Abstract
Abstract
Background
Both air pollution and temperature variability (with cold wave as an extreme form) may influence the incidence of ischaemic stroke (IS). This study aimed to examine the association between winter air pollution-cold wave sequential events (persistent air pollution followed by cold waves) and IS incidence among adults aged>=60 years.
Methods
Clinical data were sourced from the Tianjin Medical and Health Big Data Platform (covering 81 secondary/tertiary hospitals), and meteorological/air quality data were sourced from the National Meteorological Science Data Centre. Spearman rank correlation analysis was used to assess the relationships between meteorological variables (eg, 24-hour temperature decrease), atmospheric pollutants (including Air Quality Index (AQI)) and elderly IS incidence. A distributed lag nonlinear model (DLNM) was applied to analyse lagged effects of winter air pollution and cold wave sequential events on IS incidence, while a generalised additive model (GAM) was used to evaluate additive interactions between air pollution and cold waves on IS incidence.
Results
The study included 109 513 adults aged >=60 years with first-onset IS from 2016 to 2019. Eight winter air pollution-cold wave sequential events were identified over 4 years, with higher daily IS incidence during event periods (67 new cases/day), lag periods (68 new cases/day) than non-event periods (60 new cases/day). Subgroup analysis showed that among adults aged ≥80 years, proportional incidence during both events (80 to 85 years old: 1.89, 95% CI 0.52 to 3.26; 85 to 90 years old: 1.96, 95% CI 0.59 to 3.33) and lag period (80 to 85 years old: 0.90, 95% CI 0.02 to 1.78; 85 to 90 years old: 1.52, 95% CI 0.64 to 2.40) increased compared with the non-event period. Daily IS incidence was positively correlated with 24-hour temperature decreases, AQI and other air pollutants. DLNM showed that lag effects emerged 4 days post-exposure, with the highest IS risk at a 9-day lag (RR=1.122, 95% CI 0.443 to 2.838). GAM confirmed positive additive interactions between air pollution and cold waves on IS incidence (p<0.001).
Conclusion
Winter air pollution-cold wave sequential events exert a synergistic, lagged effect on IS incidence in the elderly, with adults >=80 years being the most vulnerable. The observed risk patterns and underlying mechanisms underscore the importance of integrated environmental and public health strategies to reduce IS burden in this high-risk population.
Keywords: Stroke, Climate Change, Retrospective Studies, Frail Elderly, China, Environmental Illness
STRENGTHS AND LIMITATIONS OF THIS STUDY.
The study integrated multi-dimensional clinical, meteorological and air quality data to support the first systematic analysis of winter air pollution-cold wave sequential events and elderly IS incidence.
This study used data from the Tianjin Medical and Health Big Data Platform (covering 81 hospitals in Tianjin), allowing comprehensive capture of first-onset IS cases in the elderly.
Combined use of Spearman’s rank correlation analysis, distributed lag nonlinear model and generalised additive model enabled assessing variable associations, analysing lagged effects and testing interaction effects—enhancing methodological rigour.
The observational design limits causal inference but offers valuable hypothesis-generating insights.
The study was restricted to Tianjin, making generalisation of its results to other regions difficult and limiting external validity.
Introduction
Stroke is the third leading cause of death globally according to the Global Burden of Disease (GBD) Study 2021.1 Particulate matter (PM) pollution and low temperatures are well-recognised significant environmental risk factors for ischaemic stroke (IS).2 In recent years, the global increase in extreme weather events has further highlighted climate change as a pressing public health challenge.3 Prior studies have linked IS incidence and poor prognosis to temperature-related extremes—including sudden temperature fluctuations,4 5 cold or hot waves,6 7 extreme temperatures8 9— as well as air pollution.10 11 Notably, cold periods, sudden temperature drops and high concentrations of PM, such as PM10 and PM2.5, are among the most impactful atmospheric and meteorological drivers of IS.5 12 13 Researchers have also investigated interactive effects between meteorological and environmental factors on stroke,14 15 given that air pollution levels are tightly coupled with weather patterns: air quality varies directly with changes in meteorological conditions.16 17
Amid global climate change, joint events of air pollution followed by cold waves (hereafter referred to as “sequential events”) are likely to occur more frequently.18 Mechanistically, under stagnant meteorological conditions, pollutants tend to accumulate near the ground; when a cold wave arrives, air quality improves rapidly, but this is accompanied by sharp declines in temperature.19 While existing studies have associated such joint environmental events with increased risks of respiratory and mental disorders,20 21 the impact of these sequential events on IS—particularly in vulnerable populations like the elderly—remains understudied. The elderly exhibit heightened physiological sensitivity to both cold exposure and air pollution,13 making them a key population for investigating environmental health risks. Therefore, using health data from the Tianjin Medical and Health Big Data Platform, this study aimed to investigate the association between winter air pollution-cold wave sequential events and IS incidence in adults aged >=60 years.
Methods
Study sample
Daily data of first-onset IS were extracted from the Tianjin Medical and Health Big Data Platform, which integrates electronic medical record (EMR) data from 81 hospitals (42 tertiary and 39 secondary) in Tianjin between 1 January 2016 and 31 December 2019. Tianjin is a municipality located in northern China—adjacent to Beijing to the northwest. This municipality covers a total area of approximately 12 000 square kilometres. These hospitals cover all 16 administrative districts of Tianjin (both urban and suburban areas), including nearly all public tertiary hospitals (excluding military hospitals) and one-third of secondary hospitals in Tianjin. All hospitals qualified to admit patients with acute stroke are included, ensuring comprehensive capture of first-onset IS incidence data in the elderly.
Eligible patients were adults aged >=60 years, hospitalised in Tianjin with a first-onset IS (International Classification of Diseases (ICD)10: I63) regardless of gender. Individuals with a history of stroke (either via prior stroke-related diagnoses or confirmation by previous MRI/CT imaging) before the index IS onset were excluded. Of 244 128 potential participants, 91 727 were excluded for failing to meet inclusion criteria and 42 906 for meeting exclusion criteria, resulting in a final sample of 109 513 participants. Given that the COVID-19 pandemic from 2020 to 2023 altered population health status, healthcare-seeking behaviours, as well as work and life rhythms in China, this study only analyses data collected before the pandemic (from 2016 to 2019) to ensure the generalisability of the research findings. Meteorological and air pollution data for the Tianjin region from the corresponding period of time were obtained from the National Meteorological Science Data Centre (http://data.cma.cn/). For medical records, cases with missing key variables (eg age or ICD-10 diagnosis code) were excluded during the initial screening process. The proportion of such exclusions was minimal, with no significant impact on sample size or introduction of selection bias. For meteorological and air pollution data, missing rates were <5%, and missing values were imputed using linear interpolation.
Definition of winter air pollution-cold wave sequential events
Winter air pollution-cold wave sequential events refer to the occurrence of air pollution followed by cold waves in winter (hereafter referred to as ‘sequential events’). In the Tianjin region, low-temperature conditions are primarily concentrated between November of a given year and March of the following year. Three distinct periods were defined within this winter time window: the event period, the event lag period and the non-event period.
The event period is defined as a phase where persistent air pollution is immediately followed by a cold wave, with a total duration of >=7 days. Air pollution criteria: the air quality index (AQI) exceeds 100 (per standards set by China’s Ministry of Environmental Protection), and this polluted state persists for at least three consecutive days. Subsequent to the pollution phase, a cold wave occurs—characterised by one concurrent condition: (1) a temperature drop of >=8℃ within 24 hours, or (2) a sustained decrease in daily average temperature (3–7℃ over the following 5 days). The temperature variation range for sequential cold wave events was determined with reference to the updated Chinese National Standard for Cold Wave Classification (GB/T 21 987–2022) and the winter meteorological characteristics of Tianjin. During the sequential event, a cold wave is typically accompanied by a rapid improvement in air quality.
The event lag period is defined as the consecutive days starting from the first natural day immediately after the end of an event period, with its total duration being exactly consistent with the actual number of days of the corresponding event period.
The non-event period was matched to each event period with identical duration and met the following criteria: (1) generally good air quality (AQI<=100); (2) stable temperature fluctuations (<= 12℃ within a 48-hour window); and (3) the start date of the non-event period was at least 9 days after the end of the corresponding event period (to avoid residual effects of sequential events). This matching strategy helped control for seasonal patterns, weekday/weekend effects and other time-varying confounders, thereby minimising bias arising from unequal exposure to confounding factors.
Statistical analysis
We first characterised potential confounding factors, including sex, age, comorbidities and medication usage. Medication usage was defined as the use of >=1 chronic medication (eg, cardiovascular, hypoglycaemic and primary prevention agents) within 1 year of stroke onset (online supplemental table 2). Descriptive statistics were presented as counts (%) for categorical variables and median (SD, SD) for age. Group differences in variables across non-event, event and lagged-event periods were compared using χ2 test or Wilcoxon rank-sum test. Statistical significance was set at a p-value < 0.05, and variables with significant group difference were included in subgroup interaction analyses. We assessed the normality of all continuous meteorological and air pollution variables using the Shapiro-Wilk test. Since all variables were non-normally distributed (p<0.05), Spearman’s rank correlation was used to evaluate the relationships between these environmental factors and daily stroke incidence.
Spearman’s rank correlation analysis
Spearman’s rank correlation analysis was used to examine the associations between: (1) meteorological factors and daily IS cases; (2) air pollution parameters and daily IS cases and (3) meteorological factors and air pollution concentrations. Meteorological factors of interest include: temperature drop within 24 hours (°C), relative humidity (%rh), wind speed (m/s), atmospheric pressure (Kpa), and air pollution parameters include: AQI (μg/m3) and concentrations of various pollutants (PM2.5, PM10, SO2, NO2, CO and O3). Additionally, variance inflation factor (VIF) tests were conducted to assess multicollinearity among variables.
Distributed Lag Nonlinear Models
Meteorological factors correlated with daily IS cases were included in a base regression model. The optimal model was selected using Akaike Information Criterion (AIC), with the final model incorporating relative humidity (%rh), NO2, CO and O3. We constructed cross-matrices for daily IS incidence (dependent variable) and meteorological data, fitting the model with a quasi-Poisson link function. Controlling for seasonal trends, long-term trends and day-of-the-week effects, the distributed lag nonlinear model (DLNM) model22 was employed to quantify the lagged association between 24-hour temperature drop (defined as the difference between the current day’s and previous day’s average temperature) and daily IS incidence, while adjusting for confounding by pollution-related meteorological factors. The number of daily new cases approximately followed a Poisson distribution. Natural cubic spline functions were applied to control for residual confounding from air pollution and other meteorological factors.
Generalised Additive Model
Building on the DLNM (which explored lagged effects of temperature changes on IS incidence), we used generalised additive models (GAMs) to further investigate potential interactions between 24-hour temperature drop and air pollution levels (measured by AQI) during winter air pollution-cold wave sequential events. GAMs integrate generalised linear models with additive models, enabling the analysis of nonlinear relationships without assuming linearity between dependent and independent variables. In our GAM, the dependent variable was daily IS cases, fitted with a Poisson link function. Tensor product interactions (defined via the ti function) were included to examine: (1) main effects of 24-hour temperature drop and AQI; and (2) interactive effects between these two factors. A smooth model based on tensor products was established to characterise these effects.
Results
Impact of winter air pollution-cold wave sequential events on ischaemic stroke (IS) incidence in elderly adults
Over 4 years, eight winter air pollution-cold wave sequential events were identified, seven of which lasted 7 days and one lasted 9 days. The patient selection procedure was presented in figure 1. We conducted descriptive analysis on daily new IS cases, air pollution levels (average AQI) and average daily temperature range across three periods: non-event, event and event lag periods. The event lag period had the highest cumulative new IS cases and daily new cases (3,983, 68 new cases/day), followed by the event period (3,929, 67 new cases/day); the non-event period had the lowest total cases (3,491, 60 new cases/day) (table 1), representing an absolute increase of seven to eight cases per day. The event lag period consistently had the highest total and annual average IS cases throughout the study duration (2016–2019). Notably, because the lag period corresponding to the 2017 event period was delayed until the beginning of 2018, this resulted in a decrease in the total number of cases during the lag period in 2017.
Figure 1. Flowchart of patient selection. ICD, International Classification of Diseases.
Table 1. Number of new ischaemic stroke cases in the event period, event lag period and non-event period in each time period.
| Year | Subgroup | Event period (n=3929） |
Event lag period (n=3983） | Non-event period (n=3491） |
|---|---|---|---|---|
| 2016–2019 | Average | 982 | 996 | 873 |
| 2016 | Total year | 978 | 1053 | 909 |
| Winter (from January to March） | 559 | 621 | 497 | |
| Winter (from November to December） | 419 | 432 | 412 | |
| January | -- | -- | -- | |
| February | -- | -- | 497 | |
| March | 559 | 621 | -- | |
| November | 138 | -- | 412 | |
| December | 281 | 432 | -- | |
| 2017 | Total year | 865 | 521 | 921 |
| Winter (from January to March) | 555 | 521 | 464 | |
| Winter (from November to December) | 310 | -- | 457 | |
| January | -- | -- | -- | |
| February | 555 | -- | -- | |
| March | -- | 521 | 464 | |
| November | -- | -- | 457 | |
| December | 310 | -- | -- | |
| 2018 | Total year | 1156 | 1464 | 748 |
| Winter (from January to March) | 698 | 1000 | 484 | |
| Winter (from November to December) | 458 | 464 | 264 | |
| January | 186 | 485 | -- | |
| February | 512 | -- | -- | |
| March | -- | 515 | 484 | |
| November | 458 | 464 | -- | |
| December | -- | -- | 264 | |
| 2019 | Total year | 930 | 945 | 913 |
| Winter (from January to March) | 930 | 447 | 151 | |
| Winter (from November to December) | -- | 498 | 762 | |
| January | 450 | 447 | 151 | |
| February | 480 | 295 | -- | |
| March | -- | 203 | -- | |
| November | -- | -- | 483 | |
| December | -- | -- | 279 |
Note: The symbol “--” indicates that no event periods, event lag periods or non-event periods occurred in the corresponding months. As our analysis only includes new ischaemic stroke (IS) cases occurring within these three periods, months with no such periods have no case counts associated with the study’s defined exposure/non-exposure contexts and thus their new IS incidence is not included in the table’s statistical scope.
Air pollution severity and daily temperature ranges showed period-specific patterns, consistent with the definition of air pollution-cold wave sequential events: the event period had the most severe pollution, with a mean AQI of 151.11±82.17—classified as ‘moderate pollution’ per China’s national air quality standards. The event lag period had milder pollution, with a mean AQI of 114.45±56.74 (‘mild pollution’). The non-event period maintained generally good air quality, with mean AQI consistently below 100 (the threshold for ‘good’ air quality). In terms of daily temperature ranges, the event lag period had the largest temperature fluctuation (10.03±3.86℃), followed by the event period (9.82±3.49℃). The non-event period had significantly reduced temperature variability, with mean daily ranges lower than those in the event and event lag periods.
We also compared baseline characteristics across periods. Baseline characteristics of elderly IS patients (sex, comorbidities and chronic medication use) were generally balanced across the three periods (table 2), indicating minimal confounding bias. The only statistically significant baseline difference was in median age, so we conducted age-stratified subgroup analyses to identify vulnerable subgroups (online supplemental table 1). Winter air pollution-cold wave sequential events significantly altered the age structure of IS incident populations, with the incidence core shifting from younger-old adults (notably (65,70) years) to older-old adults (>=80 years). The (65,70) years group—non-event period’s core incident subgroup (27.70% of total cases)—saw its proportional incidence (PI) drop to 25.04% during events. This reflects reduced contribution to total cases, as this group retains better physiological functions (cardiovascular regulation, temperature maintenance and immunity) than older subgroups, leading to stronger tolerance to low temperature/pollution and slower case growth. ‘Very elderly’ subgroups showed significant AR increases: (80,85) years (9.11% → 11.00%) and (85,90) years (3.15% → 4.05%), suggesting their susceptibility to the combined effects of air pollution and cold waves.
Table 2. Characteristics of the study population across different exposure periods.
| Non-event period, n=3491 |
Event period, n=3929 |
P value | Event lag period, n=3983 |
P value | |
|---|---|---|---|---|---|
| Sex | 0.334 | 0.513 | |||
| Male | 1881 (53.88%) | 2161 (55.00%) | 2116 (53.13%) | ||
| Female | 1610 (46.12%) | 1768 (45.00%) | 1867 (46.87%) | ||
| Age, years; median (SD) | 70.61 (7.32) | 71.03 (7.54) | 0.035* | 71.04 (7.65) | 0.046* |
| Hypertension | 2641 (75.65%) | 3010 (76.61%) | 0.334 | 2991 (75.09%) | 0.577 |
| Coronary heart disease | 1791 (51.30%) | 2028 (51.62%) | 0.788 | 2040 (51.22%) | 0.941 |
| Diabetes mellitus | 1184 (33.92%) | 1265 (32.20%) | 0.116 | 1291 (32.41%) | 0.168 |
| Hyperlipidaemia | 1110 (31.80%) | 1251 (31.84%) | 0.968 | 1271 (31.91%) | 0.916 |
| Arrhythmia | 743 (21.28%) | 827 (21.05%) | 0.805 | 849 (21.32%) | 0.973 |
| Atrial fibrillation | 241 (6.90%) | 261 (6.64%) | 0.656 | 289 (7.26%) | 0.554 |
| Gastritis | 449 (12.86%) | 463 (11.78%) | 0.158 | 500 (12.55%) | 0.690 |
| Any baseline medication | 3160 (90.52%) | 3553 (90.43%) | 0.897 | 3588 (90.08%) | 0.526 |
χ2 test, Wilcoxon rank-sum test or Fisher’s exact test, as appropriate.
represents p value <0.05
Delayed effects of winter air pollution-cold wave sequential events on ischaemic stroke incidence in elderly adults
The results showed correlations between environmental factors and IS cases, and lagged effects of temperature changes. Spearman’s rank correlation analysis revealed that daily new IS cases were correlated with multiple environmental variables (figure 2), with the following patterns: 24-hour temperature decrease (°C), AQI index and concentrations of pollutants (PM2.5, PM10, NO2, O3) were positively correlated with daily new IS cases. The average relative humidity and average atmospheric pressure were negatively correlated with daily new IS cases.
Figure 2. Spearman analysis of the correlation between daily new case numbers and meteorological-pollution parameters. * indicates a statistically significant difference: *p<0.05, **p<0.01, ***p<0.001. AQI, air quality index; PM, particulate matter.
A DLNM was used to quantify the lagged exposure-response relationship between 24-hour temperature decrease and daily IS incidence. Figure 3 and online supplemental table 3 simultaneously illustrate the exposure-response (24-hour temperature decrease vs IS risk) and lag-response (lag days vs IS risk) dimensions of temperature decrease-related IS risk. A statistically significant nonlinear association was observed between temperature changes at different lag days and daily IS incidence. This plot confirmed that the adverse effect of 24-hour temperature decrease on IS incidence was not immediate but exhibited a lagged pattern, with risk varying by both the magnitude of temperature decrease and the number of lag days. No significant adverse effects on daily IS incidence were observed on the day of temperature decrease (lag 0). Significant IS risk began to emerge at lag 2 days, and this risk intensified with increasing lag days— indicating a ‘dose-dependent’ lagged effect. The maximum RR of IS was observed when the 24-hour temperature decrease reached 9°C at lag 9 days (1.122, 95% CI: 0.443 to 2.838). Although the relative risk (RR) indicated an elevated risk, the 95% CI spanned 1, reflecting uncertainty in the peak risk estimate (likely due to sample size constraints at extreme temperature decrease levels). Notably, the risk showed an overall upward trend as the 24-hour temperature decrease intensified—consistent with the positive correlation observed between temperature decrease and IS incidence, and supporting the potential adverse effect of cold-related temperature fluctuations on IS risk.
Figure 3. Relative risk (RR) of ischaemic stroke (IS) incidence associated with a decrease in temperature variability within 24 hours, by magnitude and lag days. This figure illustrates the exposure-lag-response relationship, showing the RR of IS associated with a 24-hour temperature drop across different lag days. It provides a two-dimensional view of this relationship: as lag days increase, the stroke risk rises gradually (colour shifting from blue to red); meanwhile, for the 24-hour temperature drop (X-axis), a greater temperature decrease correlates with a higher stroke risk (redder colour). Notably, the highest risk (darker red area) occurs with an approximate 8–10℃ temperature drop at a 9 day lag.
Interaction patterns of winter air pollution-cold wave sequential events on ischaemic stroke incidence in elderly adults
A GAM was used to quantify the main effects of 24-hour temperature decrease and AQI on daily IS incidence in elderly adults, as well as their interactive effect. The GAM results confirmed that both 24-hour temperature decrease and AQI had statistically significant main effects on daily IS incidence with the following patterns: the exposure-response curve from the GAM further demonstrated that the adverse effect of AQI on IS incidence intensified with increasing AQI values, peaking when AQI reached approximately 150 (figure 4)—a level classified as ‘moderate pollution’ per China’s national air quality standards. Consistent with prior DLNM findings, the 24-hour temperature decrease exhibited a significant positive main effect on IS incidence. The GAM also identified a statistically significant interactive effect between 24-hour temperature decrease and AQI on daily IS incidence, and the magnitude of this interaction effect was greater than the main effect of 24-hour temperature decreases alone. This interaction pattern indicates that ambient air pollution exacerbates the cold-related IS risk in elderly adults, highlighting the synergistic harm of combined exposure to temperature drops and poor air quality during winter sequential events.
Figure 4. Joint Effects of decrease in temperature variability within 24 hours and air quality index (AQI) on ischaemic stroke (IS) incidence. (A) This figure depicts the interaction between the decrease in temperature variability within 24 hours and AQI on IS incidence, derived from the GAM. The plot shows that the risk is highest when a large temperature drop (>8°C) coincides with moderately high pollution levels (AQI around 150). This visualises the synergistic effect where the combined exposure leads to a higher risk than the sum of the individual effects. (B) The impact of temperature decrease magnitude is depicted by a curve demonstrating an approximately linear increase in IS risk corresponding to greater 24-hour temperature reductions. (C) The influence of air pollution, measured by AQI, is characterised by a curve showing an initial rise in risk with increasing AQI values, reaching a peak near 150, followed by a decline.
Discussion
This study used a regional clinical database (Tianjin Medical and Health Big Data Platform) to investigate the impact of winter air pollution-cold wave sequential events on first-onset IS in adults aged >=60 years. Our key findings included: (1) a lagged effect of sequential events on IS incidence, with peak risk at lag 9 days; and (2) a positive additive interaction between air pollution and cold waves, amplifying IS risk in the elderly. To our knowledge, this is the first study to specifically quantify the association between winter air pollution-cold wave sequential events and IS incidence in an elderly population, filling a gap in understanding of combined environmental exposures in this vulnerable group.
Prior studies18 20 have defined air pollution-cold wave ‘sequential events’ using percentile thresholds: for example, daily average PM2.5 exceeding the 95th percentile of the national average for at least 2 days followed by daily average temperature drop lower than the national fifth percentile for at least 2 days. In contrast, our definition focused on persistent air pollution (AQI >100 for >=3 days) coupled with notable temperature fluctuations (24-hour temperature drop>=8°C, or sustained 3–7°C decline over 5 days)—a framework more tailored to Tianjin’s winter meteorological characteristics (eg, frequent stagnant pollution episodes followed by cold wave-induced dispersion). A previous work18 reported that air pollution-cold wave sequential events typically last 1.0–9.7 days. We set a >=7-day event period to ensure sufficient exposure dosage. Eight winter air pollution-cold wave sequential events were identified, with seven lasting 7 days and one lasting 9 days. Our analysis focuses on the average daily number of IS cases; variations in the duration of event periods (either 7 days or 9 days) do not compromise the assessment of IS incidence. Correspondingly, the duration of the lag period and non-event period for each event was consistently set to match the duration of its respective event period. Notably, most prior studies employed 7-day lag periods (with peak IS risk detected on Day 7) and did not explore risks beyond this timeframe. In contrast, our study investigated lag periods ranging from 7 to 9 days, and via the DLNM, we identified the peak IS risk on Day 9 post-exposure (RR=1.122) at lag 9. To be noticed, the wide range of 95% CI is likely attributable to limited extreme temperature drop events and insufficient sample size for certain patterns of exposures, reducing the precision of the risk estimate.
Consistent with prior studies on environmental co-exposures,20 21 our results suggested a synergistic interaction between air pollution and cold waves in increasing IS risk. A study on winter PM2.5 dynamics in Eastern China found that haze episodes (high PM2.5) are often followed by cold waves, during which PM2.5 concentrations first rise (stagnation) then decline (dispersion)—highlighting the temporal coupling of these two factors.23 Our GAM analysis built on this observation by quantifying the interactive effect: the adverse impact of 24-hour temperature decreases on IS incidence was stronger under moderate-to-severe pollution than under good air quality. This aligns with reports that combined environmental stressors (eg, pollution+cold) exert greater health harm than the sum of individual exposures, emphasising the need to prioritise co-exposure risk in epidemiological studies.
We observed distinct patterns of independent associations for low temperature and air pollution with IS incidence. A linear positive correlation existed between 24-hour temperature decrease and daily IS incidence—consistent with the biological mechanism of cold-induced vasoconstriction and elevated blood pressure.724,26 Larger temperature drops were associated with higher risk, reflecting the cumulative physiological stress of rapid cold exposure. Air pollution was positively correlated with IS incidence that peaked at AQI of about 150 (moderate pollution) and declined thereafter. This ‘saturation’ or risk reduction may be explained by two factors27: (1) cold waves often disperse pollutants, but when pre-event pollution is extremely high, residual warmth from stagnant air masses may modestly buffer cold-related IS risk; (2) elevated pollution levels prompt elderly individuals to reduce outdoor activities and remain indoors, which simultaneously decreases exposure to both cold air and ambient pollutants—creating a potential less harmful influence on IS onset. These patterns highlight the complexity of environmental risk: individual factors (temperature, pollution) do not act in isolation, and behavioural adaptations might modulate their impact.
Furthermore, our subgroup analysis revealed a significant increase in the proportional incidence of the >=80 years group during events, which directly confirms that older-old adults are more sensitive to event-related exposures—a key finding with important mechanistic and public health implications. High susceptibility to the combined effects of air pollution and cold waves among adults aged >=80 years is consistent with age-related declines in physiological reserve (eg, impaired thermoregulation, reduced endothelial function and comorbidity burden).28 This population is less able to compensate for the synergistic stress of pollution and cold, making them the primary target for preventive interventions. Meanwhile, decreased proportional incidence of the (65,70) years group indicates that the ‘contribution’ of such group to the total incident population was notably reduced during events. Given that such a group, though elderly, retains relatively intact physiological functions (eg, cardiovascular regulation, body temperature maintenance and immunity) and proactive risk mitigation (eg, improved adherence to antiplatelet/antihypertensive medications, intentional avoidance of outdoor exposure) compared with older age groups,29 it exhibits stronger tolerance to event-related exposures (eg, low temperature, air pollution). Consequently, the growth rate of incident cases in this group was lower than that in older-old subgroups during events, leading to a ‘dilution’ of its proportional incidence. These findings emphasise the need for age-tailored public health strategies, for example, targeted outreach for adults >=80 years (eg, home temperature monitoring and medication reminders) and scaling of effective preventive behaviours (eg, indoor activity promotion).
While air pollution and cold waves have distinct physical properties, their mechanisms of inducing IS are complementary and likely synergistic,14 providing a biological basis for our observed interactive effects. Cold exposure triggers peripheral vasoconstriction, increases blood pressure and enhances platelet aggregation—creating a pro-thrombotic state that promotes IS.724,26 Cold-induced vasoconstriction may retard the clearance of inhaled pollutants, prolonging their residence time and increasing systemic absorption.30 Meanwhile, ambient pollutants (eg, PM2.5, NO2) induce systemic inflammation, oxidative stress and endothelial dysfunction.1031,33 This pre-existing inflammatory state ‘sensitises’ the vasculature to cold stress, amplifying vasoconstrictive responses and increasing the risk of plaque rupture or thrombus formation.34 For the elderly, these synergistic mechanisms are particularly detrimental: age-related declines in antioxidant capacity and vascular elasticity further exacerbate inflammation and vasoconstriction, creating a ‘double burden’ of environmental stress and reduced physiological resilience.28 Our DLNM analysis identified a lagged IS risk peak at lag 9 days. This long lag window suggests that the combined stress of air pollution and cold waves may induce cumulative physiological damage (eg, sustained inflammation and endothelial dysfunction) that manifests in delayed IS onset. This mechanistic framework explains why sequential events pose a greater IS risk than either exposure alone.
Our findings have direct implications for optimising IS prevention in elderly populations during winter. Given the 9 days of sustained lag risk, public health risk warning alerts should cover not only the sequential event period but also the subsequent 2 weeks, especially for very elderly adults with lower adaptability and tolerance during these events.29 For regions with frequent winter air pollution-cold wave events (eg, North China), interventions could include home heating subsidies to maintain indoor temperatures (reducing cold exposure), air purification support in elderly care facilities (mitigating indoor pollution) and medication adherence campaigns during high-risk periods. In terms of policy alignment, air quality policies (eg, limiting winter industrial emissions) and cold wave response plans should be integrated, as separate strategies may underestimate combined risk.
Limitations
Our study has several limitations. First, as an observational study, this research cannot establish causal relationships between winter air pollution-cold wave sequential events and IS incidence. Second, IS cases were identified solely based on the ICD-10 code I63 of diagnosis record. IS incidence analysis was restricted to first-onset episodes because for patients with prior IS, their subsequent medical visits still carry I63 codes, but these codes cannot distinguish ‘recurrent’ IS events—making it impossible to accurately capture recurrent cases and introducing measurement bias in assessing the timing of recurrent onset. Additionally, potential coding omissions may further underestimate overall IS incidence. However, such limitations primarily only underestimate the impact of winter air pollution-cold wave sequential events on IS onset and do not alter the study’s overall conclusions. Third, the absence of stroke death before hospital admission may introduce potential exposure measurement errors, and this type of non-differential measurement error typically biases results toward the null (underestimating pollution-related effects) without affecting the study’s core conclusions. Fourth, despite adjusting for confounders such as age, sex, comorbidities and medication use, our analysis lacked individual data on socioeconomic status and lifestyle factors and used regional-level air pollutant data (not individual actual exposure doses), introducing potential residual confounding and measurement bias. Finally, the analysis was restricted to Tianjin. The findings might be only suitable for cities that share similar climatic features with Tianjin, including strong continental influences, severe winter pollution and frequent cold waves. However, the results cannot be directly generalised to regions with distinct climates (eg, mild winters in southern China) or pollution profiles (eg, areas with higher SO2 concentrations), and caution is needed when extrapolating specific risk estimates to such regions.
Conclusion
Winter air pollution-cold wave sequential events exert a synergistic, lagged effect on IS incidence in the elderly, with adults >=80 years being the most vulnerable subgroup. The observed risk patterns and underlying mechanisms underscore the importance of integrated environmental and public health strategies—including early warning, targeted home heating subsidies, air purification support in elderly care facilities, medication adherence campaigns for high-risk groups and alignment of air quality policies with cold wave response plans—to reduce IS burden in this high-risk population.
Supplementary material
Acknowledgements
The authors thank Tianjin Health Medical Big Data Company for its support on construction and data management of regional stroke database; they thank Qingyue Data for their efforts in organising meteorological and environmental data from the National Meteorological Science Data Center (http://data.cma.cn/) and thank Peking University, Palan DataRx Company for providing method and technical support.
Footnotes
Funding: The Natural Science Foundation of Tianjin (21JCQNJC01560); Tianjin Nankai District Traditional Chinese Medicine Heritage Innovation and Development Demonstration Pilot Project (20240204009); Tianjin High-Level Talent Training Project in Health Industry (TJSQNYXXR-D2-098); Young Talents Cultivation Program of China Association for Acupuncture and Moxibustion (2024-2026ZGZJXH-QNRC004); China Meteorological Administration - Nankai University Atmospheric Environment and Health Research Joint Laboratory Open Fund Project (CMANKU202204); Science Project of Hebei Provincial Administration of Traditional Chinese Medicine (T2026087, T2026097).
Prepublication history and additional supplemental material for this paper are available online. To view these files, please visit the journal online (https://doi.org/10.1136/bmjopen-2024-096297).
Provenance and peer review: Not commissioned; externally peer reviewed.
Patient consent for publication: Not applicable.
Data availability free text: The data used in this study include clinical data of elderly patients with ischemic stroke sourced from the Tianjin Medical and Health Big Data Platform, as well as meteorological and air quality data from the National Meteorological Science Data Center. Due to strict data security and privacy protection requirements for healthcare data, the clinical data cannot be publicly shared. For researchers with legitimate purposes can contact the corresponding author and access the relevant data within the designated secure environment of the platform. The daily meteorological and air quality data are publicly available via the National Meteorological Science Data Center (https://data.cma.cn/).
Patient and public involvement: During the design and implementation of this study, patients or members of the public were not directly involved. However, following the publication of the study findings, we will actively disseminate the results to the public through multiple accessible channels (official research institutions’ websites, official WeChat public accounts-a widely used social media platform in China, and health science popularization lectures) to help the public, especially elderly individuals and populations at high risk of IS, understand the association between winter air pollution-cold wave sequential events and IS incidence.
Author note: Li LI, Quanxi GE, Sihan SUN and Sha YANG contributed equally, and share the first authorship.
Data availability statement
Data may be obtained from a third party and are not publicly available.
References
- 1.Feigin VL, Abate MD, Abate YH, et al. Global, regional, and national burden of stroke and its risk factors, 1990–2021: a systematic analysis for the Global Burden of Disease Study 2021. Lancet Neurol. 2024;23:973–1003. doi: 10.1016/S1474-4422(24)00369-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Steinmetz JD, Seeher KM, Schiess N, et al. Global, regional, and national burden of disorders affecting the nervous system, 1990–2021: a systematic analysis for the Global Burden of Disease Study 2021. Lancet Neurol. 2024;23:344–81. doi: 10.1016/S1474-4422(24)00038-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Hu W, Chen Y, Chen J. Short-term effect of fine particular matter on daily hospitalizations for ischemic stroke: A time-series study in Yancheng, China. Ecotoxicol Environ Saf. 2021;208:111518. doi: 10.1016/j.ecoenv.2020.111518. [DOI] [PubMed] [Google Scholar]
- 4.Ponjoan A, Blanch J, Alves-Cabratosa L, et al. Extreme diurnal temperature range and cardiovascular emergency hospitalisations in a Mediterranean region. Occup Environ Med. 2021;78:62–8. doi: 10.1136/oemed-2019-106245. [DOI] [PubMed] [Google Scholar]
- 5.Chen Z, Liu P, Xia X, et al. Temperature variability increases the onset risk of ischemic stroke: A 10-year study in Tianjin, China. Front Neurol. 2023;14:1155987. doi: 10.3389/fneur.2023.1155987. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Li G, Guo Q, Liu Y, et al. Projected Temperature-Related Years of Life Lost From Stroke Due To Global Warming in a Temperate Climate City, Asia. Stroke. 2018;49:828–34. doi: 10.1161/STROKEAHA.117.020042. [DOI] [PubMed] [Google Scholar]
- 7.Luo Y, Li H, Huang F, et al. The cold effect of ambient temperature on ischemic and hemorrhagic stroke hospital admissions: A large database study in Beijing, China between years 2013 and 2014—Utilizing a distributed lag non-linear analysis. Environ Pollut. 2018;232:90–6. doi: 10.1016/j.envpol.2017.09.021. [DOI] [PubMed] [Google Scholar]
- 8.Alahmad B, Khraishah H, Kamineni M, et al. Extreme Temperatures and Stroke Mortality: Evidence From a Multi-Country Analysis. Stroke. 2024;55:1847–56. doi: 10.1161/STROKEAHA.123.045751. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Zhou L, Wang Y, Wang Q, et al. The interactive effects of extreme temperatures and PM2.5 pollution on mortalities in Jiangsu Province, China. Sci Rep. 2023;13:9479. doi: 10.1038/s41598-023-36635-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Lv X, Shi W, Yuan K, et al. Hourly Air Pollution Exposure and Emergency Hospital Admissions for Stroke: A Multicenter Case-Crossover Study. Stroke. 2023;54:3038–45. doi: 10.1161/STROKEAHA.123.044191. [DOI] [PubMed] [Google Scholar]
- 11.Qiu H, Sun S, Tsang H, et al. Fine particulate matter exposure and incidence of stroke: A cohort study in Hong Kong. Neurology (ECronicon) 2017;88:1709–17. doi: 10.1212/WNL.0000000000003903. [DOI] [PubMed] [Google Scholar]
- 12.Bryan L, Landrigan P. PM2.5 pollution in Texas: a geospatial analysis of health impact functions. Front Public Health. 2023;11:1286755. doi: 10.3389/fpubh.2023.1286755. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Huang K, Jia J, Liang F, et al. Fine Particulate Matter Exposure, Genetic Susceptibility, and the Risk of Incident Stroke: A Prospective Cohort Study. Stroke. 2024;55:92–100. doi: 10.1161/STROKEAHA.123.043812. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Tian Y, Xiang M, Peng J, et al. Modification effects of seasonal and temperature variation on the association between exposure to nitrogen dioxide and ischemic stroke onset in Shenzhen, China. Int J Biometeorol. 2022;66:1747–58. doi: 10.1007/s00484-022-02315-0. [DOI] [PubMed] [Google Scholar]
- 15.Danesh Yazdi M, Wei Y, Di Q, et al. The effect of long-term exposure to air pollution and seasonal temperature on hospital admissions with cardiovascular and respiratory disease in the United States: A difference-in-differences analysis. Science of The Total Environment. 2022;843:156855. doi: 10.1016/j.scitotenv.2022.156855. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Shao M, Dai Q, Yu Z, et al. Responses in PM2.5 and its chemical components to typical unfavorable meteorological events in the suburban area of Tianjin, China. Sci Total Environ. 2021;788:147814. doi: 10.1016/j.scitotenv.2021.147814. [DOI] [PubMed] [Google Scholar]
- 17.Maji KJ, Ye W-F, Arora M, et al. PM2.5-related health and economic loss assessment for 338 Chinese cities. Environ Int. 2018;121:392–403. doi: 10.1016/j.envint.2018.09.024. [DOI] [PubMed] [Google Scholar]
- 18.Huang Y, Wang P, Yang Z, et al. Spatiotemporal characteristics and influencing factors for joint events of air pollution wave and cold wave in China. Environ Int. 2024;184:108475. doi: 10.1016/j.envint.2024.108475. [DOI] [PubMed] [Google Scholar]
- 19.Li Q, Zhang R, Wang Y. Interannual variation of the wintertime fog-haze days across central and eastern China and its relation with East Asian winter monsoon. Int J Climatol. 2016;36:346–54. doi: 10.1002/joc.4350. [DOI] [Google Scholar]
- 20.Huang Y, Wang Y, Zhang T, et al. Exploring Health Effects under Specific Causes of Mortality Based on 90 Definitions of PM 2.5 and Cold Spell Combined Exposure in Shanghai, China. Environ Sci Technol. 2023;57:2423–34. doi: 10.1021/acs.est.2c06461. [DOI] [PubMed] [Google Scholar]
- 21.Li H, Li M, Zhang S, et al. Interactive effects of cold spell and air pollution on outpatient visits for anxiety in three subtropical Chinese cities. Science of The Total Environment. 2022;817:152789. doi: 10.1016/j.scitotenv.2021.152789. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Gasparrini A, Armstrong B, Kenward MG. Distributed lag non-linear models. Stat Med. 2010;29:2224–34. doi: 10.1002/sim.3940. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Zhang S, Zeng G, Yang X, et al. Two leading patterns of winter PM2.5 variations in eastern China before the outbreak of cold surge and their causes. Atmos Res. 2023;287:106696. doi: 10.1016/j.atmosres.2023.106696. [DOI] [Google Scholar]
- 24.Wolf K, Schneider A, Breitner S, et al. Air temperature and the occurrence of myocardial infarction in Augsburg, Germany. Circulation. 2009;120:735–42. doi: 10.1161/CIRCULATIONAHA.108.815860. [DOI] [PubMed] [Google Scholar]
- 25.Schneider A, Panagiotakos D, Picciotto S, et al. Air Temperature and Inflammatory Responses in Myocardial Infarction Survivors. Epidemiology (Sunnyvale) 2008;19:391–400. doi: 10.1097/EDE.0b013e31816a4325. [DOI] [PubMed] [Google Scholar]
- 26.Zhou J, Xu J, Geng F, et al. Childhood co-exposure of cold stress and PM2.5 aggravates the susceptibility and severity of asthma in adulthood of mice. Environ Toxicol. 2021;36:177–84. doi: 10.1002/tox.23023. [DOI] [PubMed] [Google Scholar]
- 27.Yang C, Lei L, Li Y, et al. Bidirectional modification effects on nonlinear associations of summer temperature and air pollution with first-ever stroke morbidity. Ecotoxicol Environ Saf. 2024;272:116034. doi: 10.1016/j.ecoenv.2024.116034. [DOI] [PubMed] [Google Scholar]
- 28.Swieczkowski M, Lip GY, Kurasz A, et al. Air pollution as a trigger for ischemic stroke - retrospective analysis of over 80-million-person years of follow-up (EP-PARTICLES study) Eur J Prev Cardiol. 2024;31 doi: 10.1093/eurjpc/zwae175.126. [DOI] [Google Scholar]
- 29.Brook RD, Rajagopalan S, Pope CA, 3rd, et al. Particulate matter air pollution and cardiovascular disease: An update to the scientific statement from the American Heart Association. Circulation. 2010;121:2331–78. doi: 10.1161/CIR.0b013e3181dbece1. [DOI] [PubMed] [Google Scholar]
- 30.Koo GPY, Zheng H, Aik JCL, et al. Clustering of Environmental Parameters and the Risk of Acute Ischaemic Stroke. Int J Environ Res Public Health. 2023;20:4979. doi: 10.3390/ijerph20064979. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Pieters N, Plusquin M, Cox B, et al. An epidemiological appraisal of the association between heart rate variability and particulate air pollution: a meta-analysis. Heart. 2012;98:1127–35. doi: 10.1136/heartjnl-2011-301505. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Bang OY, Ovbiagele B, Kim JS. Nontraditional Risk Factors for Ischemic Stroke. Stroke. 2015;46:3571–8. doi: 10.1161/STROKEAHA.115.010954. [DOI] [PubMed] [Google Scholar]
- 33.Guo L, Li B, Miao J-J, et al. Seasonal variation in air particulate matter (PM10) exposure-induced ischemia-like injuries in the rat brain. Chem Res Toxicol. 2015;28:431–9. doi: 10.1021/tx500392n. [DOI] [PubMed] [Google Scholar]
- 34.Kulick ER, Kaufman JD, Sack C. Ambient Air Pollution and Stroke: An Updated Review. Stroke. 2023;54:882–93. doi: 10.1161/STROKEAHA.122.035498. [DOI] [PMC free article] [PubMed] [Google Scholar]