Associating emergency room visits with first and prolonged extreme temperature event in Taiwan: A population-based cohort study

Associating emergency room visits with

first and prolonged extreme temperature

event in Taiwan: A population-based cohort study

Yu-Chun Wang

, Yu-Kai Lin

, Chun-Yu Chuang

, Ming-Hsu Li

, Chang-Hung Chou

,

Chun-Hui Liao

, Fung-Chang Sung

⁎

Department of Bioenvironmental Engineering, College of Engineering, Chung Yuan Christian University, 200 Chung-Pei Road, Chung Li 320, Taiwan

b

Environmental and Occupational Medicine and Epidemiology Program, Department of Environmental Health, Harvard School of Public Health, 677 Huntington Ave., Boston, MA 02115, USA

c

Department of Biomedical Engineering and Environmental Sciences, National Tsing Hua University, 101, Section 2, Kuang-Fu Road, Hsinchu 300, Taiwan

d_{Graduate Institute of Hydrological & Oceanic Sciences, National Central University, 300 Chung-Da Road, Chung Li 320, Taiwan} e

China Medical University College of Life Sciences, 91 Hsueh-Shih Road, Taichung 404, Taiwan

f

China Medical University Department of Public Health, 91 Hsueh-Shih Road, Taichung 404, Taiwan

g

China Medical University Hospital Division of Psychiatry, 91 Hsueh-Shih Road, Taichung 404, Taiwan

h

Institute of Environmental Health, National Taiwan University, 17 Hsu Chou Road, Taipei 100, Taiwan

a b s t r a c t

a r t i c l e i n f o

Article history: Received 9 October 2011

Received in revised form 27 November 2011 Accepted 28 November 2011

Available online 29 December 2011 Keywords:

Emergency room visits Circulatory

Respiratory

Extreme temperature event Taiwan

The present study evaluated emergency room visit (ERV) risks for all causes and cardiopulmonary diseases associated with temperature and long-lasting extreme temperatures from 2000 to 2009 in four major cities in Taiwan. The city-specific daily average temperatures at the high 95th, 97th, and 99th percentiles, and the low 10th, 5th, and 1st percentiles were defined as extreme heat and cold. A distributed lag non-linear model was used to estimate the cumulative relative risk (RR) of ERV for morbidities associated with tempera-tures (0 to 3-day lags), extreme heat and cold lasting for 2 to 9 days or longer, and with the annualfirst extreme heat or cold event after controlling for covariates. Low temperatures were associated with slightly higher ERV risks than high temperatures for circulatory diseases. After accounting for 4-day cumulative temperature effect, the ERV risks for all causes and respiratory diseases were found to be associated with extreme cold at the 5th per-centile lasting for >8 days and 1st perper-centile lasting for >3 days. The annualfirst extreme cold event of 5th per-centile or lower temperatures was also significantly associated with ERV, with RRs ranging from 1.09 to 1.12 for all causes and from 1.15 to 1.26 for respiratory diseases. The annualfirst extreme heat event of 99th percentile temperature was associated with higher ERV for all causes and circulatory diseases. Annualfirst extreme tempera-ture event and intensified prolonged extreme cold events are associated with increased ERVs in Taiwan.

© 2011 Elsevier B.V. All rights reserved.

1. Introduction

The temperature–mortality association has been widely evaluated

worldwide for various weather conditions appearing with U-, V-, and

J-shaped relationships (Basu, 2009; Curriero et al., 2002; Huynen et

al., 2001). On the other hand, studies on the morbidity–temperature

association are limited (Knowlton et al., 2009; Kovats et al., 2004;

Lin et al., 2009; Linares and Diaz, 2008; Mastrangelo et al., 2007; Michelozzi et al., 2009; Rocklov and Forsberg, 2009; Theoharatos et

al., 2010; Tong et al., 2010). Previous studies have shown that

tem-perature has strong and acute effects on mortality from circulatory

diseases (Keatinge and Donaldson, 1995; Lin et al., 2011;

Medina-Ramon et al., 2006). The morbidities of respiratory diseases are

more likely associated with extreme heat (Kovats et al., 2004; Lin et

al., 2009; Linares and Diaz, 2008; Michelozzi et al., 2009; Rocklov

and Forsberg, 2009) and its long-lasting event (Mastrangelo et al.,

2007; Theoharatos et al., 2010).

Because of the trend of increasing extreme temperatures, the ad-ditional mortality risk associated with prolonged extreme heat and

cold has attracted attentions. (Anderson and Bell, 2009, 2011;

D'Ippoliti et al., 2010; Gasparrini and Armstrong, 2011; Hajat et al., 2006; Huynen et al., 2001; Knowlton et al., 2009; Lin et al., 2011;

Tong et al., 2010; Wang et al., 2011). Studies generally define the

temperatures at the high 99.5th, 99th, 97th, 95th, and 90th tiles as extreme heat and those at the low 10th, 5th, and 1st

percen-tiles as extreme cold (Anderson and Bell, 2009, 2011; Braga et al.,

2002; D'Ippoliti et al., 2010; Diaz et al., 2002; Hajat et al., 2002, 2006; Abbreviations: CI, confidence interval; CWB, Central Weather Bureau; DLNM, distributed

lag non-linear model; ERV, emergency room visit; Flu, influenza; NHRI, National Health Research Institute; PM10, particulate matter less than 10μm in aerodynamic

diame-ter; RR, relative risk; RH, relative humidity; TCDC, Taiwan Centers for Disease Control; TEPA, Taiwan Environmental Protection Administration; WS, wind speed.

⁎ Corresponding author at: China Medical University and Hospital, Department of Public Health, 91 Hsueh-Shih Road, Taichung 404, Taiwan. Tel.: + 886 2206 2295; fax: + 886 4 2201 9901.

E-mail address:[email protected](F.-C. Sung).

0048-9697/$– see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.scitotenv.2011.11.073

Contents lists available atSciVerse ScienceDirect

Science of the Total Environment

Lin et al., 2011; Pattenden et al., 2003; Tong et al., 2010). Studies have in-vestigated mortality risks associated with extreme temperatures lasting

for 2 days, 4 days, and more than 3 days (Anderson and Bell, 2009, 2011;

Hajat et al., 2006; Tong et al., 2010; Wang et al., 2011), but there are only

a few studies that have evaluated their association with extreme

tem-peratures lasting for 8 days and longer (Gasparrini and Armstrong,

2011; Hemon et al., 2003; Lin et al., 2011). Further, there are fewer

stud-ies on the association of morbidity with thefirst long-lasting extreme

temperature in the year. Studies conducted in Greece (Theoharatos et

al., 2010) and Italy (Mastrangelo et al., 2007) found significant

associa-tion between hospitalizaassocia-tion, emergency room visit (ERV), and longer duration of heat waves and its early occurrence in the year.

Taiwan is an island located on the western part of the Pacific

Ocean, with an area of 150 km wide and 350 km long, stretching from 22 to 25° north latitude. With subtropical climate, the island has an average mean temperature of 24 °C, but varies from north to south, and the mean daily temperature ranges from 8 °C in winter to 33 °C in summer in urban cities. Approximately 13.6 million people of the 23 million population inhabit in four major cities, namely, Taipei, Taichung, Tainan, and Kaohsiung.

How the extreme weather may trigger disease exacerbations for the population living in this subtropical island is of interest. This eco-logical study aims to evaluate the association between emergency room visit for all causes and cardiopulmonary diseases and tempera-ture change, prolonged extreme temperatempera-ture events, and the annual

occurrence of the first extreme temperature event. We used

population-based representative insurance claims data for the popula-tion residing in the four major cities to evaluate the cumulative relative risk (RR) of ERV for all causes and cardiopulmonary diseases.

2. Materials and methods 2.1. Data source

The present study used daily meteorological records obtained from the Central Weather Bureau (CWB), universal health insurance claims from the National Health Research Institute (NHRI), daily air pollution monitoring records from the Taiwan Environmental Protec-tion AdministraProtec-tion (TEPA), and daily virological surveillance data from the Taiwan Centers for Disease Control (TCDC). The study period was from 2000 to 2009. The four cities included in the present study are located from Northern to Southern Taiwan.

The CWB provided 24-hour weather data (average temperature, maximum temperature, minimum temperature, relative humidity, and barometric pressure) from 25 real-time weather monitoring

sta-tions in Taiwan (Central Weather Bureau, 2011a). The present study

used daily weather measurements from Taipei, Taichung, Tainan,

and Kaohsiung weather stations from 2000 to 2009 (Central

Weather Bureau, 2011b).

Over 96% of the 23 million population in Taiwan have been cov-ered under the Taiwan National Health Insurance program since

1996 (Bureau of National Health Insurance, 2001). The Taiwan NHRI

established a cohort with the electronic reimbursement claim re-cords, consisting of a national representative population of one

mil-lion people randomly sampled from all insured residents (Taiwan

National Health Insurance Research Database, 2011). In the year

2000, approximately 58.7% of the one million people resided in the

study cities. The dataset contained scrambled identification numbers

of citizens and information on gender, birth dates, health care re-ceived, physicians' diagnoses for outpatient visits, inpatient admis-sions and discharges, and emergency services, and medical care providers. Disease diagnoses were coded according to the 9th revision

of the International Classification of Diseases with Clinical Modification

(ICD9 CM). The records of ERV for all causes (excluding injuries and

ex-ternal causes (ICD9 CM 800–999)), circulatory diseases (ICD9 CM

390–459), and respiratory diseases (ICD9 CM 460–519) during the

study period were retrieved.

The Taiwan Air Quality Monitoring Network established by TEPA in 1993 included 74 stationary monitoring stations distributed throughout

the island (Taiwan Environmental Protection Administration, 2011;

Taiwan Governmental Information Of_{fice, 2008}). Concentrations of

am-bient air pollutants, such as particulate matters less than 10μm in

aero-dynamic diameter (PM10), nitrogen oxides (NOx), and ozone (O3), were

determined and recorded hourly at each station. The present study

analyzed the daily average data for PM10, O3, and NOx monitored at

32 general ambient stations, 13 in Taipei, 5 inTaichung, 4 inTainan, and 10 in Kaohsiung.

The TCDC launched a nationwide virus surveillance network in

1999 (Taiwan Centers for Disease Control, 2011). Information on

specimen collection and viral identification was described elsewhere

(Shih et al., 2005). Brie_{fly, nasal and/or throat samples from patients}

with one or more symptoms of respiratory tract infections, including cough, sore throat, tonsillitis, pharyngitis, pneumonia, and bronchio-lotis, were collected by sentinel physicians in communities. This study used laboratory-based viral surveillance data from 2000 to 2009 in each city, which contained information on scrambled patient

identifications, gender, birthday, residential area, dates of disease

di-agnosis and admission, results of viral isolation, and medical

care-givers providing services. Influenza A virus (Flu A), influenza B virus

(Flu B), and adenovirus (AV) were identified from viral surveillance.

Daily city-virus-specific isolation rates were calculated using the

number of positive identified respiratory infections specimens divided

by the total specimens from regional reference virology laboratories.

2.2. Definition of temperature extremes

The intensity, the duration, and the timing of extreme tempera-ture events are the key factors in evaluating their health impacts

(Anderson and Bell, 2011; D'Ippoliti et al., 2010). Daily temperatures

during the entire study period were classified into normal or extreme

heat/cold based on the city-specific daily average temperatures. A

de-tailed method was described in a previous report (Lin et al., 2011).

This study evaluated ERV risks associated with 16 extreme

tempera-tures of the 97th, 95th, 10th, and 5th percentiles lasting for 3–5,

6–8, and >8 days, and of the 99th and 1st percentiles lasting for

2–3 days and >3 days.

In addition, the annualfirst extreme temperatures of the 97th,

95th, 10th, and 5th percentiles lasting more than 2 days, and of the 99th and 1st percentiles lasting more than 1 days were coded as an extra categorical dummy variable (First) to assess adverse effects for

populations exposed to the annualfirst extreme temperature event.

2.3. City-specific relative risk estimate

Poisson regression had been widely used in the health risk studies associated with the ambient temperature and air pollutants, which used time-series numbers of events from a cohort to estimate the

rela-tive risks (McCullagh and Nelder, 1989; Thomas, 2009). Therefore, for

each city, the associations between the daily average temperatures and daily ERVs for all causes, circulatory diseases, and respiratory diseases were evaluated using a distributed lag non-linear model

(DLNM) with Poisson distribution (Armstrong, 2006; Gasparrini et al.,

2010).

This study used the natural cubic spline (NS) function for average temperature in the DLNM models, so the relative risk of ERV per 1 °C temperature change would not be favorably reported. In Taiwan,

18 °C is approximately 5th–20th percentiles of the average

tempera-ture, and 30 °C is approximately 95th–97th percentiles of the average

temperature across these cities. Therefore, the cumulative 4-day

(from 0 to 3-day lags) ERV risks and 95% confidence intervals (CI)

by comparing with the mean temperature of the city-disease-specific lowest ERV temperature (the centered value of temperature basis variables).

The previously defined extreme temperature events and the annual

first extreme temperature events were set as the categorical covariates and risks were estimated by comparing with the nonconsecutive days of extreme temperature and days of normal temperatures. A linear

rela-tionship was assumed between ERV and air pollutants, such as PM10, O3,

and NOx, with zero thresholds and 5-day maximum lag.

The model for the expected disease-specific ERV count at day (t)

in each city (c) is LogE Yct
 ¼ β0þ X5 t¼0 Xci;tþ X3 t¼0 NS Tct; 5; lag; 2   þ Extremesc tþ First c t þ NS RHc t; 4   þ NS WSc t; 5   þ NS Time;7 year = Þ þ covariates; 

where Ytcis the expected disease-specific ERV for city c on day t; β0is

the model intercept; Xi, tc represents the linear effects of air pollutants

(i = 1–3 for PM10, O3, and NOx) for city c; and NS(Ttc, 5 ; lag, 2) is the

natural cubic splines of the daily average temperature. The tempera-tures for city c were set at 5 degrees of freedom (df) and their effects were totaled for 4 days (from 0 to 3-day lags) under a 2 df lag

strati-fication. Extremestcare the categorical variables representing extreme

temperature events for city c on day t. Firsttcindicates the annual

city-specific first-occurrence extreme heat or cold event. Natural cubic

splines were also applied in the daily measurements of relative hu-midity (RH, 4 df) and wind speed (WS, 5 df). The smoother time term (Time) was set at 7 df per year. Other covariates, such as holi-days, day of the week, and the daily viral isolation rates of Flu A, Flu B, and AV were also adjusted in the models.

Sensitivity analysis was used to evaluate df, which ranged from 3

to 6 for the temperature–morbidity curves and from 0 to 2 for the

co-efficient–lag curves. Time smoothing with various df=4, 7, and 14

per year was also performed. The Akaike's information criterion was

used for model selection (Akaike, 1973).

2.4. Random-effect meta-analysis

City-specific relative risks of ERV associated with temperature,

ex-treme temperature events and the annual_{first extreme temperature}

events were further evaluated for combined effects using

meta-analysis (Viechtbauer, 2010). The restricted maximum likelihood

was set as an estimator of the amount of heterogeneity. We estimated RR and 95% CI of factors associated with ERV using exponential of

model coefficient. All data manipulation and statistical analyses

were performed using SAS version 9.1 (SAS Institute Inc., Cary, NC, USA) and Statistical Environment R 2.12.

3. Results

Table 1describes the latitude of location, the population size, and

the characteristics of temperatures and air pollutants for the four cities. Compared with other Taiwan cities, Taipei is hotter in summer but colder in winter. The numbers of days with extreme temperature, in general, are similar among these 4 cities. There were more extreme

low temperatures than extreme high temperatures (596–598 days vs.

350–356 days). Among the four cities, the PM10concentration was the

highest in Kaohsiung, and the NOx level was the highest in Taipei. The

O3levels in Tainan and Kaohsiung were likely higher than the other

cities.

From 2000 to 2009, a total of 317,343 insured population in the 4-study cities made 867,306, 54,560 and 170,005 ERVs for all causes, circulatory diseases and respiratory diseases, respectively (data not shown). The male-to-female ratio is 1.01, and 16.5% patients were

aged 65 years and above. Among all ERVs for circulatory diseases,

ce-rebrovascular disease (ICD9 CM 430–439) accounted for 40.2% and

is-chemic heart disease (ICD9 CM 410–414) for 29.6%. Among ERVs for

respiratory cares, 62.8% of visits were for pneumonia (ICD9 CM

480–486) and 20.2% for chronic airway obstruction, not else classified

(ICD9 CM 496). In addition, 83,200 biological specimens were collected for viral determinations at the same stated period.

Fig. 1shows the daily means of pooled temperatures and

cause-speci_{fic ERVs by month in the 4-study cities. The daily mean ERVs}

were the highest in February for all causes and circulatory diseases Table 1

Latitude of location, population size, and characteristics of temperatures and air pollut-ants of the four cities in Taiwan, 2000–2009.

Taipei Taichung Tainan Kaohsiung

Latitude, °N 25.0 24.2 23.1 23.0

Population in millions 6.41 2.60 1.87 2.76 Population aged 65 + years, % 9.74 8.13 10.8 9.56 Daily atmospheric environment

Average temperature, °C Mean 23.4 23.7 24.7 25.3 Minimum 8.3 8.5 9.3 11.6 25th 19.3 20.0 21.1 22.5 50th 23.9 24.9 26.0 26.4 75th 28.0 27.7 28.7 28.4 Maximum 33.0 32.0 31.7 32.0 PM10,μg/m3 Minimum 10.7 13.2 12.3 17.9 25th 31.4 38.5 42.2 45.1 50th 43.6 55.1 65.9 75.2 75th 60.1 77.6 93.0 104 Maximum 286 255 410 218 NOx, ppb Minimum 3.90 3.20 1.80 5.00 25th 23.1 19.3 13.9 18.0 50th 30.1 24.9 19.2 25.0 75th 39.5 34.0 26.2 36.2 Maximum 119 108 64.8 76.9 O3of 24 h, ppb Minimum 4.50 3.80 4.70 3.00 25th 19.0 19.0 20.7 20.0 50th 24.9 24.7 27.9 28.8 75th 31.2 31.8 36.1 38.4 Maximum 73.1 77.8 76.5 73.4

Days with extreme temperature

1th percentile 41 39 38 37 5th percentile 187 188 187 185 10th percentile 368 371 369 375 95th percentile 200 189 198 189 97th percentile 114 111 111 122 99th percentile 37 51 41 45

Fig. 1. Daily average of pooled temperatures and cause-specific emergency room visits by month in study cities during 2000–2009.

(72.6 and 4.21 visits per day, respectively), and in January for the re-spiratory diseases (18.2 visits per day).

3.1. Adverse temperature effects

The ERV risks associated with city-specific average temperatures

and extreme temperature events were estimated using DLNM after

controlling for the daily city-specific average levels of PM10, NOx,

O3, RH, WS, daily viral isolation rates for Flu A, Flu B and AV, holidays,

day of the week, and long-term trends.Fig. 2shows city-specific

asso-ciation between daily mean temperatures and ERVs. The lowest ERV was associated with an average temperature of approximately 24 °C for the four cities in general.

Fig. 3 summarizes the cumulative 4-day ERV risk estimates of

random-effect meta-analyses at temperatures of 30 °C and 18 °C. Risk patterns for low temperatures were somewhat different from those for high temperatures. In the 18 °C environment, only the

pooled ERV risks for circulatory diseases were significantly elevated

(RR = 1.07, 95% CI: 1.01_{–1.13). The ERV risks for all causes and}

respi-ratory diseases increased as the city location was at lower latitude. In the 30 °C environment, the cumulative 4-day RR of ERVs for all causes

was 1.05 (95% CI: 0.98–1.13) in the pooled estimates of the four cities.

RRs were higher in Taichung (1.13, 95% CI: 1.09–1.18) and Tainan

(1.11, 95% CI: 1.05–1.17).

3.2. Adverse effects from consecutive days of extreme temperatures

Fig. 4shows the pooled ERV risks for diseases in the 4-study cities

associated with occurrences of extreme heat and cold events by the extreme levels and durations. The ERV risk estimates for all causes and respiratory diseases increased as durations of the 5th and the 1st percentiles cold extremes increased. The RR of ERVs for all causes

was 1.18 (95% CI: 1.10–1.27) for populations exposed to the 5th

per-centile extreme temperature for 9 days or longer. The RR increased to

1.31 (95% CI: 0.94–1.84) for populations exposed to the 1st percentile

extreme temperature for 4 days or longer. The corresponding RRs of

ERVs for respiratory diseases were 1.28 (95% CI: 1.08–1.52) and

1.77 (95% CI: 0.97–3.22). Sensitivity analyses showed that the added

effects of consecutive extreme temperatures were robust to the alter-native models (data not shown).

Fig. 5shows that the ERV risks for all causes and respiratory diseases

were significantly associated with the exposure to the annual first

ex-treme cold event of the 5th and the 1st percentiles. The RRs for all

causes were 1.12 (95% CI: 1.05–1.20) and 1.09 (95% CI: 1.04–1.13),

and 1.26 (95% CI: 1.08–1.48) and 1.15 (95% CI: 1.06–1.26) for the

respi-ratory diseases. Moreover, exposure to thefirst extreme heat event of

the 99th percentile temperature was moderately associated with ERV for all causes and circulatory diseases, with RRs of 1.08 (95% CI:

1.01–1.15) and 1.23 (95% CI: 0.98–1.54), respectively.

3.3. Risks from other covariates

Exposure to PM10was significantly associated with ERVs of all

causes with a 6-day (0 to 5-day lags) cumulative RR of 1.03 (95% CI:

1.02–1.04) in Taipei and 1.01 (95% CI: 1.00–1.02) in Kaohsiung, as

PM10 increased for 1μg/m3 (data not shown). The association

be-tween PM10and ERVs for respiratory diseases was the strongest in

Taipei with a RR of 1.06 (95% CI: 1.04–1.08) as PM10increased for

1μg/m3

. In contrast, exposure to ozone, with an increase of 1 ppb, the 6-day cumulative risk of ERVs for respiratory diseases was the

highest in Taichung (RR = 1.16 (95% CI: 1.09–1.24)).

Further analysis on circulating respiratory viruses reveals that ERV risk had stronger association with Flu A than with Flu B and AV. These

associations were statistically significant only in Taichung. ERVs for

all causes were associated with per 10% Flu A isolation with a RR of

1.01 (95% CI: 1.00_{–1.01). In addition, RRs of ERVs were 1.01 (95% CI:}

1.01–1.02) for all causes and 1.03 (95% CI: 1.02–1.05) for respiratory

diseases associated with 10% increase in the isolation rate of Flu B. 4. Discussion

Single extreme temperature event has been evaluated for sudden

morbidity effects (Knowlton et al., 2009; Kovats et al., 2004; Lin et al.,

2009; Linares and Diaz, 2008; Michelozzi et al., 2009; Rocklov and

RR 10 20 30 0.6 1.0 1.4 RR 10 20 30 0.6 1.0 1.4 RR 10 20 30 0.6 1.0 1.4 RR 10 20 30 0.6 1.0 1.4 RR 10 20 30 1.0 1.4 1.8 RR 10 20 30 1.0 1.4 1.8 RR 10 20 30 0.6 1.2 1.8 RR 10 20 30 0.6 1.0 1.4 RR 10 20 30 0.6 1.0 1.4 RR 10 20 30 0.8 1.2 1.6 RR 10 20 30 1.0 1.4 1.8 RR 10 20 30 1.0 1.4 1.8

Average temperature (degree C)

Respir ator y Circulator y All causes

Taipei Taichung Tainan Kaohsiung

Fig. 2. Associations between cause-specific emergency room visits and daily average temperatures in studied cities. Relative risks were estimated using DLNM, controlling for con-secutive temperature extremes, daily city-specific averages of PM10, NOx, O3, relative humidity, wind speed, daily isolation rate for influenza A, influenza B, and adenovirus, holiday

Forsberg, 2009; Wang et al., 2011). The present study is thefirst to

evaluate how intensified long-duration extreme temperature events

and how annual _{first extreme temperature events are associated}

with ERV risks for all causes and cardiopulmonary diseases in a sub-tropical island in Southeastern Asia. The climate of the 4-study cities is somewhat hot and humid. However, associations between the mor-bidities in residents and the weather changes vary among the 4 cities.

Thefirst extreme cold temperature in a year, longer duration, and

higher intensity of extreme cold events are associated with higher ERV risks for all causes and respiratory diseases. Furthermore, heat extremes are associated with higher ERV risks for respiratory dis-eases, but not for all causes and circulatory disdis-eases, which are

more likely associated with the annualfirst extreme heat event. The

present study shows that temperature–ERV associations are very

0.9 1 1.1 1.2 1.3 Kaohsiung Tainan Taichung Taipei Pooled 0.98 1.11 1.13 1 1.05 0.6 0.8 1 1.2 1.4 Kaohsiung Tainan Taichung Taipei Pooled 0.78 0.74 1.01 0.9 0.86 0.8 1 1.2 1.4 Kaohsiung Tainan Taichung Taipei Pooled 0.97 1.11 1.05 0.92 1 0.8 0.9 1 1.1 1.2 Kaohsiung Tainan Taichung Taipei Pooled 1.07 1.02 0.95 0.89 0.98 0.8 1 1.2 1.4 1.6 Kaohsiung Tainan Taichung Taipei Pooled 0.98 1.02 1.16 1.08 1.07 0.8 1 1.2 1.4 Kaohsiung Tainan Taichung Taipei Pooled 1.16 1.08 0.96 0.93 1.02

Relative risk (95% CI)

18 degree C

18 degree C

All causes Circulatory Respiratory

Fig. 3. City-specific and pooled risk estimates for overall 4-day effects at 18 °C and 30 °C temperatures, compared with the average temperature with the lowest emergency room visits at 24 °C. Relative risks by city-specific average temperature were estimated using DLNM, controlling for consecutive temperature extremes, daily city-specific averages of PM10, NOx, O3, relative humidity, wind speed, daily isolation rate for influenza A, influenza B and adenovirus, holiday effects, day of the week, and long-term trends. The pooled

risk estimates were obtained using random-effect meta-analysis.

0.8 1 1.2 1.4 95th(3−5d) 95th(6−8d) 95th(9d+) 97th(3−5d) 97th(6−8d) 97th(9d+) 99th(2−3d) 99th(4d+) 1 1.02 1.04 0.98 1.02 1.01 1 0.95 0.6 1 1.4 95th(3−5d) 95th(6−8d) 95th(9d+) 97th(3−5d) 97th(6−8d) 97th(9d+) 99th(2−3d) 99th(4d+) 0.97 0.93 1.01 1 0.78 1.16 0.94 0.91 0.8 1.2 1.6 95th(3−5d) 95th(6−8d) 95th(9d+) 97th(3−5d) 97th(6−8d) 97th(9d+) 99th(2−3d) 99th(4d+) 1.07 1.04 1.19 1.04 1.1 1.08 1.11 1.02 0.8 1.2 1.6 2 1st(2−3d) 1st(4d+) 5th(3−5d) 5th(6−8d) 5th(9d+) 10th(3−5d) 10th(6−8d) 10th(9d+) 1 1.31 0.92 0.92 1.18 0.99 0.99 0.98 0.7 1 1.3 1st(2−3d) 1st(4d+) 5th(3−5d) 5th(6−8d) 5th(9d+) 10th(3−5d) 10th(6−8d) 10th(9d+) 1.08 1.02 0.95 0.99 1.14 0.96 0.94 0.92 0.75 1.25 1.75 2.25 2.75 3.25 1st(2−3d) 1st(4d+) 5th(3−5d) 5th(6−8d) 5th(9d+) 10th(3−5d) 10th(6−8d) 10th(9d+) 0.97 1.77 0.86 0.91 1.28 1 1.05 0.98

Relative risk (95% CI)

All causes Circulatory Respiratory

Cold extremes

Heat extremes

Fig. 4. Four city pooled relative risks of emergency room visits from all causes, circulatory diseases, and respiratory diseases associated with consecutive extreme temperatures lasting for 3–5 days, for 6–8 days, and >8 days by random-effect meta-analysis.

different from temperature_{–mortality associations (}Anderson and Bell, 2009; Gasparrini and Armstrong, 2011; Hajat et al., 2006;

Kovats et al., 2004; Lin et al., 2011).

Only a few studies have reported varying threshold temperatures

associated with morbidity.Lin et al. (2009)found that temperature

threshold ranging from 28.9 °C to 29.4 °C is critical for the hospitaliza-tion of patients with respiratory and cardiovascular diseases in New

York City (Lin et al., 2009).Kovats et al. (2004)reported that thresholds

varied with admission causes ranging from 6 °C to 24 °C in Greater

London, UK (Kovats et al., 2004). Linares and Díaz also reported in

2007 that there were more hospital admissions for residents above 75 years old, with a 34 °C as the threshold maximum temperature in

Madrid, Spain (Linares and Diaz, 2008). Tong et al. reported that 27 °C

was threshold temperature for ERV and death in Brisbane (Tong et al.,

2010).

In Taiwan, the threshold temperature varies with the disease

causes, ranging from 25 °C to 27 °C in temperature_–mortality

associa-tion (Lin et al., 2011). In the present study, the threshold ranges from

10 °C to 33 °C for the temperature–ERV association. Residents in cities

located at the lower latitude (Tainan and Kaohsiung) are more sensitive to the exposure to low temperatures. They are likely to have higher

mortality risks from circulatory and respiratory diseases (Lin et al.,

2011) and higher ERV risks for all causes and respiratory diseases. Similar

to a previous study (Kovats et al., 2004), the risks of high temperatures

are significantly associated with mortality, but not with emergency

room visits in Taiwanese population.

The intensity, the duration, and the timing of extreme tempera-ture events are the key factors for evaluating the mortality impact

of extreme weather conditions (D'Ippoliti et al., 2010). A Czech

Re-public study found an excess mortality from cardiovascular always

appears in thefirst winter cold spell for population (Kysely et al.,

2009). However, there are limited studies on the association between

morbidity and thefirst or prolonged extreme temperature in the year.

Studies on morbidity have reported that a hot humid weather that

lasts longer than 4 days is significantly associated with an increase

in hospital admissions (Mastrangelo et al., 2007). Heat waves that

occur in early summer seem to cause greater health effects in Attica,

Athens (Theoharatos et al., 2010). Our previous mortality analysis

showed that the increase of mortality risk is associated with stronger

and prolonged heat extremes, but there is no significant association

observed for prolonged extreme cold event (Lin et al., 2011;Kysely

et al., 2009). The present study has identi_{fied different patterns for}

as-sociations between ERVs and extreme temperatures. Both annualfirst

extreme heat and extreme cold events cause noticeable ERV risks. To

the best of our knowledge, this study is the_{first report that the first}

day of cold temperatures in the year has the greatest impact on ERVs. This phenomenon is probably a unique for population residing on a subtropical island. They have generally accustomed themselves

to mainly warm weather. We suspect that thefirst extreme cold

weather is associated with the interruption of medical care for chronic conditions, triggering patients with chronic cardiovascular and respira-tory conditions at risk for a greater disease exacerbation. It is probably

more important to establish a warning system for thefirst cold extreme

than heat extreme, particularly for the emergency care facilities. Short-term extreme temperatures cause greater adverse effects on mortality from circulatory diseases than the long-lasting extreme

temperature event (Lin et al., 2011). Most studies on ERV for circulatory

diseases fail to identify the positive association between temperature,

extreme temperature event and hospital admission (Kovats et al.,

2004; Linares and Diaz, 2008; Mastrangelo et al., 2007; Michelozzi et

al., 2009; Rocklov and Forsberg, 2009). On the contrary, mortality

from and morbidity for respiratory diseases are consistently sensitive to temperature changes and intensify prolonged extreme temperature

events (Basu, 2009; Kovats et al., 2004; Lin et al., 2009, 2011; Linares

and Diaz, 2008; Mastrangelo et al., 2007; Michelozzi et al., 2009;

Rocklov and Forsberg, 2009). Linares and Díaz proposed that people

die from circulatory diseases rapidly before they can be admitted to

hospitals when they are exposed to high temperatures (Linares and

Diaz, 2008). We observed a similar pattern in the population of Taiwan.

Studies have associated air pollutants, such as ozone and PM10,

with mortality from cardiovascular and respiratory diseases during

heat waves (Anderson and Bell, 2009; Hajat et al., 2006;

Medina-Ramon and Schwartz, 2007). Moreover, air pollutants like NO2, O3,

and particulate matter were also reported correlated with the

in-creased ERV of cardiopulmonary diseases (Peel et al., 2005; Tsai et

al., 2009). The present study confirmed the significant association

be-tween ambient levels of PM10 and O3 and the ERV of respiratory

diseases.

Some studies have linked the circulating respiratory viruses with

ERV (Bourgeois et al., 2006; Olson et al., 2007). However, this study

failed to identify the extraordinary risks for these respiratory viruses as we had expected. Previous studies reported that winter seems to have a higher positive rate of identifying the circulating respiratory

viruses.(Denny, 1995; Diaz et al., 2007) In Taiwan, only Flu A and

Flu B are more likely to have seasonal variation (data not shown). Cold air could enhance the susceptibility of sensitive population by increasing vasoconstriction in upper airways and depressing the

clearance mechanism of infections (Chu et al., 2006; Shephard and

Shek, 1998). Additionally, transmission of suspected infectious

bioaerosol (e.g. influenza viruses) has been reported negatively related

with ambient temperature (Lowen et al., 2007). The seasonal

tempera-tures and the biological interactions deserve further study. All causes 0.9 1.1 1.3 1st 5th 10th 95th 97th 99th 1.09 1.12 0.99 0.98 1.04 1.08 Circulatory 0.6 1 1.4 1st 5th 10th 95th 97th 99th 1 1.09 0.99 1.13 1.05 1.23 Respiratory 0.8 1.2 1.6 1st 5th 10th 95th 97th 99th 1.15 1.26 0.97 0.92 1.01 1.04

Relative risk (95% CI)

Annual first temperature extreme by definition

Fig. 5. Pooled ERV risk for all causes, circulatory diseases, and respiratory diseases as-sociated with the annual occurrence of thefirst extreme temperature event (the 99th, 97th, 95th, 10th, 5th, and 1st percentile temperatures).

Lin et al. and Gasparrini and Armstrong used DLNM to evaluate the main effects of daily temperature and the additional effects of

ex-treme heat events (Gasparrini and Armstrong, 2011; Lin et al.,

2011). They found that the general temperature explains most of

the excess risk of mortality, and that heat extremes contribute little

additional effects even if it sustains for more than 4 days. Temperature_–

ERV association is different from temperature–mortality association

(Kovats et al., 2004; Linares and Diaz, 2008; Michelozzi et al., 2009).

Wargon et al. concluded in a study that using the weather condition

to predict emergency room utilization is less practical (Wargon et al.,

2009). Our study has conflicting findings. The annual first extreme

tem-perature event and intensified long-duration extreme cold event

dem-onstrate more significant association with ERV than the daily

temperature. Therefore, improving the predictive ability of extreme temperatures and enhancing the warning system in advance to the public is important.

5. Conclusion

This population-based study used daily representative morbidity data and accurate well-matched day-to-day climate and air pollution data. Data analysis completed the concise measures. The present

study is thefirst to evaluate how the intensity, the duration and the

timing of extreme temperature events are associated with ERV for populations exposed to a subtropical climate, controlling for potential

confounders. Intensified long-duration extreme cold events are

sig-nificant conditions leading to higher ERV risks for all causes and

re-spiratory diseases. ERV risk for rere-spiratory diseases is also partially

associated with extreme heat event. The annualfirst extreme heat

event with the temperature of 99th percentile is associated with ele-vated ERV for all causes and cardiopulmonary diseases, which have to be included in the warning system as well. Therefore, establishing the predictive ability of extreme temperatures and enhancing the early warning system for the public, particularly on more vulnerable popu-lations, are important.

Authors' contribution

All authors have been involved in this study. YC Wang, YK Lin, CY Chuang, MH Li, CH Chou, CH Liao and FC Sung have designed and obtained funds. YC Wang analyzed data. YC Wang, YK Lin, and FC

Sung drafted and _{finalized the manuscript. All authors have read}

and approved thefinal version of the manuscript.

Acknowledgments

We want to thank the of_{ficers of Taiwan National Health Research}

Institute, Taiwan Environmental Protection Administration, Taiwan Centers for Disease Control, and Taiwan Central Weather Bureau for providing research data. The interpretation and the conclusions con-tained herein do not represent the views of these agencies. The pre-sent study was supported in part by the Taiwan National Science Council (NSC 99-2221-E-033-052 and NSC 99-2621-M-039-001), the China Medical University Hospital (1MS1), and the Taiwan Department of Health (DOH100-TD-B-111-004, DOH100-TD-C-111-005).

References

Akaike H. Information theory and an entension of the maximum likelihood principle. 2nd International Symposium on Information Theory; 1973. p. 267–81. Anderson BG, Bell ML. Weather-related mortality: how heat, cold, and heat waves

af-fect mortality in the United States. Epidemiology 2009;20:205–13.

Anderson GB, Bell ML. Heat waves in the United States: mortality risk during heat waves and effect modification by heat wave characteristics in 43 U.S. communities. Environ Health Perspect 2011;119:210–8.

Armstrong B. Models for the relationship between ambient temperature and daily mortality. Epidemiology 2006;17:624–31.

Basu R. High ambient temperature and mortality: a review of epidemiologic studies from 2001 to 2008. Environ Health 2009;8:40.

Bourgeois FT, Valim C, Wei JC, McAdam AJ, Mandl KD. Influenza and other respiratory virus-related emergency department visits among young children. Pediatrics 2006;118:e1–8.

Braga AL, Zanobetti A, Schwartz J. The effect of weather on respiratory and cardiovas-cular deaths in 12 U.S. cities. Environ Health Perspect 2002;110:859–63. Bureau of National Health Insurance. National Health Insurance Profile. [in Chinese]

Taipei: Bureau of National Health Insurance, Department of Health, Executive Yuan, Republic of China; 2001.

Chu YH, Wu CC, Wang HW. Effect of cooling on electricalfield stimulation and norepinephrine-induced contraction in isolated hypertrophic human nasal mucosa. Am J Rhinol 2006;20:471–5.

Curriero FC, Heiner KS, Samet JM, Zeger SL, Strug L, Patz JA. Temperature and mortality in 11 cities of the eastern United States. Am J Epidemiol 2002;155:80–7. Denny Jr FW. The clinical impact of human respiratory virus infections. Am J Respir Crit

Care Med 1995;152:S4-S12.

Diaz J, Garcia R, Velazquez de Castro F, Hernandez E, Lopez C, Otero A. Effects of ex-tremely hot days on people older than 65 years in Seville (Spain) from 1986 to 1997. Int J Biometeorol 2002;46:145–9.

Diaz A, Barria P, Niederman M, Restrepo MI, Dreyse J, Fuentes G, et al. Etiology of community-acquired pneumonia in hospitalized patients in Chile: the increasing prevalence of respiratory viruses among classic pathogens. Chest 2007;131: 779–87.

D'Ippoliti D, Michelozzi P, Marino C, de'Donato F, Menne B, Katsouyanni K, et al. The impact of heat waves on mortality in 9 European cities: results from the EuroHEAT project. Environ Health 2010;9:37.

Gasparrini A, Armstrong B. The impact of heat waves on mortality. Epidemiology 2011;22:68–73.

Gasparrini A, Armstrong B, Kenward MG. Distributed lag non-linear models. Stat Med 2010;29:2224–34.

Hajat S, Kovats RS, Atkinson RW, Haines A. Impact of hot temperatures on death in London: a time series approach. J Epidemiol Community Health 2002;56:367–72.

Hajat S, Armstrong B, Baccini M, Biggeri A, Bisanti L, Russo A, et al. Impact of high tem-peratures on mortality: is there an added heat wave effect? Epidemiology 2006;17:632–8.

Hemon D, Jougla E, Clavel J, Laurent F, Bellec S, Pavillon G. Surmortalite liee a la cani-cule d'aout 2003 en France. Bull Epidemiol Hebd 2003;45–46:221–5.

Huynen MM, Martens P, Schram D, Weijenberg MP, Kunst AE. The impact of heat waves and cold spells on mortality rates in the Dutch population. Environ Health Perspect 2001;109:463–70.

Keatinge WR, Donaldson GC. Cardiovascular mortality in winter. Arctic Med Res 1995;54(Suppl. 2):16–8.

Knowlton K, Rotkin-Ellman M, King G, Margolis HG, Smith D, Solomon G, et al. The 2006 California heat wave: impacts on hospitalizations and emergency depart-ment visits. Environ Health Perspect 2009;117:61–7.

Kovats RS, Hajat S, Wilkinson P. Contrasting patterns of mortality and hospital admis-sions during hot weather and heat waves in Greater London, UK. Occup Environ Med 2004;61:893–8.

Kysely J, Pokorna L, Kyncl J, Kriz B. Excess cardiovascular mortality associated with cold spells in the Czech Republic. BMC Public Health 2009;9:19.

Lin S, Luo M, Walker RJ, Liu X, Hwang SA, Chinery R. Extreme high temperatures and hospital admissions for respiratory and cardiovascular diseases. Epidemiology 2009;20:738–46.

Lin Y-K, Ho T-J, Wang Y-C. Mortality risk associated with temperature and prolonged temperature extremes in elderly populations in Taiwan. Environ Res 2011;111: 1156–63.

Linares C, Diaz J. Impact of high temperatures on hospital admissions: comparative analysis with previous studies about mortality (Madrid). Eur J Public Health 2008;18:317–22.

Lowen AC, Mubareka S, Steel J, Palese P. Influenza virus transmission is dependent on relative humidity and temperature. PLoS Pathog 2007;3:1470–6.

Mastrangelo G, Fedeli U, Visentin C, Milan G, Fadda E, Spolaore P. Pattern and determi-nants of hospitalization during heat waves: an ecologic study. BMC Public Health 2007;7:200.

McCullagh P, Nelder JA. Generalized linear models. Boca Raton, Florida: Chapman & Hall/CRC; 1989.

Medina-Ramon M, Schwartz J. Temperature, temperature extremes, and mortality: a study of acclimatization and effect modification in 50 United States cities. Occup Environ Med 2007;64:827–33.

Medina-Ramon M, Zanobetti A, Cavanagh DP, Schwartz J. Extreme temperatures and mortality: assessing effect modification by personal characteristics and specific cause of death in a multi-city case-only analysis. Environ Health Perspect 2006;114:1331–6.

Michelozzi P, Accetta G, De Sario M, D'Ippoliti D, Marino C, Baccini M, et al. High tem-perature and hospitalizations for cardiovascular and respiratory causes in 12 Euro-pean cities. Am J Respir Crit Care Med 2009;179:383–9.

Olson DR, Heffernan RT, Paladini M, Konty K, Weiss D, Mostashari F. Monitoring the im-pact of influenza by age: emergency department fever and respiratory complaint surveillance in New York City. PLoS Med 2007;4:e247.

Pattenden S, Nikiforov B, Armstrong BG. Mortality and temperature in Sofia and Lon-don. J Epidemiol Community Health 2003;57:628–33.

Peel JL, Tolbert PE, Klein M, Metzger KB, Flanders WD, Todd K, et al. Ambient air pollu-tion and respiratory emergency department visits. Epidemiology 2005;16:164–74. Rocklov J, Forsberg B. Comparing approaches for studying the effects of climate ex-tremes— a case study of hospital admissions in Sweden during an extremely warm summer. Glob Health Action 2009;2.

Shephard RJ, Shek PN. Cold exposure and immune function. Can J Physiol Pharmacol 1998;76:828–36.

Shih SR, Chen GW, Yang CC, Yang WZ, Liu DP, Lin JH, et al. Laboratory-based surveil-lance and molecular epidemiology of influenza virus in Taiwan. J Clin Microbiol 2005;43:1651–61.

Taiwan Centers for Disease Control. Taiwan's virus surveillance network. Available at:

http://www.cdc.gov.tw/ct.asp?xItem=31996&ctNode=948&mp=52011. Accessed Date: 2011/05/12.

Taiwan Central Weather Bureau. Illustrations of surface observation data. Available at:http:// www.cwb.gov.tw/eng/observe/real/ob_station.htm2011. Accessed Date: 2011/05/12. Taiwan Central Weather Bureau. The Central Weather Bureau online information.

Avail-able at:http://www.cwb.gov.tw/eng/index.htm2011. Accessed Date: 2011/05/12. Taiwan Environmental Protection Administration. Taiwan air quality monitoring

net-work. Available at:http://taqm.epa.gov.tw/taqm/en/PsiAreaHourly.aspx2011. Accessed date: 2011/05/12.

Taiwan Governmental Information Office. The Republic of China year book 2008: chap-ter 12 environmental protection. Taipei. Available at: http://www.gio.gov.tw/tai-wan-website/5-gp/yearbook/2008/ch12.html2008. Accessed Date: 2011/05/12. Taiwan National Health Insurance Research Database. Available at:http://w3.nhri.org.

tw/nhird/en/Data_Subsets.html#S32011. Accessed Date: 2011/05/12.

Theoharatos G, Pantavou K, Mavrakis A, Spanou A, Katavoutas G, Efstathiou P, et al. Heat waves observed in 2007 in Athens, Greece: synoptic conditions, bioclimatolo-gical assessment, air quality levels and health effects. Environ Res 2010;110: 152–61.

Thomas DC. Statistical methods in environmental epidemiology. Oxford University Press; 2009.

Tong S, Wang XY, Barnett AG. Assessment of heat-related health impacts in Brisbane, Australia: comparison of different heatwave definitions. PLoS One 2010;5:e12155. Tsai SS, Chiu HF, Wu TN, Yang CY. Air pollution and emergency room visits for cardiac arrhythmia in a subtropical city: Taipei, Taiwan. Inhal Toxicol 2009;21:1113–8. Viechtbauer W. Conducting meta-analyses in R with the metafor package. J Stat Softw

2010;36:1-48.

Wang XY, Barnett AG, Yu W, Fitzgerald G, Tippett V, Aitken P, et al. The impact of heat-waves on mortality and emergency hospital admissions from non-external causes in Brisbane, Australia. Occup Environ Med 2011 June 30.doi:10.1136/oem.2010. 062141/. [Electronic publication ahead of print].

Wargon M, Guidet B, Hoang TD, Hejblum G. A systematic review of models for forecast-ing the number of emergency department visits. Emerg Med J 2009;26:395–9.