Short-term exposure to air pollution and emergency room visits for acute respiratory symptoms among adults

Abstract

OBJECTIVE:

To examine the short-term effect of ambient air pollution on daily acute respiratory emergency room visits among adults.

METHODS:

A time-series study (June 2017–February 2019) was carried out among adults (≥18 years) visiting a multi-specialty hospital in Delhi. We evaluated the association between the daily levels of particulate matter (PM) <2.5 μm in diameter (PM_2.5) and PM <10 μm in diameter (PM₁₀), ozone (O₃), nitrogen dioxide (NO₂), carbon monoxide (CO) and sulphur dioxide and daily count of emergency room (ER) visits for acute respiratory symptoms. Generalised additive model (GAM) was used with the Poisson link function to analyse the associations for 0–1 to 0–7 lag days.

RESULTS:

A total of 69,400 ER visits were recorded, of which 2,669 were by adults due to acute respiratory symptoms. At 0–7 lag days, an increment of 1 standard deviation in NO₂ and PM_2.5 concentration was associated with a percentage increase in acute respiratory ER visits of respectively 53.0% (95% CI 30.84–78.97) and 19.5% (95% CI 4.53–36.65). During 0–7 lag days, a positive trend was observed at higher concentrations of CO (>1.86–3.28 mg/m³), while a negative significant association was observed at low concentrations of CO (<1.171 mg/m³).

CONCLUSION:

Short-term exposure to ambient NO₂ and PM_2.5 was associated with acute respiratory emergency visits of adults at lag 0–7 days.

Air pollution plays a significant role in a wide range of health problems. It has been consistently reported to be a major risk factor for both acute and chronic lung diseases.^1,2 As per the WHO, exposure to six major air pollutants: particulate matter (PM), nitrogen dioxide (NO₂), sulphur dioxide (SO₂), ozone (O₃), lead and carbon monoxide (CO) have toxic effects on human health and the environment.² Elderly persons and people with chronic lung diseases are more susceptible to the harmful effects of ambient air pollution even at low concentrations.^3–5 Long-term and short-term exposure to air pollutants has diverse impacts on humans at different concentrations.^3,6 Short-term effects of ambient air pollutants include an increase in respiratory symptoms like cough, wheezing, difficulty breathing, etc.; exacerbations of pre-existing chronic obstructive pulmonary disease, and asthma; and increases in hospital and emergency admissions.^3–5,7,8 Emergency room (ER) visits are considered as a reliable indicator to evaluate the harmful effects of increasing air pollution on health outcomes.^7,9 Acute respiratory symptoms resulting from exposure to ambient air pollution in a brief period of time are neither routinely available nor adequately reported.¹⁰ Some of the studies from Delhi, India, have reported the short-term effect of air pollution on cardiorespiratory health.^7,11–13 However, none of the studies evaluated the effect of short-term air pollution on daily emergency admissions, particularly for acute respiratory symptoms, mainly in adults living in Delhi. In the present paper, we hypothesised that the increase in daily levels of ambient air pollutants is associated with increased emergency visits among adults with acute respiratory symptoms. To study this, we performed a time-series analysis of daily concentrations of various air pollutants and daily acute respiratory ER visits among adults in a multi-specialty hospital in Delhi. The study will help us quantify the extent of the acute respiratory problems related to short-term exposure to ambient air pollution.

METHODS

Emergency room visits

In this time-series study, we recorded daily emergency visits by adults (age ≥18 years) from 1 June 2017 to 28 February 2019. To obtain a representative sample of the city’s adult population, the All India Institute of Medical Sciences (AIIMS), New Delhi, India, was selected. AIIMS is a multispecialty hospital with an adult emergency room and maximum attendance. Daily counts of ER visits to the hospital were recorded. All patients visiting the ER were screened for respiratory symptoms. Patients were enrolled if they had acute respiratory symptoms (≤2 weeks) and were residing in Delhi for at least 4 weeks. Patients were excluded if 1) they were not available because of procedures/investigations, or 2) they were unwilling and did not consent to participate in the study. Details of the protocol have been previously published.⁷

Air pollution and meteorological data

Air pollution data were provided by the Delhi Pollution Control Committee, Delhi, India. Daily 24-h average concentrations of six air pollutants: 24-h concentrations of particulate matter (PM) <2.5 μm in diameter (PM_2.5) and PM <10 μm in diameter (PM₁₀), CO, NO₂, SO₂ and 8-h concentrations of O₃; relative humidity measures and temperature were obtained from four Continuous Ambient Air Quality Monitoring Stations (CAAQMS). The four CAAQMS were located in Mandir Marg, Anand Vihar, Punjabi Bagh and Ramakrishna Puram. Details of the methodology for determining and analysing the morbidity and air pollution data collection have been previously published.⁷

Statistical analysis

Descriptive statistics, including mean ± standard deviation (SD) or median (interquartile ranges [IQR]) for continuous variables and number (percentage) for categorical data, were calculated for the characteristics of eligible adults visiting ERs, and air pollutant levels during the study. A generalised additive model (GAM) with the Poisson link function was used to assess the association between ambient air pollutant concentrations and daily acute respiratory ER visits of adults after adjusting for the day of the week, national holidays, seasons, 24-h temperature and 24-h relative humidity. The model was fitted using the following equation:

$$\begin{array}{ccc}\text{Log}(\mathrm{E}(\mathrm{Y}0))\hfill & =\hfill & \text{Intercept}+\mathrm{\beta }1({\mathrm{P}1}_{0-\mathrm{L}})+\mathrm{\beta }2({\mathrm{P}2}_{0-\mathrm{L}})\hfill \\ \hfill & \hfill & +\dots \dots \dots +\mathrm{\beta }\mathrm{x}({\mathrm{P}\mathrm{x}}_{0-\mathrm{L}})\hfill \\ \hfill & \hfill & +\mathrm{s}({\text{Temperature}}_{0-\mathrm{L}},\text{df}=3)\hfill \\ \hfill & \hfill & +\mathrm{s}({\text{Relative\hspace{0.17em}humidity}}_{0-\mathrm{L}},\text{df}=3)\hfill \\ \hfill & \hfill & +\mathrm{s}(\text{time\hspace{0.17em}of\hspace{0.17em}year},\text{df}=3)+\mathrm{\beta }\mathrm{d}(\text{DOW})\hfill \\ \hfill & \hfill & +\mathrm{\beta }\mathrm{h}(\text{Holiday})\hfill \end{array}$$

Where E(Y0) is the expected percentage of daily patient ER visits for acute respiratory symptoms on day 0 (day of visit); β1, β2……..βx are the regression coefficients for different pollutants; P1_0–L, P2_0–L, .. Px_0–L are the moving average concentrations of air pollutants for 0–L lag days, df is the degree of freedom and s() denotes the smoother based on the penalised smoothing spline for weather variables, temperature and relative humidity. ‘DOW’ represents days of the week, i.e., weekdays or weekends; ‘Holiday’ represents national holidays. βd is the coefficient for DOW and βh the coefficient for the holiday.

The lag effect of each pollutant was assessed on moving average lag structures (0–1, 0–2, 0–3, 0–4….0–7). The minimum Akaike Information Criterion (AIC) models were selected for determining the appropriate lag structure for each pollutant. Multi-pollutant models were fitted on the moving average lag structures (0–1, 0–2, 0–3, 0–4….0–7) with all six pollutants for every symptom. The final lag structure was chosen using the backward elimination method, using minimum AIC criteria values to create the final model.¹⁴ Sensitivity analysis was performed for varying degrees of freedom (df = 3 to df = 6) at the selected lag structure for the time of year and moving average of temperature and humidity. To assess the association at different CO concentrations, we performed a subset analysis based on tertiles of 24-h average CO concentrations (0.49–1.17, >1.17–1.86, >1.86–3.28 mg/m³).

All analysis was done using R v.3.6 (R Foundation for Statistical Computing, Vienna, Austria). Effect estimates are presented as percentage changes (95% confidence intervals [CIs]) in daily ER visits per 1 SD change in moving of 0–7 lag days’ concentration of various pollutants.

Ethics approval

The study protocol was approved by the AIIMS Institute Ethics Committee, New Delhi, India (Ref No. IEC/97/12/2015). All procedures performed in the present study were in accordance with the ethical standards of the Institutional Research Committee and the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. Written informed consent was provided by participants before enrolment in the study.

RESULTS

During the 1 year 8 months’ study period from 1 June 2017 to 17 February 2019, a total of 69,400 ER visits were recorded, of which 2,669 were due to acute respiratory symptoms (Figure 1). The descriptive demographic characteristics and summary of daily mean concentrations of air pollutants are given in Supplementary Table S1. The mean (min–max) daily number of patients with acute respiratory symptoms was 4 (0–15). The annual mean values of PM_2.5, PM₁₀, SO₂, NO₂, O₃ and CO were respectively 113.8 μg/m³, 253.3 μg/m³, 19.4 μg/m³, 56.4 μg/m³, 39.5 μg/m³ and 1.6 mg/m³. The median duration of symptoms was 3 days, and majority of patients reported with difficulty in breathing (95%) and cough (74%). The admission rate was higher in patients with non-respiratory comorbidities than those with respiratory illnesses (n = 762, 68.2% vs. n = 227, 20.3%). The results of the multi-pollutant model examining the cumulative effects of various pollutants using different (0–7 days) lag structures are shown in Supplementary Figure S1. The moving average of lag 0–7 days was chosen for the minimum AIC values as the best lag structure to estimate the cumulative effect of air pollutants. During 0–7 lag days, NO₂ and PM_2.5 were positively associated with the total number of adults visiting ER with acute respiratory symptoms. For every 1 SD increase in NO₂ and PM_2.5 for lag 0–7 days, there was an increase in acute respiratory ER visits by respectively 53.0% and 19.5%. CO levels showed a concentration-dependent association with acute respiratory ER visits (Figure 2). During lag 0–7 days, CO was negatively associated with ER visits of adults with acute respiratory symptoms at low concentrations (0.49–1.171 mg/m³); 1 SD increase in CO concentration was associated with 26.1% decrease in ER visits. In contrast, CO showed positive directional trend at high concentrations (>1.86–3.28 mg/m³), although this was not significant; 1 SD increase in CO concentration showed 0.9% increase in ER visits. No significant association was seen at CO concentrations of >1.17–1.86 mg/m³. Results of the sub-group analysis according to ER visits for individual acute respiratory symptoms are given in Supplementary Figure S1. An increase in NO₂ levels was significantly and positively associated with ER visits for cough, noisy breathing and difficulty in breathing, while a decrease in CO levels was negatively associated. An increase in PM₁₀ and PM_2.5 levels was significantly associated with increased ER visits for cough and difficulty in breathing, respectively. No associations were seen for SO₂ and O₃ concentrations. Sensitivity analysis was performed to detect the robustness of the model using 3, 4, 5 and 6 df for the smoothness of temperature and humidity. Sensitivity analysis showed that the effect estimates of various pollutants were still robust to the various degrees of adjustment (Supplementary Table S2), indicating that these associations are valid.

Figure 1.

Screening and enrolment of patients during the study period. ER = Emergency Room.

Figure 2.

Estimates from multi-pollutant models showing a percentage change in the percentage of ER visits of adults with acute respiratory symptoms associated with every 1 SD change in the concentration tertiles of CO during lag 0–7 days as per minimum Akaike Information Criterion at a df of 3. Data expressed as a percentage change (95% confidence interval). ER = Emergency Room; SD = standard deviation; CO = carbon monoxide; df = degree of freedom.

DISCUSSION

In this time-series study in ER visits of a multi-speciality hospital in India, we observed a significant association between concentrations of the air pollutants (NO₂, PM_2.5 and CO) and emergency visits for all acute respiratory symptoms in adults. In a subgroup analysis by individual respiratory symptoms, an increase in NO₂ increased and CO decreased the number of ER visits for cough, noisy breathing and difficulty in breathing. PM₁₀ increased the number of ER visits for cough and PM_2.5 increased the number of ER visits for difficulty in breathing. No such association was observed for SO₂ and O_3. Sensitivity analyses gave similar results, indicative of the robustness of the results. We observed a significant positive association between previous 0–7 days of exposure to NO₂ and emergency visits for acute respiratory symptoms. Results from this study are consistent with previously reported associations between NO₂ and respiratory morbidity.^1,7,8,15,16 NO₂, a marker for traffic-related pollution, is known to have a strong positive effect at higher concentrations on respiratory health via oxidative mechanisms by altering alveolar macrophage function and releasing mediators of inflammation.^1,8,17 In the present study, we observed that an increase in acute respiratory ER visits was associated with an increase in PM₁₀ and PM_2.5 levels. These findings are in line with the previous time-series studies that have established a positive association of short-term exposure to particulate matter with various indicators of respiratory morbidities.^1,8,16–18 In the present study, NO₂ had a greater effect, almost twice that due to PM_2.5 as reported in previous literature.^8,17 CO showed a concentration-dependent pattern during previous 0–7 lag days. CO showed a positive trend at higher levels but a negative association at lower levels with daily respiratory ER visits. Our results corroborate the findings of previous studies, in which brief exposure to CO showed negative associations at lower levels and positive effects at higher levels.^4,6 CO, a metabolite of heme oxygenase-1, once known as a ‘toxic molecule’ is now paradoxically shown to be anti-inflammatory¹⁹ and a bactericidal agent^20,21 at low concentrations. Together with heme oxygenase-1, CO strengthens the immune response to common infections by inhibiting the expression of inflammatory cytokines and inducing the expression of anti-inflammatory cytokines via mitogen-activated protein kinases.^4,19 As individuals inhale a mixture of pollutants, assessing the synergistic effect of multiple pollutants acting together is essential. We therefore used a multi-pollutant model in the present study, instead of a single-pollutant model. Single-pollutant models may yield results that are an overestimation or an underestimation.²² The combined effect of multiple pollutants will be an adjusted estimate of various parameters such as other pollutants, the environment and weather conditions.^22,23

We explored different lag structures for the understanding of mechanisms of short-term exposure to ambient air pollution. The best fit model was found at 0–7 lag days. Individual subsequent days have autocorrelation and thus, result in multi-collinearity in model development. We therefore opted for moving average concentrations to obtain a cumulative effect and avoid the multi-collinearity issue in model development.^1,24 Our findings agree with those of earlier studies that assessed different lag structures for respiratory outcomes using daily levels of air pollutants.^1,24 These reported that a longer lag structure is probable for acute respiratory ER visits, as the severity of illness is lower. The pattern of lag structure mainly depends on patient’s behaviour that varies according to different diseases. Respiratory exacerbations due to inflammation usually takes longer acute episodes due to cardiovascular disease, which generally require prompt immediate emergency visit or hospital admission.²⁴ The relationship between air pollution and acute symptoms in Delhi varies with seasonal changes. In our recent study, we observed that during winters, increased acute respiratory ER visits were associated with higher PM_2.5 levels in the highly polluted north-west region of Delhi, while low numbers of ER visits were associated with the moderately polluted south-west region of Delhi having low PM_2.5 levels.²⁵ WHO-air quality guidelines (AQG) have proposed city-specific interim targets for the gradual improvement in air quality in a graded manner. Delhi showed PM_2.5 concentration levels above permissible limits (60 µg/m³) for more than 60% of days in 2019. As per WHO-AQG, Delhi is classified in the interim target 1, and the proposed ambition level of attainment for PM_2.5 is 75 µg/m³ in Delhi.²⁶

Strengths of our study include the large sample size, air quality data measurements comprising six inter-related air pollutants and meteorological data, which were not available in many of the previous studies. Our study focused on recording the number of ER visits due to acute respiratory symptoms concerning air pollution, which is commonly not recorded or consulted. There were some study limitations. Patients from outpatient departments were not included, which could have given us a more comprehensive picture of the short-term respiratory effects of air pollution. We did not record comorbidities other than respiratory illnesses, which could have helped to understand the high admission rate for patients with non-respiratory comorbidities. We were not able to measure individual exposures such as the use of air conditioning or time spent in outdoor activities. However, temporal variations were taken into account by 1) excluding patients residing outside Delhi, and 2) carrying out a time series of daily pollutant levels recorded from within-city air quality monitoring stations. Based on the findings from this study, future air pollution reduction strategies and intervention studies can be planned to combat the adverse effect of short-term ambient air pollution on respiratory morbidity. A city-specific graded interim target of AQG recommended by WHO Air Quality Guidelines 2021, instead of the single country-specific interim target, would be a more practical approach to attaining the recommended air quality standards. A detailed understanding of local sources and regional meteorological factors responsible for a wide range of concentration variations at different air quality monitoring stations needs to be studied for better mitigation strategies.

CONCLUSION

Our results show that NO₂, and PM_2.5 levels are associated with an increase in emergency visits for acute respiratory symptoms among adults from Delhi, India. The effects were robust at 0–7 lag structure after adjusting for temperature and humidity.

Funding Statement

The work was supported by the Indian Council of Medical Research, New Delhi, India.