Effect of environmental exposure on the economic burden of influenza and pneumonia: case study of a cold industrial city in China

Abstract

Background

Exposure to air pollutants and low temperatures significantly impacts respiratory health. However, quantitative evidence on the economic burden attributable to these environmental exposures. This study aims to quantify the impact of air pollutants and temperature on hospitalization costs for influenza and pneumonia in Changchun, a representative cold industrial city in China.

Methods

We analyzed inpatient data (2017–2020) for influenza and pneumonia from major hospitals, alongside daily meteorological and air pollutant data. Generalized Additive Models (GAM) and Distributed Lag Non-linear Models (DLNM) were employed to assess the associations of pollutant concentrations and temperature with hospitalization costs, respectively, while controlling for confounding factors.

Results

Among 162,621 respiratory inpatients, influenza and pneumonia accounted for 75,113 cases (46.19%) and 990.21 million yuan in costs (41.29%). A 10 µg/m³ increase in PM₂.₅ and SO₂ was associated with immediate cost increases of 0.89% (95% CI: 0.66%, 1.11%) and 3.02% (95% CI: 2.03%, 4.02%), respectively. Extreme cold temperatures (below − 17.91 °C) led to a significant 6.01% (95% CI: 2.02%, 10.16%) increase in same-day costs, with a peak cumulative increase of 33.55% over lag 0–8 days. Children, the elderly, and males were identified as high-risk populations.

Conclusions

This pioneering study provides the first economic burden assessment of environmental exposures on respiratory diseases in China’s cold industrial regions, quantifying the substantial cost impacts of air pollutants and extreme cold. Our findings underscore the urgent need for targeted emission control policies, public health interventions during cold waves, and prioritized protection for vulnerable demographics to alleviate the economic burden of respiratory diseases.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-025-25713-6.

Introduction

With the rapid development of the global economy and advancing levels of industrialization, substantial alterations have been observed in air quality and climate. The human respiratory tract’s direct connection with the external environment makes it more prone to external influence than other bodily systems. Consequently, the ramifications of environmental exposure—encompassing air quality changes and climate issues—for respiratory diseases have garnered significant attention in the domain of public health. Many epidemiological studies have shown that air pollutants can infiltrate and accumulate within the human respiratory tract and alveoli, disrupting normal blood circulation and inducing respiratory diseases, potentially culminating in death [1–3]. Numerous studies in the field of international environmental epidemiology have demonstrated that exposure to air pollutants, even at relatively low concentrations, is associated with elevated morbidity and mortality rates from respiratory diseases [4, 5]. ata from the United States [6, 7] or regions characterized by specific climatic challenges, such as prolonged cold winters [8], which can exacerbate pollution levels and human vulnerability. According to the China Health Statistics Yearbook (2022), the two-week prevalence rate of respiratory diseases among Chinese residents was 74.6‰ in 2018 [9]. Thus, respiratory diseases exert a profound effect on the health status of Chinese residents.

From a health economics standpoint, alterations in the air environment not only influence the population’s health status but are also a pivotal factor in the consumption of health resources and the escalation of health expenditures [10, 11]. In 2016, healthcare expenditure on respiratory diseases in the US amounted to $17 billion [12]. In Canada, meanwhile, the direct economic costs for one asthma patient ranged from $366 to $647 while indirect costs, resulting from lost work time, vitality, and functionality, further exacerbated the overall economic burden [13]. Research in Italy indicated that chronic obstructive pulmonary disease is a costly disease, with related healthcare spending comprising 6% of total expenditures, thereby imposing a substantial economic burden [14]. In China, respiratory diseases similarly consume considerable health resources. Investigations of the cost of treatment, based on the International Classification of Diseases (ICD), have shown that expenditure on respiratory disease constitutes a significant proportion at both the national and local levels, exceeding 10% or more [15–17], thereby serving as a key driver of increased total health expenditures. Therefore, beyond monitoring incidence rates, quantifying the economic burden, particularly the hospitalization costs driven by environmental factors, is crucial for efficient health resource allocation and policy formulation.

The city of Changchun, situated in Jilin Province, lies in the heartland of Northeast China. The region’s temperate continental monsoon climate is characterized by prolonged and severe winters, featuring an average temperature of below − 11 °C and a heating season lasting up to 6 months. The combination of such prolonged low temperatures and the heating model contributes to increases in morbidity and mortality related to respiratory diseases [18, 19]. Furthermore, as a historical industrial hub, Jilin Province has grappled with a legacy of pollution and environmental degradation [20]. While some studies in Northern Hemisphere cold regions (e.g., Finland [21], Norway [22], Sweden [23] countries) have explored the health impacts of temperature on respiratory diseases, research quantifying the associated socioeconomic burden, particularly hospitalization costs, remains insufficient, especially in the context of industrial cold cities like Changchun.

Given these considerations, this study selects the city of Changchun, noted for its typical climatic and pollution features, for a case study. The objective is to analyse the effects of environmental exposure—specifically, alterations in air pollutant concentrations and temperature—on expenditures associated with the treatment of key respiratory diseases. Furthermore, we aim to furnish empirical evidence for elucidating how environmental factors influence the incidence of respiratory diseases and the related socioeconomic effects.

Methods and data

Data collection

1. Patient data

   Per the 10th edition of the ICD (ICD-10), codes J10–J18 are designated as representing influenza and pneumonia. We gather data on all admissions corresponding to those codes to seven major tertiary hospitals in Changchun, spanning January 1, 2017, to December 31, 2020. The information collected for each case includes the patient’s age, gender, admission and discharge dates, diagnosis, hospitalization costs, and other pertinent details.

2. Meteorological data

   To examine the potential influence of meteorological factors on the occurrence of influenza and pneumonia, we retrieve daily average temperature (°C) data for Changchun spanning January 1, 2017, to December 31, 2020, from the National Meteorological Science Data Centre (URL: http://data.cma.cn). This dataset serves as a direct indicator of climate change characteristics in Changchun. Furthermore, to guarantee the precision of the analytical outcomes and eliminate the potential confounding effects of other meteorological variables, data on mean atmospheric pressure and mean relative humidity (%) are concurrently gathered for the same time frame. These meteorological parameters are needed to comprehensively understand the association between meteorological conditions and respiratory illnesses.

3. Pollutant data

   Data pertaining to concentrations of six primary air pollutants—PM2.5, PM10, SO2, NO2, O3, and CO—are obtained from the China Air Quality Online Monitoring and Analysis Platform for Changchun for the period January 1, 2017, to December 31, 2020. Specifically, O3 concentration reflects the daily maximum 8-hour mean, whereas concentrations of PM2.5, PM10, SO2, NO2, and CO signify 24-h averages. To gain further insight into the unique air quality characteristics of Changchun and its national standing, we also consult the ‘Bulletin on China’s Ecological Environment Status’ from 2017 to 2020, published annually by the Ministry of Ecology and Environment. By comparing national average temperatures and concentrations of air pollutants, we can ascertain the distinctive temperature features of Changchun and pinpoint the predominant pollutants in the region. This provides a scientific foundation for our subsequent in-depth analyses.

   It should be noted that the patient data were extracted from the hospitals’ structured electronic health record systems, where the information are mandatory for completion. Meteorological and air pollutant data were obtained from official national monitoring platforms, which also provide complete daily records. Therefore, the dataset used for analysis is complete, with no missing values for the variables of interest.

Statistical analysis

We use descriptive statistical analysis to investigate the demographic profiles and healthcare expenditures of patients. Furthermore, we adjust the treatment costs associated with respiratory disease spanning 2017–2020 for inflation using the consumer price index as reported in the Statistical Yearbooks of the Jilin Provincial Bureau of Statistics, thereby enhancing data comparability. Age is categorized into three groups: children aged 0–14, adults aged 15–59, and seniors aged 60 and older, consistent with the age categorization employed in previous national censuses. Additionally, we use the Kolmogorov–Smirnov test to assess the normality of air pollutant concentrations and meteorological factors. Based on the normality assessment, we then determine the selection of either Pearson’s or Spearman’s correlation coefficient to analyse the correlation between air pollutants and meteorological factors. The correlation coefficient (R) is calculated and subjected to statistical significance testing at the α = 0.05 level. Positive and negative values of R reflect the presence of positive and negative correlations between variables, with an |r| value approaching 1 indicating a strong correlation.

Time series analysis has been extensively used in studies examining the short-term effects of exposure to various factors—including short-term weather variations, fluctuations in pollutant levels, and pollen counts—on health outcomes such as mortality and hospitalization rates [24]. The data for each period in a time series represent the cumulative effect of multiple factors and are influenced by long-term trends, cyclic patterns, and seasonal variations [25–27]. Seasonal-trend decomposition using LOESS (STL) is a method that employs LOESS smoothing to decompose a time series into three components: low-frequency trend terms, high-frequency seasonal terms, and irregular residuals [28]. We use STL decomposition to analyse trends in major air pollutant concentrations and daily hospitalization costs for respiratory diseases in Changchun, spanning January 1, 2017, to December 31, 2020. The analysis encompasses the trend component, seasonal component, and remainder component of these time series data. The expression is as follows:

where Yt represents the time series, Tt denotes the long-term trend component, St signifies the seasonal trend component, and Rt indicates the random fluctuation component.

The choice of statistical models was guided by the distinct exposure-response characteristics of air pollutants and temperature. For the analysis of air pollutants, we employed the Generalized Additive Model (GAM). This approach is well-suited as previous evidence suggests that the relationship between acute exposure to air pollutants and health outcomes is often approximately linear or log-linear over the observed concentration ranges [29, 30]. The primary complexity we aimed to capture was the lagged response over a pre-defined period, which is effectively handled within the GAM framework by including lagged terms. Conversely, for the analysis of temperature, we utilized the Distributed Lag Non-linear Model (DLNM). This decision was based on the well-established, inherently non-linear relationship between temperature and health outcomes [31] (e.g., a U-shaped or J-shaped curve), where both extreme cold and extreme heat pose risks. The DLNM is specifically designed to simultaneously model this non-linear exposure-response relationship and the distributed lag effect, which is crucial as the health impact of a temperature extreme can be immediate and also manifest over subsequent days in a complex pattern.

The GAM, introduced by Hastie and Tibshirani in 1990, builds upon the framework of the generalized linear model, which uses nonparametric functional fitting to estimate the relationship between the predictor variables and the dependent variable. GAM has been widely used in environmental epidemiology to investigate the correlation between air pollutant exposure and disease-related morbidity or mortality [32, 33]. We use GAM to analyse the effect of variations in air pollutant concentrations on hospitalization costs for respiratory diseases. Specifically, we examine the variations in hospitalization costs for respiratory diseases at LAG0 and n-day lags (lag1-n) per 10 µg/m3 increment in air pollutant concentration, termed excess risk (ER). In prior epidemiological studies analysing the correlation between environmental changes and population health changes, the choice of lag days N was often determined according to the purpose of the study, and the differences were large [34–36]. It has been observed that shorter lag periods tend to obscure the lagged effect of environmental factors, whereas longer lag periods might result in statistically insignificant or practically meaningless correlations between environmental factors and hospitalization costs. Consequently, we set the lag time N to 14 day, a period commonly used in studies on respiratory diseases to fully capture the delayed and cumulative effects from initial exposure to symptom development, healthcare-seeking behaviour, and subsequent hospitalization [37–39]. Additionally, given that hospitalization for respiratory diseases is a rare event, its distribution resembles Poisson distribution [40, 41]. Therefore, we develop a GAM model using quasi-Poisson distribution. In this model, hospitalization costs serve as the outcome variable while average daily pollutant concentrations act as predictor variables. Furthermore, the model accounts for smoothing the time trend and meteorological factors. GAM is expressed as follows:

In the equation, Yi represents the daily summary of expected hospitalization costs for respiratory diseases on day i, Ci denotes the mean concentration of air pollutants, and β is the regression coefficient estimated through the model. Time serves as the date variable. To mitigate the influence of meteorological factors on the results, we incorporate mean temperature (MT), mean relative humidity (RH), and mean barometric pressure (MB) into the model. Natural smooth spline functions (ns) are employed. df denotes the degree of freedom. The degrees of freedom for MT, RH, and MB are set to three based on previous studies while the degrees of freedom for time are determined according to the Akaike information criterion[42–44]. Previous studies have also identified an effect of weekdays and holidays on hospital admissions for respiratory diseases [44]. Therefore, to mitigate the influence of weekdays and holidays, day of the week (Dow) and holiday indicators are incorporated as independent variables. The holiday indicator was defined as a binary variable (1 for holiday, 0 for non-holiday), encompassing all official Chinese public holidays (e.g., Spring Festival, National Day, etc.) during the study period. α represents the intercept. Lastly, we conduct sensitivity analysis by adjusting the degrees of freedom for time and establishing a two-pollutant model to mitigate the effect of other pollutants. We use this approach to validate the robustness of the model [45, 46].

The distributed lag model has been widely used to study the health effects of meteorological factors. However, owing to its assumption of a linear exposure–response relationship, it is often not applicable in reality, where many exposure–response relationships (e.g., temperature–mortality) exhibit nonlinear patterns such as U-shaped or V-shaped curves, thereby limiting its applicability [31]. Addressing these issues, Armstrong proposed the DLNM in 2006 for epidemiology [47]. DLNM offers the advantage of accounting for both the lagged effect of the exposure factor and the nonlinear relationship of exposure–response. In this study, we construct a cross-basis matrix for the hospitalization expenses and temperature data of patients with respiratory diseases, with hospitalization expenses as the dependent variable. A quasi-Poisson link function is used for model fitting. By controlling for seasonality, long-term trends, and day-of-the-week effects, DLNM is employed to fit the association between temperature and hospitalization costs for respiratory diseases. Meanwhile, we construct a two-dimensional cross-basis matrix using the natural cubic spline function, with the mean temperature as the reference level. We assess the relationship between the occurrence of high temperatures (P_2.5 percentile of the temperature series) and low temperatures (P_97.5 percentile of the temperature series) at different lag periods and changes in hospitalization costs for respiratory diseases—namely, relative risk (RR) and ER—relative to the reference level. This percentile-based approach is widely used in environmental health studies to define extreme temperatures relative to the local climate [48, 49], as it reflects the population’s adaptation and the specific temperature range to which they are acclimatized, rather than relying on arbitrary absolute values. Additionally, using GAM, we set a maximum lag period of 14 days. The model is as follows:

where basis.temp denotes the cross-basis function matrix. To mitigate the effect of air pollutants and other meteorological variables, the model incorporates factors such as pollutants (pl), average RH, and MB. The definitions of the remaining variables are consistent with those used in the GAM framework. Ultimately, we conduct sensitivity analysis to ascertain the robustness of the model, with adjustments made for varying degrees of temporal freedom.

Results

Demographic characteristics and costs of patients with respiratory diseases

Between 2017 and 2020, a total of 162,621 individuals with respiratory conditions were admitted to the surveyed hospitals, incurring a total hospitalization cost of 2398.46 million yuan. Among these admissions, those attributable to influenza and pneumonia comprised the highest proportion, totalling 75,113 cases (46.19%), representing about half of the total, with their hospitalization expenses constituting 41.29% of the total. Consequently, influenza and pneumonia stand out as the respiratory disease types with the highest incidence and greatest consumption of healthcare resources. The data presented in Table 1 illustrate the distribution of admissions and costs for respiratory diseases during this period.

Table 1.

Number and cost of cases of different types of respiratory disease

Diseases	Number		Expense
	N	%	Yuan (million)	%
Acute upper respiratory infections	10,139	6.23	54.43	2.27
Influenza and pneumonia	75,113	46.19	990.21	41.29
Other acute lower respiratory	10,124	6.23	81.33	3.39
Other diseases of upper respiratory tract	23,184	14.26	280.50	11.70
Chronic lower respiratory diseases	24,010	14.76	437.94	18.26
Lung diseases owing to external agents	193	0.12	5.29	0.22
Other respiratory diseases principally affecting the interstitium	4751	2.92	118.56	4.94
Suppurative and necrotic conditions of lower respiratory tract	1433	0.88	56.13	2.34
Other diseases of pleura	3808	2.34	89.92	3.75
Other diseases of the respiratory system	9866	6.07	284.14	11.85
Total	162,621	100.00	2398.46	100.00

Male individuals comprised the majority (54.34%) of influenza and pneumonia patients, with their associated expenses constituting 58.51% of the total. The percentage of cases involving children aged 0–14 and seniors aged 60 and older was as high as 82.29%, with their expenses representing 77.00% of the total. Figure 1 depicts the characteristics of influenza cases admitted and the associated costs.

Fig. 1.

Number and cost of cases of influenza and pneumonia

A temporal decomposition analysis of hospitalization expenses for influenza and pneumonia cases in Changchun spanning from 2017 to 2020 demonstrated a fluctuating and generally upward trend, with the exception of a notable decline in expenses attributable to the 2020 epidemic. Additionally, hospitalization expenses for influenza and pneumonia exhibit seasonal variations, characterized by higher peaks during winter months (December to March) and lower troughs during summer months (June to August). Figure 2 illustrates these trends.

Fig. 2.

Time series decomposition of costs for influenza and pneumonia

Air pollutants and temperatures from 2017 to 2020

For the period 2017–2020, the average temperature in Changchun was 7.20 °C, with a maximum daily average temperature of 29.90 °C and a minimum of − 26.10 °C, yielding a temperature range of 56.00 °C. Regarding air quality, concentrations of the six primary air pollutants exhibited a downward trend from 2017 to 2020. Specifically, compared with 2017, concentrations of PM_2.5, PM₁₀, SO₂, NO₂, CO, and O3 in 2020 declined by 8.67, 24.54, 62.47, 21.01, 39.83, and 11.25%, respectively. Furthermore, compared with the national average, the average temperature in Changchun is notably lower. In terms of concentrations of major air pollutants, the average concentrations of PM_2.5 and PM₁₀ in Changchun exceeded the national average in 2017, 2019, and 2020. Specifically, the average concentration of PM_2.5 in Changchun in 2020 was 1.27 times higher than the national average. Additionally, the average concentrations of SO₂ in Changchun surpassed the national average in both 2017 and 2018. Specifically, the average concentration of SO₂ in Changchun in 2017 was 1.47 times greater than the national average. Similarly, the average concentrations of NO₂ in Changchun exceeded the national average from 2017 to 2020. The average concentration of NO₂ in Changchun in 2020 was 1.32 times higher than the national average. Conversely, the average concentrations of CO and O₃ in Changchun from 2017 to 2020 were below the national average. In summary, compared with the national average, lower temperatures represent a significant meteorological feature of Changchun. Furthermore, PM_2.5, PM₁₀, SO₂, and NO₂ are identified as the primary air pollutants in Changchun. Table 2 shows the detailed data.

Table 2.

Temperature and concentrations of major air pollutants in Changchun, 2017–2020

Variables	2017		2018		2019		2020
	Changchun	China	Changchun	China	Changchun	China	Changchun	China
Average temperature (°C)	7.05	10.39	6.92	10.09	7.65	10.34	7.18	10.25
PM2.5 (µg/m3)	46.05	43	32.23	39	37.85	36	42.06	33
PM10 (µg/m3)	81.30	75	61.19	71	65.76	63	61.35	56
SO2 (µg/m3)	26.38	18	14.63	14	10.92	11	9.90	10
NO2 (µg/m3)	40.17	31	32.43	29	33.60	27	31.73	24
CO (mg/m3)	1.18	1.7	0.77	1.5	0.74	1.4	0.71	1.3
O3 (µg/m3)	90.13	149	75.22	151	80.38	148	79.99	138

We conduct a time series decomposition analysis of the four primary air pollutants in Changchun, spanning 2017–2020. Concentrations of PM_2.5, PM₁₀, SO₂, and NO₂ demonstrate an overall decreasing trend, accompanied by distinct seasonal fluctuations. Elevated concentrations of pollutants, reaching peaks, are noted during the winter months (December to February of the subsequent year), whereas reduced concentrations, reaching troughs, are observed during the summer months (June to August), as shown in Fig. 3.

Fig. 3.

Time series decomposition of key air pollutants in Changchun

The Kolmogorov–Smirnov test results indicate that the daily average concentrations of air pollutants and temperature do not conform to normal distribution. Consequently, we use Spearman’s rank correlation coefficient to assess the correlation between key air pollutants and ambient temperature. The statistical analysis reveals that the correlations between air pollutants and ambient temperature are significant at the P < 0.05 level. Specifically, a positive correlation is observed among the major air pollutants, with correlation coefficients spanning from 0.57 to 0.87. Furthermore, the correlation coefficients between PM_2.5, PM₁₀, SO₂, and NO₂ and ambient temperature are − 0.51, − 0.32, − 0.78, and − 0.31, respectively, as shown in Fig. 4.

Fig. 4.

Correlations between meteorological factors and air pollutants

Associated effects of air pollutants on costs for influenza and pneumonia

GAM elucidates the relationships between key air pollutants and hospitalization costs for influenza- and pneumonia-related conditions. Specifically, when the concentration of PM_2.5 increases by 10 µg/m³, there is a statistically significant immediate increase in hospitalization costs of 0.89% (95% CI: 0.66%, 1.11%) on the same day, with significance persisting into the first lag day (0.37%; 95% CI: 0.15, 0.60%). Analogously, a 10 µg/m³ increase in PM₁₀ concentration results in a 0.59% (95% CI: 0.43, 0.75%) increase in same-day hospitalization costs, although the effect of PM_2.5 is more pronounced. Regarding SO₂, each 10 µg/m³ increase in concentration is associated with a 3.02% (95% CI: 2.03%, 4.02%) increase in same-day hospitalization costs, maintaining statistical significance throughout, despite a gradual decline over a 14-day lag period. Similarly, a 10 µg/m³ increase in NO₂ concentration results in a significant increase in same-day hospitalization costs of 3.10% (95% CI: 2.52%, 3.69%), with an additional peak observed at a 9-day lag (1.23%; 95% CI: 0.69%, 1.77%). Notably, the effect of SO₂ and NO₂ on hospitalization costs surpasses that of particulate matter (PM_2.5 and PM₁₀). See Table 3; Fig. 5 for details.

Table 3.

Association between changes in air pollutant concentrations and costs for influenza and pneumonia. [ER (95% CI)]

Lag days	PM2.5	PM10	SO2	NO2
lag0	0.89 (0.66, 1.11)	0.59 (0.43, 0.75)	3.02 (2.03, 4.02)	3.10 (2.52, 3.69)
lag1	0.37 (0.15, 0.60)	0.14 (− 0.02, 0.30)	2.89 (1.87, 3.91)	2.06 (1.48, 2.64)
lag2	0.01 (− 0.20, 0.23)	−0.02 (− 0.17, 0.13)	0.26 (− 0.74, 1.27)	0.95 (0.40, 1.51)
lag3	0.00 (− 0.21, 0.22)	0.13 (− 0.02, 0.28)	−0.66 (− 1.64, 0.34)	0.46 (− 0.08, 1.00)
lag4	−0.14 (− 0.36, 0.07)	−0.06 (− 0.21, 0.09)	−0.42 (− 1.42, 0.59)	0.77 (0.24, 1.31)
lag5	−0.03 (− 0.24, 0.19)	−0.05 (− 0.21, 0.10)	0.77 (− 0.21, 1.76)	0.67 (0.14, 1.21)
lag6	−0.46 (− 0.68, − 0.24)	−0.05 (− 0.20, 0.10)	−0.04 (− 1.05, 0.97)	−0.35 (− 0.88, 0.18)
lag7	−0.12 (− 0.34, 0.10)	−0.05 (− 0.21, 0.11)	−0.13 (− 1.15, 0.90)	−0.11 (− 0.64, 0.42)
lag8	−0.23 (− 0.44, − 0.01)	−0.21 (− 0.37, − 0.05)	−0.27 (− 1.27, 0.75)	0.93 (0.40, 1.47)
lag9	−0.10 (− 0.32, 0.11)	0.09 (− 0.06, 0.24)	−0.15 (− 1.14, 0.85)	1.23 (0.69, 1.77)
lag10	−0.25 (− 0.47, − 0.03)	−0.19 (− 0.34, − 0.03)	1.07 (0.09, 2.07)	1.18 (0.64, 1.72)
lag11	−0.36 (− 0.58, − 0.14)	−0.08 (− 0.23, 0.07)	0.14 (− 0.87, 1.16)	0.77 (0.24, 1.31)
lag12	−0.45 (− 0.67, − 0.23)	−0.16 (− 0.32, − 0.01)	−1.03 (− 2.03, − 0.03)	−0.08 (− 0.61, 0.45)
lag13	−0.13 (− 0.35, 0.09)	0.03 (− 0.12, 0.19)	−1.00 (− 20.00, 0.00)	0.12 (− 0.41, 0.66)
lag14	−0.05 (− 0.27, 0.17)	−0.17 (− 0.33, 0.00)	0.17 (− 0.85, 1.19)	0.15 (− 0.39, 0.68)

Fig. 5.

Association between changes in air pollutant concentrations and costs for influenza and pneumonia

The cumulative lag effect results demonstrate that, when PM_2.5 and PM₁₀ concentrations increase by 10 µg/m³, the hospitalization costs associated with influenza and pneumonia show an increase of − 0.06% (95% CI: −0.43%, 0.31%) and 0.05% (95% CI: −0.26%, 0.36%), respectively, during a lag period of 0–14 days. Neither of these increases reaches statistical significance. However, the peak cumulative lag effects are observed on the day of the increase in pollutant concentrations, resulting in statistically significant cumulative increases in hospitalization costs of 0.89% (95% CI: 0.66%, 1.11%) and 0.59% (95% CI: 0.43%, 0.75%), respectively. However, the time points exhibiting the largest cumulative lag effects were both within a lag period of 0–1 days, accompanied by cumulative increases in hospitalization costs of 3.78% (95% CI: 2.64%, 4.94%) and 3.29% (95% CI: 2.63%, 3.95%), respectively. See Table 4; Fig. 6 for further details.

Table 4.

Cumulative lag effect of changes in air pollutant concentrations on costs for influenza and pneumonia. [ER (95% CI)]

Air pollutants	Cumulative lag effect within 14 days	Maximum cumulative lag effect
		Lag days	Cumulative lag effect
PM2.5	−0.06 (−0.43, 0.31)	0	0.89 (0.66, 1.11)
PM10	0.05 (−0.26, 0.36)	0	0.59 (0.43, 0.75)
SO2	2.31 (0.79, 3.84)	1	3.78 (2.64, 4.94)
NO2	2.68 (1.72, 3.64)	1	3.29 (2.63, 3.95)

Fig. 6.

Cumulative lag effect of changes in air pollutant concentrations on costs for influenza and pneumonia. The shaded areas represent the 95% confidence intervals (95% CI)

We use statistical models to investigate the effects of various air pollutants on hospitalization expenses associated with influenza and pneumonia among individuals of different ages and genders. The findings indicate that, across different age groups, an increase of 10 µg/m³ in concentrations of PM_2.5, PM₁₀, SO₂, or NO₂ led to the most pronounced increases in hospitalization expenses among adults aged 15–59. During the maximum cumulative lag interval, the peak cumulative increases in hospitalization costs amounted to 1.86% (95% CI: 1.43%, 2.30%) for PM_2.5, 1.54% (95% CI: 1.09%, 2.00%) for PM₁₀, 10.46% (95% CI: 7.32%, 13.69%) for SO₂, and 5.94% (95% CI: 4.60%, 7.29%) for NO₂. According to gender, during the same interval, an increase of 10 µg/m³ in concentrations of these pollutants resulted in more pronounced cumulative lag effects on hospitalization costs for influenza among females than males. For females, cumulative increases in hospitalization costs amounted to 1.74% (95% CI: 1.36%, 2.12%) for PM_2.5, 1.44% (95% CI: 1.11%, 1.77%) for PM₁₀, 14.95% (95% CI: 12.52%, 17.43%) for SO₂, and 7.04% (95% CI: 5.76%, 8.33%) for NO₂. By contrast, for males, the cumulative increases were 0.92% (95% CI: 0.66%, 1.18%) for PM_2.5, 0.79% (95% CI: 0.60%, 0.98%) for PM₁₀, 10.56% (95% CI: 8.57%, 12.59%) for SO₂, and 4.44% (95% CI: 3.63%, 5.25%) for NO₂. All observed increases are statistically significant. See Tables 1, 2 and 3 for details.

Association between temperature and costs for respiratory diseases

The P_2.5 and P_97.5 percentiles of the temperature series in Changchun are − 17.91 °C and 26.81 °C, respectively. We define temperatures below − 17.91 °C as low temperatures and those above 26.81 °C as high temperatures. The results of DLNM, which analyses the association between temperature variations and hospitalization costs for respiratory diseases in Changchun, indicate that, after controlling for long-term trends, holiday trends, and pollutant factors, there is a significant association between temperature changes and hospitalization costs for respiratory diseases. However, the associations between high and low temperatures and hospitalization costs for various types of respiratory diseases exhibit variability.

A 3-D plot illustrating the exposure–response relationship between temperature and hospitalization costs for influenza and pneumonia demonstrates that, within a 14-day lag period, compared with the average temperature (7.2 °C), both low and high temperatures on the day of occurrence lead to increased hospitalization costs for influenza and pneumonia, with peak RR observed. Notably, the RR for hospitalization costs associated with low temperatures is higher than that associated with high temperatures. Over the 14-day lag period, the RR of hospitalization costs gradually diminishes over time. Furthermore, during the lag period following low temperatures, the decrease in the RR of hospitalization costs is more marked (Fig. 7).

Fig. 7.

Exposure–response relationship between costs for influenza and pneumonia and temperature

The DLNM results show that on the day of low-temperature occurrence, the hospitalization costs associated with influenza and pneumonia rise by 6.01% (95% CI: 2.02%, 10.16%). During the first 6 days of the lag period, ER exhibits a gradual decline and remains statistically significant. On day 6 of the lag, hospitalization costs due to influenza and pneumonia increase by 1.92% (95% CI: 0.16%, 3.72%). On the day of high-temperature occurrence, hospitalization costs due to influenza and pneumonia increase by 3.88%, with a 95% confidence interval ranging from − 0.61% to 8.57%. Over the 0–14 day lag period, ER undergoes a gradual decrease, although this decline is not statistically significant.

The analysis of cumulative lagged effects, examining the influence of high and low temperatures on hospitalization costs associated with influenza and pneumonia, shows a steady augmentation in the cumulative lagged effects of influenza incidence and costs during and subsequent to periods of low temperature, within a 14-day lag framework. The peak cumulative lagged effect is noted between lag 0 and 8 days, with hospitalization costs showing a cumulative increase of 33.55% (95% CI: 8.73%, 64.03%) during this timeframe. Throughout periods of high temperature and the subsequent lags, the cumulative lagged effect on hospitalization costs exhibits an initial increase followed by a decrease. Over a 14-day period, the cumulative change in hospitalization costs associated with influenza and pneumonia amounts to a decrease of 12.11% (95% CI: −32.72%, 14.81%). Nevertheless, this decrease is not statistically significant. See Table 5 for the detailed data.

Table 5.

Association and cumulative effects of high and low temperatures on costs for influenza and pneumonia at different lag days [ER (95% CI)]

Lag days	Association effect		Cumulative effect
	<−17.91 °C	>26.81 °C	<−17.91 °C	>26.81 °C
lag0	6.01 (2.02, 10.16)	3.88 (− 0.61, 8.57)	6.01 (2.02, 10.16)	3.88 (− 0.61, 8.57)
lag1	5.32 (1.78, 8.98)	3.19 (− 0.77, 7.31)	11.65 (3.84, 20.05)	7.19 (− 1.38, 16.50)
lag2	4.63 (1.52, 7.84)	2.50 (− 0.95, 6.08)	16.82 (5.43, 29.45)	9.87 (− 2.31, 23.56)
lag3	3.95 (1.24, 6.72)	1.82 (− 1.15, 4.88)	21.43 (6.77, 38.11)	11.87 (− 3.4, 29.55)
lag4	3.27 (0.93, 5.65)	1.15 (− 1.39, 3.75)	25.40 (7.83, 45.82)	13.15 (− 4.67, 34.3)
lag5	2.59 (0.58, 4.64)	0.47 (− 1.69, 2.68)	28.65 (8.60, 52.40)	13.69 (− 6.13, 37.68)
lag6	1.92 (0.16, 3.72)	−0.19 (− 2.08, 1.73)	31.12 (9.03, 57.68)	13.47 (− 7.79, 39.63)
lag7	1.25 (− 0.37, 2.91)	−0.86 (− 2.61, 0.92)	32.76 (9.09, 61.58)	12.50 (− 9.69, 40.13)
lag8	0.59 (− 1.02, 2.23)	−1.52 (− 3.30, 0.30)	33.55 (8.73, 64.03)	10.79 (− 11.84, 39.23)
lag9	−0.07 (− 1.79, 1.69)	−2.17 (− 4.13, − 0.17)	33.46 (7.92, 65.05)	8.39 (− 14.29, 37.07)
lag10	−0.72 (− 2.66, 1.26)	−2.82 (− 5.07, − 0.51)	32.50 (6.58, 64.73)	5.33 (− 17.09, 33.81)
lag11	−1.37 (− 3.59, 0.91)	−3.46 (− 6.08, − 0.78)	30.69 (4.64, 63.21)	1.68 (− 20.27, 29.68)
lag12	−2.01 (− 4.56, 0.60)	−4.11 (− 7.12, − 1.00)	28.06 (2.04, 60.70)	−2.49 (− 23.91, 24.96)
lag13	−2.65 (− 5.54, 0.32)	−4.74 (− 8.17, − 1.19)	24.66 (− 1.29, 57.43)	−7.12 (− 28.05, 19.91)
lag14	−3.29 (− 6.54, 0.07)	−5.38 (− 9.23, − 1.35)	20.56 (− 5.41, 53.66)	−12.11 (− 32.72, 14.81)

We find that low-temperature conditions have a significant effect on hospitalization costs across different age and gender groups. Specifically, during the occurrence of low temperatures, hospitalization costs for children aged 0–14 significantly increase by 9.31% (95% CI: 4.72%, 14.11%), and those for adults aged 15–59 also significantly increased by 8.59% (95% CI: 0.48%, 17.36%); both increases are statistically significant. By contrast, although hospitalization costs for seniors aged 60 and older increase by 4.35% (95% CI: −2.58%, 11.78%), this increase is not statistically significant. However, during the 5–9-day lag period following low temperatures, hospitalization costs for seniors aged 60 and older increase by 3.42% (95% CI: 0.45%, 6.47%) on lag day 5 and by 2.67% (95% CI: 0.49%, 4.90%) on lag day 9. These increases at both time points are statistically significant.

For different gender groups, the effect of low-temperature conditions also exhibits significant differences. During the occurrence of low temperatures and the subsequent 1–3-day lag period, hospitalization costs for males significantly increase, with an increase of 6.41% (95% CI: 1.12%, 11.97%) on the same day and an increase of 3.88% (95% CI: 0.31%, 7.58%) on lag day 3. For females, the increase in hospitalization costs over the entire lag period (during the occurrence of low temperatures and the subsequent 1–10-day lag period) shows a decreasing trend over time, but all increases are statistically significant. Specifically, during the occurrence of low temperatures, hospitalization costs for females increase by 7.90% (95% CI: 2.76%, 13.29%). Even on lag day 10, the increase remains at 2.51% (95% CI: 0.19%, 4.88%).

We also examine the cumulative delayed effects of low-temperature conditions on hospitalization costs for influenza and pneumonia across different age groups. The results show that the maximum cumulative delayed effects for children aged 0–14 and adults aged 15–59 occur during the 0–6-day and 0–7-day lag periods, respectively. Specifically, hospitalization costs for children aged 0–14 cumulatively increase by 41.06% (95% CI: 16.3%, 71.09%) during the 0–6-day lag period. For adults aged 15–59, there is a significant trend of increasing hospitalization costs during the 0–7-day lag period, particularly within the first 5 days of the lag period, with an increase of 38.91% (95% CI: 0.27%, 92.44%). For elderly seniors aged 60 and above, the most notable delayed effect is observed during the 9–14-day lag period, with a cumulative increase in hospitalization costs of 56.79% (95% CI: 18.33%, 107.74%) during the 0–14-day lag period.

By contrast, the effect of high-temperature conditions on hospitalization costs is relatively small. On the day of high temperatures and during the subsequent delayed periods, the increases in hospitalization costs across different age and gender groups are either not statistically significant or exhibit negative values. For different gender groups, within the 0–7-day delayed period following low temperatures, the cumulative increase in hospitalization costs for males is 31.36% (95% CI: 1.19%, 70.52%). For females, within the 0–14-day delayed period, the cumulative increase in hospitalization costs is 82.62% (95% CI: 40.85%, 136.77%). Both increases during these time periods are statistically significant. However, during and after high-temperature periods, the cumulative delayed effects on hospitalization costs for both males and females are either not significant or exhibit negative values. See appendix Tables 4–7 for details.

Sensitivity analysis

To assess robustness, we establish GAMs and DLNMs for key air pollutants, temperature, and hospitalization costs by adjusting the degrees of freedom for smoothing terms over the maximum lag period of the effects of air pollutants and temperature on hospitalization costs. The results show that the changes in hospitalization costs for influenza and pneumonia and their 95% confidence intervals are relatively small. In addition, we establish two pollutant models to exclude the influence of other pollutants. The results also show insignificant changes in hospitalization costs for influenza and pneumonia and their 95% confidence intervals, indicating high robustness. See appendix Tables 8 and 9 for details.

Discussion

Adopting a health economics perspective, this study explores the specific effects of environmental exposure factors (including air pollutant concentrations and temperature variations) on hospitalization costs for patients with influenza and pneumonia. Besides, we evaluate high-risk populations vulnerable to environmental exposure, providing a scientific basis and reference for formulating prevention and control strategies, optimizing resource allocation, and improving population health.

Among respiratory diseases affecting patients, influenza and pneumonia comprised the highest proportion of both case numbers and treatment costs among all respiratory disease cases, with this proportion reaching nearly 50% for both. Thus, influenza and pneumonia are the respiratory diseases with the highest patient volume and the greatest consumption of healthcare resources. Previous research has demonstrated that the annual number of influenza cases ranged from 84 million to 144 million between 2011 and 2019 [50], with influenza-associated all-cause excess mortality rates of 6.9 to 17.2 per 100,000 people, which translates to approximately 96,000 to 240,000 all-cause excess deaths attributable to influenza each year [51]. This severe burden has seriously affected the public health situation and has led to a large number of outpatient and inpatient cases. Hence, implementing robust preventive measures against influenza and raising awareness regarding the disease are instrumental for improving public health outcomes.

In terms of the demographic characteristics pertaining to influenza and pneumonia cases, the number of male patients exceeded that of females, comprising 58.51% of the total. This observed gender disparity, consistent with other studies [52, 53], may be attributed to a combination of behavioural, environmental, and biological factors. In the adult population, potential explanations include historically higher smoking rates among males [54, 55], as well as gender-based differences in occupational exposures [56] (e.g., a greater proportion of men working in outdoor or industrial settings with higher air pollution exposure) and healthcare-seeking behaviours. It is important to note that these are population-level associations, and individual-level data on these factors were not available in the present study. Therefore, the effective control of smoking among males, coupled with comprehensive occupational health education, is instrumental for decreasing the incidence of influenza and pneumonia among this demographic, thereby enhancing their overall health status.

Regarding age distribution, children aged 0–14 accounted for more than 50% of influenza and pneumonia cases, a finding that underscores their heightened vulnerability. Biological factors provide the most plausible explanation for this predominance in the paediatric group. Children exhibit lower nasal filtration efficiency for larger particles and demonstrate higher microflow conditions compared with adults. Furthermore, children’s immune systems and lungs are in a stage of incomplete maturation. Consequently, children are more vulnerable to the adverse effects of air pollution [57]. The slight male predominance within the paediatric cases is a common epidemiological finding for many childhood respiratory illnesses, which may be linked to immature immunological or physiological differences between boys and girls, rather than behavioural risk factors like smoking. In addition to children, seniors aged 60 and older also show a substantially high number of cases and associated costs related to influenza and pneumonia. This could be attributable to the decline in physical function and underlying medical conditions among the elderly, rendering them more susceptible to air pollution and temperature changes, ultimately resulting in a higher incidence of influenza and pneumonia. Therefore, children and the elderly constitute vulnerable populations with regard to respiratory disease. Minimizing their exposure to adverse weather conditions, ensuring the adoption of proper personal protective measures, and enhancing health management and health education are essential for reducing the incidence of influenza and pneumonia in this vulnerable group.

PM_2.5 and PM₁₀ are defined as particulate matter with aerodynamic diameters of 2.5 micrometres (µm) or less and 10 micrometres (µm) or less, respectively. Previous studies have suggested that particulate matter between 2.5 μm and 10 μm, when inhaled, can potentially enter the lower respiratory tract and affect gas exchange and other pulmonary functions [58]. Additionally, these particles can carry various pathogenic microorganisms, such as bacteria, viruses, and fungi, which can lead to the development of multiple respiratory diseases. However, while evidence on the association between particulate matter and influenza incidence is growing [59], evidence specifically linking particulate matter to the economic costs of influenza remains limited. Our study shows that when concentrations of PM_2.5 and PM₁₀ increase by 10 micrograms per cubic metre (µg/m³), within the period of the maximum cumulative lag effect, they result in cumulative increases in hospitalization costs of 0.89% (95% CI: 0.66%, 1.11%) and 0.59% (95% CI: 0.43%, 0.75%), respectively. A study of Wuhan using the cost-of-illness method also showed that if concentrations of PM2.5 and PM10 can meet the guideline values recommended by the World Health Organization—namely, an annual average concentration of 10 micrograms per cubic metre (µg/m3) for PM2.5 and 20 µg/m3 for PM10—then economic losses of 194 million yuan could potentially be mitigated annually owing to the reduction in respiratory diseases [62].Therefore, efforts should be made to reduce the generation of particulate matter and increase residents’ awareness of personal protective measures to decrease the incidence of respiratory diseases and economic losses.

Compared with particulate matter (PM_2.5 and PM₁₀), sulphur dioxide (SO₂) and nitrogen dioxide (NO₂) exert greater effects on human health and impose substantial medical burdens. Studies have demonstrated that SO₂ has high water solubility and readily reacts with water in the respiratory tract to produce sulphurous acid. Sulphurous acid is inherently unstable and subsequently undergoes dissociation to form bisulphite and sulphite derivatives. SO₂ and its derivatives can cause oxidative damage, inflammation, and apoptosis [60] in the lungs. Consequently, exposure to SO₂ might induce mitochondrial dysfunction, thereby leading to cellular dysfunction and ultimately precipitating diseases such as tracheitis, bronchial asthma, and emphysema [61, 62]. Furthermore, elevated concentrations of SO₂ increase the likelihood of additional respiratory damage or the exacerbation of respiratory conditions among the populace, thereby contributing to increases in treatment costs. As for No₂, prior studies have indicated that short-term personal exposure to NO₂ might elicit respiratory inflammation, oxidative stress, and airway hyperresponsiveness. It can also impair the function of alveolar macrophages and epithelial cells, thereby increasing the risk of lung infection [63, 64]. Our findings demonstrate that with an increment of 10 micrograms per cubic metre (µg/m³) in the concentration of SO₂ or NO₂, the peak 1-day surge in hospitalization costs for influenza and pneumonia amounts to 3.02 and 3.10%, respectively. Furthermore, during a latency period of 14 days, the highest cumulative augmentation in hospitalization costs for influenza and pneumonia reaches 3.78 and 3.29%, respectively. Hence, there is a robust correlation between variations in SO₂ and NO₂ concentrations and elevations in hospitalization costs for influenza and pneumonia. Cheng [65] found that, for the period 2014–2020, reducing SO₂ concentration to the level stipulated by the WHO’s 2021 Global Air Quality Guidelines—namely, 40 micrograms per cubic metre (µg/m³)—could have resulted in savings of 2.7743 million yuan in hospitalization costs.Therefore, efforts should be undertaken to reduce the production and emission of SO₂ and NO₂, thereby minimizing their effects on incidence rates, hospital admissions, and hospitalization costs related to respiratory diseases.

Our findings suggest that exposure to low temperatures (below − 17.91 °C) exhibits a significant correlation with elevated hospitalization costs for influenza and pneumonia, with the peak single-day increase in hospitalization costs reaching 6.01%. Research conducted in Dongguan revealed that during episodes of low temperatures, the incidence rate of respiratory diseases rose by 12% [3]. Moreover, time series decomposition revealed noticeable seasonal variations in hospitalization costs for both influenza and pneumonia, alongside concentrations of air pollutants. In particular, elevated hospitalization costs and air pollutant concentrations are observed during the winter months (December to February) with prominent peaks while troughs manifest during summer and autumn. Moreover, Changchun is situated in Northeast China, a region known for its cold climate. During winter (November to March), temperatures are low, and the period for coal-fired heating is relatively long, leading to high emissions of pollutants [66, 67]. The cold temperatures hinder the dispersion of pollutants, resulting in severe air pollution in winter. Conversely, winter’s low temperatures prolong the activity of influenza viruses in cold environments, thereby facilitating their survival and transmission and increasing the risk of influenza and pneumonia among residents. Hence, we posit that low temperatures during winter are significantly correlated with hospitalization costs owing to influenza and pneumonia. Therefore, enhancing the monitoring and reporting of extreme temperatures, as well as increasing residents’ awareness of preventive measures, is crucial for mitigating the incidence and economic burden of influenza and pneumonia.

Our study assessed the individual effects of SO₂ and low temperature. It is important to note that these two exposures are strongly correlated seasonally in our study setting (as shown in Fig. 4), which presents a methodological challenge for disentangling their independent and potential interactive effects within a single model. While the question of whether their effects are additive or synergistic is scientifically important, our primary aim was to first establish and quantify their distinct associations with hospitalization costs using robust, standard methods (GAM and DLNM) that control for confounding by time and other meteorological factors. Future studies specifically designed to investigate multi-exposure interactions, potentially employing case-crossover designs or more complex multi-pollutant models, would be valuable to elucidate the potential synergistic effects between specific air pollutants like SO₂ and meteorological stressors like extreme cold.

Limitations

This study has several limitations. First, the findings are based on data from seven tertiary hospitals, which may limit their generalizability to primary or secondary care settings. Second, we could not distinguish between viral and bacterial pneumonia, which might have different environmental triggers. Third, the inclusion of the year 2020, a period affected by COVID-19 lockdowns, may have led to an underestimation of influenza and pneumonia cases and distorted the associations between environmental factors and hospitalization costs due to altered air quality and human behaviour [68].

Conclusion

This study quantifies the impact of environmental exposure on hospitalization costs for influenza and pneumonia in a cold industrial city. Our results reveal several key findings that inform specific public health interventions: First, prioritize control of SO₂ and NO₂ emissions, as these pollutants had the most pronounced effect, with a 10 µg/m³ increase leading to immediate cost rises of 3.02% and 3.10%, respectively. This underscores the need for stringent industrial desulfurization and clean transportation policies. Second, implement public health measures against extreme cold, given that temperatures below − 17.91 °C were associated with a 6.01% same-day spike in costs. Early cold-wave warning systems and protection strategies for vulnerable populations are essential. Third, although particulate matter (PM₂.₅ and PM₁₀) showed relatively smaller effect sizes, their pervasive presence necessitates action. The observed significant immediate cost increases suggest that continued efforts in controlling particulate emissions from coal combustion and industrial processes remain crucial. Finally, focus healthcare resources on high-risk demographics—children (0–14 years), the elderly (≥ 60 years), and males—who accounted for the majority of cases and costs, enabling precise targeting of vaccination campaigns and health education. In summary, this study offers a scientific rationale and reference for developing effective prevention and control measures, optimizing resource allocation, and advancing population health levels.

Authors’ contributions

HJ: Funding acquisition, Writing-original draft, Methodology. SX and XS: Methodology, Writing-review& editing. SG and YZ: Data curation, Writing-review & editing. CM and XY: Supervision, Writing-review & editing. All authors read and approved the final manuscript.

Funding

This work was supported by Humanities and Social Science Fund of Ministry of Education of China [grant number 24YJCZH111]. The funding agencies had no role in design, analysis, interpretation, or writing of this study.

Data availability

The datasets generated and/or analysed during the current study are not publicly available due some of the data relate to internal hospital operations but are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.