of lags up to 48 h [10,16,17].
In the second stage, we assessed the associations of air pollutant mixtures with allcause and cause-specific AECs by conducting quantile g-computation models.Quantile g-computation has been regarded as a simple and efficient statistical method for evaluating the combined effect of air pollutant mixtures when all pollutants simultaneously increase each quantile, which allows the combined air pollutants to present positive, negative, and null effects simultaneously [18,19].In this study, we implemented this model by categorizing PM 2.5 , PM 10 , Ozone, NO 2 , and SO 2 into quantiles, coded as 0, 1, 2, 3, and 4, respectively.The quantile g-computation model equation is as follows (confounders Z could also be included): where ∑ p j=1 β j = ψ and each air pollutant is given a negative or positive weight.The quantile g-computation estimator of the exposure-response relationship is the sum of the regression coefficients across the included exposures [20].Afterward, the excessive relative risk (ERR) with 95% CIs expressed the associations (excessive relative risk indicated that the risk of all-cause AECs resulted from each interquartile range increment of the combined effect of the mixture of air pollutants).The formula for ERR is as follows: Apart from this, we also conducted a subgroup analysis according to sex (male, female), age group (<20 years old, 20~64 years old, ≥65 years old), and season to determine the associations in a variety of subgroups.
In the sensitivity analysis, we further generated the nonlinear and non-homogeneous quantile g-computation model to check the stability of the results.Moreover, we extracted representative subtypes of vascular diseases (i.e., myocardial infarction, acute coronary syndromes, and stroke) and respiratory diseases (pneumonia, asthma, and upper respiratory infectious), as well as those associated significantly with air pollutants, to explore the potential differences among the main diseases.All analyses were performed in R version 4.2.1, and the statistical significance was set at a two-tailed p-value < 0.05.
The combined effect of the air pollutant mixture included a relatively predominant positive effect, with the sum of positive coefficients (β) equaling 0.0376, and a relatively slight negative effect, with the sum of negative coefficients (β) equaling −0.0204 (Figure S5).PM 10 (45.5%),NO 2 (23.4%), and Ozone (13.3%) exhibited positive weights with the same direction in the overall effect of air pollutants, whereas negative weights were observed among PM 2.5 (52.6%) and SO 2 (22.4%).

Discussion
In our large-scale study in Shenzhen, China, we found that exposure to PM 2.5 , PM 10 , Ozone, NO 2 , and SO 2 was associated with an increased risk of the acute exacerbation of chronic conditions (AECs).The associations were significant in the older population and varied by time of day.These findings have important implications for developing an early warning system to mitigate pollutant-related AECs in the context of global climate change.
Our study uniquely examines the combined effects of air pollutants on all-cause and cause-specific AECs using hourly air pollutant data, which provide valuable insights for medical rescue operations.This information is beneficial for emergency departments in optimizing ambulance arrangements and reducing response time delays.By understanding the specific effects of air pollutants on AECs at an hourly level, healthcare providers can enhance their preparedness and response strategies, ultimately improving patient outcomes.The findings from our study align with a robust body of previous research, consistently demonstrating positive associations between hourly air pollutants and the all-cause and cause-specific acute exacerbation of chronic conditions (AECs) [8,9,[21][22][23][24].For example, a time-stratified study by Chen et al. in Shenzhen, China, showed that PM 10 and PM 2.5 had immediate adverse effects on hourly AECs, persisting for 5 h; meanwhile, Ozone exhibited a longer lag pattern [8].Similarly, a study in Luoyang City, China, found that PM 2.5 and PM 10 had acute health effects on all-cause morbidity at the current hours, while NO 2 and SO 2 were positively associated with all-cause morbidity at the lag period exceeding 18 h [9].Another study by Liu et al. showed that exposure to PM 2.5 significantly elevated the risk of all-cause AECs, as well as respiratory-and cardiovascular-related AECs in Chengdu, China, between 2013 and 2015 [21].Moreover, Rao et al. conducted a case-crossover study by using British Columbia Emergency Health Service data from 2010 to 2015 in Canada, which showed a positive association between PM 2.5 exposure and cardiovascular disease-related AECs, as well as a slight increase in respiratory disease-related AECs within 48 h [22].
Existing evidence primarily focuses on single-and two-pollutant models when analyzing the association [6,7,[25][26][27][28].However, it is crucial to examine the combined effects of hourly air pollutants on the risk of AECs, considering that humans are typically exposed to multiple pollutants simultaneously in the real world.In this study, we employed the quantile g-computation model to assess the combined effects of hourly air pollutant mixtures and identify the contribution of each pollutant.Previous studies used the same statistical approach [29,30], but they mainly focused on long-term air pollutants and cardiovascular morbidity and mortality.For example, Li et al. conducted a cohort study involving 31,162 individuals, demonstrating that exposure to a mixture of air pollutants increased the risk of 10-year atherosclerotic cardiovascular diseases, with PM pollutants accounting for 54% of the mixture effect, followed by Ozone at 42% [29].Another study including 56,090 person-years found that a quartile increment in the mixture of Ozone, PM 2.5 , PM 1 , and NO 2 was significantly associated with a 137% (95% CI: 130%, 144%) increase in cardiovascular risk in the Beijing-Tianjin-Hebei region during the period from 2017 to 2021 [30].
Our study revealed a positive association between exposure to hourly air pollutant mixtures and vascular-related AECs.This could be explained by air pollutants increasing the calcification of coronary arteries and the intima-media thickness of the common carotid artery, further resulting in the rupture of atheromatous plaques or the formation of thrombosis [31,32].Given that air pollution has been regarded as an unavoidable environmental disturbance for vehicle traffic and might cause traffic accidents by worsening drivers' physical and psychological conditions, another finding that exposure to a mix of air pollutants increased the risk of injury-related AECs in our study was consistent with this explanation [33].In the subgroup analysis, we observed stronger associations of cause-specific AECs in the daytime, which could be explained by the fact that people often spend more time outdoors during the day and that they were estimated based on fixed-site monitors [13].Moreover, there were no associations of air pollutants with AECs stratified by sex in our study.These results were in line with previous literature [34,35], indicating that all patients, regardless of sex, seem to be at higher risk of AECs after transient exposure to air pollution.
This study has some limitations.Firstly, we used the individual average hourly air pollution concentrations as a proxy for personal exposure.Although this approach may introduce some degree of exposure misclassification and potential underestimation of associations, it is a common limitation in most time-series studies [35,36].Secondly, we relied on prehospital diagnoses based on para-medical records due to the unavailability of data on the final clinical diagnoses.However, we defined the subtypes of AECs using the Medical Priority Dispatch System (MPDS) [37], which helps mitigate significant biases in our results.Third, due to the unavailability of hourly pollutant data from individual monitors and the absence of patients' residential addresses, we relied on aggregated pollutant data from multiple monitoring stations as provided by the National Environmental Monitor Station, rather than a single monitor.Nonetheless, this is unlikely to introduce significant measurement bias because Shenzhen, not being an industrial city, lacks major pollutant sources, and there are no significant differences in the annual concentrations of air pollutants across the various monitors.
This study also has several strengths.Firstly, the Shenzhen Ambulance Emergency Center provided high-quality data covering most hospitals in Shenzhen.The large sample size and individual-level data allowed for comprehensive statistical analyses, enhancing the validity of the study results.Secondly, the time-stratified, case-crossover study design with hourly data on air pollutants enables us to make more robust causal inferences by effectively adjusting for time-invariant confounding factors.Finally, we performed a quantile g-computation statistical strategy to examine the associations of pollutant mixtures with AECs, shedding light on the contribution of each pollutant to the risk of AECs and providing evidence for the urgent need to control specific pollutants.
Given the increasing prevalence of multiple air pollution-related AECs, the considerable cost of treatment, and the substantial adverse effects on individuals and society, local government response systems should typically provide air pollution-related information to the public, develop strategies to reduce air pollution-related risks, and allocate emergency ambulance resources equitably.Our findings benefit emergency departments in refining ambulance allocations and curtailing response time delays.

Conclusions
Our findings revealed distinct predominant lags for each pollutant.The combined effects of these pollutant mixtures were associated with an increased risk of both all-cause and cause-specific AECs.The risk increased linearly as pollutant concentrations increased, with no evident thresholds.Moreover, these associations were more pronounced in the older population and exhibited variations based on the time of day.Local governments must take the initiative to disseminate air pollution-related information to the public, develop strategies to minimize associated risks, and ensure an equitable distribution of emergency ambulance resources.Simultaneously, individual actions can also contribute significantly to diminishing the health impacts of air pollution.Future research might pivot towards examining individual behaviors and promoting personal air pollution monitoring.

Supplementary Materials:
The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/toxics11110895/s1, Figure S1.Distribution of environmental monitors in Shenzhen city, China.Figure S2.Spearman correlations among air pollutants and meteorological factors.Figure S3.Cumulative concentration-response curves for the association of all-cause ambulance emergency calls with PM2.5, PM10, Ozone, NO 2 , and SO 2 over lags 0-48 h in Shenzhen from 2013 to 2019. Figure S4.Lag structures for the associations of ambulance emergency calls with each interquartile range increase concentration in PM2.5, PM10, Ozone, NO 2 , and SO 2 over lags 0-48 h in Shenzhen from 2013 to 2019. Figure S5.Weights represent the proportion of the positive or negative partial effect of each pollutant in the quantile g-computation model.Figure S6.Sensitivity analysis: exposure-response relationship between air pollutant mixtures and all-cause ambulance emergency calls in Shenzhen from 2013 to 2019. Figure S7.Sensitivity analysis: exposure-response relationship between air pollutant mixtures and cause-specific ambulance emergency calls in Shenzhen from 2013 to 2019.Funding: This research was supported by the Sanming Project of Medicine in Shenzhen (No. SZSM201911005).The funders had no role in the study design, data collection, data analysis, data interpretation, or in the writing of the report.All authors accept responsibility for the decision to submit for publication.

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Informed consent was obtained from all participants involved in the study.The study was conducted in accordance with the Declaration of Helsinki.

Figure 1 .
Figure 1.Exposure-response relationship between air pollutant mixtures and all-cause ambulance emergency calls in Shenzhen from 2013 to 2019.The line is the exposure-response curve, and the grey error bar is the confidence interval at each percentile.p-value < 0.001 for linear relationship.

Figure 2 .
Figure 2. Exposure-response relationship between air pollutant mixtures and cause-specific ambulance emergency calls in Shenzhen from 2013 to 2019: (A) represents vascular disease (p-value <

Figure 1 .
Figure 1.Exposure-response relationship between air pollutant mixtures and all-cause ambulance emergency calls in Shenzhen from 2013 to 2019.The line is the exposure-response curve, and the grey error bar is the confidence interval at each percentile.p-value < 0.001 for linear relationship.

Figure 1 .
Figure 1.Exposure-response relationship between air pollutant mixtures and all-cause ambulance emergency calls in Shenzhen from 2013 to 2019.The line is the exposure-response curve, and the grey error bar is the confidence interval at each percentile.p-value < 0.001 for linear relationship.

Figure 2 .Figure 2 .
Figure 2. Exposure-response relationship between air pollutant mixtures and cause-specific ambulance emergency calls in Shenzhen from 2013 to 2019: (A) represents vascular disease (p-value < Figure 2. Exposure-response relationship between air pollutant mixtures and cause-specific ambulance emergency calls in Shenzhen from 2013 to 2019: (A) represents vascular disease (p-value < 0.001 for linear relationship); (B) represents respiratory disease (p-value < 0.001 for linear relationship); (C) represents reproductive disease (p-value < 0.001 for linear relationship); (D) represents injury (p-value < 0.001 for linear relationship).The line is the exposure-response curve, and the grey error bar is the confidence interval at each percentile.

Table 1 .
Baseline characteristics in Shenzhen from 2013 to 2019.
Abbreviations: PM 2.5 , particulate matter 2.5 µm in diameter; PM 10 , particulate matter 10 µm in diameter; NO 2 , nitrogen dioxide; SO 2 , sulfur dioxide.For categorical variables, number (n) and percentage (%) are presented.For continuous variables, mean (SD) is presented.a Others: ambulance emergency calls of poison, mental, digestion, and endocrine.b Time of day: daytime is from 08:00 to 19:00 for one day; nighttime is from 20:00 to 07:00 the next day.

Table 2 .
Associations of a mixture of air pollutants evaluated using quantile g-computation with all-cause and cause-specific ambulance emergency calls.

Table 3 .
Associations of a mixture of air pollutants evaluated using quantile g-computation with all-cause and cause-specific ambulance emergency calls stratified according to age, sex, season, and time of day.Abbreviations: ERR, excessive relative risk; 95% CI, 95% confidence interval.Associations of an interquartile range (IQR) increase in PM 2.5 , PM 10 , Ozone, NO 2 , and SO 2 exposure with ambulance emergency calls in multipollutant models.Models were adjusted for hourly temperature, hourly relative humidity, holiday, day of week, and calendar year.