Int J Biometeorol (2014) 58:941–948 DOI 10.1007/s00484-013-0677-4

ORIGINAL PAPER

Characterizing the effect of summer temperature on heatstroke-related emergency ambulance dispatches in the Kanto area of

Chris Fook Sheng Ng & Kayo Ueda & Masaji Ono & Hiroshi Nitta & Akinori Takami

Received: 25 March 2013 /Revised: 26 April 2013 /Accepted: 26 April 2013 /Published online: 23 May 2013 # ISB 2013

Abstract Despite rising concern on the impact of heat on with a combined relative risk (RR) of 1.22 (95 % confidence human health, the risk of high summer temperature on interval, 1.03–1.44), increasing to 1.49 (1.42–1.57) at peak heatstroke-related emergency dispatches is not well under- AT. When linear exposure was assumed, combined RR was stood in Japan. A time-series study was conducted to examine 1.43 (1.37–1.50) per degree Celsius increment. Overall asso- the association between apparent temperature and daily ciation was significant the first few times when median AT heatstroke-related ambulance dispatches (HSAD) within the was initially exceeded in a particular warm season. More than Kanto area of Japan. A total of 12,907 HSAD occurring from two-thirds of these initial hot days were in June, implying the 2000 to 2009 in five major cities—, , , harmful effect of initial warming as the season changed. Risk Kawasaki, and —were analyzed. Generalized ad- increase that began early at the fairly mild perceived temper- ditive models and zero-inflated Poisson regressions were used ature implies the need for early precaution. to estimate the effects of daily maximum three-hour apparent temperature (AT) on dispatch frequency from May to Keywords Apparent temperature . Heat . Heatstroke . September, with adjustment for seasonality, long-term trend, Emergency ambulance dispatch . Time-series . Japan weekends, and public holidays. Linear and non-linear expo- sure effects were considered. Effects on days when AT first exceeded its summer median were also investigated. City- Introduction specific estimates were combined using random effects meta-analyses. Exposure-response relationship was found to Hot weather is an environmental hazard with well-documented be fairly linear. Significant risk increase began from 21 °C impact on human health. Rising temperature that leads to more hot days (Intergovernmental Panel on Climate Change 2007; World Meteorological Organization 2012)isexpectedtoex- Electronic supplementary material The online version of this article acerbate future heat-related mortality and morbidity (Cheng et (doi:10.1007/s00484-013-0677-4) contains supplementary material, which is available to authorized users. al. 2008; Doyon et al. 2008;Goslingetal.2009; Knowlton et al. 2007; Martin et al. 2012). The causal association between * : : C. F. S. Ng ( ) K. Ueda H. Nitta high ambient temperature and premature mortality has been Environmental Epidemiology Section, Center for Environmental Health Sciences, National Institute for Environmental Studies, widely established (Armstrong et al. 2011; Anderson and Bell 16-2 Onogawa, Tsukuba, Ibaraki 305-8506, Japan 2011;Goldbergetal.2011; Kim et al. 2006). Heat exposure e-mail: [email protected] has also been linked to increases in cardiovascular and respi- ratory hospital admissions (Lin et al. 2009; Michelozzi et al. M. Ono Center for Environmental Health Sciences, National Institute for 2009;Wichmannetal.2011), emergency ambulance trans- Environmental Studies, Tsukuba, Ibaraki, Japan ports (Alessandrini et al. 2011; Bassil et al. 2011;Dolneyand Sheridan 2006;Goldenetal.2008;Schafferetal.2012), and A. Takami climate-sensitive diseases (Alexander 2013; Bambrick et al. Regional Atmospheric Environment Section, Center for Regional Environmental Research, National Institute for Environmental 2008;Brayetal.2011; Hashizume et al. 2007). These heat- Studies, Tsukuba, Ibaraki, Japan related health consequences constitute considerable burden in 942 Int J Biometeorol (2014) 58:941–948 terms of reduced quality of life, higher cost of healthcare including surveillance and prevention, and loss of productivity in the economic sense (Bambrick et al. 2008). Managing these ’ heat-related health risks remains a challenge even in today s Saitama highly developed nations.

In Japan, average summer temperature from June to Tokyo August has been trending upwards since tracking first began Chiba in 1898. An all-time high was documented in the summer of Kawasaki Yokohama 2010, with average temperature higher than the 1971– 2000 historical average by 1.64 °C (Japan Meteorological Association 2010). This particular episode of heat spell was thought to be responsible for the record high heatstroke- related deaths (Tokyo Metropolitan Government 2012)and emergency transports in many cities (National Institute for Environmental Studies 2012). More recently, Japan recorded a surge in the number of heatstroke-related ambulance dispatches (HSAD) in July 2012 at an increase of 17 % country-wide over the same period in the previous year Fig. 1 Study areas (shaded regions) and monitoring stations (stars). Dark outline denotes inland prefecture boundary, while lighter outline (Fire and Disaster Management Agency 2012). The growing denotes city boundary incidence of heat-related disorders is garnering public concerns and at the same time brings increasing attention to period from 2000 to 2009. In Japan, life support ambulances the current prevention strategies (Hoshi and Inaba 2007). are organized by the fire departments. For monitoring of Short-term effect of summer heat on mortality has been ambulance response system, ambulance staffs follow up with previously established in Japan in a multi-region study covering the emergency physicians to keep a record of diagnosis for all Tokyo (Chung et al. 2009). However, the impact of summer transported patients (Abe et al. 2009). Determination of heat- temperature on HSAD has not been assessed in the epidemio- stroke diagnosis was according to the emergency physicians logical context. The pattern and magnitude of this association based on the first presentation of patients at the emergency remain unexplored in Japan. In addition, because HSAD cor- departments. respond to a more responsive indicator of heat effect in a Hourly data for ambient and dew point temperature mea- population due to the stronger association with high tempera- sured at the monitoring stations in each city (Fig. 1) were ture in comparison to mortality (Alessandrini et al. 2011), there acquired from the Japan Meteorological Agency. For Saitama is an interest to understand the time course of the effect of high city, the monitoring station in the northwest was selected ambient temperature on HSAD, particularly when such associ- instead of the one in Tokyo to capture the slightly different ation becomes significant. This information can be helpful climate due to the mountainous feature in the northwest region. when evaluating the relevance of existing prevention guide- Kawasaki and Yokohama shared the same monitoring station lines. Given these motivations, we conducted a population- given the close distance. To measure temperature exposure, we based multi-city time-series study using a 10-year monitoring opted for apparent temperature, an index of perceived temper- data to evaluate the association between ambient temperature ature which incorporates the effect of humidity. This tempera- and HSAD in the Kanto metropolitan area of Japan. ture index can better capture response on health, and has been commonly used to investigate the health impact of weather and air pollutants (Basu 2009; Chung et al. 2009; Kovats and Hajat Material and methods 2008;O’Neill et al. 2005). Hourly apparent temperature was calculated using the formula: apparent temperature=−2.653 + Study region and data (0.994 × temperature) + (0.0153 × dew point temperature2) (Kalkstein and Valimont 1986; Steadman 1979). We computed The five study sites are located in the densely populated Kanto the maximum 3-hour average of apparent temperature as the major metropolitan area. They include the Tokyo metropolis and daily measure of temperature exposure (hereinafter referred to four surrounding designated cities—Saitama, Chiba, Kawasaki, as AT) (Wichmann et al. 2011). and Yokohama—covering approximately 3,257 km2 (1,258 square miles) in area and 19.6 million in population as of 2005 Statistical analysis (Fig. 1). Daily case series of HSAD from May to September were We first examined the lag distribution of AT in association collected from the fire departments in each city over a 10-year with the daily HSAD in each city to determine a suitable Int J Biometeorol (2014) 58:941–948 943 length of temperature exposure. Delayed effect of exposure, temperaturecutoffsrangingfrom21°Cto31°Cwereattempted if any, was included as an average in the regression model. separately for Tokyo to understand the risk of dispatches at lower Assuming daily dispatch counts followed an overdispersed temperature threshold. Poisson distribution (Alessandrini et al. 2011; Basu et al. Analyses were performed in R 2.14.2 (R Foundation for 2005), a semi-parametric generalized additive model Statistical Computing, Vienna, Austria) and Stata 11.2 (StataCorp (GAM) was used to explore the exposure–response relation- LP, College Station, TX, USA) at 5 % statistical significance level. ship based on a pooled dataset (Hastie and Tibshirani 1990; Peng et al. 2006). To investigate possible non-linear associ- ation, AT was smoothed using thin plate regression spline Results with the degree of freedom determined via generalized cross validation method. Unlike other knot-based smoothing, this Summary statistics for the study population and HSAD technique removes the limitation of having to choose knot according to study location are provided in Table 1.On locations (Wood 2006). average, about 48 % of the total HSAD occurred in We estimated the effects of AT assuming linear and non- August and 37 % in July, with roughly a third of the cases linear temperature exposure. When AT was assumed linear, aged 65 years and older. Given the proximity, the selected we estimated the relative risk (RR) for the association be- cities share fairly similar weather characteristics. Temperature tween 2-day average AT (maximum 3-hour) and daily dis- was usually the highest in August, followed by July and patches in each city using GAM, controlling for seasonality September (Fig. 2). Relative humidity ranged between and longer time trend via the cubic regression splines of 60 and 80 % for most of the warm season, except during month and year with 3 degrees of freedom (DF) each. The May when it was slightly lower. Frequency of HSAD began to sensitivity of the RR estimates to the choice of DF was increase when daily AT exceeded 25 °C, peaking at around examined. Confounding effects of public holidays and 38 °C (Fig. 3). Such pattern was particularly apparent in the weekends were accounted for using dummy variables. Tokyo metropolis given the large population. Because the time series of emergency dispatches contained A previous-day lag of AT (lag 1) was evident in all cities an excess number of zero counts, especially during early except Chiba (Fig. 4). We did not observe any harvesting summer, zero-inflated Poisson (ZIP) models were also fitted effect, which was denoted by a dip below unity RR before for comparison (Bassil et al. 2011). Robust standard errors returning to one. This single-day delayed effect was included were chosen to correct for overdispersion (Cameron and in the subsequent effect estimation using an average of the Trivedi 2009). The logit portion of the ZIP model contained current- and previous-day AT (lag 0 and 1). The exposure– AT and a monthly indicator as independent variables; the response curve based on pooled data for all study locations Poisson portion added the yearly, public holiday, and week- showed a fairly linear shape beyond 15 °C (Fig. 5). Due to the end dummy variables. City-specific estimates were then very low number of HSAD at low temperature, association in combined using random-effects meta-analyses in the second the low temperature region has a large margin of error and was stage to obtain a regional mean effect (Berkey et al. 1995). sensitive to the choice of smoothing methods. To assess the non-linear rate of ambulance dispatches, AT Assuming a nonlinear rate of HSAD, combined RR as- was smoothed as natural spline with five DF using a distributed sociated with a degree Celsius increase in AT was close to lag non-linear model (Gasparrini et al. 2010) within the GAM null and insignificant at low temperature. The combined RR framework with the same set of covariates. An estimate of RR estimates gradually rose to significance starting with an RR was computed for each city and then combined using random- of 1.22 (95 % confidence interval, 1.03–1.44) at 21 °C to effects meta-analyses for every degree increment from the 1.40 (1.24–1.58) at 27 °C, before slowing down to 1.49 minimum to maximum AT in the five cities. The RR estimates (1.42–1.57) for AT around 40 °C (Fig. 6). When a linear rate were tested against various DF for the smoothing of AT. is assumed, the combined RR estimate for HSAD associated The rate of dispatches on days when AT first exceeded the with 1 °C increase in the 2-day average of AT was 1.43 median of 29 °C in a given warm season was estimated by (1.37–1.50), with a corresponding I-squared of 94.3 % including an indicator for the events in the GAM model (Table 2), denoting heterogeneity of effects across cities. together with the 2-day average AT exposure. This median Weekends were significantly associated with HSAD at a AT cutoff was substantiated by preliminary exploratory find- combined RR of 1.09 (1.01–1.17) and I-squared of ings that showed increasing dispatch frequency starting from 55.4 %. Public holidays appeared to have no effect on around 30 °C for the study area. Effects of the above-median dispatch frequency. Effect estimates from the ZIP models AT were estimated for each city up to the first ten median- were comparable, despite the slightly wider confidence exceeding events within a given warm period. The city- intervals. specific estimates were subsequently combined to obtain the Combined RR estimates for days when AT initially regional mean effects. Besides the median ATcutoff, different exceeded the median of 29 °C during a given warm period 944 Int J Biometeorol (2014) 58:941–948

Table 1 Characteristics of study locations and heatstroke-related ambulance dispatches from May to September, 2000–2009

City Area (km2) Population in Total dispatches Annual dispatches per Age over 64 Average heatstroke-related emergency 2005 (‘000) (10-years) 100,000 population yearsa (%) ambulance dispatches per day

May Jun Jul Aug Sep

Saitama 217.5 1,176 1,035 8.8 29.2 0.05 0.25 1.07 1.75 0.23 Chiba 272.1 924 1,128 12.2 33.9 0.03 0.18 1.49 1.67 0.28 Tokyo 2187.7 12,577 7,392 5.9 31.9 0.37 1.48 8.81 11.38 1.91 Kawasaki 142.7 1,327 1,113 8.4 26.5 0.03 0.17 1.41 1.64 0.35 Yokohama 437.4 3,580 2,239 6.3 28.0 0.04 0.44 2.62 3.71 0.45 Total 3257.4 19,584 12,907 6.6 31.1 0.52 2.53 15.39 20.16 3.22 a Based on 2006–2009 were significant up to the first nine events, with borderline in the Kanto area of Japan. It improves the current understand- significance when ten events were considered (Fig. 7). ing of the magnitude and characteristics of risk increase in Estimated RR for the first time median AT was surpassed HSAD which can be ascribed to high temperature in the was the largest at 2.53 (95 % CI, 1.57–4.09) and gradually region. An important finding of the study is the risk increase decreased to about 1.32 (95 % CI, 1.12–1.55) for surpasses of heatstroke-related medical dispatches that began at a tem- up to the ninth time (represented with a single dummy perature level typically not perceived as deleterious to human variable). More than two thirds of these events occurred in health. According to the current prevention guideline by the June (70.8 %), while the rest were in July (21 %) and May Japan Society of Biometeorology (2013), a heatstroke warn- (8.2 %) (Table 3). These daily records of above-median AT ing is recommended when the wet bulb globe temperature included both single-day events and continuous episodes (WBGT) reaches 25–28 °C (about 28–31 °C of maximum 1- that lasted a few days. When different AT cutoffs were hour temperature, supplementary material Fig. S2). Findings defined for Tokyo (besides the summer median), results here however showed a significant risk increase that began at showed significant risk increase of HSAD for cutoff as 21 °C (or 21.5 °C of maximum 1-hour temperature) with low as 25 °C, while peaking at 29 °C (Fig. S1). These effect about 22 % increase of HSAD for a degree increase in the 2- estimates were fairly invariable to the DF of smoothing day average of AT. Estimated risk increase quickly escalated splines for AT. to about 40 % at around 27 °C (25.5 °C of daily 1-hour maximum temperature) and subsequently hovered between 40 % and 50 % for higher apparent temperature. When linear Discussion exposure effect was assumed, the estimated risk increase was

This study examined the association between apparent tem- perature and heatstroke-related emergency medical transports Temperature (°C)

Maximum 3-h Apparent Temperature Maximum 1-h Temperature

Fig. 2 Monthly median and the interquartile range (vertical line)of Fig. 3 Annual average heatstroke-related ambulance dispatches by daily maximum 3-h apparent temperature and maximum 1-h tempera- each degree of the maximum 3-h apparent temperature in five cities, ture for all study locations, 2000–2009 2000–2009 Int J Biometeorol (2014) 58:941–948 945

Saitama Chiba Tokyo Kawasaki Yokohama RR

Lag (day)

Fig. 4 Lag distribution for the delayed effects of daily maximum 3-h apparent temperature in each city for a degree Celsius increase above the median of 29 °C based on linear temperature exposure. Shaded region denotes 95 % confidence interval around 40 % (i.e., 38 % or 43 % depending on the modeling Analysis of delayed temperature effect showed a lack of strategy). harvesting effect in the emergency dispatch time-series, Another key finding of the study relates to the risk suggesting high AT harms the general population in the increase of HSAD that began quite early in the summer. region, not just the physically frail or heat-sensitive sub- Notwithstanding that most of the dispatches occurred in groups. Harvesting hypothesis in the context of this study August in tandem with the peaking of temperature, the risk describes the bringing forward of imminent emergency dis- increase of HSAD began as early as June and July as patch incidences following exposure to high temperature. If indicated by the findings of significant effect estimates for this is true, a negative association should be observed be- days when median AT (29 °C or the equivalent maximum 1- tween previous-day exposure and dispatches on a later day hour temperature of 26.8 °C) was initially exceeded during due to the pool of susceptible individuals that has been the warm season. In other words, high AT that initially reduced by an earlier exposure (Zanobetti and Schwartz surpassed the summer median within the May–September 2008). This concept was originally applied to mortality period is associated with higher frequency of HSAD in the time-series and has been thoroughly explained (Braga et Kanto area. This association persists for the first few events al. 2002; Pattenden et al. 2003; Schwartz et al. 2004). up to the ninth, implying the impact of early high tempera- Nonetheless, given a third of the HSAD in this study aged ture. Further analysis in Tokyo revealed this association was 65 years and above, early countermeasures should actively significant even at a lower AT cutoff of 25 °C. Although we reach out to this vulnerable subgroup as they are usually the could not replicate the findings in other cities due to sample most affected by adverse meteorological conditions due to size restriction, results imply the burden of rising tempera- the reduced tolerance to heat stress (Yeo 2004), their ten- ture and humidity towards June as the season changes, dency to refrain from air conditioning use possibly due to which justify early precautions against heatstroke when economic or personal preference, and the possibility that warm weather is forecasted in early summer. RR Log Ambulance Dispatch

Maximum 3-h Apparent Temperature (°C) Maximum 3-h Apparent Temperature (°C)

Fig. 5 Exposure–response curve for the adjusted effect of daily max- Fig. 6 Combined estimates of relative risk for each degree increment imum 3-h apparent temperature on heatstroke-related ambulance dis- of maximum 3-h apparent temperature (2-day average) assuming non- patches for all five cities, 2000–2009. Dotted lines denote 95 % con- linear temperature predictor smoothed using natural spline with 5 fidence interval degrees of freedom 946 Int J Biometeorol (2014) 58:941–948

Table 2 Estimates of relative risk and the corresponding 95 % Variable GAM quasi-Poisson Zero-inflated Poisson confidence intervals for heat- stroke-related ambulance dis- RR 95 % CI RR 95 % CI patches associated with 1 °C in- crease in the 2-day average Apparent temperature maximum 3-h apparent temper- Saitama 1.36 (1.33, 1.40) 1.30 (1.23, 1.38) ature, and the estimates for the Chiba 1.37 (1.34, 1.41) 1.36 (1.28, 1.45) effects of weekends and public holidays. All estimations were Tokyo 1.43 (1.40, 1.45) 1.40 (1.36, 1.44) based on linear temperature Kawasaki 1.54 (1.50, 1.58) 1.43 (1.33, 1.54) exposure Yokohama 1.47 (1.44, 1.50) 1.40 (1.35, 1.46) Combined city estimatesa 1.43 (1.37, 1.50) 1.38 (1.35, 1.42) RR relative risk, CI confidence Other covariatesa interval Weekends 1.09 (1.01, 1.17) 1.08 (1.01, 1.15) a City estimates combined using Public holidays 1.04 (0.77, 1.40) 1.02 (0.78, 1.34) random-effects meta-analysis some may not be aware of a heat alert or emergency in oppressively hot days with heat alerts and heat emergencies broadcast media (Dolney and Sheridan 2006; Kondo et al. (Dolney and Sheridan 2006). Other studies that examined the 2012; Kovats and Hajat 2008). effect of summer heat on emergency medical transports for Although most studies investigating the adverse impact of specific types of diseases also reported a similar pattern. A heat have focused on mortality in the past, a few recent studies time-series study in the Emilia-Romagna region of Italy doc- have looked into the non-fatal heat-related illnesses using umented a significant association of apparent temperature and ambulance response call and emergency dispatch data. A emergency dispatches for non-traumatic, cardiovascular, and comparable study that looked into the heat illness-related respiratory diseases (Alessandrini et al. 2011). The study ambulance response calls in Toronto, Canada reported 29– reported a greater increase of dispatches on days when mean 31 % increase of calls for every degree increase in maximum apparent temperature exceeded 30 °C. temperature during summer (Bassil et al. 2011). The percent- In Japan, as part of the effort to curb heat-related disor- age increase however reduced to 16 % when humidex index ders, authorities have implemented early heat alert warning was used as the exposure variable. A different study systems to protect community health (Martinez et al. 2011). encompassing 47 prefectures in Japan reported large correla- Forecast of heat stress index based on the WBGT and tion between daily highest air temperature and ambulance information on the hazard levels of heat-related illnesses transports stratified by heatstroke cases in August 2009 are provided online by the Ministry of the Environment (Miyatake et al. 2012). Elsewhere in Phoenix, United States, and the National Institute of Environmental Studies a 6-year data revealed more heat-related medical dispatches (National Institute for Environmental Studies 2012). The during times of high human comfort indices (Golden et al. current recommended heat warnings are based on a fixed 2008). A spatial analysis of ambulance response calls over a 4- set of criteria derived from the WBGT which is the preferred year period in Toronto showed a 10 % increase of calls on risk indicator (Japanese Society of Biometeorology 2013). This guideline, which suggests a cutoff of WBGT 25 °C for

Table 3 Monthly distribution of the first ten times when daily maxi- mum 3-h apparent temperature exceeded 29 °C during warm season, 2000–2009

City May June July

Saitamaa 2.0 (20) 8.0 (80) 0 (0) Chiba 0.5 (5) 7.6 (76) 1.9 (19) Tokyo 0.8 (8) 7.4 (74) 1.8 (18) Fig. 7 Estimates of relative risk for heatstroke-related ambulance Kawasaki 0.4 (4) 6.2 (62) 3.4 (34) dispatches on days when the median daily maximum 3-h apparent Yokohama 0.4 (4) 6.2 (62) 3.4 (34) temperature of 29 °C was initially exceeded for up to the first ten times during warm period in a given year. The first vertical line for each city Average 0.82 (8.2) 7.08 (70.8) 2.1 (21.0) represents the first surpass of temperature measurement within a warm period, the second vertical line represents the first and second sur- Percentage in parenthesis passes within the same warm period, and so on a For 2005–2009 Int J Biometeorol (2014) 58:941–948 947 heatstroke warning, may not capture the vulnerabilities which temperature: a time series analysis in Emilia-Romagna, Italy. – can occur at lower temperature during the early warming of Environ Res 111(8):1192 1200 Alexander P (2013) Association of monthly frequencies of diverse weather as shown in this study. A gradual system with pro- diseases in the calls to the public emergency service of the city gressive temperature cutoffs for the triggering of warnings in of Buenos Aires during 1999–2004 with meteorological variables each month or at least the early period of summer would likely and seasons. Int J Biometerol 57(1):83–90 be better in protecting the susceptible individuals with slower Anderson GB, Bell ML (2011) Heat waves in the United States: mortality risk during heat waves and effect modification by heat wave characteristics rate of acclimatization to summer heat as the season changes. in 43 U.S. Communities. Environ Health Perspect 119(2):210–218 Nevertheless, more information on the nature of heatstroke Armstrong BG, Chalabi Z, Fenn B, Hajat S, Kovats S, Milojevic A, cases, such as events that occur indoors, outdoors, during Wilkinson P (2011) Association of mortality with high tempera- exercise and so on, is needed to understand the different sub- tures in a temperate climate: England and Wales. J Epidemiol Commun Health 65(4):340–345 groups so as to identify suitable measures that capture the Bambrick H, Dear K, Woodruff R, Hanigan I, McMichael A (2008) The different patterns of heat susceptibility. impacts of climate change on three health outcomes: temperature- Classification of heatstroke in our data relies on the first related mortality and hospitalisations, salmonellosis and other bacte- diagnosis by clinicians upon initial presentation. These early rial gastroenteritis, and population at risk from dengue. Garnaut Climate Change Review. http://www.garnautreview.org.au/ diagnoses may eventually change for some patients and be- CA25734E0016A131/WebObj/03-AThreehealthoutcomes/$File/03- cause the information was not captured in our data, causal A%20Three%20health%20outcomes.pdf. Accessed 18 March 2013 interpretation should take into consideration this limitation. Bassil KL, Cole DC, Moineddin R, Lou W, Craig AM, Schwartz B, Another issue common to the population-based environmen- Rea E (2011) The relationship between temperature and ambu- lance response calls for heat-related illness in Toronto, Ontario, tal epidemiological studies is the method of measuring per- 2005. J Epidemiol Commun Health 65(9):829–831 sonal exposure which typically assumes population in a spe- Basu R (2009) High ambient temperature and mortality: a review of cific area to have the same temperature exposure based on the epidemiologic studies from 2001 to 2008. Environ Health 8:40 closest monitoring station. This may not reflect the variability Basu R, Dominici F, Samet JM (2005) Temperature and mortality among the elderly in the United States: a comparison of epidemi- of exposure given the spatial differences and personal suscep- ologic methods. Epidemiology 16(1):58–66 tibility. However, the use of monitoring data as proxy to Berkey CS, Hoaglin DC, Mosteller F, Colditz GA (1995) A random- personal exposure is often preferred in population-based stud- effects regression model for meta-analysis. Stat Med 14(4):395–411 ies since preventive measures based on real-time surveillance Braga ALF, Zanobetti A, Schwartz J (2002) The effect of weather on respiratory and cardiovascular deaths in 12 US cities. Environ data is often more pragmatic than those that require monitor- Health Perspect 110(9):859–863 ing of personal exposure. Bray JE, Mosley I, Bailey M, Barger B, Bladin C (2011) Stroke public In conclusion, we found significant association of apparent awareness campaigns have increased ambulance dispatches for – temperature and HSAD at temperature below the current stroke in Melbourne, Australia. Stroke 42(8):2154 2157 Cameron AC, Trivedi PK (2009) Microeconometrics using stata. Stata warning guideline. This association began rather early in the Press, College Station warm season, particularly in June. Heat-related risk increases Cheng CS, Campbell M, Li Q, Li GL, Auld H, Day N, Pengelly D, Gingrich that begin early at temperatures below the warning threshold S,KlaassenJ,MacIverD,ComerN,MaoY,ThompsonW,LinH have strong repercussions on public health as well as the (2008) Differential and combined impacts of extreme temperatures and air pollution on human mortality in south-central Canada. Part II: future planning of emergency care services, and as such, should be estimates. Air Qual Atmos Health 1(4):223–235 given consideration when strategizing for prevention. Chung JY, Honda Y, Hong YC, Pan XC, Guo YL, Kim H (2009) Ambient temperature and mortality: an international study in four Acknowledgments This work was a part of the Development of capital cities of . Sci Total Environ 408(2):390–396 Seamless Chemical Assimilation System and its Application for Dolney TJ, Sheridan SC (2006) The relationship between extreme heat Atmospheric Environmental Materials (SALSA) supported by a grant- and ambulance response calls for the city of Toronto, Ontario, in-aid for the Research Program on Climate Change Adaptation (RECCA) Canada. Environ Res 101(1):94–103 by the Ministry of Education, Culture, Sports, Science and Technology Doyon B, Belanger D, Gosselin P (2008) The potential impact of climate (MEXT) and the Ministry of the Environment (MOE) in Japan. The views change on annual and seasonal mortality for three cities in Quebec, expressed in this article are those of the authors and do not necessarily Canada. Int J Health Geogr 7:23. doi:10.1186/1476-072X-7-23 represent the views and policies of the aforementioned organizations. Fire and Disaster Management Agency, Ministry of Internal Affairs and Communications (2012) Heatstroke-related emergency am- bulance dispatches report for summer of 2012 (in Japanese). http://www.env.go.jp/chemi/heat_stroke/ic_rma/24a2/mat01.pdf. Accessed 11 March 2013 References Gasparrini A, Armstrong B, Kenward MG (2010) Distributed lag non- linear models. Stat Med 29(21):2224–2234 Goldberg MS, Gasparrini A, Armstrong B, Valois MF (2011) The short- Abe T, Tokuda Y, Ohde S, Ishimatsu S, Nakamura T, Birrer RB (2009) term influence of temperature on daily mortality in the temperate The relationship of short-term air pollution and weather to ED climate of Montreal, Canada. Environ Res 111(6):853–860 visits for asthma in Japan. Am J Emerg Med 27(2):153–159 Golden JS, Hartz D, Brazel A, Luber G, Phelan P (2008) A biomete- Alessandrini E, Zauli Sajani S, Scotto F, Miglio R, Marchesi S, orology study of climate and heat-related morbidity in Phoenix Lauriola P (2011) Emergency ambulance dispatches and apparent from 2001 to 2006. Int J Biometeorol 52(6):471–480 948 Int J Biometeorol (2014) 58:941–948

Gosling S, McGregor G, Lowe J (2009) Climate change and heat- adaptation to climate change. Int J Environ Res Public Health related mortality in six cities Part 2: climate model evaluation and 8(12):4563–4581 projected impacts from changes in the mean and variability of Michelozzi P, Accetta G, De Sario M et al (2009) High temperature and temperature with climate change. Int J Biometeorol 53(1):31–51 hospitalizations for cardiovascular and respiratory causes in 12 Hashizume M, Amstrong B, Hajat S, Wagatsuma Y, Faruque AS, European cities. Am J Respir Crit Care Med 179(5):383–389 Hayashi T, Sack DA (2007) Association between climate variability Miyatake N, Sakano N, Murakami S (2012) The relation between and hospital visits for non-cholera diarrhoea in Bangladesh: effects ambulance transports stratified by heat stroke and air temperature and vulnerable groups. Int J Epidemiol 36(5):1030–1037 in all 47 prefectures of Japan in August, 2009: ecological study. Hastie TJ, Tibshirani RJ (1990) Generalized additive models. Environ Health Prev Med 17(1):77–80 Chapman and Hall, New York National Institute for Environmental Studies (2012) Heatstroke pre- Hoshi A, Inaba Y (2007) Prediction of heat disorders in Japan. vention information website by the Ministry of the Environment. Global Environmental Research. http://www.airies.or.jp/ http://www.nies.go.jp/health/HeatStroke/spot/2010/ attach.php/6a6f75726e616c5f31312d31656e67/save/0/0/11_1- 2010report.pdf (in Japanese). Accessed 11 March 2013 07.pdf. Accessed 10 March 2013 O’Neill MS, Hajat S, Zanobetti A, Ramirez-Aguilar M, Schwartz J Intergovernmental Panel on Climate Change (IPCC) (2007) Climate (2005) Impact of control for air pollution and respiratory epi- change 2007: synthesis report. Contribution of Working Group I, demics on the estimated associations of temperature and daily II and III to the Fourth Assessment Report of the Intergovermental mortality. Int J Biometeorol 50(2):121–129 Panel on Climate Change. Geneva, Switzerland: IPCC. http:// Pattenden S, Nikiforov B, Armstrong BG (2003) Mortality and tem- www.ipcc.ch/publications_and_data/publications_ipcc_fourth_as- perature in Sofia and London. J Epidemiol Commun Health sessment_report_synthesis_report.htm. Accessed 11 March 2013 57(8):628–633 Japan Meteorological Association (2010) Japan hit by the hottest Peng RD, Dominici F, Louis TA (2006) Model choice in time series summer in more than 100 years. http://ds.data.jma.go.jp/tcc/tcc/ studies of air pollution and mortality. J R Statist Soc A 169:179– news/press_20100910.pdf. Accessed 24 April 2013 198 Japan Society of Biometeorology (2013) Guide to preventing heatstroke Schaffer A, Muscatello D, Broome R, Corbett S, Smith W (2012) in everyday life. Version 2. http://www.med.shimane-u.ac.jp/assoc- Emergency department visits, ambulance calls, and mortality jpnbiomet/pdf/shishinVer2.pdf (in Japanese). Accessed 11 March associated with an exceptional heat wave in Sydney, Australia, 2013 2011: a time-series analysis. Environ Health 11(1):3. doi:10.1186/ Kalkstein LS, Valimont KM (1986) An evaluation of summer discom- 1476-069X-11-3 fort in the United-States using a relative climatological index. B Schwartz J, Samet JM, Patz JA (2004) Hospital admissions for heart Am Meteorol Soc 67(7):842–848 disease: the effects of temperature and humidity. Epidemiology Kim H, Ha JS, Park J (2006) High temperature, heat index, and 15(6):755–761 mortality in 6 major cities in . Arch Environ Occup Steadman RG (1979) The assessment of sultriness. Part II: effects of Health 61(6):265–270 wind, extra radiation and barometric pressure on apparent tem- Knowlton K, Lynn B, Goldberg RA, Rosenzweig C, Hogrefe C, perature. J Appl Meteorol 18:874–885 Rosenthal JK, Kinney PL (2007) Projecting heat-related mortality Tokyo Metropolitan Government (2012) Heatstroke mortality in sum- impacts under a changing climate in the New York City region. mer of 2012. http://www.fukushihoken.metro.tokyo.jp/kansatsu/ Am J Public Health 97(11):2028–2034 oshirase/24necchusyou.html (in Japanese). Accessed 11 March Kondo M, Ono M, Nakazawa K, Kayaba M, Minakuchi E, Sugimoto 2013 K, Honda Y (2012) Population at high-risk of indoor heatstroke: Wichmann J, Andersen Z, Ketzel M, Ellermann T, Loft S (2011) the usage of cooling appliances among urban elderlies in Japan. Apparent temperature and cause-specific emergency hospital ad- Environ Health Prev Med 18(3):251–257. doi:10.1007/s12199- missions in greater Copenhagen, Denmark. PLoS One 012-0313-7 6(7):e22904. doi:10.1371/journal.pone.0022904 Kovats RS, Hajat S (2008) Heat stress and public health: a critical Wood S (2006) Generalized addictive models: an introduction with R. review. Ann Rev Publ Health 29:41–55 Chapman and Hall/CRC, Florida, pp 154–160 Lin S, Luo M, Walker RJ, Liu X, Hwang SA, Chinery R (2009) World Meteorological Organization (2012) WMO statement on the Extreme high temperatures and hospital admissions for respirato- status of global climate in 2011. http://www.wmo.int/pages/prog/ ry and cardiovascular diseases. Epidemiology 20(5):738–746 wcp/wcdmp/documents/1085_en.pdf. Accessed 11 March 2013 Martin SL, Cakmak S, Hebbern CA, Avramescu ML, Tremblay N Yeo TP (2004) Heat stroke: a comprehensive review. AACN Clin (2012) Climate change and future temperature-related mortality Issues 15(2):280–293 in 15 Canadian cities. Int J Biometeorol 56(4):605–619 Zanobetti A, Schwartz J (2008) Mortality displacement in the associ- Martinez GS, Imai C, Masumo K (2011) Local heat stroke prevention ation of ozone with mortality: an analysis of 48 cities in the plans in Japan: characteristics and elements for public health United States. Am J Respir Crit Care Med 177(2):184–189