Short-term air pollution exposure and risk of respiratory pathogen infections: an 11-year case-crossover study in Guangzhou, China

Abstract

Background

Limited epidemiological evidence exists on the relationship between short-term exposure to air pollutants and respiratory pathogen infections. This study investigates the association between short-term air pollution exposure and respiratory pathogen infections in Guangzhou, southern China.

Methods

A time-stratified case-crossover study design was applied. Data from 96,927 patients with suspected respiratory pathogen infections between 2013 and 2023 were collected. The daily air pollutant concentration is obtained from the local environmental monitoring station. Logistic regression was used to assess the effect of air pollutant exposure included in the equation on the risk of respiratory pathogen infection. Generalized additive models were used to analyze the relationship between pollutant exposure and hospital visits, adjusting for potential confounders such as temperature and precipitation. Sub-group analysis was performed to estimate the reliability of the correlations among the subgroups.

Results

The logistic regression model shows that PM_2.5, NO₂ and CO are included in the variable equation. Single-pollutant models indicate that there is a significant association between short-term exposure to NO₂ and CO and an increased risk of hospital visits for respiratory infections, especially on lag day 0, while PM_2.5 shows a non-linear relationship. In the multi-pollutant model, for each unit increase in NO₂, the risk of hospital visits increased by 11.66%, and for CO, the risk increased by 0.64%. Subgroup analysis showed the effects were more pronounced in minors (< 18 years), while no significant gender differences were observed. Additionally, CO and NO₂ interacted with PM_2.5, amplifying the risk of infection.

Conclusion

This large-scale epidemiological study demonstrates significant associations between short-term air pollutant exposure and respiratory infections, particularly highlighting the risks of NO₂ and CO exposure. The findings underscore the critical need for strengthening air quality monitoring and protection strategies in rapidly urbanizing regions, with special attention to vulnerable populations such as minors. These results provide evidence-based support for enhancing environmental health policies in metropolitan areas to better protect public health through improved air quality standards and early warning systems.

Introduction

Respiratory tract infection has become a common acute disease in children and adults, bringing huge economic and health burden to the general population [1, 2]. In 2019, for upper respiratory tract infections, cases in China were estimated at 2.55 million (95% UI: 2.29–2.86) [3]. In recent years, the incidence of respiratory tract infection has continued to rise and has become one of the major public health problems affecting the health of Chinese people [4]. According to a 3-year study in South China, the prevalence of upper respiratory tract infection alone reached 7.86% [5]. Respiratory infections can occur in almost all age groups, increasing the psychosocial burden of patients and the risk of mental health disorders.

Despite the significant burden of these diseases, there are no universally effective strategies to prevent and control the occurrence of respiratory infections. However, early identification and limiting specific exposure factors, especially those air pollutants that can increase the risk of infection, may be an effective way to reduce the incidence of respiratory infections [6]. Many epidemiological studies have shown a significant association between short-term exposure to air pollution and increased hospital visits due to respiratory pathogen infections [7, 8]. In addition, traffic-related air pollutants such as nitrogen dioxide (NO₂) and carbon monoxide (CO) have also been shown to significantly increase the risk of respiratory infections, especially for susceptible populations such as children and the elderly [7, 9, 10].

However, while multiple studies have explored the relationship between air pollutants and respiratory infections, the association between short-term exposure and daily visits for respiratory infections remains uncertain. Some studies have shown that PM_2.5, NO₂, sulfur dioxide (SO₂) and other pollutants may increase susceptibility to pathogen infection by damaging respiratory mucosa and inhibiting the immune system [11]. However, the findings are inconsistent. For example, while some studies have found that PM_2.5 is strongly associated with the risk of respiratory infections [12, 13], other studies have shown that SO₂ and O₃ play a more critical role in respiratory viral infection [14]. These inconsistent results may be attributed to differences in air pollution levels, inherent challenges in the study design, and complex infection patterns in different regions.

China, as the largest low—and middle-income country, still has a high level of air pollution despite improvements in air quality in recent years [6]. Yao et al. [15] noted that the national all-cause premature mortalities attributable to all air pollutants was 1.35 million, accounting for 17.2% of reported deaths in China. Guangzhou, as the largest national central city in southern China, not only faces complex changes in air pollutants but also has changing patterns of pathogen infection. Although there is evidence of a strong link between air pollutants and the prevalence of respiratory infections, to our knowledge, no studies have fully explored the relationship between air pollutants and the daily risk of hospital visits for respiratory infections at the urban level in China.

In this time-stratified case-crossing study, we comprehensively and quantitatively examined whether short-term exposure to air pollutants is associated with an increase in the number of hospital visits for respiratory infections in Guangzhou, using both single-pollutant and multi-pollutant models, as well as to explore any potential modifying components. The aim is to provide scientific basis for future public health policy making.

Method

Study design

We conducted a time-stratified case-crossover study to explore the association between short-term air pollution and hospital visits due to respiratory pathogens (Fig. 1). This study design has been widely used in the analysis of acute outcomes caused by short-term exposure, especially for the study of environmental pollutants. To ensure comparability and minimize potential selection bias, we employed a time-stratified referent selection approach where control days were selected from the same calendar month and year as the case day, matching on the same day of the week [16]. This bidirectional control selection strategy identified 3–4 control days for each case day, both before and after the case day within the same month, effectively controlling for seasonal patterns and time trends [16]. For instance, if a case occurred on a Monday, all other Mondays within the same calendar month served as control days, excluding public holidays and days with missing air pollution data to prevent potential confounding from activity patterns. In cases where a selected control day coincided with another case day for the same individual, it was excluded to prevent overlap. Since the comparison of both case and control groups were based on the same patient, potential confounding factors, such as age, sex, lifestyle and genetic factors, have been effectively controlled for. This matching strategy ensures that cases and controls are comparable in terms of seasonal factors, day-of-week effects, and long-term trends, while maintaining independence between exposure measurements.

Fig. 1

Flowchart of study design. Note: PM_2.5, Particulate Matter less than 2.5 μm; PM₁₀, Particulate Matter less than 10 μm; CO, Carbon monoxide; SO₂, Sulfur dioxide; NO₂, Nitrogen dioxide; O₃, Ozone

Data collection

We obtained the detection results of pathogen specific IgM fluorescent antibodies in 96,927 patients with suspected respiratory pathogen infection (SRI) from the First Affiliated Hospital of Guangzhou Medical University from January 2013 to October 2023. The inclusion criteria were: (1) patients presenting with symptoms suggestive of respiratory infection, including but not limited to cough, fever (> 37.5 °C), dyspnea, and/or chest pain; (2) age ≥ 0 years; and (3) willingness to provide respiratory specimens for pathogen testing, and the exclusion criteria were: (1) patients with incomplete medical records; (2) those who withdrew consent for sample collection; and (3) cases where sample quality was inadequate for testing. (4) Patients were unable to match information about air pollutants. Based on these criteria, we excluded specimens enrolled before December 2, 2013, and after February 15, 2023 in this study (time span to match air pollutant data).

The participants’ demographic data and clinical characteristics, such as age and gender, were extracted directly from the hospital’s Laboratory Information System (LIS) and electronic medical records. No additional questionnaires or surveys were administered specifically for this study. All data were collected as part of routine clinical care during the study period. In addition, all serum samples were tested for respiratory pathogens using indirect immunofluorescence assay (Vircell, S.L, Spain) after diagnosis. Data on suspected respiratory pathogen infections include nine common pathogens, Legionella pneumophila (LP), Mycoplasma pneumoniae (MP), Coxiella burnetii (COX), Chlamydia pneumoniae (CP), adenovirus (ADV), respiratory syncytial virus (RSV), influenza A (IFA), influenza B (IFB), and parainfluenza viruses (PIVS).

According to the purpose of our experiment, the respiratory pathogens included in the detection are divided into two categories, which are:

• Bacterial pathogens: Legionella pneumophila, Mycoplasma pneumoniae, Coxiella burnetii and Chlamydia pneumoniae.

• Viral pathogen: Adenovirus, Respiratory syncytial virus Influenza A, Influenza B, and Parainfluenza viruses.

Briefly, patient sera were diluted 1:40 and incubated with antigen-coated slides for 30 min at 37 °C. After washing, fluorescein-labeled anti-human IgM was added and incubated for 30 min. Slides were examined under a fluorescence microscope by two independent, experienced technicians. Samples were considered positive if they exhibited specific fluorescence at a titer ≥ 1:40.

Air pollution data

We obtained daily air pollution data from Guangdong Department of Ecology and Environment (http://gdee.gd.gov.cn/), including 24-h average concentrations of CO, NO₂, PM_2.5, PM₁₀, SO₂, and 8-h mean O₃ concentration in Guangzhou from December 2, 2013 to February 15, 2023. Data from multiple monitoring stations were averaged to represent city-wide exposure levels. Each case and control period were matched with corresponding daily air pollution data. While primarily using daily averages, we conducted quality control for hourly measurements, where values exceeding three standard deviations were verified against nearby stations, and performed sensitivity analyses comparing daily average versus maximum hourly concentrations to account for short-term pollution spikes.

Statistical analysis

Sensitivity analysis was used to compare the characteristics of included and excluded subjects. The median (1st quartile, 3rd quartile) was used to describe the age of the participants because the data presented a distribution bias. The distribution of sex and age groups (stratified by 18 years) was reported by frequency (ratio), and differences between groups were compared by chi-square test. Spearman correlation analysis was used to calculate the correlation between air pollutants, that is, if the correlation coefficient is greater than or equal to 0.60, it is closely related.

According to the results of logistics regression analysis (Supplementary Material Table 5), three common air pollutants (PM_2.5, CO, NO₂) and two infection modes of respiratory pathogens (bacterial pathogens and viral pathogens) were selected to study the relationship between air pollution and daily hospital visits caused by corresponding pathogens. In this study, we only combined the effects of three common air pollutants (CO, PM_2.5, and NO₂) because the underlying biological mechanisms between short-term exposure to these air pollutants and respiratory pathogens infection risk are well established. Air pollutant data show that PM_2.5, CO and NO₂ are the most common air pollutants in Guangzhou. However, we still consider the confounding effects of O₃_8h, SO₂ and PM₁₀ on the multi-pollutant model.

In this matched case-crossing study, we used a series of conditional logistic regression analyses with a single predictor to estimate the rough effects of CO, PM_2.5, and NO₂ exposure on respiratory pathogen-mediated hospital visits on different lag days.

In addition, we used a generalized additive model (GAM) to explore the nonlinear relationship between pollutant concentration and the risk of infection by respiratory pathogens (expressed as an adjusted odds ratio), while using a linear model to fit possible trends in the linear relationship between pollutant concentration and infection risk. The generalized additive model selection is based on the high value of R²_adj [17], smaller Akaike’s Information Criterion (AIC) [18], and generalized cross validation (GCV) [19]. The following selected log-linear generalized additive model reads:

$${Y}_{t}\sim Gamma\left({\mu }_{t},\nu \right);$$

$$\text{Log}{\mu }_{t}=\alpha +\beta {X}_{t,l}+ns\left(t,df=5\right)+ns\left(Tem{p}_{t,l},df=2\right)+ns\left(Pre{c}_{t,l},df=2\right)+ns\left(Wind{s}_{t,l},df=2\right)+ns\left(Sea{s}_{t,l},df=2\right)+ns(Wee{k}_{t,l},df=2)$$

where Y_t is the observed daily SRI cases (at day t), following Gamma distribution with mean µ_t and dispersion parameter 1/v (here v can be obtained by maximum likelihood estimation), a is the intercept, β is the long-relative rate of SRI cases related to per unit increase of air pollutants. X_t,l represents the air pollutants concentration at day t with lag l ranging from 0 to 7 days [20]. ns is the non-parametric smooth B-spline function of calendar time (t), temperature (Temp_t,l), precipitation (Prec_t,l), windspeed (Winds_t,l), seasons (Seas_t,l) and weekday (Week_t,l). This basic model, incorporating the linear function of air pollutants concentration and the smoothed B-spline functions of time, weather conditions and other factors, offers a framework to characterize non-linear and nonmonotonic links between daily SRI cases and the factors mentioned above [21].

Ethical approval

The study protocol was approved by the Institutional Review Board of the First Affiliated Hospital of Guangzhou Medical University (approval number: Medical Ethics Review 2016 No. 73). Written informed consent was obtained from all patients or their legal guardians.

Results

Descriptive characteristics of participants

A total of 96,927 participants who met the inclusion criteria for suspected respiratory pathogen infection and underwent examination at the hospital were included in the analysis. The baseline characteristics of the study population are presented in Table 1. Notably, the number of patients diagnosed with a bacterial pathogen was significantly higher than those diagnosed with a viral pathogen. Furthermore, across all respiratory pathogens, including when categorized as bacterial or viral infections, the positive detection rate was higher among females compared to males (P < 0.001). Adolescents (< 18 years of age) were also more likely to test positive for respiratory pathogens compared to adults, with significantly higher positive detection rates for both bacterial and viral pathogens (P < 0.001). The number of people infected with LP, MP, COX, CP, ADV, RSV, IFA, IFB and PIVS were 1,100, 13,354, 418, 197, 658, 687, 430, 1,665 and 1,701, respectively (Supplementary Material Table 4).

Table 1 Descriptive characteristics of the participants

The average 24-h median concentrations of air pollutants in Guangzhou, along with the interquartile ranges (1st and 3rd quartiles), were as follows: 32.00 (22.00, 48.00) μg/m³ for PM_2.5, 50.00 (36.00, 73.00) μg/m³ for PM₁₀, 11.00 (8.00, 15.00) μg/m³ for SO₂, 0.90 (0.80, 1.00) mg/m³ for CO, 42.00 (33.00, 55.00) μg/m³ for NO₂ and 83.00 (48.00, 120.00) μg/m³ for O₃_8h (Supplementary Material Table 2). Spearman correlation analysis among air pollutants is depicted in Fig. 2.

Fig. 2

Heat map of correlation between air pollutants and meteorological indicators. Correlation coefficient is greater than or equal to 0.60, the two are highly correlated (Supplementary Table 6)

The analysis revealed that PM_2.5 is strongly related to PM₁₀, SO₂, CO and NO₂, while PM₁₀ also showed a high correlation with SO₂ and NO₂ (Spearman’s correlation coefficient, rₛ > 0.6, P < 0.001). To avoid multicollinearity among highly correlated air pollutants, the multi-pollutant model incorporated Spearman correlation analysis to account for the relationship between air pollutants and meteorological data.

Single pollutant model results

Following logistic regression analysis, we conducted a statistical screening of the pollutant variables. Ultimately, PM_2.5, NO₂ and CO were included in the final model, and single-pollutant models were established with different lag days for each corresponding pollutant to predict hospital visits due to respiratory pathogen infections (Table 2 and Fig. 3). The trends of air pollutant concentration, meteorological index data and the number of researchers included over time were shown accordingly (Supplementary Material Fig. 3).

Table 2 Results of conditional logistic regression analysis for the association between air pollutants and respiratory pathogen infections

Fig. 3

Association of different lag days for short-term exposure to PM_2.5, NO₂ and CO with hospital admissions for suspected respiratory pathogen infection. Note: Error bars represent 95% confidence intervals of ORs. Lag day 0, the day of hospital admission; Lag day 1, previous 1 day; Lag day 2, previous 2 day; Lag day 3, previous 3 day; Lag day 4, previous 4 day; Lag day 5, previous 5 day; Lag day 6, previous 6 day; Lag day 7, previous 7 day; Lag day 2–3, 2-days moving average of lag day 2 and lag day 3; Lag day 3–4, 2-days moving average of lag day 3 and lag day 4; Lag day 2–4, 3-days moving average of lag day 2, lag day 3, and lag day 4. Pathogens infection, infected to any pathogens; Bacterial pathogens infection, infected to at least one of the LP, MP, COX, CP; Viral pathogens infection, infected to at least one of the ADV, RSV, IFA, IFB, PIVS

As shown in Fig. 3, all three pollutants demonstrated a certain risk effect on hospital visits for respiratory pathogen infections, particularly on lag day 0 (the day of hospital admission). The odds ratios (ORs) with 95% confidence intervals (CIs) for PM_2.5, NO2 and CO on lag day 0 were 1.003 (1.002, 1.004), 4.465 (3.970, 5.021), and 1.023 (1.022, 1.025), respectively (Supplementary Table 7). Compared to viral pathogens, PM_2.5 and NO₂ exhibited a higher risk effect on infections caused by bacterial pathogens, whereas CO did not demonstrate a similar trend. Notably, for bacterial infections, lag day 0 showed the highest risk effect for all pollutants (Supplementary Table 8).

In addition, while PM₁₀ and SO₂ exposures showed some positive or negative effects on both bacterial and viral pathogen infections, O₃_8h had no significant impact on any lag day, except for lag days 3–4, where it was modestly associated with viral pathogen infections (P = 0.004) (Supplementary Table 9). Despite this, PM₁₀, SO₂, and O₃_8h were not included as variables in the logistic regression equation.

Generalized additive model results

Figures 4 and 5 show the exposure–response curve of the day’s mean (lag day 0) PM_2.5, NO₂ and CO concentrations in relation to hospital visits for suspected respiratory pathogen infections.

Fig. 4

Generalized additive and linear models of lag day 0 concentrations of PM_2.5, CO and NO₂ for hospitalized patients with suspected respiratory pathogen infection. Note: a, c, e, the lag day 0 concentration of PM_2.5, NO₂, CO and the ORs value of the corresponding nonlinear relationship; b, d, f, the lag day 0 concentration of PM_2.5, NO₂, CO and the ORs value of the corresponding linear relationship. The blue line represents the nonlinear relationship fitted by the generalized additive model, the red line represents the relationship fitted by the linear model, and the light blue region represents the 95% confidence interval of the nonlinear light system, adjusted for the ORs according to the average daily temperature, average daily precipitation, average daily wind speed, season, and weekend

Fig. 5

Analysis of subgroup association between gaseous air pollutants (CO, part-a and NO₂, part-b) and risk of infection with suspected respiratory pathogens

After adjusting for potential confounding factors, including average daily temperature, precipitation, wind speed, as well as seasonal variation and weekends, the risk associated with PM_2.5 exposure showed a pronounced non-linear pattern, with higher risks at both lower and higher concentrations. While NO₂ and CO also demonstrated slight non-linear relationships, their overall trends predominantly showed positive associations with hospital visits for suspected respiratory pathogen infections, especially at higher concentrations.

In the multi-pollutant model, where we assessed the effects of individual pollutants per unit increase in concentration, higher PM_2.5 levels were associated with a slight reduction in the risk of hospital visits. Specifically, for each 1 µg/m³ increase in PM_2.5, the odds of a hospital visit decreased by approximately 0.49% (OR = 0.9951, 0.9943–0.9958, P < 0.001). However, the evidence was not sufficient to conclude that PM_2.5 significantly reduces the risk of suspected respiratory pathogen infections. On the other hand, higher concentrations of NO₂ and CO were associated with increased hospital visit risk, with the odds increasing by 11.66% per unit increase in NO₂ (OR = 1.1166, 1.057–1.180, P < 0.0001) and 0.64% per unit increase in CO (OR = 1.0064, 1.0056–1.0071, P < 0.0001).

The model also revealed that some additional factors influenced the risk. While average daily temperature and precipitation were not significantly associated with hospital visits (P > 0.05), average daily wind speed had a notable effect, with a slight increase in risk per unit increase in wind speed (P < 0.001). Seasonal effects were also observed, with spring showing a modest reduction in disease risk (P = 0.0185) compared to autumn (the reference category), while the effects of summer and winter were not statistically significant. Furthermore, working days versus weekends did not have a significant impact on hospital visits for suspected respiratory pathogen infections. The detailed results of the multi-pollutant model are presented in Supplementary Table 10.

Stratified analysis

For stratified analyses, we employed adjusted models controlling for potential confounders including daily temperature, precipitation, wind speed, seasonal variation, and day of the week. The same adjustment strategy was applied across all subgroup comparisons to maintain consistency and enable valid between-group comparisons. Figures 5 and 6 illustrate the robustness of the association between lag day 0 levels (per standard deviation increment) of CO, NO₂ and PM_2.5 across various subgroups and the risk of infection with suspected respiratory pathogens. The correlation between PM_2.5 and CO exposure and infection risk was consistently higher in adolescents (under 18 years) compared to adults. For PM_2.5, the odds ratios (ORs, 95% CI) were 1.013 (0.996–1.031) for adolescents versus 0.997 (0.987–1.008) for adults. Similarly, for CO, the ORs were 1.074 (1.054–1.094) for adolescents and 1.060 (1.049–1.072) for adults. However, there was no significant difference between the male and female.

Fig. 6

Analysis of subgroup association between particulate air pollutants (PM_2.5) and risk of infection with suspected respiratory pathogens

Interestingly, the positive association between NO₂ exposure and the risk of infection was significantly stronger on weekends compared to weekdays [OR = 1.197 (1.136–1.261) vs. OR = 1.028 (1.019–1.037)]. At the same time, the positive association between NO₂ and the risk of infection with suspected respiratory pathogens increased after 2019 compared to those included before 2019 [1.050 (1.032–1.068) vs. 1.028 (1.019–1.037)]. Notably, these patterns have not been observed in PM_2.5 and CO.

Moreover, we found that CO and NO₂ may modify the effect of PM_2.5 related infection, as higher concentrations of CO and NO₂ (above median, 0.8 mg/m³ for CO and 42 μg/m³ for NO₂) resulted in statistically different PM_2.5 risk estimates compared to lower concentrations (P < 0.001).

Discussion

This study systematically evaluated the relationship between short-term exposure to PM_2.5, NO₂, and CO and the risk of infection with suspected respiratory pathogens using a time-stratified case-crossover design. In the single-pollutant model, we observed a significant increase in the risk of hospitalization associated with higher concentrations of these pollutants, particularly on lag day 0. A distinct effect pattern emerged, with PM_2.5 and NO₂ being more closely linked to bacterial infections than viral ones. Using a generalized additive model, after adjusting for confounders such as daily temperature, precipitation, wind speed, seasonality, and weekends, we identified a nonlinear relationship between PM_2.5 exposure and infection risk: the risk was higher at lower concentrations, decreased as concentrations rose, and increased again at higher levels. In contrast, NO₂ and CO showed a different relationship with infection risk, with an 11.66% increase in odds per unit rise in NO₂ and a 0.64% increase per unit rise in CO. Subgroup analysis confirmed the robustness of these associations, especially for PM_2.5, NO₂, and CO exposure (per standard deviation increments), and the risk of respiratory infections. Notably, the association between PM_2.5 and CO exposure and infection risk was stronger in adolescents (< 18 years) compared to adults, with no significant gender differences. Additionally, the risk associated with NO₂ exposure was more pronounced in patients exposed to higher PM_2.5 levels, suggesting a possible interaction between pollutants. A similar pattern was observed in patients visiting the hospital on weekends and in those enrolled after 2019. These findings provide further evidence supporting the link between acute air pollution exposure and respiratory pathogen infection, helping to fill gaps in current epidemiological knowledge.

The relationship between outdoor air pollutants and respiratory pathogens remains a key area of interest in academic research, and our study aligns with numerous findings from both domestic and international studies, while also highlighting notable distinctions. For instance, Wang et al. [7] conducted a study in Shanghai, demonstrating that a 10 µg/m³ increase in PM_2.5 concentration was linked to a 0.37% rise in respiratory infection hospitalizations, which closely mirrors the 0.3% increase observed in our study on lag day 0. Similarly, research from yang et al. [8] found that on lag day 0, an interquartile increase in PM_2.5 was associated with a 3.08% rise in influenza-like illness at national level.

Internationally, similar patterns have been identified. A study from South Korea [12] also showed that short-term exposure to elevated PM_2.5 levels significantly increased hospitalization rates for respiratory infections, particularly among children, who appeared especially vulnerable. These findings emphasize the global public health burden posed by PM_2.5 as a fine particulate matter. However, the nonlinear relationship between PM_2.5 and infection risk observed in our study contrasts with the linear trends seen in other countries. For example, a nationwide study by Dominici et al. [13] in the United States found that that PM_2.5 exposure consistently increased infection risk, particularly at higher concentrations, suggesting regional variations in pollutant sources and effects.

In southern Chinese cities like Guangzhou, PM_2.5 primarily originates from traffic emissions, secondary organic aerosols, and industrial sources, whereas northern cities such as Beijing see higher contributions from coal burning and industrial emissions during the winter. Northern PM_2.5 is more likely to contain toxic elements like heavy metals and sulfur dioxide, which, under cold and still atmospheric conditions, which are easy to accumulate under low temperature and calm wind conditions, causing acute damage to respiratory mucosa [22,23,24]. In contrast, due to the warm and humid climate in Guangzhou, pollutants are easy to diffuse in the atmosphere, and the proportion of organic carbon and sulfate in PM_2.5 is high, but its acute toxicity is relatively low, which is more manifested by chronic effects on respiratory tract [25,26,27]. PM_2.5 particles may form more complex chemical reactions after contact with respiratory tract, leading to different infection risk modes. These regional differences suggest that when studying the effects of air pollution on respiratory health, it is necessary to consider local pollutant composition and climatic conditions and take targeted public health interventions.

Compared to particulate matter, gaseous pollutants may directly impact respiratory susceptibility by damaging the respiratory mucosa and impairing immune function. Studies have shown that NO₂, as one of the main markers of traffic pollution, can cause damage to respiratory epithelial cells through oxidative stress response, thereby increasing the chances of bacterial and viral invasion. This finding is consistent with prior literature. For example, Chen et al. [28] demonstrated that NO₂ exposure can significantly increase the risk of respiratory infection in patients with chronic obstructive pulmonary disease (COPD) by triggering an inflammatory response. Similarly, our study is the first in southern China to document a significant effect of CO and NO₂ on hospitalization rates for respiratory pathogen infections using a time-stratified cross-analysis method. This mirrors findings from a study [29] in America, which reported that NO₂ exposure significantly elevated respiratory infection rates in children. The pronounced effect of CO exposure can be explained by its ability to inhibit hemoglobin’s oxygen-binding capacity, leading to hypoxia in the respiratory mucosa and weakening the immune defense [30, 31]. Liu et al. [30] further confirmed that short-term CO exposure was significantly associated with the incidence of acute respiratory infections.

Our findings revealed that PM_2.5 and NO₂ exposure showed stronger associations with bacterial infections compared to viral infections. This difference may be explained by several mechanisms. First, bacterial pathogens (0.2–2 μm) are typically larger than respiratory viruses (0.02–0.3 μm), potentially leading to different deposition patterns when interacting with air pollutants [32]. The larger bacterial cells may more readily form aggregates with PM_2.5 particles in airways. Additionally, gaseous pollutants like NO₂ can cause direct damage to respiratory epithelial cells and impair mucociliary clearance, which may preferentially benefit bacterial adherence and invasion. Studies have also shown that air pollutants can alter the expression of pattern recognition receptors and antimicrobial peptides, which are crucial in bacterial defense, potentially explaining the stronger association observed with bacterial infections [33].

Our findings suggest that the presence of elevated PM_2.5 levels appeared to be associated with heightened health risks from gaseous pollutants. As shown in our subgroup analysis, the risk of infection associated with NO₂ and CO exposure was notably higher in the presence of high PM_2.5 concentrations, suggesting these pollutants may have combined effects on respiratory health.

A study from Zhu et al. [34] demonstrated that emergency department visits for respiratory diseases were 1.19% (95%Cl, 0.53%-1.85%) with combined exposure to multiple air pollutants in Chengdu, China. Similarly, research from Colombia [35] supported the existence of a synergistic effect between NO₂ and PM_2.5, while SO₂ showed a similar effect (although not found in our study). These interactions occur because different pollutants can chemically react in the atmosphere to form new harmful substances, while individual pollutant exposure can exacerbate oxidative stress and inflammation. For instance, nitrogen dioxide-induced airway inflammation may increase the permeability of PM_2.5, allowing it to penetrate deeper into the lungs, thereby worsening respiratory health outcomes [36, 37]. This evidence highlights that the combined effect of gaseous pollutants and particulate matter may be a critical factor in elevating infection risks. Future research should aim to dissect the mechanisms behind these interactions across different concentrations and population groups, ultimately providing more targeted public health policies.

Identifying more vulnerable subgroups exposed to air pollution can help reduce the burden of respiratory disease with minimal public health resources. Our study also underscores the importance of identifying vulnerable subgroups to minimize the public health burden, particularly among minors. Adolescents, as shown in our findings, appear more susceptible to respiratory pathogen infections following short-term exposure to CO and PM_2.5. This aligns with existing literature indicating that adolescents are more vulnerable to air pollutants due to their developing respiratory systems and higher metabolic demands [38]. In addition, due to the immature immune system, the airway is more susceptible to inflammation and infection when exposed to air pollutants [38, 39].

This study has several notable strengths as follows. First, this is the first large-scale investigation (n = 96,927) in southern China to examine the association between short-term exposure to air pollutants and daily risk of hospitalization for respiratory pathogen infections. The large sample size enhances the study’s statistical power and allows for more reliable and robust conclusions. Secondly, this study adopts a time-stratified and cross-sectional study design, which effectively controls potential confounding factors within individuals, making the research results more reliable. Additionally, stratified analyses revealed important subgroup differences, particularly highlighting the increased susceptibility of minors to respiratory infections under air pollution exposure, as well as the interactions between PM_2.5 and NO₂/CO. These findings provide valuable insights for the development of targeted public health policies aimed at protecting different demographic groups.

Despite its strengths, this study has some important limitations that warrant discussion. First and foremost, all patients were from a single hospital, the first affiliated hospital of Guangzhou Medical University, and the study was confined to one region. This geographical limitation may affect the generalizability of the findings to other regions, where differences in pollutant concentration and composition may result in varying health outcomes. Furthermore, the chemical composition and toxicity of air pollutants can vary substantially across different geographical locations due to diverse emission sources, meteorological conditions, and atmospheric chemistry. Thus, the conclusions drawn may not fully apply to areas with different pollution levels or climatic conditions. Future multi-center studies across different geographical regions would be valuable to validate and expand upon our findings.

In summary, this study identifies a significant association between short-term exposure to NO₂ and CO and an increased risk of hospitalization for respiratory pathogen infections, with each unit increase associated with 11.66% and 0.64% elevated odds respectively, particularly at lag day 0. These findings support implementing stricter emission controls and establishing low-emission zones around vulnerable populations in urban areas. Given the observed synergistic effects between pollutants, we recommend developing comprehensive air quality management approaches that include early warning systems and enhanced indoor air quality standards, especially in facilities serving minors. Public health policies should prioritize protecting susceptible groups through targeted interventions and strengthened public education during high-pollution periods. Future research should further explore pollutant interactions and their long-term health effects on different population groups to develop more precise preventive measures.