Associations between air pollution and pediatric pneumonia incidence in Ulaanbaatar, Mongolia: age and sex susceptibility

Abstract

Background

Childhood pneumonia is a major public health issue in Mongolia and other low- and middle-income countries (LMIC). Identifying modifiable environmental risk factors and vulnerable populations is essential for developing effective prevention strategies. We investigated whether short-term exposure to ambient air pollutants was associated with daily outpatient visits for pneumonia among children under 5 years of age in Ulaanbaatar, Mongolia, and assessed whether age and sex modified these associations.

Methods

We conducted a time-series analysis from 2014 to 2019 using daily outpatient visits for pneumonia among children under 5 years of age and daily concentrations of particulate matter with an aerodynamic diameter ≤2.5 μm (PM_2.5), particulate matter with an aerodynamic diameter ≤1 μm (PM₁), particulate matter with an aerodynamic diameter ≤10 μm (PM₁₀), sulfur dioxide (SO₂), nitrogen dioxide (NO₂), and ozone (O₃)in Ulaanbaatar. Associations between ambient air pollutants and daily pneumonia outpatient visits were estimated using generalized additive models (GAMs) with a quasi-Poisson distribution. The models estimated relative risks (RRs) per interquartile range (IQR) increase in pollutant concentrations while adjusting for long-term trends, seasonality, temperature, humidity, day of the week (DOW), holidays, and raw coal ban policy. We examined pollutant effects across different lag periods, assessed effect modification by age and sex, and evaluated the robustness of associations using multi-pollutant models.

Results

A total of 48,488 outpatient visits for pneumonia were recorded. At lag day 2, each IQR increase in pollutant concentration was associated with an increased risk of outpatient visits: PM_2.5 [RR =1.057; 95% confidence interval (CI): 1.015–1.100], PM₁ (1.065; 1.025–1.106), SO₂ (1.136; 1.023–1.261), and NO₂ (1.383; 1.108–1.726). PM₁₀ and O₃ showed positive but non-significant associations. We observed effects based on age and sex where infants under 1 year old demonstrated significant effect on pneumonia outpatient visits at lag day 0 for NO₂ (RR =1.565; 95% CI: 1.161–2.110) and PM₁ (RR =1.098; 95% CI: 1.046–1.152) and male exhibited a greater effect size for PM_2.5, PM₁ and NO₂. In multi-pollutant models, PM_2.5 and NO₂ remained positively associated with pneumonia outpatient visits after adjustment for co-pollutants, whereas the associations for PM₁ were attenuated.

Conclusions

Short-term exposure to ambient air pollution, particularly NO₂, PM₁, SO₂ and PM_2.5, was associated with increased risk of daily outpatient visits for clinically diagnosed pneumonia among children under 5 years of age in Ulaanbaatar. The consistency of PM_2.5 and NO₂ associations in multi-pollutant models suggests that these pollutants may represent important independent contributors to childhood pneumonia risk. Boys and infants appeared to be more susceptible, highlighting vulnerable populations that may benefit from targeted air pollution mitigation strategies.

Highlight box.

Key findings

• Short-term exposure to PM_2.5, PM₁, SO₂, and NO₂ was significantly associated with increased outpatient visits for pneumonia among children aged ≤5 in Ulaanbaatar.

• PM_2.5 and NO₂ showed the most consistent associations, and these associations generally remained after adjustment for co-pollutants in multi-pollutant models.

• Boys and infants under 1 year old were more susceptible to air pollution-related pneumonia risks.

What is known and what is new?

• Previous studies from the following countries and regions—China, Korea, Taiwan, and Vietnam have demonstrated associations between ambient air pollution and pediatric pneumonia. Evidence from highly polluted cold-climate cities is lacking.

• This study provides the first evidence from Mongolia that short-term increases in ambient air pollution are associated with increased outpatient visits for childhood pneumonia. This study was the first to examine sex and age as susceptibility factors in relation to pediatric pneumonia and air pollution.

What is the implication, and what should change now?

• These findings highlight the need for continued strengthening of air pollution control strategies in Ulaanbaatar, with particular attention to reducing exposure to PM_2.5 and NO₂ from major urban emission sources.

• Reducing ambient air pollution exposure may help decrease the burden of childhood pneumonia-related outpatient visits, particularly among vulnerable populations such as infants and young children.

• Based on this study, public health interventions particularly those that prioritize vulnerable groups (i.e., infants and boys) during periods of severe winter air pollution should be considered to reduce pneumonia risk.

Introduction

Respiratory diseases, including pneumonia, bronchitis, and asthma are major health concerns for children in Mongolia (1,2). Pneumonia is a major cause of serious illness in young children, resulting in missed school, lower quality of life, hospitalizations and mortality (3). Thus, understanding factors that increase the risk for pneumonia are important for developing public health measures. Prior research supports that in addition to pathogen virulence, environmental factors, such as meteorological conditions and air pollution, can impact the incidence and severity of pneumonia (4). In support of this, the incidence of overall respiratory infections often peak during colder months. In Mongolia, this also coincides with a period of peak air pollution levels. However, evidence remains limited regarding whether elevated levels of air pollution also contribute to increased pneumonia rates, particularly in children.

Air pollution is a major global issue. Ulaanbaatar, Mongolia’s capital, is one of the most polluted cities in the world (5,6). About half of Mongolia’s population lives in Ulaanbaatar, many in “Ger” areas that rely on coal (before May 2019) and coal briquettes (after May 2019) for heating and cooking. Due to lack of proper infrastructure much of this coal for heating is not filtered thus leading to direct exposure to individuals. Furthermore, local traffic and industry add to this air pollution exposure. Finally, Ulaanbaatar is located in a valley with mountains on both sides and often experiences temperature inversions that trap pollutants, especially in winter.

Research has highlighted the adverse effects of ambient air pollution on children’s health. Fine particulate matter particulate matter with an aerodynamic diameter ≤2.5 μm (PM_2.5) inhalation can lead to oxidative stress and inflammation in the respiratory system (7-9). Children are particularly vulnerable to air pollution effects due to a combination of reasons. This is because these exposures occur during periods of lung, immune and metabolic systems development. In addition, they have higher ventilation rates, are principal mouth breathers limiting nasal filtration, have a high lung surface area to body weight ratio, and are regularly engaged in more outdoor activities (10). Thus, early-life exposure to air pollution is a significant concern in children as it can result in impaired lung function and a heightened risk of respiratory illnesses (11,12).

Previous studies conducted in Mongolia have reported on effects of air pollution on children’s health. Long-term air pollution exposure has been associated with decreased lung function (13), increased respiratory symptoms and illnesses (14-16). Additionally, maternal air pollution exposure can impact children’s health via associated adverse birth outcomes (17) with additional potential to impact post-natal lung development and immune responses (18,19). Despite the relatively high incidence of childhood pneumonia rates in Ulaanbaatar, no studies have explicitly investigated the effect of air pollution on pneumonia incidence. Hence, we sought to evaluate the impact of air pollution on pneumonia among children under 5 years old, the age group most vulnerable to pneumonia. Given the concomitant elevations in both air pollution levels and pneumonia incidence in colder months, we sought to assess associations between day-to-day variations in pollution levels and daily number of outpatient visits for clinically diagnosed pneumonia at various lag days. This short time scale allowed us to adjust for long-term confounding from time trend including season, day of week, holiday, and weather conditions. Our hypothesis is that an increase in daily average concentrations of air pollutants would be associated with an increase in pneumonia incidence on the same day or several days later. We further hypothesize that child age and sex may modify the susceptibility to the effect of air pollution on pneumonia. We present this article in accordance with the STROBE reporting checklist (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1940/rc).

Methods

Study design

We conducted a time-series study to examine the short-term associations between ambient air pollution and daily outpatient visits for pneumonia among children aged <5 years in Ulaanbaatar, Mongolia. The study covered the period from January 2014 through December 2019. The study period was selected based on the availability of routinely collected pneumonia surveillance data and corresponding air pollution and meteorological measurements. All eligible daily observations available from January 2014 through December 2019 were included.

Data on daily pneumonia

Data on daily pneumonia cases were sourced from the Health Development Center of Mongolia (HDC), the national agency responsible for collecting and managing health statistics from all public and private health facilities in Mongolia. The HDC maintains a centralized health database that integrates routinely reported data from both inpatient admissions and outpatient visits, submitted through the country’s electronic health information system, which enables standardized, nationwide surveillance (20,21). The study period spanned from January 2014 to December 2019, totaling 2191 days of data collection. Pneumonia cases were identified using International Classification of Diseases, 10th edition (ICD-10) code J18. We selected J18 because it represents the majority of clinically diagnosed pneumonia cases in routine healthcare records where the causative organism is not specified. Cases coded under other pneumonia-related ICD-10 categories (J12–J17) were not included in the dataset. Information collected included the date of visit, age, and primary diagnosis. Cases were aggregated by day to calculate daily counts of pneumonia visits. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Data on air pollution exposure

National Agency of Meteorology and Environmental Monitoring and the Department of Outdoor Air Pollution Reduction set up 15 fixed-site air pollution stations in Ulaanbaatar to monitor air quality located based on area population density (22). Air pollutants are measured and recorded hourly at the 15 stations, including PM_2.5, particulate matter with an aerodynamic diameter ≤10 μm (PM₁₀), particulate matter with an aerodynamic diameter ≤1 μm (PM₁), nitrogen dioxide (NO₂), sulfur dioxide (SO₂), ozone (O₃). We averaged concentrations of air pollutants from the 15 monitoring stations as the proxy for the general exposure for all populations. Additionally, meteorological data, including temperature, relative humidity (RH), and wind speed, were gathered. These data sets were used to calculate 24-hour averages for both air pollutants and meteorological conditions.

Statistical analysis

We used generalized additive models (GAMs) incorporating a quasi-Poisson distribution feature to assess the association between an air pollutant and daily outpatient visits for pneumonia. We incorporated several covariates in the main model: (I) a smooth function of calendar time with a basis dimension of 42 (6 years × 7 basis dimensions per year) to control for long-term and seasonal trends; (II) natural smooth functions with 6 degree of freedom for the present-day (day of outpatient visit) mean temperature and the present-day wind speed to control for potential nonlinear confounding effects of weather conditions; (III) a categorical variable for day of the week (DOW) to account for the variation of outpatient visits within a week; (IV) a categorical variable for raw coal ban policy to account for potential impact of the policy; (V) a categorial variable for holiday to control for changes in outpatient visit behavior; and (VI) cubic regression splines function of RH. To explore lag patterns, we evaluated single-day lags from 0 to 7 days and moving average (MA) exposure windows using the same models. For instance, MA02 represented the average concentration over the present and previous 2 days (3-day exposure window), and MA04 represented the average concentration over the present and previous 4 days (5-day exposure window). The core model was established as follows:

$$LogE\left(Y_t\right)=Z_t+s(Rh,bs{=}^{\prime }c{r}^{\prime })+s(t,df)+s(WS,df)+s(Temp,df)+factor\left(rawcoalban\right)+factor\left(dow\right)+factor\left(holiday\right)$$

where $E\left(Y_t\right)$ denotes the estimated daily outpatient visits for pneumonia on day t; $Z_t$ indicates the pollutant concentrations on same day; s() indicates a smoothing function; Rh is relative humidity; bs=‘cr’ specifies a cubic regression spline basis for smoothing; t is calendar time; df refers to degrees of freedom; WS is wind speed; Temp is ambient temperature; factor(rawcoalban) is a variable for raw coal ban policy; factor(dow) is a variable for DOW; and factor(holiday) is an indicator for public holidays.

Missing data were handled using complete-case analysis. Observations with missing values in the outcome, exposure, or covariates included in a given model were excluded from that specific analysis. No imputation was performed.

Because several air pollutants were moderately to strongly correlated, particularly PM_2.5 with PM₁ (r=0.82) and SO₂ with NO₂ (r=0.73), single-pollutant models were specified as the primary analyses (Table S1). Two-pollutant and multi-pollutant models were subsequently conducted as sensitivity analyses to evaluate the robustness of pollutant-specific associations to adjustment for co-pollutants. In the two-pollutant models, the pollutant of interest was adjusted for one co-pollutant at a time to assess whether the observed association was robust to adjustment for individual co-pollutants. In the multi-pollutant models, multiple pollutants were simultaneously included in the regression equation to evaluate associations under combined pollutant exposure. Model 1 included PM₁, PM_2.5, and NO₂; Model 2 included PM_2.5, PM₁₀, NO₂, and SO₂; and Model 3 included PM_2.5, PM₁₀, NO₂, SO₂, and O₃.

We conducted additional sensitivity analyses to examine associations at longer lag periods (lag 8–lag 14) using the same multi-pollutant model specifications used in the primary analyses restricted to lag 0–lag 7 effects. We also varied the degrees of freedom for the smooth functions from 1 to 8 to evaluate the robustness of the results to different levels of temporal smoothing.

The relative risk (RR) and 95% confidence intervals (CIs) of pneumonia outpatient visits were estimated per interquartile range (IQR) increase in each air pollutant. We plotted the exposure-response relationship curves between each of the 6 air pollutants and outpatient visits for pneumonia by adding a natural spline function with 3 df for each pollutant term in the above main model.

We conducted stratified analysis by age (0–1 vs. 1–5 years) and gender (male vs. female) using the same model structures as the primary analysis models. All analyses were performed with the R statistical software program (version 4.4.1).

Results

Baseline pneumonia rates and air pollution assessments

Table 1 summarizes descriptive statistics of this study. Between January 2014 and December 2019, a total of 48,488 outpatient visits with a diagnosis of pneumonia were recorded among children under 5 years old. On average, there were 22.2 pneumonia outpatient visits daily, with 6.7 cases/day for children <1 year old and 15.5 cases/day for children aged 1–5 years. Based on sex, there were 11.7 and 10.5 daily averaged pneumonia cases were recorded for boys and girls respectively.

Table 1. Summary statistics for daily outpatient visits for pneumonia among children aged <5 years, meteorological variables, and ambient air pollutants.

Variable	Mean	SD	Min	Q1	Median	Q3	IQR	Max
Pneumonia outpatient visits per day	22.2	26.1	0	3	16	32	29	250
By age
<1 years old	6.7	8.3	0	1	5	10	9	75
1–5 years old	15.5	18.6	0	2	10	22	20	183
By sex
Male	11.7	13.7	0	2	8	17	15	133
Female	10.5	12.9	0	1	7	15	14	117
Air pollutant concentrations (24-h average, µg/m3)
PM2.5	82.4	79.2	10.1	24.2	48.9	117.9	93.6	404.3
PM10	150.2	82.1	43.0	88.9	129.7	192.2	103.2	710.4
PM1	72.7	81.5	2.1	12.2	35.5	109.9	97.6	425.9
NO2	37.6	17.2	11.1	24.9	32.9	47.9	22.9	119.3
SO2	33.0	30.2	8.0	12.7	21.3	41.1	28.4	283.4
O3	29.7	18.5	1.6	14.5	26.8	41.9	27.3	131.9
Weather conditions (24-h average)
Temperature (℃)	0.6	14.1	−30.3	−13.3	2.5	12.9	26.2	34.3
Relative humidity (%)	55.3	15.5	14.8	43.5	56.2	68.6	25.1	89.0
Wind speed (m/s)	2.0	0.9	0.6	1.5	1.8	2.2	0.7	9.9

IQR, interquartile range; Min, minimum; Max, maximum; NO₂, nitrogen dioxide; O₃, ozone; PM₁, particulate matter with an aerodynamic diameter ≤1 μm; PM_2.5, particulate matter with an aerodynamic diameter ≤2.5 μm; PM₁₀, particulate matter with an aerodynamic diameter ≤10 μm; Q1, first quartile1; Q3, third quartile; SO₂, sulfur dioxide; SD, standard deviation.

The average daily air pollution concentrations were 37.6 µg/m³ for NO₂, 33.0 µg/m³ for SO₂, 29.7 µg/m³ for O₃, 72.7 µg/m³ for PM₁, 82.4 µg/m³ for PM_2.5, and 150.2 µg/m³ for PM₁₀. Regarding meteorological factors, the daily mean temperature recorded was 0.6 ℃, the daily mean RH was 55.3%, and the daily mean wind speed was 2.0 m/s.

Association between air pollution exposure and pneumonia

Table 2 presents the associations between ambient air pollutant concentrations and daily outpatient visits for pneumonia in single-pollutant models across different lag periods. Overall, significant positive associations were observed for PM_2.5, PM₁, SO₂, and NO₂. The strongest effects were generally observed at shorter lag periods (lag 0–lag 2). At lag 0, NO₂ showed the largest effect estimate, with an IQR increase associated with a 56.3% increase in daily pneumonia outpatient visits (RR =1.563; 95% CI: 1.211–2.017). At lag 2, PM_2.5, PM₁, and SO₂ were significantly associated with increased outpatient visits, with RR estimates of 1.057 (95% CI: 1.015–1.100), 1.065 (95% CI: 1.025–1.106), and 1.136 (95% CI: 1.023–1.261), respectively. MA models showed consistent associations, with MA02 (average concentration over the present and previous 2 days, corresponding to a 3-day exposure window) significantly associated with PM_2.5, PM₁, and NO₂, and MA04 (average concentration over the present and previous 4 days, corresponding to a 5-day exposure window) significantly associated with PM₁. Associations for PM₁₀ and O₃ were generally positive but did not reach statistical significance across the examined lag periods. The lag-specific associations are shown in Figure 1.

Table 2. Relative risks (95% CIs) of daily outpatient visits for pneumonia associated with an interquartile range increase in air pollutant concentrations at different lag days in single-pollutant models among children aged <5 years in Ulaanbaatar.

Exposure window	PM2.5	PM10	PM1	NO2	SO2	O3
Lag 0	1.059 (1.016, 1.104)†	1.014 (0.981, 1.048)	1.066 (1.024, 1.110)†	1.563 (1.211, 2.017)†	1.034 (0.918, 1.164)	1.231 (0.971, 1.561)
Lag 1	1.019 (0.978, 1.062)	0.986 (0.955, 1.018)	1.036 (0.996, 1.077)	1.135 (0.901, 1.430)	1.049 (0.935, 1.177)	1.173 (0.932, 1.476)
Lag 2	1.057 (1.015, 1.100)†	1.004 (0.974, 1.036)	1.065 (1.025, 1.106)†	1.383 (1.108, 1.726)†	1.136 (1.023, 1.261)†	0.904 (0.719, 1.137)
Lag 3	1.036 (0.997, 1.077)	0.979 (0.949, 1.010)	1.044 (1.006, 1.084)†	1.005 (0.812, 1.245)	0.995 (0.891, 1.112)	1.046 (0.836, 1.310)
Lag 4	0.983 (0.945, 1.022)	0.985 (0.954, 1.016)	0.979 (0.942, 1.017)	1.070 (0.855, 1.337)	1.002 (0.898, 1.118)	0.912 (0.720, 1.155)
Lag 5	1.018 (0.979, 1.058)	0.993 (0.963, 1.025)	1.007 (0.969, 1.046)	1.049 (0.843, 1.305)	1.118 (1.000, 1.251)†	0.896 (0.706, 1.136)
Lag 6	1.003 (0.964, 1.043)	0.993 (0.964, 1.023)	1.003 (0.966, 1.042)	0.972 (0.790, 1.197)	1.059 (0.947, 1.183)	0.931 (0.737, 1.175)
Lag 7	0.998 (0.960, 1.039)	0.999 (0.968, 1.030)	1.031 (0.993, 1.071)	0.970 (0.786, 1.197)	1.040 (0.927, 1.167)	0.917 (0.729, 1.153)
MA02	1.059 (1.007, 1.113)†	0.993 (0.956, 1.032)	1.076 (1.027, 1.128)†	1.390 (1.060, 1.823)†	1.140 (0.999, 1.301)	1.035 (0.798, 1.341)
MA04	1.060 (0.995, 1.129)†	0.973 (0.926, 1.021)	1.071 (1.011, 1.135)†	1.352 (0.962, 1.901)	1.112 (0.938, 1.318)	1.005 (0.743, 1.360)

^†, the statistically significant estimates. For single-day lag analyses, pollutant exposure was defined as the 24-hour average concentration for the corresponding lag day. CI, confidence interval; MA02, the moving average concentrations on the present and previous 2 days; MA04, the moving average concentrations on the present and previous 4 days; NO₂, nitrogen dioxide; O₃, ozone; PM₁, particulate matter with an aerodynamic diameter ≤1 μm; PM_2.5, particulate matter with an aerodynamic diameter ≤2.5 μm; PM₁₀, particulate matter with an aerodynamic diameter ≤10 μm; SO₂, sulfur dioxide.

Figure 1.

Lag-specific relative risks of pneumonia outpatient visits associated with ambient air pollutant exposure in single-pollutant models among children aged <5 years in Ulaanbaatar. NO₂, nitrogen dioxide; O₃, ozone; PM₁, particulate matter with an aerodynamic diameter ≤1 μm; PM_2.5, particulate matter with an aerodynamic diameter ≤2.5 μm; PM₁₀, particulate matter with an aerodynamic diameter ≤10 μm; SO₂, sulfur dioxide.

Exposure-response relationships

We also wanted to consider the relationship between single pollutant exposure levels and their relationship to pneumonia outpatient visits to determine if there are threshold effects. Figure 2 illustrates the exposure-response relationships between ambient air pollutants and childhood pneumonia. The exposure-response association for PM_2.5 showed a generally increasing pattern, with RR increasing progressively at higher concentrations. The association appeared relatively flat at lower concentrations but became more pronounced above approximately 200 µg/m³. PM₁₀ showed a comparatively weak and near-linear positive association across the exposure range, with RR increasing gradually with increasing concentration. For PM₁, the exposure-response curve demonstrated a nonlinear pattern. RR increased from the lower exposure range, reaching a maximum at approximately 250–300 µg/m³, followed by a gradual decline at higher concentrations, although the estimated risk remained above the reference level. NO₂ also showed a nonlinear exposure-response relationship, with risk increasing at lower concentrations, approaching the reference level around 50–60 µg/m³, and then increasing more substantially at higher concentrations. SO₂ exhibited an approximately inverted U-shaped association, with RR increasing at lower concentrations and reaching a maximum at approximately 50–70 µg/m³, followed by a gradual decline toward the reference level at higher concentrations. In contrast, O₃ showed a generally increasing exposure–response relationship, although the estimated RR remained below 1 at the lowest concentrations and crossed the reference level at approximately 30 µg/m³.

Figure 2.

Concentration-response curves for ambient air pollutants at lag 0 and daily outpatient visits for pneumonia among children aged <5 years in Ulaanbaatar, Mongolia. NO₂, nitrogen dioxide; O₃, ozone; PM₁, particulate matter with an aerodynamic diameter ≤1 μm; PM_2.5, particulate matter with an aerodynamic diameter ≤2.5 μm; PM₁₀, particulate matter with an aerodynamic diameter ≤10 μm; SO₂, sulfur dioxide.

Effect modification by age and sex

We assessed modification of air pollution and pneumonia responses based on age and sex (Table 3). Among children younger than 1-year, significant associations were observed for PM_2.5, PM₁, and NO₂. The strongest association was observed for NO₂ at lag 0 (RR =1.565; 95% CI: 1.161–2.110), while PM_2.5 was also significantly associated at lag 0 (RR =1.069; 95% CI: 1.016–1.126). PM₁ showed a consistent association across all examined lag periods, with RR estimates ranging from 1.080 to 1.098 for single-day lags and remaining significant in both MA02 and MA04 models (RR =1.126 and 1.109, respectively). Among children aged 1–5 years, significant associations were observed for PM_2.5, PM₁, SO₂, NO₂, and O₃ at specific lag periods, while PM₁₀ showed no significant associations. The strongest associations in this age group were observed for NO₂ at lag 0 (RR =1.448; 95% CI: 1.134–1.848) and lag 2 (RR =1.427; 95% CI: 1.131–1.801). When stratified by sex, significant associations were observed for PM_2.5, PM₁, NO₂, and O₃ among boys, while PM_2.5, PM₁, SO₂, and NO₂ showed significant associations among girls. Overall, boys exhibited larger effect estimates for PM_2.5, PM₁, and NO₂, particularly at shorter lag periods (Table 3).

Table 3. Relative risks (95% CIs) of daily outpatient visits for pneumonia associated with an interquartile range increase in air pollutant concentrations at different lag days among children aged <5 years in Ulaanbaatar, by age and sex.

Exposure window	PM2.5	PM10	PM1	NO2	SO2	O3
<1 years old
Lag 0	1.069 (1.016, 1.126)†	1.033 (0.991, 1.076)	1.098 (1.046, 1.152)†	1.565 (1.161, 2.110)†	1.085 (0.931, 1.265)	1.014 (0.757, 1.358)
Lag 1	1.040 (0.989, 1.094)	1.002 (0.963, 1.042)	1.086 (1.037, 1.137)†	1.253 (0.934, 1.680)	0.985 (0.848, 1.144)	0.929 (0.695, 1.242)
Lag 2	1.033 (0.984, 1.086)	1.000 (0.962, 1.039)	1.080 (1.032, 1.130)†	1.307 (0.987, 1.730)	1.063 (0.925, 1.222)	0.860 (0.647, 1.144)
MA02	1.056 (0.993, 1.123)	1.001 (0.955, 1.050)	1.126 (1.065, 1.190)†	1.399 (0.994, 1.968)	1.037 (0.873, 1.232)	0.866 (0.627, 1.198)
MA04	1.039 (0.962, 1.122)	0.961 (0.905, 1.020)	1.109 (1.036, 1.187)†	1.259 (0.824, 1.922)	0.958 (0.769, 1.192)	0.847 (0.583, 1.231)
1–5 years old
Lag 0	1.054 (1.008, 1.102)†	1.006 (0.972, 1.042)	1.050 (1.005, 1.096)†	1.448 (1.134, 1.848)†	1.012 (0.894, 1.145)	1.329 (1.035, 1.707)†
Lag 1	1.009 (0.966, 1.055)	0.979 (0.946, 1.013)	1.011 (0.969, 1.054)	1.087 (0.853, 1.385)	1.075 (0.953, 1.212)	1.294 (1.015, 1.650)†
Lag 2	1.067 (1.023, 1.113)†	1.007 (0.974, 1.040)	1.057 (1.014, 1.101)†	1.427 (1.131, 1.801)†	1.164 (1.045, 1.295)†	0.920 (0.721, 1.173)
MA02	1.059 (1.004, 1.117)†	0.990 (0.950, 1.031)	1.051 (0.999, 1.107)	1.383 (1.039, 1.840)†	1.182 (1.031, 1.355)†	1.114 (0.846, 1.466)
MA04	1.069 (0.999, 1.144)	0.978 (0.929, 1.030)	1.050 (0.986, 1.118)	1.390 (0.970, 1.992)	1.179 (0.987, 1.407)	1.078 (0.782, 1.486)
Male
Lag 0	1.069 (1.023, 1.117)†	1.020 (0.985, 1.055)	1.063 (1.019, 1.110)†	1.636 (1.254, 2.133)†	1.042 (0.919, 1.181)	1.275 (1.002, 1.622)†
Lag 1	1.034 (0.990, 1.079)	0.983 (0.950, 1.017)	1.050 (1.008, 1.094)†	1.215 (0.953, 1.550)	1.026 (0.908, 1.159)	1.199 (0.941, 1.526)
Lag 2	1.049 (1.006, 1.094)†	0.994 (0.962, 1.027)	1.069 (1.027, 1.113)†	1.306 (1.035, 1.650)†	1.110 (0.994, 1.240)	0.913 (0.718, 1.162)
MA02	1.064 (1.010, 1.121)†	0.983 (0.944, 1.024)	1.091 (1.038, 1.146)†	1.398 (1.050, 1.859)†	1.103 (0.959, 1.268)	1.057 (0.804, 1.388)
MA04	1.063 (0.994, 1.136)	0.960 (0.912, 1.011)	1.088 (1.024, 1.157)†	1.321 (0.923, 1.890)	1.085 (0.907, 1.299)	0.955 (0.695, 1.313)
Female
Lag 0	1.046 (0.995, 1.100)	1.005 (0.967, 1.044)	1.068 (1.021, 1.118)†	1.454 (1.116, 1.892)†	1.009 (0.880, 1.158)	1.160 (0.884, 1.524)
Lag 1	1.002 (0.956, 1.050)	0.990 (0.955, 1.027)	1.019 (0.975, 1.065)	1.059 (0.815, 1.375)	1.076 (0.945, 1.225)	1.142 (0.879, 1.483)
Lag 2	1.065 (1.019, 1.114)†	1.016 (0.982, 1.052)	1.061 (1.016, 1.107)†	1.513 (1.179, 1.943)†	1.166 (1.038, 1.310)†	0.893 (0.688, 1.159)
MA02	1.057 (0.999, 1.118)	1.006 (0.963, 1.051)	1.061 (1.006, 1.119)†	1.456 (1.074, 1.975)†	1.185 (1.022, 1.374)†	1.010 (0.752, 1.356)
MA04	1.061 (0.988, 1.140)	0.988 (0.935, 1.043)	1.050 (0.984, 1.121)	1.397 (0.951, 2.052)	1.139 (0.941, 1.380)	1.068 (0.759, 1.504)

^†, the statistically significant estimates. For single-day lag analyses, pollutant exposure was defined as the 24-hour average concentration for the corresponding lag day. CI, confidence interval; MA02, the moving average concentrations on the present and previous 2 days; MA04, the moving average concentrations on the present and previous 4 days; NO₂, nitrogen dioxide; O₃, ozone; PM₁, particulate matter with an aerodynamic diameter ≤1 μm; PM_2.5, particulate matter with an aerodynamic diameter ≤2.5 μm; PM₁₀, particulate matter with an aerodynamic diameter ≤10 μm; SO₂, sulfur dioxide.

Multi-pollutant and sensitivity analyses

Given the potential for mixed pollutant exposures as modifiers of childhood pneumonia rates, we also considered 2 pollutant models across the different lag days. The associations between PM_2.5 and pneumonia outpatient visits remained significant at lag 0, lag 2, and MA02 after adjusting for PM₁₀, SO₂, and O₃, but only at lag 2 after adjusting for PM₁ and NO₂ (Table 4). The effect of PM₁ on the outcome remained generally consistent and statistically significant across most lags, particularly when adjusted for PM₁₀, SO₂, and O₃, with the strongest and most consistent associations observed at lags 0, lag 2, and MA02. The effect of NO₂ remained consistently significant at lags 0, lag 2, and MA02 across all adjustment models, with the strongest associations observed at lag 0, particularly when adjusted for PM₁₀ and SO₂. For SO₂, after adjustment for other pollutants, the association at lag 2 remained statistically significant.

Table 4. Relative risks (95% CIs) for daily pneumonia outpatient visits among children aged <5 years in Ulaanbaatar associated with an interquartile range increase in air pollutant concentrations in two-pollutant models.

Pollutant	Model	Lag 0	Lag 1	Lag 2	MA02	MA04
PM2.5	Adjusted for PM10	1.073 (1.020, 1.129)†	1.015 (0.973, 1.060)	1.056 (1.014, 1.099)†	1.056 (1.004, 1.112)†	1.057 (0.991, 1.127)
	Adjusted for PM1	1.035 (0.983, 1.090)	1.003 (0.961, 1.046)	1.051 (1.010, 1.094)†	1.045 (0.993, 1.099)	1.046 (0.982, 1.115)
	Adjusted for SO2	1.071 (1.021, 1.123)†	1.017 (0.974, 1.062)	1.056 (1.015, 1.099)†	1.058 (1.006, 1.114)†	1.058 (0.992, 1.129)
	Adjusted for NO2	1.036 (0.986, 1.088)	1.005 (0.964, 1.048)	1.054 (1.013, 1.097)†	1.048 (0.996, 1.102)	1.044 (0.979, 1.112)
	Adjusted for O3	1.068 (1.024, 1.115)†	1.021 (0.980, 1.064)	1.057 (1.016, 1.100)†	1.061 (1.009, 1.115)†	1.060 (0.995, 1.129)
PM1	Adjusted for PM2.5	1.066 (1.024, 1.110)†	1.016 (0.974, 1.059)†	1.061 (1.021, 1.102)†	1.060 (1.010, 1.113)†	1.053 (0.993, 1.117)
	Adjusted for PM10	1.072 (1.025, 1.122)†	1.034 (0.993, 1.076)†	1.064 (1.024, 1.106)†	1.075 (1.025, 1.128)†	1.069 (1.008, 1.133)†
	Adjusted for SO2	1.070 (1.025, 1.116)†	1.035 (0.994, 1.078)	1.064 (1.025, 1.106)†	1.077 (1.026, 1.129)†	1.070 (1.009, 1.134)†
	Adjusted for NO2	1.049 (1.003, 1.097)†	1.024 (0.984, 1.065)	1.062 (1.022, 1.103)†	1.068 (1.018, 1.120)†	1.058 (0.998, 1.121)
	Adjusted for O3	1.074 (1.031, 1.119)†	1.038 (0.998, 1.079)	1.065 (1.025, 1.106)†	1.078 (1.028, 1.130)†	1.072 (1.012, 1.136)†
NO2	Adjusted for PM2.5	1.437 (1.085, 1.904)†	1.069 (0.844, 1.352)	1.408 (1.129, 1.757)†	1.364 (1.040, 1.790)†	1.351 (0.962, 1.897)
	Adjusted for PM10	1.676 (1.254, 2.241)†	1.116 (0.880, 1.416)	1.387 (1.111, 1.732)†	1.378 (1.049, 1.810)†	1.346 (0.957, 1.892)
	Adjusted for PM1	1.426 (1.089, 1.866)†	1.097 (0.870, 1.383)	1.440 (1.154, 1.796)†	1.398 (1.068, 1.832)†	1.396 (0.994, 1.959)
	Adjusted for SO2	1.668 (1.265, 2.198)†	1.126 (0.881, 1.440)	1.379 (1.105, 1.722)†	1.393 (1.054, 1.841)†	1.342 (0.952, 1.893)
	Adjusted for O3	1.615 (1.269, 2.057)†	1.139 (0.904, 1.436)	1.369 (1.096, 1.709)†	1.383 (1.054, 1.815)†	1.333 (0.947, 1.875)
SO2	Adjusted for PM2.5	0.939 (0.818, 1.077)	1.022 (0.900, 1.150)	1.139 (1.027, 1.264)†	1.125 (0.985, 1.284)	1.105 (0.932, 1.310)
	Adjusted for PM10	1.013 (0.886, 1.157)	1.041 (0.926, 1.171)	1.136 (1.023, 1.261)†	1.135 (0.994, 1.296)	1.108 (0.935, 1.314)
	Adjusted for PM1	0.968 (0.854, 1.098)	1.041 (0.928, 1.168)	1.147 (1.035, 1.272)†	1.141 (1.000, 1.301)†	1.128 (0.952, 1.336)
	Adjusted for NO2	0.897 (0.778, 1.033)	1.000 (0.888, 1.127)	1.123 (1.012, 1.247)†	1.099 (0.961, 1.256)	1.065 (0.896, 1.265)
	Adjusted for O3	1.043 (0.924, 1.178)	1.053 (0.938, 1.182)	1.120 (1.008, 1.245)†	1.129 (0.989, 1.288)	1.097 (0.924, 1.301)

^†, the statistically significant estimates. For single-day lag analyses, pollutant exposure was defined as the 24-hour average concentration for the corresponding lag day. CI, confidence interval; MA02, the moving average concentrations on the present and previous 2 days; MA04, the moving average concentrations on the present and previous 4 days; NO₂, nitrogen dioxide; O₃, ozone; PM₁, particulate matter with an aerodynamic diameter ≤1 μm; PM_2.5, particulate matter with an aerodynamic diameter ≤2.5 μm; PM₁₀, particulate matter with an aerodynamic diameter ≤10 μm; SO₂, sulfur dioxide.

Multi-pollutant analyses generally supported the positive associations observed for PM_2.5 and NO₂. PM_2.5 remained positively associated with pneumonia outpatient visits at lag 0, 2, and 3 in Models 2 and 3, with the strongest estimates at lag 3 (RR =1.096, 95% CI: 1.039–1.156 in Model 2; RR =1.097, 95% CI: 1.040–1.156 in Model 3). NO₂ showed positive associations at lag 0 in all three models, with RRs ranging from 1.349 (95% CI: 1.025–1.776) to 1.773 (95% CI: 1.309–2.401). PM₁ was not consistently associated with pneumonia outpatient visits, with a statistically significant association observed only at lag 7 in Model 1 (RR =1.055, 95% CI: 1.005–1.107). PM₁₀ showed inverse associations at lag 1–3 in Models 2 and 3. O₃ was positively associated with pneumonia outpatient visits at lag 0 in Model 3 (RR =1.367, 95% CI: 1.079–1.731). Overall, the associations for PM_2.5 and NO₂ remained generally robust after adjustment for multiple co-pollutants (Table 5, Table S2).

Table 5. Lag-specific relative risks (95% CIs) of pneumonia outpatient visits associated with ambient air pollutants among children aged <5 years in Ulaanbaatar: multi-pollutant models.

Model	Pollutant	Lag 0	Lag	Lag 2	Lag 3
Model 1	PM1	1.040 (0.989–1.093)	1.035 (0.985–1.088)	1.043 (0.993–1.095)	1.043 (0.993–1.095)
	PM2.5	1.011 (0.958–1.067)	0.991 (0.938–1.047)	1.011 (0.958–1.067)	1.026 (0.973–1.082)
	NO2	1.349 (1.025–1.776)†	1.075 (0.819–1.412)	1.211 (0.931–1.576)	0.838 (0.648–1.084)
Model 2	PM2.5	1.064 (1.007–1.124)†	1.035 (0.979–1.094)	1.062 (1.006–1.121)†	1.096 (1.039–1.156)†
	PM10	0.968 (0.927–1.011)	0.958 (0.918–0.999)†	0.950 (0.911–0.990)†	0.938 (0.899–0.979)†
	NO2	1.655 (1.227–2.232)†	1.194 (0.885–1.610)	1.340 (1.006–1.785)†	1.024 (0.773–1.356)
	SO2	0.865 (0.744–1.006)	1.015 (0.876–1.175)	1.055 (0.928–1.200)	0.961 (0.835–1.106)
Model 3	PM2.5	1.068 (1.011–1.128)†	1.038 (0.982–1.097)	1.062 (1.006–1.121)†	1.097 (1.040–1.156)†
	PM10	0.971 (0.930–1.014)	0.959 (0.919–1.000)†	0.950 (0.911–0.990)†	0.938 (0.899–0.979)†
	NO2	1.773 (1.309–2.401)†	1.245 (0.919–1.687)	1.337 (0.999–1.789)	1.038 (0.778–1.386)
	SO2	0.863 (0.742–1.003)	1.012 (0.874–1.172)	1.056 (0.928–1.200)	0.960 (0.834–1.105)
	O3	1.367 (1.079–1.731)†	1.219 (0.961–1.546)	0.988 (0.779–1.254)	1.054 (0.833–1.334)

^†, the statistically significant estimates. For single-day lag analyses, pollutant exposure was defined as the 24-hour average concentration for the corresponding lag day. CI, confidence interval; NO₂, nitrogen dioxide; O₃, ozone; PM₁, particulate matter with an aerodynamic diameter ≤1 μm; PM_2.5, particulate matter with an aerodynamic diameter ≤2.5 μm; PM₁₀, particulate matter with an aerodynamic diameter ≤10 μm; SO₂, sulfur dioxide.

Sensitivity analyses using longer lag periods (lag 8–lag 14) showed generally attenuated and less consistent associations compared with the primary lag window (lag 0–lag 7). In the multi-pollutant models, PM_2.5 remained positively associated with the outcome at lag 10 (RR, 1.006–1.007; P=0.017–0.019), while most other pollutant estimates were close to the null. The detailed results for longer lag periods are presented in Table S3.

Discussion

This study represents the first investigation of the short-term association between ambient air pollution and clinically diagnosed pneumonia at outpatient visits among children under 5 years of age in Ulaanbaatar, Mongolia. During the study period (January 2014 to December 2019), we observed a significant burden of pneumonia outpatient visits among children, with a total of 48,488 recorded visits. Short-term exposure to air pollutants (PM_2.5, PM₁, SO₂, and NO₂), assessed using different lag periods and MA exposure periods, were associated with increased daily pneumonia outpatient visits. Overall, this suggests that short term increases in air pollution exposure may impact the rates of pneumonia outpatient visits in children and represent an area of needed public health focus in highly polluted cities such as Ulaanbaatar.

Our findings are generally consistent with previous studies from East and Southeast Asia reporting positive associations between ambient air pollutants and childhood pneumonia hospitalizations or emergency visits. A nationwide cohort study in Korea found that short-term exposure to NO₂ and SO₂ was associated with increased pneumonia risk, whereas associations with PM₁₀, O₃, and other pollutants were less consistent (23). Similarly, a study among children under 5 years of age in Ningbo, China reported increased risks of pneumonia and asthma hospitalization associated with PM_2.5, PM₁₀, SO₂, and NO₂ exposure (24). A systematic review and meta-analysis of the short-term exposure to ambient air pollution and pediatric pneumonia hospitalization rates found that each 10 μg/m³ increase in PM₁₀ and PM_2.5 was associated with excess risk of pneumonia hospitalization of 1.5% and 1.8%_, respectively. The corresponding values per 10 ppb increment of gaseous pollutants were 2.9% for SO₂, 1.7% for O₃, and 1.4% for NO₂ (25). Studies from Taiwan and Vietnam have also reported associations between particulate matter, particularly PM_2.5 and PM₁, and pediatric pneumonia-related healthcare utilization (26-28). A study conducted in Shanghai, China, reported that each 10 μg/m³ increase in NO₂ was associated with a 0.5% increase in pediatric hospital admissions for Mycoplasma pneumonia at lag days 1 and 2, while fine particulate matter was associated with a 0.3% increase in admissions at lag day 1 (29). In our study, the lag-specific analyses showed that associations were most evident during the early lag period, particularly for PM_2.5 and PM₁ at Lag 0–2, and generally attenuated at later lags (Figure 1). This temporal pattern is consistent with the short-term effects reported in previous studies and suggests that the impact of air pollution on childhood pneumonia may occur primarily within several days of exposure. Overall, these findings highlight the potential contribution of ambient air pollution to childhood pneumonia worldwide.

The exposure-response analyses demonstrated heterogeneous and predominantly nonlinear relationships between ambient air pollutants and childhood pneumonia. PM_2.5 and PM₁₀ showed generally increasing associations with pneumonia risk, consistent with previous studies (30,31). However, the shape of the associations varied across pollutants, with some curves appearing relatively flat at lower concentrations and others showing stronger increases or attenuation at higher levels. These differences may reflect variation in pollutant sources, composition, exposure distributions, and potential nonlinear responses to increasing exposure. The apparent attenuation at the highest concentrations for PM₁ and SO₂ should be interpreted cautiously because fewer observations and wider CIs at extreme exposure levels may contribute to greater statistical uncertainty.

The effect estimates observed in our study were relatively large compared with several previous reports. This difference may partly reflect the historically high levels of ambient air pollution in Ulaanbaatar. During the study period, the mean 24-hour concentrations of PM₁ (72.7 μg/m³), PM_2.5 (82.4 μg/m³), PM₁₀ (150.2 μg/m³), SO₂ (33.0 μg/m³), and NO₂ (37.6 μg/m³) indicated a heavily polluted urban environment, with substantial day-to-day variability. For comparison, a study among children under 5 years of age in Ningbo, China, reported mean concentrations of 40.2 μg/m³ for PM_2.5, 63.5 μg/m³ for PM₁₀, 13.0 μg/m³ for SO₂, and 39.7 μg/m³ for NO₂. Thus, the mean concentrations of PM_2.5, PM₁₀, and SO₂ in Ulaanbaatar were approximately 2.0, 2.4, and 2.5 times higher, respectively, than those reported in Ningbo, whereas NO₂ concentrations were broadly similar between the two settings (24). In particular, wintertime concentrations of particulate matter frequently exceeded both the Mongolian National Air Quality Standards and the World Health Organization (WHO) Air Quality Guidelines, while NO₂ commonly exceeded WHO guidelines. Such high and variable pollutant concentrations may partly explain the stronger associations observed in our study compared with those reported in cities with lower ambient pollution levels (32,33). Overall, this suggests that air pollution and improvements in air pollution could be a modifiable environmental exposure that might improve childhood pneumonia. Additionally, we also need to consider how these health impacts vary by local pollution levels, composition, climate, and population vulnerability, which is why country and city specific studies are needed. Furthermore, future studies should examine local factors specific to Ulaanbaatar, including pollution sources, spatial variability in exposure, and population characteristics, may influence the risk of childhood pneumonia.

Another important consideration is the strong seasonal relationship between respiratory infections and air pollution in Ulaanbaatar. Respiratory viruses, including respiratory syncytial virus and influenza viruses, contribute substantially to childhood lower respiratory tract disease and frequently circulate during winter months (34,35). These periods coincide with the highest air pollution concentrations in Ulaanbaatar. Although our models accounted for temporal trends, seasonality, temperature, humidity, DOW, and policy-related factors, we did not have individual-level information on viral infections. Therefore, residual confounding by seasonal respiratory virus activity cannot be completely excluded. However, experimental and epidemiological evidence suggests that air pollution may also increase susceptibility to respiratory infections through biological mechanisms including airway inflammation, oxidative stress, impaired epithelial barrier function, and altered immune responses (36-38). Thus, air pollution may act not only as a correlated seasonal exposure but also as an independent contributor that increases vulnerability to respiratory infection.

In the subgroup analyses, the effects of air pollutants on pediatric pneumonia were significant in both girls and boys, with boys generally exhibiting a larger effect of PM_2.5, PM₁ and NO₂ at lag 0 and girls showing a delayed response to SO₂. Previous studies have demonstrated inconsistent results in terms of sex difference. For instance, Duan et al. found that the effect of PM_2.5 on hospitalization for pneumonia was stronger in boys in Shijiazhuang, China (39). Conversely, other studies have reported greater susceptibility among girls for particulate matter and SO₂ exposure, while some observed no substantial differences by sex (24,40,41). Differences between studies may reflect variation in exposure levels, pollutant composition, biological susceptibility, and population characteristics. The observed greater effect estimates among boys in our study should therefore be interpreted cautiously but may suggest differences in respiratory development, immune responses, or exposure patterns based on sex. Further clarification could improve risk stratification for public health measures.

Age specific analysis indicated that infants under 1 year old were more susceptible to PM₁ and NO₂. This increased vulnerability may be explained by several physiological characteristics, including smaller airway diameter, immature immune function, higher respiratory rates, and reduced ability to compensate for airway inflammation (10,11). Previous studies have similarly reported stronger associations between particulate matter exposure and pneumonia-related hospital admissions among infants and younger children (42,43). In children aged 1–5 years, associations were observed with a broader range of pollutants, including PM_2.5, SO₂, NO₂, and O₃ at specific lag periods. Differences in exposure patterns between age groups, including increased mobility and time spent outdoors among older children, may contribute to these findings.

Though our initial analysis focused on individual air pollutants affecting pneumonia outpatient visits in children under 5, it is also important to consider air pollutant interactions and if these modify effects. In two-pollutant models, NO₂ and PM₁ associations remained strong after adjusting for co-pollutants. This suggests these might be stronger factors in pneumonia outpatient visits associations in children. This observation is consistent with prior literature where short-term impact of PM₁ on childhood pneumonia was greater compared to PM_2.5 and PM₁₀ (44). Potential reasons for why PM₁ might have more of an impact could be due to its aerodynamic size that allows it to reach alveolar region of the lung where pneumonias occur and once there induce oxidative stress and impair host immune defense (45). Additionally, NO₂ exposure disrupts lung epithelial integrity and reduces macrophage phagocytic responses, facilitating pathogen invasion (46). Though not directly tested in our proposed studies, these mechanistic studies could suggest how PM₁ and NO₂ might lead to increased childhood pneumonia.

However, the multi-pollutant models provided a more cautious interpretation. PM_2.5 remained positively associated with pneumonia outpatient visits, particularly at lags 2 and 3, while NO₂ showed the strongest association at lag 0. In contrast, the association for PM₁ was substantially attenuated after simultaneous adjustment for multiple pollutants, with limited evidence of an independent effect. O₃ and SO₂ also showed little consistent evidence of independent associations. PM₁₀ showed inverse associations at lags 1–3, which should be interpreted cautiously given its correlation with PM_2.5 and the potential for multicollinearity or statistical suppression (47-49). Overall, the multi-pollutant results provide stronger evidence for independent contributions of PM_2.5 and NO₂, while the association observed for PM₁ in single- and two-pollutant models may partly reflect its correlation with other particulate pollutants.

The sensitivity analyses showed that associations at longer lag periods (lag 8–lag 14) were generally weaker and less consistent than those observed during the primary lag window (lag 0–lag 7). Although some associations remained statistically significant at specific lags, these findings did not show a consistent pattern across pollutants or model specifications. This suggests that the observed associations were primarily concentrated within the earlier exposure period, with limited evidence for persistent effects at longer lags.

This study has several strengths. First, it provides one of the first evaluations of short-term air pollution exposure and childhood pneumonia outpatient visits in Ulaanbaatar, a city with some of the highest particulate pollution levels globally. Second, the time-series design allowed evaluation of multiple lag periods and cumulative exposure windows, providing insight into the timing of pollutant effects. Third, we adjusted for important temporal and meteorological confounders, including long-term trends, seasonality, temperature, humidity, DOW, holidays, and air pollution-related policy changes. Finally, we examined susceptibility differences by age and sex and evaluated pollutant robustness using multipollutant models.

Despite these strengths, the generalizability of these findings may be limited because the study was conducted in Ulaanbaatar, where ambient air pollution levels are particularly high, and exposure patterns may differ from those in other cities. The findings may therefore be most applicable to young children living in similarly polluted urban environments.

In addition, several limitations should be considered. First, exposure assessments were based on ambient monitoring stations rather than individual-level measurements, which may have resulted in exposure misclassification because children may experience different pollution exposures depending on residential location, indoor environments, daily activities, and time spent outdoors. Second, indoor air pollution was not assessed. During the study period, many households in Ulaanbaatar relied on coal burning stoves for heating and cooking, particularly in ger districts, where indoor concentrations of particulate matter can be substantially higher than outdoor levels. Indoor exposure may therefore have contributed to childhood pneumonia risk (15). Third, pneumonia diagnoses were identified using clinical diagnoses and ICD-10 codes without microbiological confirmation, which may lead to outcome misclassification and potential bias. Finally, the subgroup analyses by age and sex should be interpreted with caution because the smaller sample sizes within individual strata may have reduced statistical power to detect associations and increased the uncertainty of effect estimates. Although these analyses provide important insights into potential differences in susceptibility, future studies with larger populations are needed to further evaluate whether age- and sex-specific differences in the effects of air pollution on childhood pneumonia are consistent. In addition, because multiple pollutants, lag periods, subgroup analyses, and model specifications were evaluated, some statistically significant findings may have occurred by chance. The results should therefore be interpreted based on the overall pattern of associations rather than isolated statistically significant estimates.

Conclusions

This multi-year study demonstrates that short-term exposure to ambient air pollution, particularly PM_2.5, PM₁, SO₂, and NO₂, was associated with increased childhood pneumonia outpatient visits in Ulaanbaatar. Associations for PM_2.5 and NO₂ remained robust in multi-pollutant models, suggesting independent effects of these pollutants. Stronger associations were observed among infants and, for some pollutants, boys. These findings add to the growing evidence that reducing air pollution, especially emissions from combustion sources, may help decrease the burden of childhood pneumonia in highly polluted urban settings.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.