Heat Exposure Following the Rainy Season Is Associated With an Increased Risk of Cardiovascular Emergency Among the Elderly in Japan

Abstract

Background

Despite the impact of heat exposure caused by global warming, few studies have investigated the hourly effects of heat exposure and the risk of cardiovascular disease (CVD) in the elderly. We examined the associations between short‐term heat exposure and the risk of CVD in the elderly in Japan and evaluated possible effect‐measure modifications by rainy seasons that occur in East Asia.

Methods and Results

We conducted a time‐stratified case–crossover study. The study included 6527 residents in Okayama City, Japan, aged ≥65 years who were transported to emergency hospitals between 2012 and 2019 for the onset of CVD during and a few months after the rainy seasons. We examined the linear associations between temperature and CVD‐related emergency calls for each year and for hourly preceding intervals before the emergency call during the most relevant months. Heat exposure during 1 month after the end of the rainy season was associated with CVD risk; the odds ratio (OR) for a 1° C increase in temperature was 1.34 (95% CI, 1.29–1.40). When we further explored the nonlinear association by using the natural cubic spline model, we found a J‐shaped relationship. Exposures 0 to 6 hours before the case event (preceding intervals 0–6 hours) were associated with CVD risk, particularly for the preceding interval 0 to 1 hour (OR, 1.33 [95% CI, 1.28–1.39]). For longer periods, the highest risk was at preceding intervals 0 to 23 hours (OR, 1.40 [95% CI, 1.34–1.46]).

Conclusions

Elderly individuals may be more susceptible to CVD after heat exposure during the month after the rainy season. As shown by finer temporal resolution analyses, short‐term exposure to increasing temperature can trigger CVD onset.

Nonstandard Abbreviations and Acronyms

AIC

Akaike Information Criterion

JMA

Japan Meteorological Agency

PM_2.5

particulate matter <2.5 μm in diameter

Clinical Perspective.

What Is New?

• We examined the association between hourly heat exposure and the risk of cardiovascular disease in a time‐stratified case‐crossover study, evaluating potential effect‐measure modifications by the rainy season in Japan.

• Heat exposure during 1 month after the end of the rainy season was associated with a higher risk of cardiovascular emergency calls in the elderly population, with a J‐shaped relationship.

• The hourly time lag demonstrated that a higher risk was observed particularly in the preceding interval 0 to 1 hour, as well as 0 to 23 hours before the emergency call.

What Are the Clinical Implications?

• The emergency medical response and public health policy for short‐term heat exposure in the period after the rainy season should include heat mitigation measures, particularly for the elderly.

Climate change and global warming are some of the biggest issues in our society. ¹ , ² In the summer (June–August) of 2020, the seasonal global average surface temperature (ie, the average of the near‐surface air temperature over land and the sea surface temperature) was 0.41 °C higher than the 1981 to 2010 average and 0.75 °C higher than the 20th‐century average. ³ In East Asia, cold‐related mortality is projected to decrease from 7.4% to 8.7% in 2010 to 2019 to 3.7% to 5.9% in 2090 to 2099 as a result of intense warming, and heat‐related mortality is projected to increase moderately to 2.5% to 3.2%. ² In Japan, the annual mean temperature is on an upward trend at a rate of 1.30 °C/100 years, which is higher than the world annual mean temperature deviation (0.74 °C/100 years). ³

One potential adverse health effect of global warming is an increased incidence of cardiovascular disease (CVD), which is a major cause of hospital emergency visits and a leading cause of death worldwide. ⁴ Previous studies reported positive nonlinear associations between high ambient temperatures and CVD, ⁵ , ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ including myocardial infarction. ⁷ , ⁸ , ⁹ A recent study showed that high temperatures increased the risk of morbidity caused by arrhythmias and cardiac arrest (1%–6%). ¹³ Another study in the United States showed a higher risk of CVD‐related emergency department visits by elderly people after 5 to 6‐days with lagged extreme heat exposure  in the summer. ¹¹ The vulnerability of the elderly to heat exposure is mainly because of a combination of impaired thermoregulatory capacity to respond to heat and additional comorbidities. ¹⁵

While some previous studies have examined the effects of heat on CVD by focusing on maximum or mean daily temperature or long heat exposure periods, ⁶ , ⁸ , ¹¹ , ¹⁶ , ¹⁷ evidence on the nonlinear hourly effects of heat exposure on CVD risk in the elderly is sparse, ⁷ , ¹⁰ , ¹⁸ despite the growing concern for extreme heat waves. Additionally, as shown in meteorological findings, during the rainy season, there are some environmental factors that may affect health outcomes, including high temperatures, increased precipitation, high humidity, and low pressure observed with diurnal variation. ² , ⁵ , ¹⁹ , ²⁰ , ²¹ To further understand the effect of heat on CVD, the influence of these factors cannot be ignored, and flooding associated with precipitation would also influence CVD. ²² , ²³ , ²⁴ Southern‐central Japan and South Korea are affected by the East Asian monsoon during summer. As the rainy season ends, the number of extreme heat days increases up until late August by anomalous anticyclones throughout the troposphere over East Asia. ²⁵ , ²⁶ , ²⁷ The North Pacific High extends northwest, around Japan, bringing hot and sunny conditions after the end of the rainy season.^20, Because the transition is characterized by an increase in sunshine duration and temperature, and a decrease in precipitation, ²⁶ , ²⁷ it would be interesting to evaluate the possible adverse health effects of heat exposure during this season from a theoretical as well as a public health perspective. To our knowledge, however, no studies have addressed this important issue including a geo‐climatic approach.

Therefore, we aimed to examine the association between heat exposure and the risk of CVD in the elderly and to evaluate possible effect‐measure modifications by the rainy season. In so doing, we considered the effects of precipitation and sunshine duration by stratifying on rainy season. We used hourly meteorological data from the period spanning 2012 to 2019 in Okayama City, Japan.

METHODS

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

Study Design

Case‐Crossover Design During and a Few Months After the Rainy Season in Elderly

Our study was designed as a time‐stratified case‐crossover. The case‐crossover design is a variant of the case–control design and uses cases only; the design allows the elimination of time‐invariant confounders and seasonal and diurnal trends and provides an approach to studying the effects of exposure on the occurrence of acute events. ²⁸ In this design, an individual's exposure before the event is compared with their exposure at a period that is the same time, same day of the week, same month, and same year but when the disease did not occur (control time). A day on the same day of the week in the same month in which the outcome occurred was defined as the control. When there were multiple candidate control days on the event day of the week, the same days of the week within the same month were defined as control days. All control days were used in a 1:3 or 1:4 case‐to‐control ratio based on the length of the month, which were used depending on the date of case occurrence.

Study Area and Participants

The study subjects were selected from elderly residents in Okayama City, in the western part of Japan who had been transported to the emergency department by ambulance between January 2012 and December 2019 (before the COVID‐19 pandemic). The study region, located in the southern part of Okayama Prefecture, included both urban and rural communities with a population of 708 853 individuals (including 180 784 aged ≥65 years) in an area of 789.9 km² (in October 2017; Figure S1). Anonymized electronic data on all emergency ambulance calls in Okayama City during the study period were obtained from the Ambulance Division of the city's Fire Department. There were 159 506 sudden illness cases recorded in which residents had been transported by ambulance to emergency rooms during the study period. Previous studies have reported that elderly individuals have a greater risk of adverse cardiovascular effects after heat exposure, ⁵ , ⁶ , ⁸ , ¹¹ , ¹⁵ , ¹⁶ , ²⁹ and that elderly individuals are expected to be particularly sensitive to future high temperatures. ² , ³⁰ , ³¹ Therefore, in this study, we restricted participants to those aged ≥65 years who had been transported to the emergency department because of the onset of CVD.

This study was approved by the Institutional Review Committee of the Graduate School of Medicine, Dentistry and Pharmaceutical Sciences at Okayama University (No. 2003‐029). This was an observational study using administrative data; thus, the requirement for written informed consent was waived.

Meteorological Data and Air Pollution Data

We obtained hourly outdoor temperature, relative humidity, and barometric pressure data from the weather station in the city of Okayama, which is managed by the Japan Meteorological Agency (Figure S1), and there were no missing data during the study period. We also obtained hourly measurements of the mean atmospheric concentration of suspended ambient air particulate matter <2.5 μm in diameter (PM_2.5) from monitoring stations in the city of Okayama during the study period. The 3 monitoring stations for PM_2.5 measurements in Okayama provide coverage of the entire city as well as 20‐km buffers. These data were provided by the Okayama City Environmental Conservation Division from the Okayama Prefectural Government. We calculated city‐representative hourly average concentrations of PM_2.5 from hourly concentrations recorded at each monitoring station. There were no missing PM_2.5 data during the study period.

The Japanese rainy season is determined by the Japan Meteorological Agency (Table S1). The method for defining the end of the rainy season each year by the Japan Meteorological Agency are as follows: after 2 or more days of rain, the rainy season front is moving northward; the previous day and current day are sunny; and the weekly weather forecast is sunny (or partly cloudy) for at least 5 days. First, a preliminary report is released on the day when the rainy season was expected to end. The weather conditions during the period from May to August are comprehensively reviewed thereafter to determine a fixed date for the end of midsummer, around September. ²⁰ The fixed dates for the end of the rainy season vary each year. We adopted these dates and stratified them every month after the rainy season.

Health Outcomes

CVD type included in the data set from the Ambulance Division of the Okayama City Fire Department was diagnosed by the primary physician at the emergency hospital and then coded using the International Classification of Diseases, Tenth Revision (ICD‐10). We classified the following diseases as health outcomes: CVD (ICD‐10; I10–69), cardiac failure (I50), ischemic heart disease (I20–25), and arrhythmia (I44–49). Patients with prehospital out‐of‐hospital cardiac arrest (I46) were included in the arrhythmia (I44–49) classification. Prehospital deaths (such as cardiogenic out‐of‐hospital cardiac arrest) were not included in this study. Because the exact time of onset was unavailable, we used the time of the emergency call as the onset of the disease.

Statistical Analysis

Linear Model

We conducted conditional logistic regression to estimate odds ratios (ORs) and 95% CIs for every 1 °C increase in temperature for each health outcome throughout the study period. The meteorological data of mean barometric pressure, relative humidity, and PM_2.5 concentration were adjusted as covariates because exposure to PM_2.5 was associated with the risk of CVD‐related hospital admission in previous studies. ³² , ³³

Potential Effect Modifier as Stratified by the Rainy Seasons

To examine the possible effect modification by the rainy seasons, we conducted stratified analysis by the rainy season as follows: during the rainy season (reference), and the effects of 1 month, 2 months, and 3 months after the end of the rainy season. As mentioned above, we estimated ORs and 95% CIs for every 1 °C increase in temperature for each health outcome.

Analysis Separately for Each Year

We also examined the association between temperature and health outcomes separately for each year, because the rainy season was not constant every year, and there were unusually long periods of extremely hot days with temperatures exceeding 35 °C after the rainy season ended.

In particular, Okayama City experienced extensive flood damage during the 2018 rainy season (June 28–July 8), and there was a need to consider the impact of this flooding. ³⁴

Analysis of Hourly Time Preceding Intervals Before the Emergency Call

We further examined the association between temperature and CVD onset by applying hourly preceding time intervals before the emergency call occurred during the month after the end of the rainy season. In keeping with the hypothesis that the hourly temporal effect for the risk of CVD onset after heat exposure is more effective than when there is a delay of several days, we included finer time divisions for shorter preceding intervals in our models, including the following time periods: 0 to 1, 0 to 2, 0 to 3, 0 to 4, 0 to 5, and 0 to 6 hours; and 0 to 11, 0 to 23, and 24 to 47 hours. In this model, the relative humidity, barometric pressure, and PM_2.5 concentration were also adjusted at the same preceding intervals.

Nonlinear Model as Cubic Spline Curves

Finally, we used the natural cubic spline model for temperature and CVDs during the month after the end of the rainy season, because as shown in many temperature–CVD studies, the exposure–response curves are likely nonlinear (either U‐shape or J‐shape). ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹² , ¹⁴ For the spline model, we conducted degrees of freedom of 7 for mean daily temperature during and a few months after the rainy seasons. The meteorological data of mean barometric pressure, relative humidity, and PM_2.5 concentration were adjusted as covariates. The linearity of the association was examined by comparing the Akaike Information Criterion (AIC) between linear and nonlinear models and examining the shape of the plot. ⁷ , ¹⁰ , ¹⁴ , ³⁵ The AIC is a likelihood‐based model selection statistic with lower values indicating a better fit of the underlying data.

Sensitivity Analysis

First, sensitivity analysis was performed by additionally adjusting for national holidays. In the previous studies, national holidays were associated with higher risk of CVDs. ³⁶ Furthermore, when a nonlinear association will be found, the parameters in characterizing the nonlinear association between temperature and CVD were tested in the sensitivity analyses. Thus, we compared the AIC of the nonlinear model with the linear model. We calculated the difference of AIC between linear and nonlinear models (ΔAIC = AIC of spline model–AIC of linear model). We considered nonlinear associations with ΔAIC was negative and the curve was J or U shape. ⁷ , ³⁵ In addition, we constructed a spline curve with degrees of freedom of 4 for mean daily temperature during and a few months after the rainy seasons as a sensitivity analysis.

All analyses were performed using Stata SE, version 16.1 (Stata Corp, Texas). A 2‐sided P value of <0.05 was considered significant.

RESULTS

During the study period, a total of 34 639 CVD cases were diagnosed during emergency hospital visits in Okayama City, Japan. Table 1 shows the characteristics of CVD‐related emergency visits in a total of 6527 people aged ≥65 years. The mean age was 80.7 years, and cardiac failure and ischemic heart disease accounted for ≈20% and  ≈15% of CVD emergency visits, respectively. The duration of the rainy season varied annually and exceeded 1 month in some years (Table S1). The number of emergency hospital visits was highest during the rainy season (n=1895).

Table 1.

Type of Cardiovascular Emergency Visits of Adults Aged ≥65 Years During and After the Rainy Season in Okayama City, Japan (2012–2019)

	Total (n=6527)	During the rainy season (n=1895)	After the rainy season
			1 mo after (n=1556)	2 mo after (n=1485)	3 mo after (n=1591)
Individual‐level characteristics
Age, y, mean±SD	80.8±8.3	80.7±8.4	80.6±8.2	80.6±8.3	81.3±8.2
Sex, male, n (%)	3176 (48.7)	938 (49.5)	777 (49.9)	719 (48.4)	742 (46.6)
Type of emergency hospital visits, n (%)
Cardiac failure, ICD‐10: I50	1107 (17)	336 (17.7)	253 (16.3)	233 (15.7)	285 (17.9)
Ischemic heart disease, I20–25	946 (14.5)	269 (14.2)	227 (14.6)	227 (15.3)	223 (14)
Arrhythmia, I44–49	463 (7.1)	143 (7.6)	118 (7.6)	107 (7.2)	95 (6)
Atrial fibrillation and flutter, I48, n (%)	207 (3.2)	62 (3.3)	45 (2.9)	52 (3.5)	48 (3)
Hypertensive heart disease, I10–16	371 (5.7)	110 (5.8)	71 (4.6)	91 (6.1)	99 (6.2)
Pulmonary embolism, I26	32 (0.5)	8 (0.4)	12 (0.8)	5 (0.3)	7 (0.4)
Cerebrovascular disease, I60‐69	3367 (51.6)	962 (50.8)	816 (52.4)	762 (51.3)	827 (52)
Hemorrhagic stroke, I60, I61, I69.0, I69.1	835 (12.8)	211 (11.1)	179 (11.5)	203 (13.7)	242 (15.2)
Ischemic stroke, I63, I69.3	1723 (26.4)	508 (26.8)	444 (28.5)	374 (25.2)	397 (25)
Transient ischemic attack, G45.9	277 (4.2)	82 (4.3)	62 (4)	60 (4)	73 (4.6)

ICD‐10 indicates International Statistical Classification of Diseases, Tenth Revision.

Table 2 shows the meteorological data and concentration of atmospheric air pollution on the day of the CVD event. Mean temperature was highest during 1 month after the end of the rainy season, whereas the mean of both relative humidity and concentration of PM_2.5 were highest during the rainy season. As expected, mean barometric pressure was lowest during the rainy season. In Table S2, we show the results of the conditional logistic regression analysis for each health outcome. During the month after the end of rainy season, the odds per 1 °C increment in temperature were ≈30% to 40% higher for each health outcome. These associations were statistically significant. No positive associations were found between temperature and CVD outcomes during the rainy season. However, after the rainy season, some of the ORs were slightly higher than null. Similar patterns for stroke (cerebral infarction, cerebral hemorrhage, and transient ischemic attack) were found (Table S3). In Table 3, we show the results of the stratified analysis by the rainy season. When we analyzed the data separately for each year, similar patterns were observed (Figure 1); the ORs for CVD were significantly higher than null during the month after the end of rainy season in every year except for 2016, and the highest OR was observed for 2019 (1.70 [95% CI, 1.49–1.94]). Conversely, the results from other time periods fluctuated and no clear patterns were observed.

Table 2.

Meteorological Data on the Day of Cardiovascular Emergency Visits of Adults Aged ≥65 Years in Okayama City, Japan (2012–2019)

	Total (n=6527)	During the rainy season (n=1895)	After the rainy season
			1 mo after (n=1556)	2 mo after (n=1485)	3 mo after (n=1591)
Meteorological variable
Temperature, °C, mean±SD	16±9.1	24.9±3.4	29.6±3.1	27±3.7	21.9±3.9
Min	−6.1	13.2	22.0	15.6	10.9
Max	36.9	34.8	36.9	36.9	32.9
Relative humidity,  mean±SD	64.6±17.7	73.3±15.6	68.5±13.6	69.8±14.7	68.7±16.7
Min	12	22	24	30	21
Max	100	99	98	99	99
Barometric pressure, hPA, mean±SD	1014.3±7.2	1006.6±4.6	1007.2±4.3	1009.2±4.5	1014±5.7
Min	980.5	986.2	980.5	983.1	982.0
Max	1035.0	1018.1	1017.9	1021.1	1026.9
PM2.5, μg/m3, mean±SD	15.4±9.8	18.4±10.5	17±9.0	16.9±10.5	13.1±7.7
Min	−6.5	−0.7	−5.0	0.7	−1.0
Max	80.3	63.0	51.0	59.7	53.3

hPA indicates hectopascal; and PM_2.5, particulate matter <2.5 μm in diameter.

Table 3.

Adjusted Odds Ratios^* and 95% CIs for Cardiovascular Disease Onset per 1 °C in Increment Temperature Stratified by Rainy Seasons in Okayama City, Japan (2012–2019)

Subgroup	N with/without outcome	OR (95% CI); P	P for exposure within strata of the rainy seasons
During the rainy season	1895/4632	Reference
1 mo after	1556/4971	1.34 (1.3–1.39)	P<0.001
2 mo after	1485/5042	1.11 (1.08–1.15)	P<0.001
3 mo after	1591/4936	1.12 (1.09–1.15)	P<0.001

OR indicates odds ratio.

^*Adjusted for relative humidity, barometric pressure, and particulate matter <2.5 μm in diameter.

Figure 1. Adjusted odds ratios for cardiovascular disease onset per 1 °C increments of temperature during and after the rainy seasons among those aged ≥65 years, Okayama, Japan 2012 to 2019 by year.

Horizontal bars indicate 95% CIs.

Figure 2 shows the results of the association between temperature and CVD onset, by hourly preceding time intervals before the emergency call during the month after the end of the rainy season. Within the range of 0 to <6 hours (preceding interval 0–6), the OR was the highest <1 hour before CVD onset (1.33 [95% CI, 1.28–1.39]), and the associations were slightly attenuated for the 2 to 6 hours. In the longer intervals, ORs tended to be higher; the highest observed OR was 1.40 (95% CI, 1.34–1.46) for a time of <23 hours (preceding interval 0–23).

Figure 2. Adjusted odds ratios and 95% CIs for cardiovascular disease onset per 1 °C increments in temperature in the hours before the corresponding emergency call during the month after the rainy season in adults aged ≥65 years in Okayama, Japan (2012–2019).

Horizontal bars indicate 95% CIs.

In Figure 3, we present natural cubic spline curves for each period. During 1 month after the end of the rainy season, the nonlinear association between heat exposure and CVDs per 1 °C in increment temperature was J‐shaped, suggesting that the risk was even markedly higher above 30 °C. However, it was inversely U‐shaped during other periods during the rainy season and 3 months after. Rather, heat exposure may have potentially contributed to the suppression.

Figure 3. Odds ratios and their 95% CIs for cardiovascular disease onset per 1 °C increments in temperature during and after the rainy seasons among those aged ≥65 years, Okayama, Japan 2012 to 2019.

Splines have a single inferior knot; up to 7 inferior knots per spline were allowed. Reference temperature (odds ratio=1) was mean temperature. This nonlinear model was adjusted for relative humidity, barometric pressure, and particulate matter <2.5 μm in diameter. Solid lines represent odds ratios, and dashed lines represent 95% CI.

In sensitivity analysis, we additionally adjusted for national holidays as a covariate and found similar results as in the main analysis for the association between heat exposure and CVDs during the same periods (Table S4, Figure S2). Furthermore, in each period, because nonlinear associations were observed, we calculated the difference of AICs between linear and nonlinear models (ΔAIC). The AIC of the spline model was smaller than that of the linear model (Table S5). When we constructed a spline curve with degrees of freedom of 4 as a sensitivity analysis, we found similar patterns (Figure S3).

DISCUSSION

In this case‐crossover study, we examined the health effects of heat exposure during and several months after the rainy season in elderly residents in Okayama City, Japan. Stratification by the rainy season, which is a characteristic climatic occurrence in East Asia, ²⁰ , ²⁵ , ²⁶ , ²⁷ was performed for the assessment of the effects of heat exposure. To the best of our knowledge, this is the first study that examines associations between hourly heat exposure and the risk of CVD, and that evaluates potential effect‐measure modifications by the rainy season. Our results show that heat exposure during the month after the rainy season, which is defined annually by the Japan Meteorological Agency, ²⁰ was significantly associated with a higher risk of emergency calls for CVD in the elderly population. In the periods measured in this study, high temperatures and little precipitation were recorded with minimal annual deviations and were caused by the extension of the Pacific High in our study area. The additional examinations of the hourly preceding intervals demonstrated that a higher risk of CVD was especially observed during the 0‐to‐1 hour preceding interval, as well as 0 to 23 hours before the emergency call.

In this study, the association between temperature and CVDs decreased during the rainy season and 2 or 3 months after the end of the rainy season, with an inverted U shape (Figure 3). A careful interpretation is needed for the results of linear models because of possible model misclassification. Regarding the protective effect of heat exposure on CVD risk during the rainy season, a previous study from Japan reported that, among healthy elderly, the number of steps taken per day decreased exponentially with increasing precipitation, ³⁷ which may well explain why our subjects were less likely to be exposed to ambient heat temperature during the rainy season because they tended to stay indoors. By contrast, during 1 month after the end of the rainy season, we found a J‐shaped pattern, which is closer to a linear pattern. In a recent study from Germany, physical activity among the elderly increased with higher solar radiation and increased sunshine duration, ³⁸ and our subjects may more likely be exposed to heat without ensuring adequate heat acclimatization during 1 month after the end of the rainy season. In addition, even if the elderly people stay indoors on hot days, they may be at a high risk of developing indoor heatstroke because of the lack of or an inappropriate use of cooling appliances. This may partly explain the reason that an inverted U‐shape was not observed during 1 month after the end of the rainy season.

Furthermore, the elderly are reportedly prone to intravascular dehydration and CVD, such as myocardial infarction and paroxysmal atrial fibrillation, because of the effects of hemoconcentration, partly because they tend to have many comorbidities. ¹⁵ Regarding the effect of the 23‐hour delay in risk (Figure 2), CVD delay in risk may be influenced by these comorbidities and changes in living arrangements in the Japanese elderly population. Disability onset in older adults is associated with solitary and nonspouse households ³⁹ and moreover, it is often subclinical and asymptomatic. These factors tend to lead to inappropriate indoor temperature control and poor adherence to medication, which may delay detection of symptoms when they manifest, thus possibly explaining the preceding intervals that were observed before the onset of the disease. In the year‐specific analysis, we did not observe an increased CVD risk after the end of the rainy seasons only in 2016. Although the reason is unclear, heavy rain and lightning in Okayama City during the daytime of August 15 to 16, 2016 flooded roads, caused power outages, and halted public transportation, ⁴⁰ which may have affected CVD‐related emergency.

Our finding that the risk of CVD is higher 0 to 1 hour before onset is consistent with previous findings in other countries. For example, a previous study in England and Wales reported that the risk of myocardial infarction increased 1 to 6 hours after heat exposure and reduced at longer intervals, and these associations were only observed above the 20 °C threshold. ⁷ Another study in New York in the United States reported that when the temperature increased from 11 °C to 27 °C, the risk of myocardial infarction increased by 1.9% (95% CI, 1.2%–2.5%) in 1 hour. ¹⁰ Until now, there has been a lack of studies investigating this effect in Asia.

The risk of acute‐onset CVD following extreme heat exposure could be explained by the following pathophysiological perspectives. (1) Thermal stress with heat exposure leads to physiological responses that aim to change the central body temperature, which together lead to water loss and dehydration, increased skin blood flow, sweating, cardiac strain, elevated heart and breathing rate, vasodilatation, and increased or decreased coagulation. These changes can cause autonomic imbalances of the heart, increase local arterial pressure, induce systemic inflammatory response syndrome, and impair clotting responses. ⁴¹ (2) The hyperthermia‐impaired vascular endothelium induces occlusion of the arterioles and capillaries (microvascular thrombosis) or excessive bleeding (consumption coagulation), and multiorgan system failure, including CVD. ⁴² These disruptions to hemostasis potentially predispose vulnerable individuals to increased serum cholesterol levels, atherosclerotic plaque rupture, and subsequent myocardial infarctions. ⁹ , ⁴² (3) According to a seasonal analysis of acute coronary syndrome using optical coherence tomography in 6 countries including Japan (75% of the participants were Japanese), plaque erosion was more common than plaque rupture in summer, and thrombosis was caused by increased stress and erosion in blood vessels from hemoconcentration caused by dehydration at the site of local intimal damage. ⁴³ (4) Furthermore, patients who have heart failure with reduced ejection fraction might not be able to compensate for the increased circulating demand because of heat exposure. ³⁰ (5) The hourly increase in heat temperature had a significant lag effect on decreased blood pressure, with a lag time of 0 to 5 hours, ¹⁸ which may explain the increased risk of CVDs.

A primary strength of this study is the study design, which was a time‐stratified case‐crossover design that allowed us to control long‐term time‐trends, seasonality, day of the week, as well as unmeasured or even unknown individual‐level characteristics that do not vary within a month (eg, occupation, socioeconomic status, and pre‐existing CVD). ²⁸ In addition, because we accumulated both hourly meteorological and emergency call data, our findings will be useful for the implementation of potential and timely precautions or countermeasures for heat exposure. Although a few studies examined hourly associations between heat exposure and the risk of ischemic heart disease, ⁷ , ¹⁰ no studies to our knowledge have examined such an hourly association for the risk of CVD. Because temperature changes may rapidly cause circulatory failure, such as hypotension, ¹⁸ it would be significant to examine the effect of hourly temporal changes for CVD. Because the study period of the present study was before the COVID‐19 pandemic, our findings are not distorted by possible exposure misclassification because of the “stay home” campaign during the pandemic. ⁴⁴ Meanwhile, although rainfall is recognized as one of the mechanisms to reduce particulate pollution, we observed higher PM_2.5 concentration during the rainy season but lower following. The end of the rainy season is the summer possibly in the lowest PM_2.5 emissions. In addition, the decrease in PM_2.5 levels following rain lasted 3 to 6 days with the rainfall thresholds. ⁴⁵ The impact of rainfall washout may persist on a daily or weekly basis after the rainy season; therefore, we adjusted for hourly PM_2.5 as a confounding factor.

There are some limitations that must be considered. First, data on the exact time of onset of CVD were unavailable, and we used the emergency call time as a substitute. Prehospital delay for symptoms of potential acute coronary disease has been reported among the elderly, those lacking knowledge, those with gradual onset of symptoms, and those who are uninsured. ⁴⁶ , ⁴⁷ Thus, elderly subjects who live alone would be less likely to act on their symptoms, and a longer time lag could occur. However, emergency support for the elderly based on community‐based integrated care has been introduced in Japan, and Okayama City has established an emergency reporting system for the elderly living alone as a welfare service in the home. ⁴⁸ Furthermore, health care services are highly accessible because ambulance transport is free and the universal health insurance system covers all citizens. ⁴⁹ Thus, we believe that the lag between onset and emergency calls would be negligible. However, the onset of cardiac failure had a variety of time frames, and the median time was 2 hours from onset to arrival at the hospital via ambulance transport in the Tokyo metropolitan area. ⁵⁰ The type of onset in our study was presumed to be acute pulmonary edema; however, in cases of chronic exacerbation, median lags of >1 hour from the time of symptom onset to the emergency call may have occurred. In addition, a report from Shenzhen, China, which has a slightly more subtropical climate than our study region, stated that increased precipitation and temperature were associated with longer ambulance response times (measured in hours). ⁵¹ Therefore, it should be noted that analysis by hourly stratification may result in differential misclassification of outcomes and possible underestimation.

The second limitation is the possibility for misclassification of health outcomes of interest. Because the diagnosis process is standardized and performed by well‐trained medical personnel, even if there are disease names that are suspected to be incorrect, CVD diagnostic errors are thought to be few and obvious misclassification would be nondifferential. Because prehospital deaths were not included in this study, our findings may be underestimated. Third, information on indoor or personal temperature was unavailable, and we used ambient temperature instead. This could have introduced a random‐measurement error. However, because exposure misclassification would be nondifferential, we expect that it could have induced a bias towards the null. Fourth, we did not include patients who arrived at the hospital by private vehicle or walking; we might not be able to generalize our findings to all CVD‐related emergencies, because patients arriving at the hospital by these situations could also be triggered by heat exposure. Finally, although we adjusted for PM_2.5, neither inhalable PM₁₀ ³³ , ⁵² nor desert dust ⁵³ , ⁵⁴ were adjusted for in the analyses.

Despite several limitations associated with the use of an administrative data set, the present study suggests that heat exposure increases the risk of CVD in the elderly during the month after the end of the rainy season. In addition, short‐term exposure to increasing temperature, as shown by finer temporal resolution analyses, can trigger CVD onset. With the increasing number of extremely hot days occurring after the end of the rainy season and because co‐exposure to COVID‐19 and heat may increase the burden of disease in affected individuals, ³⁰ further research is needed. Our study results support that the emergency medical response and public health policy for short‐term heat exposure in the period after the rainy season should include heat mitigation measures, such as appropriate air conditioning, particularly for the elderly.

Sources of Funding

The authors received the following financial support for the research, authorship, and/or publication of this article: ES is supported by the Japan Society for the Promotion of Science (JSPS KAKENHI Grant Numbers JP20K10471, JP19KK0418, JP18K10104, and JP20K10499). SK is supported by JSPS KAKENHI Grant Number JP18K10104. The sponsor was not involved in the study design, collection, analysis, or interpretation of data, the writing of the report, or the decision to submit the paper for publication.

Disclosures

None.

Supporting information

Tables S1–S5

Figures S1–S3