Extreme Cold Events and Pediatric Asthma Outpatient Visits: The Modifying Role of Ambient Carbon Monoxide and Sulfur Dioxide Authors: Xing Gong, Li Wei, Yunshu Xu, Changdi Xu, Yi Zhang, Fengxia Sun, Man Tian, Liya Wang, Hengxue Wang, Ming Ge, Feng Liu, Lilin Xiong, Wei Pan Published: 2026 ABSTRACT AbstractPurposeExtreme cold exposure may trigger respiratory morbidity in children, but whether concurrent ambient air pollution modifies cold-related risks of pediatric asthma remains unclear. This study evaluated associations between extreme cold events and pediatric asthma outpatient visits and assessed whether these associations differed according to ambient pollutant concentrations.Patients and methodsWe conducted a time-series study of 62,167 pediatric asthma outpatient visits among children diagnosed with asthma at a tertiary children’s hospital in Nanjing, China, from 2019 to 2024. Daily asthma visits were linked with meteorological variables and ambient concentrations of PM10, PM2.5, NO2, SO2, CO, and O3. Extreme cold events were defined as ≥2, ≥3, or ≥4 consecutive days with daily mean temperature ≤4.5°C (the 10th percentile of the study-period distribution). Quasi-Poisson distributed-lag non-linear models were used to estimate cumulative relative risks over lag 0–14 days. Pollutant-related effect modification was evaluated using stratified analyses and interaction measures.ResultsOverall, extreme cold events were not significantly associated with pediatric asthma outpatient visits. However, stronger associations were observed during periods with higher CO and SO2 concentrations. For lag 0–14 days, cumulative relative risks for ≥2-day cold events were 2.60 (95% CI: 1.13–5.95) in the higher-CO stratum and 3.41 (95% CI: 1.15–10.11) in the higher-SO2 stratum. Similar patterns were observed for longer cold-event definitions. Multiplicative interactions were identified for both CO and SO2, whereas additive interaction evidence was clearer for CO. Other pollutants showed less consistent modification patterns.ConclusionThe association between extreme cold events and pediatric asthma outpatient visits may depend on concurrent gaseous pollutant conditions. Relatively elevated CO and SO2 may help identify cold-event periods with greater pediatric asthma outpatient burden.Keywords: pediatric asthma outpatient visits, extreme cold events, ambient air pollutants, distributed-lag non-linear models, effect modification Introduction Climate change has become a defining public health challenge of the twenty-first century.1 The Sixth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC) indicates that, despite continued global warming, short-term cold extremes will persist across many mid-latitude regions and may even intensify in some areas.2 A global modeling study covering 750 locations across 43 countries estimated that non-optimal temperatures were associated with 5.08 million deaths annually during 2000–2019, with cold accounting for most of this burden.3 Children are a particularly vulnerable subgroup in this context.4 This vulnerability may reflect age-related differences in thermoregulation, higher minute ventilation relative to body mass, smaller airway caliber, and the continuing development of respiratory and immune systems.5–7 Asthma is one of the most common and burdensome chronic respiratory diseases in childhood. Recent estimates from the Global Burden of Disease Study 2021 indicate that asthma continues to impose a substantial global burden, with approximately 260 million cases across all ages in 20218 In children and young people, clinical reviews have estimated that around 14% are affected by asthma.9 Beyond its high prevalence, childhood asthma also imposes considerable healthcare and societal burdens, contributing to outpatient visits, emergency department utilization, hospitalizations, medication use, and asthma-related school absenteeism10,11 Cold weather can aggravate respiratory symptoms in people with asthma, while airway cooling and drying provide a plausible pathway for bronchoconstriction in susceptible individuals.12,13 However, epidemiological evidence linking ambient cold exposure to pediatric asthma morbidity remains inconsistent. Positive associations have been reported in several Chinese studies, and estimated effects vary according to cold-spell definition, duration, lag window, and age group.14,15 In contrast, studies from other settings have suggested that temperature–asthma associations may be modified by local climate, seasonality, air pollution, and aeroallergen profiles16,17 with some pediatric emergency-department studies observing stronger associations with warmer rather than colder temperatures.18 This heterogeneity underscores the need for region-specific assessments of cold-related pediatric asthma risk. Ambient air pollution is an important co-exposure that may modify cold-related risks for pediatric respiratory outcomes. Recent multi-pollutant analyses in Chinese cities have associated particulate matter (PM2.5, PM10), nitrogen dioxide (NO2), ozone (O3), sulfur dioxide (SO2), and carbon monoxide (CO) with pediatric asthma morbidity, including asthma incidence, outpatient visits, and hospitalizations.19–21 Pediatric time-series evidence has also begun to examine temperature–pollution joint effects. In Hefei, higher PM2.5 and PM10 concentrations strengthened the association between cold exposure and childhood asthma hospitalization, whereas a study in Jinan found that low apparent temperature modified the association between NO2 exposure and pediatric asthma visits.22,23 During cold-wave episodes, increased heating demand, combustion-related emissions, stagnant atmospheric conditions, and reduced pollutant dispersion may jointly alter urban air-pollution profiles, thereby increasing the likelihood of concurrent exposure to low temperature and elevated ambient pollutants.24–26 Cold and air pollution also converge on plausible pathways involving epithelial dysfunction, oxidative stress, and airway inflammation.27–30 However, pediatric asthma-specific joint effects across multiple pollutants remain insufficiently characterized, particularly for outpatient outcomes.31 We hypothesized that associations between cold events and pediatric asthma outpatient visits would differ according to concurrent ambient pollutant levels. To test this hypothesis, we conducted a six-year time-series study in Nanjing, China, using daily pediatric asthma outpatient counts linked with meteorological and ambient air-pollution data. We first estimated overall cold-event associations across three duration thresholds and then assessed whether six ambient pollutants modified these associations. Within the pollutant-stratified analyses, we further quantified multiplicative and additive interaction measures and model-based attributable outpatient burden. Methods Study Setting and Population The present study was conducted in Nanjing, the capital of Jiangsu Province. Nanjing experiences a humid subtropical climate, characterized by hot summers and cool winters. The population of the city is consistently over 9 million. Anonymous electronic outpatient records of children diagnosed with asthma were obtained from a tertiary children’s hospital in Nanjing that serves the city and surrounding regions. The study spanned 2,192 consecutive days from January 1, 2019, to December 31, 2024. A total of 62,167 pediatric asthma outpatient visits were included in the analysis. Only records from Nanjing residents were included. All asthma diagnoses were coded according to the 10th revision of the International Classification of Diseases (ICD-10 code J45) and extracted by trained medical record reviewers. The daily counts of asthma outpatient visits were aggregated by date and further stratified by sex (boys and girls) and age group (<6 years or ≥6 years). Information on BMI, asthma severity, asthma control status, clinical phenotypes, and endotypes was unavailable; therefore, these characteristics could not be described or evaluated in subgroup analyses. Environmental Exposure Data The daily relative humidity and the 24-hour average temperature were obtained from the ERA5-Land reanalysis dataset (0.1° × 0.1°).32 The ChinaHighAirPollutants (CHAP) dataset (1 km × 1 km) was used to determine the daily average concentrations of ambient air pollutants, including PM10, PM2.5, NO2, SO2, CO, and maximum daily 8-hour average ozone (O3–8h).33–36 For each environmental variable, grid cells covering the administrative area of Nanjing were extracted and spatially averaged to generate a city-level daily exposure series. These city-level daily meteorological and air-pollution estimates were then matched to the daily aggregated pediatric asthma outpatient counts by calendar date. The environmental exposure series was complete for all 2,192 study days, and no exposure days were imputed. Daily measurements of aeroallergens, including pollen and mold, and information on indoor air pollution, household heating and ventilation, and personal exposure patterns were unavailable and therefore were not included in the statistical models. Definition of Extreme Cold Events Extreme cold events were defined using a relative temperature threshold to account for regional acclimatization. Specifically, cold events were identified as consecutive days during which daily mean temperature was at or below the 10th percentile (P10) of the study-period temperature distribution. In the six-year dataset, P10 corresponded to 4.5 °C. We then constructed three duration-based event indicators, requiring at least two, three, or four consecutive days at or below this threshold. This duration-graded definition is consistent with previous cold-spell studies in China and allowed us to examine whether associations differed by event duration.14 Statistical Analysis The statistical analyses followed a sequence of related comparisons. We first examined whether daily temperature and cold events lasting at least 2, 3, or 4 consecutive days were associated with the number of pediatric asthma outpatient visits, while accounting for seasonal and long-term time patterns, relative humidity, day of the week, public holidays, and the COVID-19 period. We then compared cold-event associations on days with relatively higher and lower concentrations of each of the six ambient pollutants to assess whether air pollution changed the relationship between cold events and asthma visits. Formal interaction measures were estimated for each pollutant, whereas subgroup analyses focused on CO and SO2, for which the pollutant-conditioned patterns were most consistent. We also estimated the model-based number and proportion of outpatient visits associated with cold events within each pollutant stratum. Finally, we repeated the analyses using alternative model specifications, pollutant cutoffs, and exclusion of the COVID-19 period to assess the robustness of the findings. Detailed model specifications are provided below. All analyses were performed in R version 4.6.0, using the dlnm and splines packages.37,38 Daily counts of asthma outpatient visits were modeled with a quasi-Poisson distribution to accommodate overdispersion.39 Statistical significance was assessed at the two-sided α = 0.05 level. The technical specifications are described below. Overall Temperature and Cold-Event Associations To model the temperature-asthma association with adequate flexibility, a distributed-lag non-linear model (DLNM) for daily mean temperature was constructed.40 The DLNM incorporated a natural cubic spline (three internal knots positioned at the 25th, 50th, and 75th percentiles of the temperature distribution) for the exposure-response dimension and a natural cubic spline (three internal knots placed at equally spaced log-scale lag values) for the lag dimension. The fundamental model was structured as follows: where Yt denotes the daily asthma visit count on day t; α is the intercept; cb(Tt,l) is the cross-basis function of daily mean temperature Tt, spanning the exposure-response dimension and the lag dimension l (l = 0, 1, …, 14); ns (·) denotes a natural cubic spline; timet is calendar time, modeled with 7 degrees of freedom (df) per year over the 6-year study period to control for long-term and seasonal trends; RHt is daily mean relative humidity, modeled with 3 df; and DOWt, Holidayt, and COVIDt are categorical indicators for day of the week, public holidays, and the COVID-19 pandemic period (1 January 2020 to 31 December 2022), respectively. Pollutant-Stratified Analyses To assess the potential of ambient air pollutants to modify the association between cold events and asthma, we conducted pollutant-stratified DLNM analyses as an effect-modification screen. For each pollutant, daily concentrations were dichotomized at the 75th percentile (P75), defining days at or above P75 as the higher-concentration stratum and days below P75 as the lower-concentration stratum. The observed P75 values were approximately 72.43 μg/m3 for PM10, 38.80 μg/m3 for PM2.5, 38.14 μg/m3 for NO2, 7.77 μg/m3 for SO2, 0.81 mg/m3 for CO, and 136.62 μg/m3 for O3. The P75 cutoff was selected as a relative within-city contrast because WHO 2021 24-h guideline values41 were either frequently exceeded (for example, PM2.5) or never exceeded (CO and SO2) in the study data, making guideline thresholds unsuitable as common stratification cut points. The DLNM model structure described above was refitted within each pollutant stratum. The cold-event indicator (binary; coded 1 on event days and 0 otherwise) was incorporated as a linear-exposure cross-basis term to estimate cumulative relative risks (cumRRs) over lag windows of 0–1, 0–7, and 0–14 days. The lag 0–14 window was prespecified as the primary cumulative window. Shorter windows were reported for completeness. Subgroup analyses were conducted by sex (boys and girls) and age group (<6 and ≥6 years). For CO and SO2, pollutant-stratified cold-event associations were estimated separately within each subgroup. These analyses focused on cumulative associations over lag 0–14 days and were considered exploratory. Interaction Analyses Formal interaction between cold events and pollutant levels was evaluated on multiplicative and additive scales using a quasi-Poisson DLNM, and the model was specified as follows: where Coldt denotes the binary cold-event indicator; Pt indicates whether the pollutant concentration on day t was at or above its 75th percentile; and cb(Coldt,l) and cb(Coldt × Pt,l) represent the cross-basis functions for the cold-event main effect and its interaction with higher-pollutant exposure over lag 0–14 days, respectively. Four joint-exposure categories were derived from the same model. For each pollutant and cold-event definition, interaction estimates were constructed using the lower-pollutant, non-cold category as the common reference. RR11 denoted the cumulative relative risk for cold-event days in the higher-pollutant stratum, RR10 denoted the cumulative relative risk for cold-event days in the lower-pollutant stratum, and RR01 denoted the cumulative relative risk for higher-pollutant days in the absence of cold events, with RR00 set to 1. Multiplicative interaction was assessed using the interaction relative risk (IRR),42 whereas additive interaction was evaluated using the relative excess risk due to interaction (RERI).43 These two interaction indices were calculated as follows: The 95% confidence intervals and P values for the IRR and RERI were obtained from 10,000 Monte Carlo draws based on the estimated regression coefficients and their variance-covariance matrices. Attributable Burden To estimate the model-based outpatient burden associated with cold events within each pollutant stratum, we calculated the number of attributable cases (AC) and the attributable fraction (AF). For each stratum and cold event duration, we calculated the AF and AC as follows: where cumRR is the cumulative relative risk; Nobserved is the total number of observed asthma visits on cold event days within that stratum. We obtained 95% confidence intervals for AC and AF using Monte Carlo simulation. Sensitivity Analyses Several sensitivity analyses were conducted to evaluate the robustness of the primary higher-CO and higher-SO2 findings. First, the degrees of freedom for the relative humidity spline were varied from 3 df in the main model to 4, 5, and 6 df. Second, the number of log-spaced knots in the lag-response function was varied from the main-model specification of three knots to two, four, and five knots. Third, because the COVID-19 pandemic may have altered healthcare-seeking behavior, respiratory infection circulation, and exposure patterns, the higher-CO and higher-SO2 cold-event analyses were repeated after excluding the COVID-19 period. Finally, the pollutant-stratified analyses were repeated using P50 and P80 cutoffs to assess sensitivity to alternative concentration thresholds. P50 provided a median split with balanced strata, whereas P80 represented an upper-tail cutoff while retaining sufficient observations for model estimation. Sensitivity analyses focused on lag 0–14 cumulative RRs in the higher-pollutant strata. Results Descriptive Characteristics A total of 62,167 pediatric asthma outpatient visits were recorded over 2,192 consecutive days from January 1, 2019, to December 31, 2024. The daily median number of visits was 27, with an interquartile range of 17 to 37. Boys accounted for 62.32% of visits, and children aged <6 years accounted for 50.93% of visits. During the study period, the daily mean temperature averaged 17.39 °C, and the 10th percentile of daily mean temperature was 4.5 °C. Based on this threshold, 210 days were classified as part of ≥2-day cold events, 182 days as part of ≥3-day events, and 167 days as part of ≥4-day events. The median daily concentrations of PM10, PM2.5, NO2, O3, SO2, and CO were 49.97 μg/m3, 26.23 μg/m3, 26.73 μg/m3, 100.43 μg/m3, 6.00 μg/m3, and 0.68 mg/m3, respectively. Notably, daily CO and SO2 concentrations remained below the WHO 2021 24-h guideline values throughout the study period. Descriptive characteristics are shown in Table 1. Table 1Descriptive Characteristics of Pediatric Asthma Outpatient Visits, Ambient Air Pollutants, Meteorological Conditions, and Extreme Cold Events in Nanjing, China, 2019–2024CharacteristicValueOutcome Total pediatric asthma visits, n62,167 Daily median (IQR)27 (17–37)Subgroup characteristics, n (% of total visits) Boys38,741 (62.32%) Girls23,426 (37.68%) Age <6 years31,660 (50.93%) Age ≥6 years30,507 (49.07%)Air pollutants, daily median (IQR) PM10 (μg/m3)49.97 (34.42–72.43) PM2.5 (μg/m3)26.23 (17.81–38.80) NO2 (μg/m3)26.73 (18.41–38.14) SO2 (μg/m3)6.00 (5.00–7.77) CO (mg/m3)0.68 (0.56–0.81) O3 (μg/m3)100.43 (70.00–136.62)Weather, daily mean (SD) Temperature (°C)17.39 (9.25) Relative humidity (%)71.80 (15.51)Extreme cold-event days, n ≥2-day events210 ≥3-day events182 ≥4-day events167 Descriptive Characteristics of Pediatric Asthma Outpatient Visits, Ambient Air Pollutants, Meteorological Conditions, and Extreme Cold Events in Nanjing, China, 2019–2024 Association Between Extreme Cold Events and Pediatric Asthma Outpatient Visits We first evaluated the overall association between ambient temperature and pediatric asthma outpatient visits. The cumulative exposure-response curve over lag 0–14 days showed a J-shaped pattern, with the minimum morbidity temperature estimated at 6.36 °C (Figure 1A). Around the cold-event threshold of 4.5 °C, the cumulative relative risk was close to unity, suggesting no clear increase in pediatric asthma visits at this temperature threshold in the overall population. Consistently, the lag-response curves for ≥2-, ≥3-, and ≥4-day cold events showed only modest deviations from a relative risk of 1.0 across lag days 0 to 14 (Figure 1B). The lag 0–14 cumulative relative risks were 1.13 (95% CI: 0.94–1.35), 1.11 (95% CI: 0.94–1.32), and 1.06 (95% CI: 0.89–1.26) for ≥2-, ≥3-, and ≥4-day cold events, respectively. Full estimates across lag windows are provided in Supplementary Table S1. Overall temperature-asthma association and lag response of extreme cold events (A) Overall cumulative relative risk (RR) and 95% confidence interval (CI) for pediatric asthma outpatient visits across daily mean temperature over lag 0–14 days. The red dashed vertical line indicates the 10th percentile temperature threshold (P10 = 4.5 °C) used to define extreme cold events, and the gray dotted vertical line indicates the MMT (6.36 °C). (B) Cumulative RRs and 95% CIs across lag days 0–14 for extreme cold events lasting ≥2, ≥3, or ≥4 consecutive days at or below 4.5 °C. The shaded areas represent 95% CIs; the horizontal dashed line indicates RR = 1. Abbreviations: CI, confidence interval; RR, relative risk. Modification of Cold-Event Effects by Ambient Air Pollutants Stratified analyses indicated that cold-event associations differed according to ambient CO and SO2 concentrations (Figure 2). Using lag 0–14 days as the primary cumulative window, cold-event associations were consistently elevated in the higher-CO stratum across all three duration definitions, with cumulative RRs of 2.60 (95% CI, 1.13–5.95), 2.41 (95% CI, 1.08–5.38), and 2.36 (95% CI, 1.07–5.22) for ≥2-, ≥3-, and ≥4-day cold events, respectively. By contrast, none of the corresponding associations reached statistical significance in the lower-CO stratum. A similar pattern was observed for SO2. In the higher-SO2 stratum, the lag 0–14 cumulative RRs were 3.41 (95% CI, 1.15–10.11), 3.40 (95% CI, 1.06–10.97), and 3.06 (95% CI, 0.98–9.60) for ≥2, ≥3, and ≥4-day cold events, respectively, whereas none of the corresponding associations reached statistical significance in the lower-SO2 stratum. The other pollutants did not demonstrate a consistent lag 0–14 modification pattern. Full pollutant-stratified estimates across lag windows are provided in Supplementary Table S2. Effect modification of cold-event associations by ambient CO and SO2. Cumulative relative risks (RRs) and 95% confidence intervals (CIs) for pediatric asthma outpatient visits associated with extreme cold events over lag 0–14 days, stratified by higher (≥75th percentile) and lower (<75th percentile) ambient pollutant concentrations. (A) Results stratified by carbon monoxide (CO). (B) Results stratified by sulfur dioxide (SO2). Points indicate cumulative RRs, horizontal lines indicate 95% CIs, and the vertical dashed line indicates RR = 1. *, P < 0.05. Attributable Burden To translate these relative risks into potential healthcare burden, we estimated model-based attributable outpatient visits for cold events within pollutant strata. Model-based attributable estimates suggested a potentially substantial outpatient burden during cold events in the higher-CO and higher-SO2 strata. In the higher-CO stratum at lag 0–14 days, 1007.57 (95% CI, 143.31–1,356.10) pediatric asthma outpatient visits during ≥2-day cold events were attributable to cold events within that stratum, corresponding to an attributable fraction of 61.47% (95% CI, 8.74%–82.74%). For ≥3-day cold events, the attributable case count was 797.64 (95% CI, 159.26–1,106.74), with an attributable fraction of 58.44% (95% CI, 11.67%–81.08%); for ≥4-day cold events, the attributable case count was 749.58 (95% CI, 67.69–1,049.16), with an attributable fraction of 57.62% (95% CI, 5.20%–80.64%). For SO2, attributable case counts were 786.38 (95% CI, 118.21–998.58), 654.03 (95% CI, 3.32–838.81), and 580.48 (95% CI, −1.36–773.51) for ≥2-, ≥3-, and ≥4-day events, respectively. Table 2 presents the detailed AC and AF estimates, and Supplementary Table S3 provides the corresponding estimates across all pollutant strata. Table 2Attributable Cases of Pediatric Asthma Outpatient Visits Associated with Cold Events, by Higher (≥75th Percentile) versus Lower (<75th Percentile) Ambient Carbon Monoxide and Sulfur Dioxide StrataPollutantStratumCold-Event DurationAttributable CasesAttributable Fraction (%)COHigher (≥P75)≥2-day1007.57 (143.31, 1356.10)61.47 (8.74, 82.74)≥3-day797.64 (159.26, 1106.74)58.44 (11.67, 81.08)≥4-day749.58 (67.69, 1049.16)57.62 (5.20, 80.64)Lower (