Introduction

Chronic obstructive pulmonary disease (COPD) is a heterogeneous lung disease characterized by chronic respiratory symptoms due to airway and alveolar abnormalities, leading to persistent airflow obstruction1. As a chronic disease that poses a significant threat to human health, COPD not only adversely impacts patients’ quality of life but also has a high morbidity and mortality rate. Moreover, it continues to impose a heavy economic burden on families and society1,2,3. Extensive research has shown that the incidence of COPD exhibits a distinct seasonal pattern, with higher rates observed in winter and spring compared to summer4,5. This seasonal variation suggests that changes in various meteorological factors may be closely related to the incidence of COPD6,7. Temperature indicators, such as daily average temperature, diurnal temperature range (DTR), and temperature nodes reflecting cold and heat effects, have been commonly utilized to investigate their relationship with the occurrence of respiratory disorders7,8,9,10,11,12,13,14,15. Numerous studies have confirmed that low temperature and low humidity are associated with acute exacerbations of COPD11,12,16,17. For instance, Chen et al. (2018) reported that the ambient temperature risk of COPD mortality in China was primarily attributed to moderate cold temperatures18. Moreover, DTR is capable of reflecting the magnitude of temperature fluctuations within a single day, which may exert a unique impact on the condition of COPD patients. Studies in Changchun and Taiwan of China have both identified its adverse effects on COPD hospitalizations10,15. Compared with daily average temperature, DTR better captures the variability of temperature, and this variability may place greater stress on the respiratory systems of COPD patients, thereby influencing their risk of hospitalization. However, despite these findings, the majority of recent research has focused on examining the connection between a single meteorological component and several illnesses, like COPD, within the same region. This approach is unable to fully reveal the comprehensive influence of multiple meteorological elements on COPD. The climate of northwest China is unique due to its low humidity and significant daily temperature variations, which may significantly impact the hospitalization of COPD patients. Despite the potential importance of this relationship, studies on the association between COPD and climatic variables in this specific region are limited. Conducting such research can provide a crucial scientific basis for the development of appropriate health protection measures in the local area. In this study, we employed a Quasi-Poisson distribution generalized additive model combined with a distributed lag non-linear model to systematically elucidate the relationship between temperature, DTR, and relative humidity and hospital admissions for COPD. Furthermore, we defined and examined the impact of high and low levels of each meteorological indicator on COPD admissions in the entire population and distinct subgroups, offering a more comprehensive analysis than previous studies.

Methods

Hospitalization data of COPD patients

Qingyang City, in the eastern Gansu Province of northwest China, is encircled by mountains on three sides and has hilly terrain. It has a temperate monsoon climate with concurrent rainy and hot seasons, distinct seasons, and scarce, uneven precipitation. Located at the edge of the monsoon region in the continental interior, Qingyang is highly sensitive to climate change and has a fragile ecology, making it an ideal setting for studying the impact of meteorological factors on chronic obstructive pulmonary disease (COPD) incidence.

By the end of 2019, Qingyang had a permanent population of 2.2788 million, a population density of 80.37 people per square kilometer, and a relatively low economic level with a per capita production value of 32,690 yuan. The low population density minimizes interference from population factors in COPD research, while the low economic level may affect medical resource distribution and lifestyle, both of which are potentially associated with COPD incidence.

Overall, Qingyang City’s unique geography, climate, population, and economic conditions make it an optimal site for investigating the relationship between meteorological factors and COPD incidence, enabling a more precise assessment of meteorological impacts.

This study collected the data of COPD patients hospitalized in all tertiary and secondary hospitals in Qingyang city from January 2015 to December 2019, including their gender, age, main diagnosis and International Classification of Diseases 10th revision (ICD-10) discharge diagnosis code and other relevant information. We screened all individuals with a diagnosis of COPD and ICD-10 code J40–44 for inclusion in this study.

Data on air pollutants and meteorological factors

This study’s meteorological data originate from the China Meteorological Data Sharing System, which is the China Meteorological Administration’s authoritative service platform for opening up common meteorological data resources. The daily average temperature (°C), maximum temperature (°C), minimum temperature (°C), daily average pressure (hPa), daily average relative humidity (%), and average wind speed (m/s) were among the data that we extracted. DTR is the difference between the day’s highest and lowest temperatures, and the air pollution data (including PM2.5, PM10, SO2, NO2, O38h, CO) during the same period were obtained from the national urban air quality real-time release platform.

Statistical models

The number of daily COPD hospitalizations is tiny in comparison to the city’s population, and because the daily counts roughly follow a Poisson distribution, quasi-Poisson regression is a better fit for managing overdispersion in the study. We attempted negative binomial modeling (via mgcv::gam(family = nb)), but it performed poorly under combinations of complex smooth terms, while quasi-Poisson regression remained stable and yielded reliable results, hence our choice of the latter model. Thus, we integrated the distributed lag nonlinear model (DLNM) with the generalized additive model (GAM) with quasi-Poisson distribution allowing for overdispersion to investigate the effect of three meteorological characteristics on COPD hospital admissions. With the number of daily hospital admissions as the dependent variable, we created cross-basis matrices for COPD admissions and meteorological data separately and fitted them with quasi-Poisson link functions. To account for them, additional covariates were included to the model. Referring to a previous study19,38, PM10, SO2 and NO2 were included in the model as air pollutants. Finally, we obtained the basic model of the relationship between ambient temperature and COPD hospitalization as follows:

log (Yt) =α + βTEMt, l + ns(PM10t, df) + ns(SO2t, df) + ns(NO2t, df) + ns(RHt, df) + ns(pt, df) + ns( Timet, df) + Dow + Holiday.

Where Yt is the number of inpatients on day t, α is the model intercept, TEMt, l is a two-dimensional matrix of daily average temperature obtained by using the “cross-basis” function in the distributed lag nonlinear model, β is the coefficient of the matrix, l is the lag days, PM10t, SO2t and NO2t are the concentration of air pollutants on day t, RHt is the relative humidity on day t, Pt is the atmospheric pressure on day t. Time represents the time variable used to control for potential confounding effects from secular and seasonal trends, while Holiday and Dow represent the holiday and day of the week effects, respectively. ns is the natural cubic spline function and df is the degree of freedom. The degrees of freedom of variables in the model were determined according to the minimum principle of Quasi-likelihood for Akaike’s information criterion (Q-AIC)20. In the natural cubic spline function, the degrees of freedom of temperature, humidity and three air pollutants were 4. The degrees of freedom for the time variable were 7 / year. Similar to the above formula, we used DTR and RH as independent variables to analyze their relationship with COPD admission, and the model formula is as follows:

log (Yt) =α + βDTRt, l + ns(PM10t, df) + ns(SO2t, df) + ns(NO2t, df) + ns(RHt, df) + ns(TEMt, df) + ns(pt, df) + ns( Timet, df) + Dow + Holiday.

log (Yt) =α + βRHt, l + ns(PM10t, df) + ns(SO2t, df) + ns(NO2t, df) + ns(TEMt, l, df) + ns(pt, df) + ns( Timet, df) + Dow + Holiday.

Combined with previous studies and the minimum Akaike information criterion21,22,23,24, in this study, 21 days were selected as the maximum lag days when analyzing the relationship between temperature and DTR and COPD admission, and 7 days were selected as the maximum lag days when analyzing relative humidity. We employed two types of lag structures: single-day lags (ranging from lag0 to lag21) and multi-day lags (ranging from lag0-1 to lag0-21). The effects of the single-day lag structure are denoted by lag0, lag1, …, lag21. For instance, lag0 represents the impact of meteorological indicators on the same day, while lag21 indicates the impact of meteorological indicators on the 21st day. The effects of the multi-day lag structure are represented by lag0-1, lag0-2, …, lag0-21. For example, lag0-1 corresponds to the impact of the 2-day moving average of meteorological indicators for the current day and the previous day, and lag0-21 corresponds to the impact of the 21-day moving average of meteorological indicators for the current day and the preceding 21 days.

In this study, the results of the single-day and cumulative risk assessments for COPD hospitalizations were presented as relative risk (RR) values and their 95% confidence intervals (CIs). Additionally, the median (P50) of the three meteorological factors was used as the reference value, and P95 and P5 were used as nodes with high and low effects of each meteorological factor to calculate the relative risk (RR) at different lag times, allowing us to examine the impact of their specific effect value on COPD admissions. SPSS 26.0 software was used to clean the data, describe it, and perform Spearman correlation analysis. R 4.3.1 was used to do regression analysis on the distributed lag nonlinear model. P-values < 0.05 were considered statistically significant. We used the "nlme, mgcv, dlnm, splines, tsModel, reshape2, ggplot2, and lubridate" packages for all statistical analysis in R version 4.3.1.

All methods in this study were carried out in accordance with relevant guidelines and regulations and informed consent has been obtained from all subjects for this study.

Results

Summary of COPD hospital admissions and meteorological indicators in Qingyang city

Figure 1 presents the time series analysis of COPD hospital admissions, meteorological variables, and air pollutant levels in Qingyang City from January 2015 to December 2019. The data reveal that the number of COPD inpatients exhibited significant fluctuations, with a marked increase during the winter months (November through January of the following year). The daily average temperature showed a parallel trend to relative humidity but an inverse relationship with daily average atmospheric pressure. Additionally, peak concentrations of air pollutants were predominantly observed during the winter and spring seasons.

Fig. 1
figure 1

Time series diagram of COPD hospital admissions, meteorological indicators and air pollutants in Qingyang city from January 2015 to December 2019.

Full size image

Table 1 provides a detailed breakdown of COPD admissions in Qingyang City from January 2015 to December 2019. The total number of admissions was 12,841, with 5249 female patients and 7592 male patients. Among these, 4156 were adults under the age of 65, while 8685 were seniors aged 65 and above. On average, there were 7 admissions per day, comprising 4 men and 3 women. Specifically, 2 admissions were adults under 65, and 5 were seniors aged 65 and above.

Table 1 Descriptive analysis of the number of hospitalized COPD patients, meteorological factors and air pollutants in Qingyang city.
Full size table

Meteorologically, the average diurnal temperature range (DTR) was 12.73 °C, with an average relative humidity of 65.62% and an average daily temperature of 10.38 °C. The mean highest and lowest recorded temperatures were 17.69 °C and 4.72 °C, respectively. The average air pressure was 893.00 hPa, and the average wind speed was 1.83 m/s.

Correlation analysis between COPD hospital admissions and meteorological indicators

Table S1 shows that the number of COPD hospital admissions is negatively correlated with temperature and relative humidity, and positively correlated with diurnal temperature range (DTR) and atmospheric pressure. The Spearman correlation coefficients are -0.347, -0.151, 0.084 and 0.228, all of which are statistically significant (p < 0.01). Given the complex interactions often existing between meteorological parameters and air pollutants, our study found that, except for O38h, PM2.5 has a relatively high correlation with other pollutants (Table S1). In contrast, the correlation coefficients of PM10, SO2, NO2 with the three core meteorological parameters are all below 0.5. According to the suggestion of Mukaka in reference25, a correlation coefficient below 0.5 indicates a low correlation between two variables, while a coefficient exceeding 0.5 indicates a moderate or high correlation. This result implies that when controlling the combined effects of meteorological variables and pollutants, the distributed lag nonlinear model can effectively distinguish the independent health effects of each environmental factor.

Effects of different meteorological indicators on COPD admission

Given that the impact of meteorological factors on individuals is influenced not only by the exposure level on the day of interest but also by exposure levels over preceding days or even an extended period, and considering that the relevant effects often manifest several days after exposure26, it is essential to account for both single-day and cumulative risks. Therefore, we employed temperature, diurnal temperature range, and relative humidity to fit the DLNM model with the number of COPD admissions in Qingyang City from January 2015 to December 2019, and obtained the RR value and 95% confidence interval of the influence of meteorological parameters on COPD admissions under two lag structures. Accordingly, we generated contour maps, three-dimensional plots, and cumulative exposure-response curves to illustrate the impact of the three meteorological factors on COPD admissions, as shown in Fig. 2. The three-dimensional plots in Fig. 2 suggest that the three meteorological factors have a nonlinear relationship with COPD admissions and change with lag days.

Fig. 2
figure 2

Contour map, three-dimensional map and cumulative exposure response curve of the influence of three meteorological factors on COPD admission.

Full size image

The upper left contour plot illustrates that the risk of COPD admission associated with lower temperatures (ranging from approximately -14.3 ℃ to 11.7 ℃) progressively increased with extended lag time, particularly between lag6 days and lag21 days. Conversely, the risk of COPD hospitalization linked to higher temperatures (12.3–26.3 ℃) exhibited a gradual decline starting from lag0 day, with this negative effect persisting for only 5 days. By lag6 days, a protective effect began to emerge and intensified with increasing lag time. The most pronounced protective effect was observed at lag21 days when the temperature was 25.3 ℃, with a relative risk (RR) value of 0.918 (95%CI: 0.851, 0.991). The most pronounced detrimental effect was observed at lag21 days with a temperature of -5.7 ℃, with RR of 1.143 (95%CI:1.058, 1.234). The exposure response curve located at the upper right corner indicates that there is an inverse J-shaped correlation with COPD admission and temperature. Specifically, temperature exerts a negative impact on COPD admission between -11.7 °C and 11.3 °C, while it exhibits a protective effect between 12.3 °C and 30.3 °C. The cumulative lagged impact of temperature on COPD admissions was most significant at -1.7 °C, with an RR value of 2.010 (95%CI:1.730, 2.335) at that point.

For DTR ranges of 0–4 °C and 4–10 °C, respectively, the contour map in the middle left illustrates that the harmful effect was more pronounced during lag12-lag21 days and lag0-12 days. The more notable harmful effect started to show on the day of lag (lag0 day) and lasted nearly the entire lag period when the DTR range was 15–25 °C. The maximum effect occurred on lag 0 days when the DTR was 19 °C, with an RR value of 1.038 (95%CI:1.015, 1.061). The exposure-response curve in the center right shows an M-shaped relationship between DTR and COPD admissions. DTR had a negative impact on COPD admission while it was between 14 °C and 24.5 °C, with its cumulative lag effect peaking at 21.7 °C, where the RR was 1.542 (95%CI:1.215, 1.959).

The contour map on the lower left illustrates that the negative effect is more visible when the relative humidity is less than 71% and occurs on the lag day (lag0). Two peaks of harmful effects were observed when the relative humidity was 0, specifically at lag0 days and lag5 days. The most significant harmful effect was noted at lag5 days with a relative humidity of 2%, with RR of 1.199 (95%CI:1.000, 1.437). The exposure-response curve in the lower right reveals a W-shaped relationship between relative humidity and COPD admissions. A negative impact on COPD admissions is observed when relative humidity ranges from 0 to 71%, while a positive impact is seen between 72% and 91%. The cumulative lag effect on COPD admissions is most substantial when the relative humidity is 0, with an RR of 2.142 (95%CI:1.106, 4.150).

High and low effect of meteorological factors had different lag effects on COPD admission

Figure 3; Table 2 demonstrate the single lag effect plot and cumulative effect lag table for each meteorological factor’s high and low effect response to COPD admission, respectively.

Fig. 3
figure 3

Single-day lag effects of high and low effect values of three meteorological factors on COPD admission.

Full size image
Table 2 Cumulative lag effects of high and low effect values of three meteorological factors on COPD admission.
Full size table

Under the two lag structures, the cold effect (P5 = -5.5 °C) began to exhibit statistically significant harmful effects at lag 16 days (single-day lag) and lag 0–18 days (cumulative lag), with both effects progressively increasing as the number of lag days increased. The highest relative risks (RR) were observed at lag 21 days and lag 0–21 days, with values of 1.142 (95%CI: 1.058, 1.234) and 1.867 (95%CI: 1.624, 2.148), respectively. Conversely, Heat effect (P95 = 24.4 °C) had statistically significant protective benefits on COPD admission starting at lag15 days (single-day lag) and lag0–16 days (cumulative lag), with both effects increasing as the number of lag days increased. The lowest RRs were recorded at lag0-21 days and lag21 days, with values of 0.560 (95%CI:0.494, 0.634) and 0.919 (95%CI:0.855, 0.987), respectively.

Low DTR (P5 = 3.1 °C) showed no statistically significant impact on COPD admission in any of the two lag structures. High DTR (P95 = 21.7 °C) showed the greatest harmful effect on COPD admission on the exposure day (lag0 day) in the single-day lag structure, with an RR value of 1.030 (95%CI:1.006, 1.054). This harmful effect progressively diminished as the number of lag days increased, persisting up to lag16 days. According to the cumulative lag structure, the harmful effect of high DTR for COPD admission gradually intensified from lag0-1 day and peaked at lag0-21 days, with an RR value of 1.542 (95% CI:1.215, 1.959).

Low humidity (P5 = 38%) had a negative impact on COPD admission, and the largest harmful effect observed at lag0 days and lag0-7 days, with RR values of 1.083 (95%CI:1.016, 1.154) and 1.239 (95%CI:1.116, 1.374), respectively. In the single-day lag structure, high humidity (P95 = 91.2℃) had a significantly protective effect on COPD admission only on lag1 day, with an RR of 0.960 (95%CI:0.931, 0.989). In the cumulative lag structure, the protective effect of high humidity on COPD admission was strongest at lag0-2, with the RR value of 0.893 (0.818, 0.975).

Subgroup analysis of the high and low effects of each meteorological factor

Figure 4; Table 3 depict the impacts of cold and heat effects on COPD admissions in various age and gender subgroups in Qingyang city using two lag structures. In both lag structures, the adverse impact of cold effect on different age and gender subgroups increased with the number of lag days, whereas the impact of heat effect was the opposite. Specifically, the cold effect was protective for males during lag0-lag2 days and lag0-1 days to lag0-5 days, and for individuals aged 65 and above during lag0-lag4 days and lag0-1 days to lag0-7 days, respectively. In the single lag structure, the greatest harmful effect of cold effect appeared at lag21 days for men, women, and the senior people above the age of 65, with RRs of 1.115 (95%CI:1.007, 1.233), 1.180 (95%CI:1.048, 1.328), and 1.191 (95%CI:1.089, 1.303), respectively. For adults under the age of 65, it was observed in lag9 days, with RR of 1.070 (95%CI:1.000, 1.144). The heat effect on COPD admission in each subgroup was statistically significant at partial lag time, indicating a protective factor. In the cumulative lag structure, the protective impact grew with the number of lag days. The most significant impacts of cold and heat effect on each subgroup were observed at lag0-21 days. At lag0-21 days, the cold effect exhibited its most pronounced detrimental impact on men, women, adults under 65 years of age, and seniors aged 65 and above, with RR values of 1.825 (95%CI:1.519, 2.192), 1.937 (95%CI:1.560, 2.406), 1.741 (95% CI: 1.373, 2.209), and 1.918 (95% CI: 1.625, 2.264).

Fig. 4
figure 4

The impact of cold and heat effects on COPD hospitalization in different age and gender subgroups under the single-day lag mode.

Full size image
Table 3 The impact of cold and heat effects on COPD hospitalization in different age and gender subgroups under the cumulative lag structure.
Full size table

Figure 5; Table 4 present the impacts of high and low DTR on COPD admissions across various age and gender subgroups in Qingyang City under two lag structures. Low DTR did not significantly affect COPD admission in any of the subgroups in any lag structure, while high DTR did not significantly affect COPD admission in the female subgroup. The effect of high DTR on COPD admission in the single lag structure peaked on the lag day (lag0 day) for all subgroups but women, and then progressively declined as the number of lag days increased. With RR values of 1.037 (95%CI:1.006, 1.069), 1.049 (95%CI:1.008, 1.092), and 1.033 (95%CI:1.005, 1.062), respectively, high DTR had the largest harmful effect on the male subgroup, the adults under the age of 65 and the senior people above the age of 65 on the lag day (lag0 day). In the cumulative lag structure, the harmful effect of high DTR on COPD admissions was significant throughout the entire lag period for the male subgroup and individuals aged 65 and above. This harmful effect progressively increased with the number of lag days, reaching its peak at lag0-21 days. However, the harmful effect on the adults under the age of 65 increased initially, then declined. On the day of the largest effect size, the RR values for the male subgroup, adults under the age of 65, and individuals aged 65 and above were 1.662 (95%CI:1.219, 2.265), 1.508 (95%CI:1.051, 2.163), and 1.684 (95%CI:1.269, 2.234), respectively.

Fig. 5
figure 5

The effects of high and low DTR on COPD hospitalization in different age and gender subgroups under the single-day lag mode.

Full size image
Table 4 The effects of high and low DTR on COPD hospitalization in different age and gender subgroups under the cumulative lag structure.
Full size table

Figure 6; Table 5 depict the impacts of high and low humidity on COPD admissions in various age and gender subgroups in Qingyang City using two lag structures. High humidity did not significantly impact COPD admissions in males and adults under 65 years of age but exhibited a protective effect on females and seniors aged 65 and above at specific lag times. In the single-day lag structure, the largest harmful effects of low humidity on the male and senior people above the age of 65 showed at lag0 days, with RR values of 1.121 (95% CI:1.031,1.218) and 1.129 (95% CI:1.048, 1.217), respectively. The cumulative lag structure showed that the harmful effect of low humidity on COPD admission in each subgroup rose with lag days, with the largest effect size occurring at lag 0–7 days. The RR values for the male, female, the adults under the age of 65 and the senior people above the age of 65 were 1.278 (95%CI:1.116, 1.464), 1.179 (95%CI:1.002, 1.386), 1.244 (95%CI:1.035, 1.495), and 1.289 (95%CI:1.141, 1.456).

Fig. 6
figure 6

The effects of high and low relative humidity on COPD hospitalization in different age and gender subgroups under the single-day lag mode.

Full size image
Table 5 The effects of high and low relative humidity on COPD hospitalization in different age and gender subgroups under the cumulative lag structure.
Full size table

Sensitivity analysis

We ran a number of sensitivity tests to ensure the model calculations were stable. We altered the humidity variable degrees of freedom (df = 3,5,6) for the temperature and DTR analyses, as well as the temperature variable degrees of freedom (df = 3,5,6) for the relative humidity study. Furthermore, the model’s time variable degrees of freedom (df = 6,8,9), maximum lag days (lag = 22,23,24), and pollutant degrees of freedom (df = 3,5,6) were modified, as illustrated in Figures S1, S2, and S3. We discovered that the RR values of the three climatic parameters on COPD admission in Qingyang City did not vary significantly, indicating that the model was relatively stable.

Discussion

Our study identified a nonlinear relationship between temperature and COPD admissions in Qingyang, characterized by a cumulative exposure-response curve that exhibited an inverse J-shape. Specifically, lower temperatures were identified as a risk factor for COPD admissions, with the relative risk (RR) progressively increasing with the number of lag days. This finding is in line with numerous studies that have extensively verified the impact of low temperatures on the increased risk of COPD admissions. Conversely, higher temperatures demonstrated a protective effect that intensified with increasing lag time. Notably, our findings indicated that both temperature extremes had RR values exceeding 1 within specific lag time ranges. We observed that the risk associated with higher temperatures manifested earlier but was short-lived and diminished over time. In contrast, exposure to lower temperatures was associated with a delayed but prolonged increase in COPD admission risk, which escalated with time. Consistent with our findings, a study conducted in New York City reported that for each 1 °C increase in temperature above the threshold of 29 °C, the risk of COPD hospitalization increased by 7.6%27. Another study encompassing 12.5 million older persons in 213 cities and counties in the United States found that the chance of hospitalization for COPD increased by 4.7% for each 10-degree Fahrenheit increase in ambient temperature28. This may indicate that both low and high temperatures raise the probability of COPD admission under various lag time circumstances. Research has indicated that extreme temperatures, both low and high, can elevate the risk of COPD-related mortality among individuals aged 65 and older12. Numerous studies have documented an increased risk of COPD admissions associated with low temperatures. The underlying mechanisms may involve a dual pathway: firstly, the reduction in lung function and respiratory defense mechanisms at lower temperatures, and secondly, the heightened airway inflammation resulting from viral infections in the lower respiratory tract16,29,30,31. Likewise, this study demonstrated the impact of greater temperatures on COPD admissions, which is consistent with the results of another study that examined the incidence of disease during heat waves using the heat index32. The cause could be excessive blood flow moving to the bottom of the skin as the body temperature rises, resulting in higher cardiovascular pressure33. On the day with the largest effect size, women were more negatively impacted by the cold effect than men were, and older people over 65 were more negatively impacted than those under 65. However, the protective effect of heat effect was just opposite to that of cold effect. Some studies have suggested that women have lower thermogenesis in cold environments34, and aged individuals have diminished muscle mass and a weak vasoconstrictor response to cold35, which may lead to a greater harmful effect of cold effects on these two subgroups of people.

According to some studies, DTR is positively connected with respiratory disease admissions and COPD mortality36,37, In this study, the greater the value of DTR, the earlier the harm effect appeared, and the chance of COPD hospitalization rose as DTR grew. Previous studies have identified a U-shaped relationship between DTR and daily mortality in Guangzhou, China38, as well as a U-shaped association between DTR and respiratory disease mortality on the plateau of northwest China39. Our cumulative exposure response curves for DTR and COPD admission were similar, but more M-shaped. We discovered that low DTR had no statistically significant influence on COPD admission, whereas high DTR had a definite negative impact on the overall population. This is consistent with past study findings10,15. For age and gender subgroups, a previous study has revealed that the relative risk of men is larger than that of women when DTR is increased, with the greatest impact in older people over 6515. Our findings revealed that the detrimental effects of high DTR on men were statistically significant compared to those on women. Specifically, on the day when the effect size was the largest, the harmful impact of high DTR on individuals over 65 years of age was greater than that on adults under 65. This is highly consistent with the conclusions of the above study. However, a study examining the effect of DTR on mortality in the northwestern plateau of China suggests that although men are more susceptible to extremely high DTR than women, the susceptibility of people under the age of 75 to DTR is higher than that of people over 6539. Another similar study conducted in Guangzhou suggested that women and the elderly were more sensitive to DTR than men and young adults40. This variation might be caused by various climatic conditions and geographic environments, as well as population adaptation. Specifically, differences in climatic conditions may lead to varying sensitivities to DTR among populations in different regions. For instance, the climatic conditions in the northwestern plateau region of China are relatively extreme, with large diurnal temperature ranges, which may result in a different adaptability to DTR among the local population compared to those in plain areas. In addition, the geographical environment can also affect the sensitivity of populations to DTR. For example, altitude is closely related to changes in DTR, and areas with higher altitudes typically have larger DTRs, which may make populations living in these areas more sensitive to DTR41. The adaptability of populations may also be a significant factor contributing to these differences. People in different regions may gradually develop adaptability to local climates due to long-term exposure to specific climatic conditions42. For example, the study in Guangzhou indicated that women and the elderly are more sensitive to DTR, which may be related to the physiological characteristics and social behaviors of these groups.

It is unclear how DTR influences the admission of patients with COPD, but it may be related to the pathophysiological responses of human respiratory epithelial cells and immune cells to short-term temperature fluctuations43,44. These reactions may vary depending on an individual’s health status and age, thereby resulting in different sensitivities to DTR among various subgroups of the population.

There are few studies looking specifically at the impact of humidity on hospitalizations for COPD. Among the numerous studies examining the effect of temperature on respiratory morbidity, it has been found that high humidity significantly increases the risk of developing COPD under low-temperature conditions45. Our investigation in Qingyang revealed that the risk of COPD admissions increased as relative humidity decreased, with a distinct "W"-shaped exposure-response relationship observed between relative humidity and COPD admissions. Low humidity had a negative impact on COPD admission, which was observed across all subgroups of the population. On the day with the largest effect size, men were more negatively impacted by low humidity than women were, and older people over 65 were more negatively impacted than adults under the age of 65. High humidity was found to have a protective effect, albeit limited to women and individuals aged 65 and older. This finding is consistent with prior research indicating that low humidity is associated with an increase of COPD exacerbations11, which our study supports.

The observed differences in the impact of humidity on COPD may stem from a variety of factors. First, climatic and geographical conditions vary across regions, influencing population sensitivity to humidity. For example, in Lanzhou, China, the dry climate and indoor heating during winter reduce indoor humidity, increasing the risk of respiratory diseases46. A similar situation may exist in Qingyang, where dry air may impede pathogen removal, thereby increasing the risk of hospitalization for COPD47. Second, humidity interacts with other environmental factors such as temperature and air pollution, affecting the onset and exacerbation of COPD. The previous study conducted in Lanzhou has shown that both cold and humid or cold and dry environments can increase the risk of outpatient visits for COPD patients46. High humidity may reduce the oxygen concentration in the air, increase air density, inhibit airflow, and raise the concentration of allergens, thereby increasing the risk of COPD48. Lastly, differences in study design and sample size may lead to variations in research outcomes. Future research should employ larger samples and more rational designs to further elucidate the relationship between humidity and COPD hospitalizations.

Similar to previous environmental epidemiological studies, our study still has several limitations that should be acknowledged. First, although we collected objective meteorological parameters of the environment, these data cannot accurately reflect the individuals’ real exposure levels. Factors such as the time spent outdoors, safety precautions, interior air conditioning, and humidification devices can all influence the actual level of exposure, particularly for climatic elements. Second, this study was limited to Qingyang City, and the findings may not be applicable to other places due to differences in geographical position, climatic conditions, and population sensitivity. Additionally, we did not adjust for other potential confounders, such as air pollution levels, socioeconomic factors, or healthcare access, which could influence the results.

In addition to these study-specific limitations, the Distributed Lag Nonlinear Model (DLNM) employed in our analysis, despite being a powerful tool for capturing the complex nonlinear and lagged relationships between exposure factors and health outcomes, is associated with risks of model misspecification. At the model parameter level, there are no universal standards for selecting the basis type, node distribution, and maximum lag time, and different settings can lead to substantial discrepancies in fitting results and effect interpretations, complicating the balance between underfitting and overfitting49,50. Covariate handling also presents challenges, as some covariates may exhibit nonlinear and lagged characteristics similar to the core exposure factors, yet there is no standardized approach for their modeling; incorrect covariate selection can severely distort the accurate estimation of exposure effects49,50. Furthermore, when choosing an estimation method, the traditional parametric approach and the emerging penalized regression method each have their own merits and drawbacks, with no clear-cut guidelines regarding their appropriate application scenarios. Future research should strive to develop objective model selection criteria, create more intelligent parameter optimization algorithms, and enhance comparative studies of various methods to facilitate the more robust application of the DLNM in environmental health research.

In conclusion, the DLNM model was employed for the first time to quantitatively examine the association of various meteorological parameters and their specific effect sizes with COPD hospital admission in Qingyang, and additional analyses were conducted by age and gender. This study’s content is more comprehensive than that of other studies in the same field. It gives local residents a basis for taking appropriate precautions based on meteorological indicators, and it also holds some inspirational significance for relevant departments in terms of preventing and treating hospitalizations due to COPD.

Conclusions

Our study identified associations between temperature, diurnal temperature range (DTR), and relative humidity with COPD admissions. There are varying degrees of lag effects between the cold effect and high DTR on COPD admission. However, heat effect and high humidity appeared to have protective effects. The senior people above the age of 65 were more susceptible to adverse weather conditions, with cold effect having a more pronounced impact on women. High DTR and low humidity had opposing impacts on COPD admissions. By examining the relationships between these three meteorological factors and daily COPD admissions, our study provides insights that can effectively guide the protection of high-risk populations and offer a scientific basis for relevant departments to develop prevention and control strategies.