Ambient Temperature and Stroke Risk Among Adults Aged 18–64 Years

Stroke mortality rates increased among younger adults from 2010 to 2020,¹ among whom known cardiovascular risk factors (eg, hypertension) are becoming more prevalent, leading to a heavier burden.² Stroke can have disproportionate effects on younger adults because of unfavorable prognosis and deteriorated life quality (eg, physical disability).

Environmental factors, including ambient temperature, are not covered by stroke prevention guidelines for younger adults, possibly because of limited evidence. The few studies for younger adults typically included a small number of individuals and generated heterogeneous results.

In this study, we focused on younger adults and evaluated short-term associations between temperature and stroke in the contiguous United States.

METHODS

We used a time-stratified case-crossover design by selecting the control date on the same day of the week as the case within the same calendar month, to control for day of week, seasonality, and long-term trends, in addition to time-invariant confounders.

We obtained the cohort data (2016–2022) from the Blue Cross Blue Shield (BCBS) Association—the largest U.S. commercial insurance provider. We identified individuals aged 18 to 64 years with the following characteristics: 1) they had inpatient/outpatient stroke claims (International Statistical Classification of Diseases and Related Health Problems, 10th Revision codes: I60-I69); 2) they were covered by BCBS continuously in the month of stroke and the prior month; and 3) they had no stroke claims in the 28 days before the current stroke. We extracted residential zip code and assigned Social Deprivation Index (SDI) and exposure values based on zip code. We calculated zip code-level population-weighted daily mean temperature and dewpoint temperature based on data from LandScan Global and Daymet (version 4).

Considering spatial heterogeneity of temperature associations, we classified the contiguous United States into 9 climate regions³ and used 2-stage models: 1) for each region, we regressed case/control status on moving averages of individually assigned zip code-level temperature at lag 0 to 6 days (natural cubic splines, with knots at climate region-specific 25th, 50th, and 75th percentiles) using conditional logistic regression; and 2) we pooled region-specific exposure-response curves using random-effects meta-analysis, and determined minimum morbidity temperature (MMT) as the temperature with the minimum estimated OR between the 1st and 99th percentile of temperature.

We conducted stratified analyses by stroke subtype (hemorrhagic [I60-I62], ischemic [I63-I64]), age, sex, race-ethnicity, and SDI tertiles (with the first tertile representing the lowest deprivation). Race and ethnicity were imputed by BCBS and were available for approximately one-third of beneficiaries, so we considered stratified analysis thereof exploratory. We evaluated heterogeneity by Q-statistics.

We conducted sensitivity analyses as follows: 1) placing knots of temperature splines at the 10th, 75th, and 90th percentiles; 2) using different lagged periods of lag 0 to 3 or 0 to 13 days; and 3) using claims with vs without known race and ethnicity. This study was exempt from Institutional Review Board approval.

RESULTS

We identified 1,051,267 individuals with stroke, of whom 56.9% were aged 55 to 64 years and 49.9% were female.

For the pooled exposure-response curve for the entire population, we found that the MMT was the first percentile, indicating no cold associations. The OR gradually increased until approximately the 60th percentile and remained relatively stable beyond (Figure 1). The OR at the 95th percentile (reference: MMT) was 1.30 (95% CI: 1.11–1.51). We observed significant heterogeneity across regions (P < 0.001). The climate-region-specific patterns for Northwest, Upper Midwest, and Northern Rockies and Plains regions were top-bottom flipped U-shaped, whereas the other regions’ curves were roughly similar to the pooled curve.

FIGURE 1. Exposure-Response Curves Between Daily Mean Temperature at Lag 0 to 6 Days and Risk of Stroke.

Exposure-response curves between daily mean temperature at lag 0 to 6 days and risk of stroke among the entire population (A) and by stroke type, age, sex, race, ethnicity, and Social Deprivation Index (B). The curves used the first percentile of temperature as the reference temperature, which was the minimum morbidity temperature of the pooled curve among the entire population. The 2 vertical dashed lines represented the 1st and 99th percentiles of temperature. Climate regions were ordered by average of daily mean temperature (2016–2022) from high to low.

The exposure-response curves by stroke subtype and sociodemographic characteristics had similar patterns to the overall curve, except for Hispanic (Figure 1). Heat associations were stronger for ischemic strokes and among those aged 18 to 54 years, female, non-Hispanic Black individuals, and the most socially-deprived tertile. However, heterogeneity was only significant for sex (P = 0.048). The OR at the 95th percentile was 1.42 (95% CI: 1.16–1.73) among females (male: 1.23; 95% CI: 1.03–1.46).

Our estimates were robust in sensitivity analyses.

DISCUSSION

Short-term exposure to high temperature was associated with an increased stroke risk in adults aged 18 to 64 years in the contiguous United States with females being more vulnerable. Our study is novel in its focus on younger adults, who are under-represented in existing studies without age restriction.⁴ These studies often suggested a cold association, and the findings on high temperatures were inconsistent. However, given that stroke occurred mostly in older adults, these findings were more representative of older adults. Other studies evaluating younger populations reported inconsistent findings. A Canadian study and an Israeli study observed heat associations with stroke morbidity among individuals aged <65 years and 50 to 74 years, respectively.^5,6 However, there are several studies showing cold-related (but not heat-related) risk or no associations in younger adults.⁷ Differences in study design and model specification preclude detailed comparisons.

Biologically, both cold and heat may increase stroke risk. Cold exposures may cause plaque formation and rupture.⁴ Heat exposures could lead to hyper-viscous blood via dehydration and stimulate the blood coagulation system, increasing the risk of formation of blood clots.⁴ There are potential mechanisms for differential associations by age. First, older adults are vulnerable to cold because of degenerative cerebral circulation functions, body temperature regulation, and immune functions. Second, profiles of risk factors (eg, hypertension, which might be temperature-related) and stroke etiology (eg, cardioembolism) are different by age.⁸ Third, individual temperature exposures and physical heat/cold stress could be different by age because of differential time-activity patterns and physical activity levels.

Heterogeneity by sex among younger adults is possible because of differential physiological conditions and risk factors (eg, factors unique to younger females like pregnancy).⁹ It is worth noting that the sex-specific findings of previous studies had no age restriction, and they were more representative of older adults. Considering the observed marginally significant heterogeneity by sex, our findings should be interpreted cautiously.

STUDY LIMITATIONS.

We had no information on individual-level socioeconomic status, the effect of which might be different from zip code-level SDI. We did not evaluate stroke severity (eg, fatal vs nonfatal) because of information unavailability. Furthermore, our findings were based on commercially insured individuals, and so they may not be generalizable to uninsured or publicly insured populations.

CONCLUSIONS

The curves highlighted risk from both extreme and moderate heat, which should be considered in early warning systems and health care management. Furthermore, potential heterogeneous vulnerability informs targeted stroke prevention and management strategies in females.

What is the clinical question being addressed?

Is short-term exposure to high ambient temperature associated with an increased risk of stroke among younger adults aged 18 to 64 years, a usually overlooked population?

What is the main finding?

The association varied by sociodemographic factors, with females appearing to be more vulnerable.