Joint effects of PM_2.5 pollution and cold waves exposure on population health risks in Shijiazhuang, China

Abstract

Against the backdrop of global climate change, joint exposure events involving cold waves and PM_2.5 pollution pose a serious threat to public health. However, research on the joint effects of these factors on mortality risk among residents, particularly in North China, remains limited. This study employed a “joint exposure” analytical framework, utilizing mortality data, meteorological records, and PM_2.5 concentration data from Shijiazhuang City between 2017 and 2024. Events where cold waves and PM_2.5 pollution occurred simultaneously were defined as joint exposures. Twelve exposure combinations were constructed using three cold waves thresholds (T1–T3) and four PM_2.5 pollution thresholds (P1–P4). A distributed lag nonlinear model assessed their risks for non-accidental mortality, circulatory disease mortality, and respiratory disease mortality. The findings showed that joint exposure to cold waves and PM_2.5 was significantly associated with increased mortality risks. The relative risks for non-accidental mortality and circulatory/respiratory disease mortality peaked at the P4T1 combination, reaching 2.55 (95% CI: 1.82, 3.57), 2.62 (95% CI: 1.76, 3.92), and 2.34 (95% CI: 1.18, 4.65), respectively. Interaction analysis revealed a significant synergistic effect, with a synergy index of 1.54 (95% CI: 1.05, 2.03) for the P3T2 combination, indicating significant super-additivity. Attribution analysis indicated a substantial disease burden attributable to the joint events, with the attributable fraction for non-accidental mortality reaching 60.8% under the P4T1 combination, corresponding to 883 attributable deaths. Subgroup analysis revealed greater vulnerability among the elderly (≥ 75 years), males, and individuals with lower education levels. Joint exposure to cold waves and PM_2.5 pollution was significantly associated with elevated mortality risk among Shijiazhuang residents. It was recommended that such joint environmental events be incorporated into public health early warning systems to implement targeted health protection measures.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-55632-4.

Introduction

As global climate change intensifies, the frequency of extreme weather events has increased significantly, posing a threat to public health¹; IPCC². Current research predominantly focuses on extreme heat, while the understanding of cold waves health risks remains insufficient, particularly in developing countries³. However, multiple epidemiological studies confirm that cold-related deaths outnumber heat-related deaths in most nations⁴. A global time-series study⁵ indicates that between 2000 and 2019, 9.43% of all annual deaths worldwide were associated with non-optimal temperatures. Of these, 8.52% were attributable to cold exposure, while heat exposure accounted for a lower proportion (0.91%). Another study revealed that 14.3% of non-accidental deaths between 2013 and 2015 were potentially related to non-optimal temperatures, with 11.6% attributed to cold exposure and 2.7% to heat exposure⁶.

Due to varying adaptive capacities among populations across different climate zones, no unified definition of cold waves currently exists³. Existing definitions primarily rely on absolute/relative temperature thresholds, duration, or the magnitude of temperature drops. Chen compared the impacts of different cold wave definitions, proposing that China’s optimal fitting method combines varying cold thresholds with duration³. Furthermore, epidemiological evidence indicates cold waves increase all-cause mortality and the incidence of respiratory and circulatory diseases, with higher risks in mid-to-high latitudes than low latitudes⁷.

Since the 1990s, the global average temperature rise has stalled, entering the so-called “global warming hiatus”. During this period, cold waves events over land areas increased due to the atmospheric circulation pattern of Arctic warming and continental cooling⁸. In 2010, the average cold waves frequency in South Korea and Japan increased by 51.44%. Future cold waves frequency in mid-latitude regions is unlikely to decrease⁵. Cold wave–related excess mortality is concentrated in Asia, particularly South and East Asia⁵. Since the 21st century, accelerated warming in the Arctic and mid-to-low latitudes has weakened the polar jet stream while strengthening the subtropical jet stream. This has caused cold air to accumulate and intensify over northern East Asia, exacerbating cold waves in the region⁹. Concurrently, the 2021 Global Burden of Disease study identified airborne PM pollution as the world’s most significant health threat, surpassing 87 other disease burden risk factors¹⁰, particularly in developing and middle-income countries^11,12. In China, persistent haze-related pollution has affected major cities like Beijing, Taiyuan, Shijiazhuang, and Jinan since 2007, with over 70% of the world’s ten most polluted cities concentrated in this region University¹³. Following the implementation of China’s Action Plan for Air Pollution Prevention and Control (2013–2017), the Blue Sky Defense Campaign (2018–2020), and the Action Plan for Continuous Improvement of Air Quality (2022–2025), air quality has significantly improved¹⁴. However, as of 2019, PM_2.5 levels in most Chinese cities remained 3–5 times higher than the World Health Organization (WHO) reference value (5 µg/m³)¹⁵. In 2020, 21 Chinese cities, including Shijiazhuang, Anyang, and Taiyuan, had poor air quality¹⁶, constituting a persistent public health burden.

More complexly, cold waves and PM_2.5 pollution often occur simultaneously during winter, forming “joint pollution-cold events”. The influence of meteorological conditions on pollutant concentrations means these two exposures are not independent. While recent studies have explored the interactive effects of temperature and PM_2.5 on health, conclusions remain inconsistent. Multiple investigations indicate that co-exposure to cold and PM_2.5 may heighten health risks, including cardiovascular disease, respiratory disorders, and diabetes¹⁷. Research¹⁸ reveals a positive synergistic effect between cold waves and pollutants like PM_2.5 on hospitalization risks for anxiety patients, with stronger interactions observed during prolonged cold waves. Huang found such joint events increase health risks for all-cause, respiratory, and circulatory mortality, leading to synergistic excess mortality risks¹⁹. Hu examined the effects of temperature and PM_2.5 on myocardial infarction incidence and mortality in Beijing under different effect modification strategies. In China, most related studies have focused on Central China or ozone heatwaves, with no prior analysis of the joint impact of PM_2.5 and cold waves events on mortality in North China.

Therefore, to address this research gap, this study employs a “joint exposure” analytical framework. Rather than treating cold waves and PM_2.5 pollution as independent exposure factors to examine their interaction, it directly defines their simultaneous occurrence as a distinct exposure event, differing from single-exposure analyses²⁰. This study will systematically compare health risks under various combinations of cold waves and PM_2.5 pollution levels. It aims to assess the impact of these joint events on mortality risk among residents, providing new scientific evidence and decision support for understanding and preventing such joint environmental health risks.

Materials and methods

Study area

Shijiazhuang City (located between 37°27' and 38°47' north latitude and 113°30' and 115°20' east longitude) is situated in North China, in the central-southern part of Hebei Province, spanning the Taihang Mountains and the North China Plain. It features a warm temperate continental monsoon climate, with a permanent population of 11.2466 million and a total area of 15,848 square kilometers. The study period spans from 2017 to 2024.

Data sources and definitions

Mortality data is sourced from the cause-of-death surveillance system of the Shijiazhuang Center for Disease Control and Prevention, covering daily death records from 2017 to 2024. Classification follows the International Classification of Diseases, 10th Revision (ICD-10): Non-accidental Mortality (A00-R99); Circulatory disease Mortality (I00-I99); Respiratory disease Mortality (J00-J99)²¹.

Meteorological data were obtained from the Shijiazhuang Meteorological Bureau, including daily average temperature, maximum temperature, minimum temperature, and relative humidity. PM_2.5 pollutant data were sourced from the national-level monitoring stations of the Shijiazhuang Environmental Protection Monitoring Center. All daily records were complete, with no missing data for mortality, meteorological variables, or PM_2.5 concentrations during the study period.

Four PM_2.5 pollution thresholds (P1–P4) and three cold wave temperature thresholds (T1–T3) were defined (see Table 1). For each combination of a PM_2.5 threshold(Pi)and a cold wave threshold(Tj), we defined three mutually exclusive exposure variables at the daily level: PM_2.5 only, cold wave only, and joint event. A three-day sliding window was applied: an event was considered to occur if at least two of three consecutive days exceeded the corresponding threshold.Reference category: For each threshold combination(Pi, Tj), the reference group consisted of days with PM_2.5 below P_i and daily mean temperature above T_j. These were days free of both the defined PM_2.5 event and the defined cold wave event under that specific combination Exposure states were categorized into four classes: no event (0), PM_2.5 pollution only (1), cold waves only (2), and joint cold waves-PM_2.5 event (3)^22–26.

Table 1.

Definitions of PM_2.5 pollution thresholds (P1–P4) and cold wave thresholds (T1–T3).

Threshold	Definition	Criterion	Value
P1	PM2.5 pollution (Moderate)	AQI “Moderate Pollution” threshold	≥ 115 µg/m³
P2	PM2.5 pollution	92.5th percentile of cold-season PM2.5 concentration	≥ 131 µg/m³
P3	PM2.5 pollution (Heavy)	“Heavy Pollution” threshold	≥ 150 µg/m³
P4	PM₂.₅ pollution	97.5th percentile of cold-season PM2.5 concentration	≥ 188 µg/m³
T1	Cold wave	10th percentile of daily mean temperature	-0.4 °C
T2	Cold wave	7.5th percentile of daily mean temperature	-1.3 °C
T3	Cold wave	5th percentile of daily mean temperature	-2.3 °C

Statistical analysis

This study evaluates the lagged effects of cold wave events, PM_2.5 pollution events, and their joint events on mortality risk among residents. The analysis was confined to the cold season (November to March annually). A distributed lag nonlinear model based on quasi-Poisson regression was employed. Three mutually exclusive binary exposure indicators were defined at the daily level: PM_only (PM_2.5 pollution only), cold_only (cold wave only), and joint (joint event), with days having no event as the reference category. For each indicator, an independent cross-basis function was constructed to capture its exposure-specific lagged effects. All three cross-basis terms were entered simultaneously into a single quasi-Poisson model, sharing the same reference category. A 21-day lag period was set to capture the overall effect of joint events, with knots placed at lags 3, 7, and 14 days. The model is defined as follows:

  

Where  is the expected number of deaths on day t;  denotes the distributed lag cross-basis function; is a natural cubic spline;  is daily average relative humidity;  controls for long-term trend and seasonality (df = 4*6 per year);  and are day−of−week and holiday indicators, respectively, α is the intercept; γ and δ are the regression coefficients^24,27. The model reports single-day lagged relative risks (RR, 95% CI) for each exposure type and cumulative relative risks (CRR, 95% CI) over lag days 0–21.

To assess the additive interaction between cold waves and PM_2.5 pollution on mortality risk, the following three metrics were calculated using individuals unexposed to both cold waves and PM_2.5 pollution as the reference category. The statistical significance of additive interaction was determined using Wald tests (significance level α = 0.05)²⁸.

Relative Excess Risk due to Interaction (RERI):

Attributable Proportion due to Interaction (AP):

Synergy Index (S):

To quantify the public health impact of the joint events, calculate the attributable fraction (AF) and attributable deaths (AN)²⁹:

The Quasi-Akaike Information Criterion (Q-AIC) was used to evaluate model fit for 12 defined thresholds, with the definition yielding the smallest mean Q-AIC value designated as the optimal threshold definition. To identify susceptible populations, stratified analyses were conducted by age, gender, and educational attainment. Subgroup models were constructed, and the joint CRR was calculated for each subgroup.

To test model robustness, sensitivity analyses were performed by varying the degrees of freedom for relative humidity (4–6), lag days (14, 28), and long-term trend over time (4 × 4, 4 × 8). In addition, we re-estimated the cumulative relative risks (CRR) using a uniform reference category (PM_2.5 < 115 µg/m³ and temperature > − 0.4 °C). To address the mortality surge during the COVID-19 policy transition period (December 2022 – February 2023), we excluded this period and winsorized daily deaths at the 99th percentile.

Statistical analyses were performed using R 4.2.2 with the dlnm, zoo, splines, ggplot2, dplyr, tidyr, lubridate, stringr, and purrr packages.

Ethical statement

This study was approved by the Institutional Ethical Review Board of Shijiazhuang Center for Disease Control and Prevention(#2026-04).All methods were carried out in accordance with the ethical standards and relevant guidelines of the Institutional Ethical Review Board of the Shijiazhuang CDC, including the Declaration of Helsinki. The dataset included the age, gender, educational level, date of death, and International Statistical Classification of Diseases and Related Health Problems, 10th Revision diagnosis code of each deceased individual. No personal identifiers such as names, ID card numbers, or residential addresses were collected or used in the analysis.Therefore, the mortality data were all anonymized, and thus the requirement for informed consent was waived.

Results

A total of 1,210 observation days were included during the study period. As shown in Table 2, the total number of non-accidental mortality was 57,189, with an average of 47.26 deaths per day. The daily average deaths from circulatory diseases and respiratory diseases mortality were 26.23 and 6.81, respectively. Demographically, females constituted 42.56% of deaths, individuals aged 75 years and older accounted for 60.55%, and those with a junior high school education or lower comprised 76.02%. The average PM_2.5 concentration was 80.73 µg/m³, the average temperature was 4.25 °C, and the average relative humidity was 49.9%.

Table 2.

Descriptive Analysis of Mortality Data, Meteorological Factors and Environmental Factors.

Parameter	Counts	X ± SD	Min / Median / Max
Non-accidental Mortality	57,189	47.26 ± 23.86	18.00 / 42.00 / 232.00
Circulatory disease Mortality	31,738	26.23 ± 13.59	8.00 / 24.00 / 130.00
Respiratory disease Mortality	8246	6.81 ± 5.50	0.00 / 6.00 / 54.00
Age,0-64yrs	11,511	9.51 ± 5.97	0.00 / 8.00 / 56.00
Age,65-74yrs	11,153	9.22 ± 5.43	0.00 / 8.00 / 45.00
Age,75 + yrs	34,628	28.62 ± 15.24	9.00 / 25.00 / 137.00
Low, edu	43,473	35.93 ± 20.77	11.00 / 32.00 / 205.00
High, edu	13,716	11.34 ± 6.55	0.00 / 11.00 / 38.00
Male	32,849	27.15 ± 14.21	7.00 / 24.00 / 145.00
Female	24,340	20.12 ± 10.82	3.00 / 18.00 / 98.00
PM2.5 (µg/m3)	/	80.73 ± 59.82	9.57 / 63.60 / 444.57
Tmean	/	4.25 ± 5.86	-10.8 / 3.40 / 25.2
Tmin	/	-0.84 ± 5.59	-15.7 / -1.80 / 17.1
Tmax	/	9.73 ± 6.69	-6.3 / 9.00 / 33.5
Tdiff	/	9.88 ± 3.80	0.70 / 9.80 / 24.20
Relative humidity (%)	/	49.90 ± 20.21	8.00 / 48.00 / 100.00

SD, Standard deviation; PM_2.5, Particulate matter with aerodynamic diameter ≤ 2.5 μm;

Table 3 shows the occurrence frequency of joint events defined differently during the study period. The proportion of joint event days relative to total study days ranged from 6.0% in P1T1 to 1.2% in P4T3, revealing a clear trend of decreasing event frequency as PM_2.5 concentration thresholds increased and cold waves definitions became stricter.

Table 3.

Distribution of Exposure Days under Different PM_2.5 and Cold Waves Threshold Combinations.

Exposure Combination Definition	Joint exposure days	PM2.5-only days	Cold waves-only days	Reference days (no event)	Frequency of joint exposure (%)
P1T1	73	165	219	753	6
P1T2	54	184	162	810	4.5
P1T3	37	201	108	864	3.1
P2T1	59	120	233	798	4.9
P2T2	48	131	168	863	4
P2T3	34	145	111	920	2.8
P3T1	41	76	251	842	3.4
P3T2	34	83	182	911	2.8
P3T3	23	94	122	971	1.9
P4T1	25	29	267	889	2.1
P4T2	23	31	193	963	1.9
P4T3	15	39	130	1026	1.2

Exposure Combination Definition: Combination of PM_2.5 pollution thresholds (P1 - P4) and cold waves temperature thresholds (T1 - T3); Joint Exposure Days: Number of times joint events occurred; PM_2.5 only Days: Number of days with PM_2.5 pollution events only; Cold Waves Only Days: Number of days with cold waves events only; Reference days : days with neither a PM_2.5 pollution event nor a cold wave event. Frequency of Joint Exposure: Percentage of joint events relative to all exposure events.

Figures 1, 2 and 3 show the 21-day single-day lagged relative risk curves for the joint exposure to cold waves and PM_2.5 on non-accidental mortality, circulatory disease mortality, and respiratory disease mortality. For three mortality outcomes, the single-day lagged RR rises rapidly during lag days 0–3, then enters a phase of small, stable fluctuations, followed by a slight increase again around lag day 14. Overall, the lagged effect pattern exhibits a short-term elevation, a mid-term plateau, and a late-term modest rebound, suggesting that the health hazards of joint exposure have both immediate acute effects and sustained lagged effects (Figures 4 and 5).

Fig. 1.

21-day lag response curve of joint event exposure on non-accidental mortality.

Fig. 2.

21-day lag response curve of joint event exposure on circulatory disease mortality.

Fig. 3.

21-day lag response curve of joint event exposure on respiratory disease mortality.

Fig. 4.

Cumulative relative risks (95%CI) of different joint event combinations for different causes of mortality.

Fig. 5.

Interaction indices (RERI/AP/S) of three types of mortality under different joint event combinations.

Research findings indicate that joint exposure events involving PM_2.5 and cold waves significantly increase the risk of non-accidental mortality. As shown in Fig. 4 and Fig. 6, detailed data are presented in Tables S4-S6. Under 12 defined threshold combinations, the joint exposure CRR demonstrated statistical significance, ranging from 1.72 (95% CI: 1.39, 2.13) to 2.55 (95% CI: 1.82, 3.57). In some definitions, the risk showed an increasing trend with stricter thresholds, such as from P2T2 (CRR = 1.72) to P3T2 (CRR = 1.97) and P4T1 (CRR = 2.55). However, the CRR for the strictest combination, P4T3, was 1.75, lower than that for P4T1, suggesting that the exposure-risk relationship is not a simple linear gradient. This may be influenced by reduced sample size, increased statistical uncertainty, or the presence of actual nonlinear effects. From a public health burden perspective, the attributable fraction (AF) for non-accidental mortality ranged from 42.2% to 60.8% across the 12 joint exposure definitions, corresponding to 368 to 1,696 attributable deaths (AN) during joint event days.

Fig. 6.

Attributable proportion and attributable number of three types of mortality associated with joint event exposure.

This study found that joint exposure to PM_2.5 and cold waves also significantly increased the risk of circulatory disease mortality. Across 12 definitions, the joint exposure CRR ranged from 1.84 (95% CI: 1.42, 2.37) to 2.62 (95% CI: 1.76, 3.92), all statistically significant. Although overall risk levels exceeded those of single exposures, risk did not exhibit a strictly linear increase with threshold severity. For instance, the highest risk occurred under P4T1 (CRR = 2.62), while the CRR for the most stringent definition, P4T3, decreased to 1.83—lower than some moderately stringent combinations. This nonlinearity may relate to reduced sample sizes under extreme conditions and saturation of the exposure-response relationship. In terms of attributable burden, the attributable fraction (AF) for circulatory disease mortality ranged from 44.8% to 61.8% across the 12 definitions, corresponding to 220 to 994 attributable deaths (AN) during joint event days.

The study also revealed that joint exposure to PM_2.5 and cold waves substantially increased respiratory disease mortality risk. Under most threshold combinations, the joint exposure showed a markedly elevated CRR, with the highest risks observed in P4T1 (CRR = 2.34, 95% CI: 1.18, 4.65) and P4T2 (CRR = 2.33, 95% CI: 1.15, 4.73). However, the risk did not continue to increase with stricter thresholds. The CRR for the most stringent definition combination, P4T3, decreased to 1.48 (95% CI: 0.56, 3.87), but this reduction was not statistically significant.This suggests increased uncertainty in risk estimation under extreme exposure conditions, potentially due to reduced event numbers or the presence of a true effect plateau. From a public health burden perspective, the attributable fraction (AF) for respiratory disease mortality ranged from 26.5% to 57.3% across the 12 definitions, corresponding to 38 to 152 attributable deaths (AN) during joint event days.

The study evaluated the interactions among three types of mortality at different joint exposure thresholds using RERI, AP, and S indices. Analysis results revealed additive interactions for non-accidental deaths and circulatory disease mortality across multiple exposure combinations (Fig. 5,detailed data are presented in Tables S7-S9). Non-accidental mortality showed additive interactions at P1T2 (RERI = 0.52, 95% CI: 0.07, 0.98; P = 0.003), P2T2 (RERI = 0.51, 95% CI: 0.07, 0.95; P = 0.001), and P3T2 (RERI = 0.69, 95% CI: 0.13, 1.24; P < 0.001). Cardiovascular disease mortality showed additive interactions in seven combinations (P1T1, P1T2, P2T1, P2T2, P3T1, P3T2, and P4T1; P < 0.05), with the most significant interactions observed in the P1T2 and P3T2 combinations (P < 0.001). Respiratory disease mortality exhibited a significant negative interaction only in the P4T3 combination (RERI = -1.71, 95% CI: -4.08, 0.65; P = 0.008). Results for the AP indicator were largely consistent with RERI, showing corresponding positive or negative interactions in significant combinations. The S index also indicated synergistic effects (S > 1) or antagonistic effects (S < 1) in some significant combinations.

This study analyzed the health impacts of joint exposure across different populations through stratified analysis, as shown in Fig. 7, detailed data are presented in Tables S11-S17. In the age stratification, individuals aged 75 years and older exhibited the highest sensitivity to joint exposure, with significantly elevated joint exposure CRR across all threshold combinations (P < 0.01), peaking at P4T1 (CRR = 2.85, 95% CI: 1.95, 4.17). The 65–75 age group exhibited the next highest joint exposure risk, remaining statistically significant in most combinations, though the risk did not consistently increase under some stricter definitions. The 0–64 age group showed the lowest joint exposure risk, with most combinations lacking statistical significance. Overall, the joint risk did not increase continuously as exposure thresholds became stricter, suggesting a nonlinear exposure-response relationship. Stratified analysis by educational attainment revealed significantly higher joint exposure CRR among low-educated individuals compared to high-educated individuals, with a markedly heavier attributable burden. Interaction analysis showed significant positive additive interactions (RERI > 0, P < 0.05) under multiple definitions (e.g., P1T2, P2T2, P3T2) in low-educated individuals, whereas interactions in high-educated individuals were predominantly negative or non-significant. However, under the strictest exposure definition (P4T3), both groups exhibited negative interactions, and the joint exposure risk decreased compared to P4T1, further suggesting a nonlinear exposure-response relationship. Gender-stratified analysis revealed significantly elevated mortality risk in men under joint exposure, with positive additive interactions (RERI > 0, P < 0.05) observed across multiple definitions (P1T2, P2T2, P3T2). Women also exhibited significant joint exposure risks in most combinations, with positive additive interactions observed in certain definitions (e.g., P1T2, P2T2, P3T2; RERI > 0, P < 0.05).

Fig. 7.

Cumulative Relative Risks of Joint Events by Demographic Characteristics and Exposure Definitions.

However, under the strictest exposure definition (P4T3), both genders showed negative interactions (male: RERI = -0.62, P = 0.028; female: RERI = -0.50), and the joint exposure risk decreased compared to P4T1, further supporting a nonlinear exposure-response pattern. These findings indicate that older adults and those with lower education levels are more vulnerable to joint exposure to PM_2.5 and cold waves, with health effects exhibiting significant heterogeneity across age, education level, and gender.

To examine model robustness, we performed sensitivity analyses by varying the degrees of freedom for relative humidity (4–6), lag days (14, 28), and time trends (4 × 4, 4 × 6, 4 × 8). The Q-AIC values identified P4T2 as the optimal definition for non-accidental and circulatory mortality, and P1T2 for respiratory mortality (Table S10). Key estimates remained stable across parameter settings; for non-accidental mortality (P4T2), lag-0 RR remained ~ 1.07 and lag-0-21 CRR ~ 2.15. Using a uniform reference category (PM₂.₅ <115 µg/m³, temperature >–0.4 °C) did not change the CRR ranking (Table S18). Excluding the COVID-19 policy-transition period (Dec 2022–Feb 2023) or winsorizing daily deaths at the 99th percentile left the main estimates largely unchanged (Table S19). Overall, the sensitivity analyses confirm the robustness of our findings (Table 4).

Table 4.

RR and CRR of three types of mortality associated with joint exposure under different parameter adjustments.

model choices	Non-accidental Mortality		Circulatory disease Mortality		Respiratory disease Mortality
	lag 0 (RR, 95% CI)	lag0-21 (CRR, 95% CI)	lag 0 (RR, 95% CI)	lag0-21 (CRR, 95% CI)	lag 0 (RR, 95% CI)	lag0-21 (CRR, 95% CI)
df for RH:3	1.07(0.99, 1.16)	2.15(1.52, 3.05)	1.06(0.97, 1.17)	2.12(1.40, 3.21)	1.06(0.95, 1.19)	1.53(1.00, 2.35)
df for RH:4	1.07(0.99, 1.15)	2.15(1.52, 3.05)	1.06(0.97, 1.16)	2.12(1.40, 3.20)	1.06(0.95, 1.19)	1.53(1.00, 2.35)
df for RH:5	1.07(0.99, 1.16)	2.15(1.52, 3.06)	1.06(0.97, 1.17)	2.12(1.40, 3.21)	1.06(0.94, 1.19)	1.53(1.00, 2.34)
df for RH:6	1.07(0.99, 1.16)	2.16(1.52, 3.06)	1.06(0.97, 1.17)	2.12(1.40, 3.22)	1.06(0.94, 1.19)	1.53(1.00, 2.34)
Lag = 21	1.07(0.98, 1.15)	2.15(1.52, 3.05)	1.06(0.97, 1.17)	2.12(1.40, 3.21)	1.06(0.94, 1.19)	1.53(1.00, 2.35)
Lag = 14	1.03(0.96, 1.11)	1.53(1.17, 2.00)	1.06(0.97, 1.15)	1.64(1.19, 2.25)	1.01(0.91, 1.13)	1.16(0.82, 1.64)
Lag = 28	1.12(1.03, 1.23)	3.04(1.97, 4.68)	1.11(1.00, 1.24)	2.90(1.73, 4.84)	1.11(0.98, 1.26)	2.37(1.43, 3.92)
df for date:4*6	1.02(0.98, 1.08)	2.15(1.52, 3.05)	1.05(0.99, 1.11)	2.12(1.40, 3.21)	0.98(0.89, 1.09)	1.53(1.00, 2.35)
df for date:4*4	1.10(1.01, 1.21)	2.05(1.45, 2.91)	1.10(0.99, 1.22)	2.16(1.46, 3.22)	1.16(1.01, 1.33)	1.88(1.28, 2.76)
df for date:4*8	1.06(1.00, 1.13)	1.80(1.32, 2.46)	1.06(0.98, 1.15)	1.65(1.12, 2.44)	1.06(0.95, 1.19)	1.25(0.80, 1.95)

Discussion

This study assessed the joint effects of cold waves and PM₂.₅ on mortality risk in Shijiazhuang. Results indicate that joint exposure was associated with increased risks of non-accidental, circulatory, and respiratory mortality. Health effects tended to be more pronounced with higher PM₂.₅ levels and more severe cold waves, suggesting synergistic interactions. Elderly individuals, males, and those with lower educational attainment were more vulnerable.

The study found that joint exposure to cold waves and PM_2.5 pollution was associated with elevated risks of non-accidental and cardiovascular mortality across all defined threshold combinations. A clear dose-response relationship emerged with increasing pollution concentrations and cold intensity, with risks for non-accidental, cardiovascular, and respiratory mortality peaking at P4T1. Previous studies have also identified joint effects of air pollution and cold waves on all-cause mortality, non-accidental mortality, cardiovascular mortality, respiratory mortality²⁶, metabolic syndrome (MetS)³⁰, or mental disorders³¹. Huang²⁴ found in East China that the relative risk of all-cause mortality from joint events reached 2.07 under the strictest definition. In contrast, Shijiazhuang City in this study showed a CRR of 1.75 (1.10, 2.80) under the strictest definition, while the P4T1 definition yielded the highest CRR (CRR = 2.55, 95% CI: 1.82, 3.57). This discrepancy may relate to Shijiazhuang’s geography and energy mix: on one hand, daily PM_2.5 concentration variations are more pronounced than in the Yangtze River Delta region; on the other, winter coal-fired heating significantly increases emissions. This likely contributes to higher health risks²³.Ning reported that cold waves and PM_2.5 exposure were associated with significantly increased IHD hospitalization rates in both high-altitude (Xining) and low-altitude (Xinxiang) regions. Xining exhibited a synergistic effect where both factors jointly contributed to IHD episodes, while Xinxiang showed stronger individual effects of cold waves and PM_2.5 exposure compared to Xining¹⁷. He reported that exposure to cold waves accompanied by high Air Quality Index (AQI) levels carried a higher relative risk than cold waves exposure alone³¹.There is also the pollutant-temperature exposure model proposed with a pre-specified functional form, which suggests a synergistic effect between PM_2.5 and cold temperature, consistent with the findings of this study³².

However, synergistic effects exist between the two, though the underlying pathophysiological mechanisms remain to be elucidated. Experimental studies have suggested³³that short-term exposure to PM_2.5exacerbates cardiovascular and respiratory damage: After being phagocytosed by macrophages, PM_2.5induces reactive oxygen species (ROS) production from mitochondria and NOX2, activating NLRP3 inflammasomes and TLR4/NF-κB signaling pathways. This leads to the release of inflammatory mediators into the circulation, triggering inflammation and oxidative stress, which in turn excites the sympathetic nervous system. affecting both cardiovascular and respiratory systems to worsen pathological alterations. Concurrently, prolonged cold exposure triggers cold stress in the body, causing peripheral vasoconstriction, increasing cardiac workload, inducing inflammation and pyroptosis, and resulting in myocardial injury³⁴. Both PM_2.5 and cold stress may induce pulmonary oxidative stress and pro-inflammatory responses; studies suggest their joint action may elevate interleukin-6 and interleukin-8 levels. These findings suggest that PM_2.5 and cold stress may have a synergistic effect in inducing inflammation, which may be associated with increased disease incidence³⁵. Such pathways appear to be more pronounced during joint exposure, which may help explain the observed associations in cardiovascular, cerebrovascular, and respiratory mortality.

Studies evaluating multiple PM_2.5 and cold waves exposure combinations for interactions with three mortality categories found significant additive interactions for non-accidental mortality across multiple thresholds. Notably, cardiovascular disease mortality showed the strongest interaction in the P1T2 and P3T2 combinations (P < 0.001). Qiu’s study reported an interaction between cold temperatures and air pollutants on the risk of hospitalization for chronic obstructive pulmonary disease, with men exhibiting greater sensitivity³⁶. Xu’s investigation into myocardial infarction found no significant synergistic interaction between cold waves and PM_2.5³⁷. He studied schizophrenia patients in Hefei and reported a synergistic effect between cold waves and high AQI in males, with a relative excess risk in interaction (RERI) of 0.018 (95% CI: 0.005, 0.030³¹. A Guangzhou study reported significant interactions between PM_2.5 and low temperatures for mortality in non-accidental deaths, cardiovascular deaths, and among the elderly³⁸. Furthermore, AP and S-index results provide additional quantitative evidence supporting the existence and directionality of additive interactions which may lay the groundworkfor establishing health early warning systems based on joint exposures. Future research should integrate longer-term time series data with individual exposure information to further validate interaction patterns between cold waves and PM_2.5 across different regions and health outcomes, thereby optimizing synergistic warning thresholds and public health intervention strategies.

Numerous epidemiological studies have reported that cold waves and PM_2.5 are independently associated with higher mortality risks. This study found that under stricter extreme temperature thresholds (T3), the health impacts on non-accidental mortality and circulatory disease mortality were more apparent. A global study spanning diverse climatic zones and socioeconomic/demographic characteristics³⁹ reported that extreme cold was associated with an increased risk and burden of all-cause cardiovascular disease mortality; A case-crossover study covering 323 Chinese cities⁴⁰ found cold waves were associated with a 4.24% increased risk. A random-effects meta-analysis revealed that a 1 °C decrease in temperature was associated with a 1.6% increase in cardiovascular disease-related mortality, with the most pronounced effects observed for coronary heart disease mortality (RR 1.015;95% CI 1.011, 1.019) and the incidence of aortic aneurysms and dissections (RR 1.026; 95% CI 1.021, 1.031)⁴¹. Exposure to varying cold temperatures was associated with increased risks of cardiovascular disease mortality or hospitalization⁴². A study across 31 provincial capitals in China³ indicated that cold waves in Xining increased the risk of non-accidental deaths, with an RR of 1.44 (1.15, 1.81). A Lanzhou study⁴³ confirmed that extreme cold (-12 °C) also increases IHD hospitalization rates. However, another Jiangsu study³⁷ found no interaction between cold waves and myocardial infarction, potentially due to differences in study design, statistical analysis methods, and research domains. Zhou also observed a predominantly positive interaction between PM_2.5 pollution and extreme cold, though with lower statistical significance⁴⁴. A meta-analysis of studies spanning 18 years showed the strength of the relationship, which is consistent with the finding that air pollution exposure is associated with increased risks of all-cause CVD mortality and morbidity^45,46. A European cohort study⁴⁷ further indicated that long-term exposure to low levels of PM_2.5 was associated with higher mortality, with a 5 µg/m³ increase in PM_2.5 associated with a 1.13-fold (1.11, 1.16) relative risk (RR) increase in mortality. A study in Beijing²⁰ suggests that cold waves may exacerbate the impact of cardiovascular morbidity and mortality under conditions of elevated PM_2.5 concentrations.

However, compared to the strong synergistic effects observed between heatwaves and pollutants²⁷, this study found that the interaction between cold waves and PM_2.5 was not significant in certain populations, and the risk associated with PM_2.5 exposure alone also lacked statistical significance in most cases. This phenomenon can be explained by behavioral patterns and environmental factors during cold waves: People spend most of their time indoors during cold waves and are more directly exposed to indoor sources of PM_2.5, which are less dispersed compared to outdoor sources. This partially affects PM_2.5 penetration and human exposure levels⁴⁸. Furthermore, in regions with centralized heating, indoor thermal comfort is maintained, partially offsetting cold stress by altering exposure patterns. However, coal emissions from heating sources increase winter air pollutant concentrations, and since heating significantly impacts PM_2.5 levels, this elevates outdoor PM_2.5 concentrations, thereby altering the actual exposure-response relationship^49–51. This may have weakened the interaction between cold waves and PM_2.5. This finding also aligns with inconsistent reports in the literature regarding the health effects of PM_2.5. A study in the Madrid region of Spain demonstrated that exposure to PM_2.5 is associated with health outcomes including respiratory and cardiovascular diseases, lung cancer, diabetes, and dementia⁵². These discrepancies may stem from variations in population characteristics, climate, and pollution composition across studies, as well as differences in individual exposure levels, susceptibility, and physiological responses.

Climate change has been associated with more frequent extreme temperature events, which may threaten vulnerable populations, particularly in low-income regions³⁹. This study found that among all demographic groups, older adults (≥ 75 years) appear to have the highest sensitivity to the joint effects of cold waves and PM_2.5, with their mortality risk showing a significant increase with exposure intensity. This heightened vulnerability stems primarily from physiological decline: reduced thermoregulatory function, higher prevalence of underlying health conditions, and diminished organ compensatory capacity. Furthermore, disparities in socioeconomic status, healthcare accessibility, and health behaviors among older adults may increase their sensitivity to environmental exposures. Mortality risk and attributable burden were significantly higher among individuals with lower education levels compared to those with higher education, suggesting environmental health inequities. Lower education levels correlate with poorer socioeconomic conditions, potentially manifesting in substandard housing with inadequate heating and ventilation, limited knowledge of health protection measures, and delayed access to medical resources. Regarding gender disparities, while men have higher risks and stronger interactions in joint events, they face greater risks from single pollutant exposures during isolated events. Conversely, women show heightened sensitivity to cold waves exposure. These differences stem from physiological and behavioral factors. Physiologically, women possess relatively weaker thermoregulatory capacity, higher airway reactivity, and increased metabolic burden during cold exposure, contributing to the health effects of cold waves⁵³. Behaviorally, men engage in more outdoor work and exhibit higher smoking rates^54,55, which may lead to higher cumulative exposure doses and elevated risks from pollutants.

This study has several strengths: It innovatively employs a “joint exposure” analytical framework, overcoming the limitations of traditional approaches that treat cold waves and PM_2.5 pollution as independent factors. For the first time, it defines the simultaneous occurrence of both as an independent exposure event. By comparing multiple combinations of health risks, it fills a gap in research on joint environmental health risks. Focusing on Shijiazhuang City, it clarifies the synergistic role of joint exposure in elevating mortality risks among residents, reveals the pattern of effect intensification with increasing pollution and cold waves severity, and identifies elderly individuals, males, and those with lower education levels as vulnerable subgroups, providing a scientific basis for risk prevention and control. Of greater practical value, the study proposes establishing a “cold waves-PM_2.5” joint early warning mechanism and strengthening protections for vulnerable populations in response to the increasing frequency of extreme events under climate change. These recommendations not only inform environmental health decision-making in Shijiazhuang and northern Chinese cities but also lay the groundwork for improving climate change health impact assessments and guiding adaptation policies.

Limitations of this study include the following: First, the inability to obtain precise individual-level exposure data based on exact mortality locations or indoor exposure information. Consequently, exposure assessments relied on fixed-site environmental data, failing to reflect actual microenvironmental exposure levels. Second, potential diagnostic or coding errors regarding causation cannot be entirely ruled out. Additionally, the study design employed an ecological time-series analysis, which inherently has limitations in causal inference. Future research could integrate mobile monitoring, individual behavioral logs, and prospective cohort designs to more precisely evaluate the health effects of joint exposures.

Conclusion

This study shows that joint air pollution and cold wave events in Shijiazhuang are associated with a significantly increased risk of resident mortality.The elevated mortality risk associated with the joint effects of PM_2.5 and cold waves varies across different subpopulations, which helps identify vulnerable subgroups that could be prioritized for targeted interventions. Therefore, comprehensively elucidating the underlying health impact mechanisms is important for designing targeted public health strategies. Our findings support the integration of such joint environmental threats into public health early warning and emergency response systems to help safeguard public health.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Mengna Li: Conceptualization, Software, Validation, Writing – Original Draft; Feng Li: Conceptualization, Software, Validation, Writing – Original Draft; Siyuan Chen: Methodology, Formal analysis; Yishan Ding: Data Curation; Fengge Chen: Project administration, Resources; Qiang Shen: Supervision; Lei Bao: Supervision; Jinsha Ma: Resources; Yi Ji: Funding acquisition; Hui Kang: Writing – Review & Editing, Data Curation, Validation.

Funding

This work was supported by the Medical Science Research Project of Hebei Province [grant number 20251134].

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.