We collected data on outpatient visits for CRS during January 1, 2018, to December 31, 2019, using the electronic medical records of the Department of Otolaryngology-Head and Neck Surgery at Wuhan Tongji Hospital. Daily meteorological data, encompassing daily mean temperature (°C), precipitation (mm), and relative humidity (%), were sourced from the National Meteorological Information Center of China¹ for the same period. Diagnoses of CRS during outpatient visits were performed by qualified otolaryngologists at Tongji Hospital. We reviewed medical records to identify and include patients diagnosed with CRS, adhering to the CRS diagnostic criteria outlined in the European Position Paper on Rhinosinusitis and Nasal polyps (2020) definition (15). Participants were restricted to residents of Wuhan, excluding individuals residing outside the city. For each patient, we collected demographic and clinical characteristics. The data used in this study were anonymized prior to analysis, therefore individual consent was not required. This processing method has been approved by the Ethics Committee of Tongji Hospital of Tongji Medical College, Huazhong University of Science and Technology (TJ-C20190311) and was conducted in accordance with the Declaration of Helsinki.

A descriptive analysis was conducted on the demographic data of CRS, including age and type of season, as well as meteorological factors such as daily mean temperature, relative humidity, and precipitation. A time series diagram was plotted to illustrate the trends of these variables over time. Spearman correlation analysis was employed to assess the relationships between meteorological factors and the number of daily outpatient visits for CRS. To evaluate the exposure-response relationships between meteorological factors and outpatient visits for CRS, a generalized additive model (GAM) following a negative binomial distribution, along with a distributed lag non-linear model (DLNM), was utilized. This model accounted for several confounding factors, including meteorological variables, day of the week (DOW), and holidays. The model is defined as follows:

log [ ( Y t ) ] = β 0 + ∑ cb [ ( meteoro log ical var iables , d f 1 ) , ( lag , d f 2 ) ] + s ( doy , d f 3 ) + s ( t , d f 4 ) + factor ( dow )

Where E(Y_t) represents the adjusted daily outpatient visits on day t; t is the day of the study period from 2018 to 2019 (1,2, … 730); β 0 is the intercept; ^cb() denotes the cross-basis function in the distributed lag models, which describes the exposure-response and lag-response associations with their corresponding degrees of freedom; s () indicates a thin plate penalized spline function; and DOY represents the day of the year (1, 2, …, 365). Meteorological variables, including daily mean temperature, relative humidity, and precipitation were controlled in the models to account for potential confounding effects. In addition to the year-round analysis, a stratified analysis was performed by age groups (<18 years, 18–45 years, 45–65 years, and >65 years) and by season, specifically the warm season from April through October and the cold season from November through March (16).

The current analysis was conducted on a daily scale, with the maximum period for delayed effects determined to be 30 days (16–18). Different degrees of freedom were applied to all cross-basis and spline functions to assess the robustness of the models. The regression coefficient (R²) and Akaike’s Information Criterion (AIC) were utilized for model selection. All associations were reported as the relative risk (RR) of a specified percentile value of a particular meteorological variable, accompanied by the corresponding 95% confidence interval (CI), in comparison to a pre-defined reference value. For instance, the RR for temperature was calculated as the ratio of the likelihood of daily visits when an individual was exposed to a specific temperature value, relative to the probability of daily visits when that individual was exposed to the reference temperature value. All statistical analyses were performed using the dlnm (19) and gam packages (20) in R software version 3.4.0.

During the study period, 14,259 outpatient visits for CRS were recorded. Among these cases, 7,335 CRS outpatient visits (51.44%) occurred during the cold season, while 6,924 CRS outpatient visits (48.56%) took place during the warm season. The recorded values of the daily mean temperature, mean relative humidity (RH) and daily total rainfall were 17.53°C, 76.67% and 1.28 mm, respectively. Detailed information is shown in Table 1.

Time trends for daily outpatient visits related to CRS and associated meteorological variations are illustrated in Figure 1. A significant seasonal trend in temperature variations is evident, with maximum temperatures consistently observed during the peak summer months of July and August, while minimum temperatures are recorded in the winter season, particularly in January and February. Notably, the recorded cases of CRS exhibit minor peaks during this same period. Furthermore, the graph indicates an increasing trend in outpatient visit volume for chronic rhinosinusitis as temperatures decreased.

Spearman’s correlations between the number of outpatient visits for CRS and various meteorological variables were presented in Supplementary Figures S1–S3. A negative correlation is observed between CRS and daily mean temperature (r = −0.26, p < 0.01), indicating a potential inverse relationship (Supplementary Figure S1). An age-stratified analysis revealed that the outpatient visit rate for CRS among children (under 18 years) and young adults (18–45 years) showed a negative correlation with the daily average temperature (Supplementary Figure S1). Specifically, an increase in outpatient visit rates was observed when temperatures decreased, suggesting that lower temperatures are associated with heightened outpatient visits in these age groups. Furthermore, no significant correlations were identified between daily mean temperature and CRS outpatient visit rates in middle-aged and older adult populations (Supplementary Figure S1). Seasonal stratification analysis indicated an inverse correlation between daily mean temperature and outpatient visits among CRS population during the warm season (Supplementary Figure S2), while no significant association was identified in the cold season (Supplementary Figure S3).

In constructing the ultimate statistical model, we considered outpatient visits due to CRS from 2018 to 2019, totaling 14,259 cases. To capture the temporal characteristics of the outpatient visits data, the model incorporated autoregressive terms, representing the daily outpatient visits for CRS from 1 to 6 days prior. Additionally, the model allocated six degrees of freedom for the day of the year and four degrees of freedom for the number of days during the study period. This approach aimed to more accurately reflect and simulate the seasonal fluctuations and long-term trends present in the data.

Our results illustrated the lag-response associations for daily mean temperature of −1.8°C (1st percentile), relative to a certain reference value of 32.9°C for CRS, and the exposure-response associations between temperature and CRS cumulative to its corresponding delayed timeframes (8 days; Figure 2). Compared to the maximum reference temperature of 32.9°C, extremely low temperature (−1.8° C) was associated with an increased risk of CRS outpatient visits from the day of exposure through the 8th day (Figure 2A). Furthermore, Figure 2B depicts the cumulative relative risks (RRs) of exposure to a specific temperature value across the entire range compared to the predefined reference value, aggregating all contributions up to 8 days, along with their 95% confidence intervals. A significant correlation was observed for temperatures ranging from −1.8°C to 28°C. Above 28°C, the risk of CRS outpatient visits decreased gradually, with extreme temperatures (−1.8°C, 1st percentile) associated with an RR of 1.946 (95% CI 1.273–2.973). Stratified analyses revealed seasonal modulation of outpatient risks for CRS (Figure 3). During warm seasons, extreme cold (11.7°C, 1st percentile) was associated with an increased risk of same-day visits, a reduced risk of visits 1–2 days post-heatwave, and a subsequent rise in risk after 3 days. The relative risk attributable to extreme cold was 10.531 (95% CI 3.216–34.482; Figures 3A,B). In the cold season, extreme cold (−2.5°C, 1st percentile) was linked to a higher risk of CRS outpatient visits for up to 12 days, compared to the reference temperature (21.8°C), and the extreme temperature was associated with a relative risk of 2.240 (95% CI 1.033–4.859; Figures 3C,D).

We found a distinct sensitivity to temperature variations between children (under 18 years; Figures 4A,B) and young adults (18 ~ 45 years; Figures 4C,D). The risk of CRS outpatient visits demonstrates a significant correlation with daily mean temperature, a relationship that is modulated by temporal lag and cumulative exposure. In both age groups, the lag-response curves exhibit an inverse M-shaped configuration, with the under-18 age group displaying a milder response to temperature extremes. Beginning on the day of cumulative exposure, an increase in temperature is associated with a stepwise reduction in the risk of CRS-related outpatient visits across both demographic groups.