Background

Pneumonia is a common but potentially life-threatening disease with a high incidence around the world [1]. It is considered a leading infectious cause of hospitalization and death with rising health care costs [2]. The annual incidence of pneumonia is around 2.5 cases per 1000 adults with almost 20% requiring intensive care and 2% deaths. In terms of management, the cost of inpatient care is 25 times that of outpatient care [3].

Air pollution still poses a major health threat worldwide. It has been reported that over 90% of the world’s population live in areas where the air pollution level exceeds the World Health Organization (WHO) guideline limits [4]. Air pollution has raised serious concerns regarding environment and public health [5]. According to WHO, ambient air pollution accounted for 4.2 million deaths in 2016, which represented 7.6% of all deaths worldwide [4]. The adverse effects of air pollution have been investigated on various diseases [6,7,8].

The air pollution can increase the onset risk of pneumonia, including both hospitalization [9, 10] and outpatient visit [11]. In addition, particulate matter (PM) can worsen the prognosis of pneumonia patients; according to Chen et al. [12], it was associated with increased risk of invasive respiratory and/or vasopressor support and in-hospital mortality.

Although several studies systemically reviewed the effects of particulate matter (PM) on hospital admissions for respiratory diseases, such as asthma and chronic obstructive pulmonary disease (COPD) [13, 14], few have directly investigated the effects of air pollutants on pneumonia-specific hospital admission or emergency room (ER) visit. To summarize the existing evidence and provide a quantitative answer to the above concerns, we performed a systematic review and meta-analysis for the association between short-term exposure of air pollutants and hospital admission or ER visit for pneumonia.

Methods

Literature search strategy

The literature search was performed using PubMed, Embase, and Web of Science for studies on the association between short-term exposure to air pollution and hospital admission or ER visit for pneumonia up to April 10, 2020. The search included keywords related to air pollution (PM2.5, PM10, SO2, NO2, CO, and O3) and hospital admission or ER visit for pneumonia. The search strategy is detailed in Supplementary Table 1, Additional file 1. After removing duplicates, two researchers independently screened the titles and abstracts of all records to identify potentially eligible studies. Then, a full-text review was performed to determine the final inclusion according to eligibility criteria. In cases of disagreement, a consensus was reached by discussion.

Inclusion and exclusion criteria

Studies were included if they: (1) were original studies published in peer-reviewed journals, (2) investigated the short-term effects (defined as those occurring up to 5 days prior to the hospital admission or ER visit) of air pollutants on hospital admission or ER visit for pneumonia, (3) provided sufficient information to calculate regression estimates and 95% confidence intervals (CIs), (4) used time-series or case-crossover study design, and (5) were published in English. Exclusion criteria were: (1) reviews, commentaries, or editorials; (2) in vitro or in vivo studies; (3) studies on children only; (4) studies under special conditions (e.g., high and low temperature) without overall estimates; or (5) studies on combined outcomes with other respiratory diseases. If there were overlapping data, only the most recent and comprehensive data were included in the meta-analysis.

Data extraction

Data were extracted independently by two researchers and discrepancies were resolved by consensus. The following information was extracted from each study: name of the first author, publication year, study setting, study design, number of cases, percentages of male and elderly patients, air pollutants studied, outcome level, and study results. If there were multiple lag estimates for the same exposure, only one estimate was selected to prevent over-representation of a single study in the meta-analysis. For multiple lag estimates, a priori lag selection protocol devised by Atkinson et al. was used with the following priorities: (1) the lag that the author focused on in the abstract or stated a priori, (2) the lag with the most statistical significance (positive or negative), and (3) the lag that showed the largest effect estimate (positive or negative) [15].

Quality assessment

Due to the lack of validated scales for quality assessment of time-series and case-crossover studies, we adapted the quality assessment approach developed by Mustafic et al. [16]. Three components were assessed: (1) pneumonia diagnosis, where one point was given if the diagnosis of pneumonia was coded according to the International Classification of Diseases (ICD) or based on medical records; (2) the air pollutant measures, where one point was given if measurements were performed at least daily with less than 25% missing data; (3) adjustment for confounders, where one point was given if an adjustment for either long-term trends, seasonality, or temperature was made; a second point was given if an additional adjustment was performed either for humidity or day of the week; and a third point was given if a further adjustment was made for influenza epidemics or holidays. Studies that achieved maximum points for all three components were regarded as good quality, whereas those that achieved no points in any of the three components were regarded as low quality; the remaining were regarded as intermediate quality.

Statistical analysis

We used % increase with 95% CIs as a measure of effect size. To pool the results, all estimates were standardized to an increase of 10 μg/m3 of PM2.5 and PM10 concentrations; 1 ppm of CO; 10 ppb of SO2, NO2, and O3. To transform the estimate, the following equation was used: odds ratio (OR)standardized = ORincrement(10)/increment(original) [17]. As the authors of the original articles adjusted for the time-varying confounders, we extracted the adjusted values. Statistical significance was analyzed by Z-test and a p-value < 0.05 was considered statistically significant. Heterogeneity between studies was assessed by a chi square-based Q test and I2 test. A random-effects model (DerSimonian-Laird method) was applied to consider the heterogeneity within and between studies and to give a more conservative estimate of statistical confidence [18]. Publication bias was assessed using funnel plot and Begg’s test [19]. Sensitivity analysis using the leave-one-out method was performed to assess the stability of results. Subgroup analysis was also conducted per region. In addition, meta-analyses for combining the results with the same lag day were performed. All statistical analyses were performed using R software (version 3.6.0; R Foundation for Statistical Computing, Vienna, Austria). This review followed the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) guidelines [20].

Results

The study selection process is summarized in Fig. 1. A total of 1334 records were identified from the three databases and 396 duplicates were excluded. After removing 835 studies during title and abstract screening, 103 were selected for full-text review, and 82 studies were excluded for the following reasons: reviews or letter (n = 4), different outcomes (n = 3), combined outcomes (n = 31), irrelevant studies (n = 10), not providing short-term effects (n = 3), not providing overall effects (n = 10), conducted in children (n = 1), unable to extract data (n = 1), and overlapping studies (n = 19). Finally, 21 studies were included for meta-analysis [9, 10, 21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39].

Fig. 1
figure 1

Flow diagram of study selection

Full size image

The characteristics of included studies are shown in Table 1. Of the 21 remaining studies, 14 were time-series studies and 7 were case-crossover studies. The United States of America was the country where the research was most conducted (n = 7), followed by China (n = 4). Most studies used the ICD code for pneumonia diagnosis; almost 60% of the studies examined hospital admission, while the rest examined ER visit. The number of cases in each study ranged from 325 to 4.2 million. In terms of study quality, about 43% of studies were regarded as good quality, while the rest were intermediate quality.

Table 1 Characteristics of included studies
Full size table

Among the air pollutants analyzed, PM2.5, PM10, NO2, and CO were associated with an increased risk of hospital admission or ER visit for pneumonia (Fig. 2). For PM2.5 and PM10, a 10 μg/m3 increase was associated with a 1.0% (95% CI: 0.5–1.5; I2 = 70%) and 0.4% (95% CI: 0.2–0.6; I2 = 49%) increase in hospital admission or ER visit for pneumonia, respectively. In addition, every 1 ppm increase of CO was associated with 4.2% (95% CI: 0.6–7.9; I2 = 85%) increase in hospital admission or ER visit for pneumonia. For every 10 ppb increase of NO2, SO2, and O3 increased pneumonia-specific hospital admission or ER visit by 3.2% (95% CI: 1.3–5.1; I2 = 60%), 2.4% (95% CI: -2.0-7.1; I2 = 75%), and 0.4% (95% CI: 0–0.8; I2 = 48%), respectively.

Fig. 2
figure 2

Forest plots of the association between air pollutants and hospital admission or emergency room visit for pneumonia. a PM2.5. b PM10. c SO2. d NO2. e CO. f O3

Full size image

The publication bias was assessed using the funnel plot and Begg’s test (See Supplementary Figure 1, Additional file 1) and no evidence of publication bias was found (P > 0.05 for all analyses). Sensitivity analysis was performed by sequentially excluding each study. With the exception of CO, all pollutants, which obtained statistical significance in the main analyses, showed similar results, indicating that no individual study significantly affected the pooled results. The ranges of increase were 0.9–1.1% for PM2.5, 0.4–0.5% for PM10, and 2.4–4.0% for NO2. When excluding the three largest studies, which had more than 1 million cases [10, 37, 38], the pooled % increase of hospital admission or ER visit per 10 μg/m3 increase in PM2.5 and PM10 and 10 ppb in O3 was 1.1% (95% CI: 0.6–1.6; I2 = 46.5%), 0.5% (95% CI: 0.3–0.7; I2 = 0%), and 5.5% (95% CI: -1.1-2.2; I2 = 56%), respectively.

As there were six and five studies for PM2.5 conducted in North America and East Asia, respectively, we performed a subgroup analysis. Analysis on East Asia showed that every 10 μg/m3 increase of PM2.5 was associated with a 1.2% (95% CI: 0.3–2.0) increase in hospital admission or ER visit for pneumonia, whereas the effect estimate for North America was considerably smaller than that for East Asia; yet the confidence intervals still exhibited considerable overlap (0.3, 95% CI: − 0.4-1.0). For PM10, East Asia and North America showed a 0.3% (95% CI: 0.1–0.4) and 0.4% (95% CI: 0.3–0.6) increase for every 10 μg/m3 increase, respectively; for O3, the corresponding values were 0.9% (95% CI: − 0.5-2.4) and 0.2% (95% CI: − 0.8-1.2). In terms of SO2, NO2 and CO, there were no available studies conducted on North America.

When combining results with the same lag day, PM2.5 and PM10 showed significant associations with pneumonia-specific hospital admission or ER visit for all lag days (lag 0 to lag 5) and the largest association was observed for lag 3 and lag 5, respectively (See Supplementary Table 2, Additional file 1). The % increase range per 10 μg/m3 increase for PM2.5 at lag 3 and PM10 at lag 5 was 1.0% (95% CI: 0.4–1.6) and 0.4% (95% CI: 0.2–0.6), respectively. For NO2 and CO, the largest association was observed for lag 2 (% increase: 2.1, 95% CI: 0.1–4.2) and lag 5 (% increase: 29.8, 95% CI: 0.8–67.2), respectively. On the contrary, SO2 and O3 levels did not show significant associations for all lag days (lag 0 to lag5), and the % increase range for SO2 and O3 was − 0.3 to 1.9 and − 1.0 to 0.7, respectively.

Discussion

In this meta-analysis, we demonstrated a significant association between air pollutants (PM2.5, PM10, NO2, and CO) and hospital admission or ER visit for pneumonia, although no such association was identified regarding SO2 and O3. Except for CO, the sensitivity analyses yielded similar results, demonstrating the robustness of the results. In a subgroup analysis by region, PM2.5 increased hospital admission or ER visit for pneumonia in East Asia but not in North America.

PM has been associated with cardiovascular hospitalization in several meta-analysis studies [8, 40]. PM2.5 was reported to even increase cardiovascular mortality by approximately 0.4 to 1.0% for every 10 μg/m3 [41]. Associations have also been reported between PM and other respiratory diseases, including asthma and COPD [13, 14]. It was shown that asthma- and COPD-specific hospital admission increased by almost 2% for every 10 μg/m3 increase in PM2.5. Our findings are also consistent with a recent meta-analysis on short-term PM and pneumonia in children, which used similar meta-analytic methods, including time-series and case-crossover studies, and according to Nhung et al., a 10 μg/m3 increment of PM2.5 and PM10 was associated with a 1.8 and 1.5% increase in pneumonia hospital admission, respectively [42]. The higher estimates in children compared to our results may be attributed to their increased inhalation per body weight and immature immune systems, rendering them more susceptible to infections [43, 44].

Several studies have suggested that PM is related to inflammation. According to Gordon et al., PM2.5 is linked to an increase in pro-inflammatory cytokines (interleukin-1, interleukin-6, and tumor necrosis factor-α) and Th1-type cytokines (interleukin-12 and interferon-γ) [45]. Elevated levels of white blood cells, C-reactive protein, and von Willebrand factor, which are involved in systemic inflammation, have also been observed after PM exposure [46, 47].

There are two possible mechanisms that mainly account for the increased risk of pneumonia induced by PM: altered immunity and oxidative stress, both of which are closely linked to inflammation. The respiratory system possesses multiple tiers of immunity to defend against harmful airborne particles and microorganisms [48]. However, despite its complex protective mechanisms, several studies have shown that PM2.5 exposure can damage the mucociliary system [49], suppress alveolar macrophage uptake [50], and impair microbial clearance [51], whereby it can enhance pneumococcal adherence to airway epithelial cells [52]. Although PM2.5, which can be inhaled more deeply into the lungs, is considered more harmful to health than PM10, the latter can also cause pneumonia in a similar manner [41].

Oxidative stress is another factor believed to play an important role in the pathogenesis of PM-induced pneumonia. Several studies have shown that acute exposure to PM triggers pulmonary oxidative stress. PM can directly generate reactive oxygen species on the surface, alter mitochondrial function, dysregulate antioxidant enzymes (e.g., superoxide dismutase), increase other oxidases in the lungs (e.g., inducible nitric oxide synthase), and activate metabolic enzyme activity including cytochrome P450s and glutathione S-transferase [53, 54]. These responses can cause pulmonary oxidative damage, which induces an inflammatory process in the lungs [55].

Along with PM, NO2, a free-radical gaseous component of indoor and outdoor air pollution, was associated with increased risk of pneumonia hospital admission or ER visit in this study. As NO2 is linked to nitrosative stress in the lungs, it can lead to airway injury [56]. Furthermore, animal studies have shown that exposure to NO2 results in increased susceptibility to both bacterial and viral infections [57, 58], which explains the positive association between NO2 and pneumonia.

CO was another air pollutant associated with pneumonia. According to Ghio et al., CO can trigger proinflammatory responses in the airways [59]. In addition, previous studies have reported the association between CO and other respiratory diseases, including asthma and COPD [13, 60, 61]. However, in the sensitivity analysis, the association did not remain significant, especially when studies of Duan et al. [9], Phosri et al. [30], and Santus et al. [34] were excluded separately, possibly because the studies showed strong associations between CO and pneumonia in addition to having considerable weight. Therefore, caution is required to interpret the association between CO and pneumonia-specific hospital admission or ER visit.

Subgroup analysis by region showed that an increase of PM2.5 was associated with risk of hospital admission or ER visit for pneumonia in East Asia but not in North America. The two regions are known to have remarkably different concentrations of PM2.5 with East Asia generally having a higher exposure to PM2.5 and its consequent health burden than North America [62]. In addition, regional differences could be accounted for by the variability in composition and toxicity of air pollutants and/or variations in population susceptibility. As there have been insufficient studies investigating the direct effects of PM2.5 on pneumonia in different countries, further research is required on the effects of air pollutants between different regions.

We performed meta-regression with the confounder-adjusting method, which is assessed on a 0 to 3 scale by the third component of quality assessment (described in the method section). Except for PM10, no association between air pollutants and pneumonia was affected by the confounder-adjusting method. Although meta-regression analysis showed that the confounder-adjusting method affected the pooled estimates for PM10, most studies on PM10 yielded the same score (2 point); therefore, the result was not confirmative.

This meta-analysis has some limitations that should be considered when interpreting the results. First, included studies used the air pollutant levels obtained from monitoring stations rather than personal exposures. Second, considerable heterogeneity was observed. Third, due to the lack of information from individual studies, some potential factors, which could affect the risk of pneumonia-specific hospital admission or ER visit (e.g., patients’ age or comorbidities), could not be adjusted.

Conclusions

To our knowledge, this is the first systematic review and meta-analysis to evaluate the acute effects of air pollutants on hospital admission or ER visit for pneumonia. By combining the inconsistent findings of several studies, this study revealed the associations between short-term exposure of air pollutants and pneumonia-specific hospital admission or ER visit, especially for PM and NO2. Based on the results, stricter intervention policies regarding air pollution and programs for protecting human respiratory health should be implemented.