Effect of extreme hot and cold weather on cause-specific hospitalizations in Sweden: A time series analysis

(1)

Environmental Research 193 (2021) 110535

Available online 30 November 2020

Effect of extreme hot and cold weather on cause-specific hospitalizations in Sweden: A time series analysis

Osvaldo Fonseca-Rodríguez, Ph.D.

^a^,^b^,^*

, Scott C. Sheridan, Ph.D.

^c

, Erling H¨aggstr¨om Lundevaller, Ph.D.

^b

, Barbara Schumann, Ph.D.

^a^,^b

aDepartment of Epidemiology and Global Health, Umeå University, 901 85, Umeå, Sweden

bCentre for Demographic and Ageing Research, Umeå University, 901 87, Umeå, Sweden

cDepartment of Geography, Kent State University, Kent, OH, 44242, USA

A R T I C L E I N F O Keywords:

Cardiovascular hospitalizations Respiratory hospitalizations Spatial synoptic classification Hot weather

Cold weather Sweden

A B S T R A C T

Considering that several meteorological variables can contribute to weather vulnerability, the estimation of their synergetic effects on health is particularly useful. The spatial synoptic classification (SSC) has been used in biometeorological applications to estimate the effect of the entire suite of weather conditions on human morbidity and mortality. In this study, we assessed the relationships between extremely hot and dry (dry tropical plus, DT+) and hot and moist (moist tropical plus, MT+) weather types in summer and extremely cold and dry (dry polar plus, DP+) and cold and moist (moist polar, MP+) weather types in winter and cardiovascular and respiratory hospitalizations by age and sex. Time-series quasi-Poisson regression with distributed lags was used to assess the relationship between oppressive weather types and daily hospitalizations over 14 subsequent days in the extended summer (May to August) and 28 subsequent days during the extended winter (November to March) over 24 years in 4 Swedish locations from 1991 to 2014. In summer, exposure to hot weather types appeared to reduce cardiovascular hospitalizations while increased the risk of hospitalizations for respiratory diseases, mainly related to MT+. In winter, the effect of cold weather on both cause-specific hospitalizations was small; however, MP+ was related to a delayed increase in cardiovascular hospitalizations, whilst MP+ and DP + increased the risk of hospitalizations due to respiratory diseases. This study provides useful information for the staff of hospitals and elderly care centers who can help to implement protective measures for patients and residents. Also, our results could be helpful for vulnerable people who can adopt protective measures to reduce health risks.

1. Introduction

The effect of heat and cold on cardiovascular and respiratory mortality has been studied extensively in a wide range of different geographic regions (Chen et al., 2017; Scovronick et al., 2018; Urban and Kysely 2018; Yang et al., 2018), including Sweden (Fonseca-Ro- dríguez et al., 2020; Rockl¨ov et al. 2011, 2014). Other studies have addressed the role of weather variability on total and cause-specific mortality before and during industrialization in Sweden (Astrom et al., 2016; Rocklov et al., 2014; Schumann et al., 2013).

Nevertheless, fewer studies have reported the effect of temperature on morbidity outcomes (e.g., hospitalizations) compared to studies on mortality impacts (Son et al., 2014). Furthermore, most studies on

morbidity have focused on the effect of temperature without considering other weather variables, showing partly opposing results (Kovats and Hajat 2008; Ma et al., 2011; Michelozzi et al., 2009; Phung et al., 2016;

Sheridan and Lin 2014; Turner et al., 2013; Urban et al., 2014; Urban and Kysely 2018; Wang et al., 2009; Wichmann et al., 2013).

Apart from ambient temperature, other variables such as relative humidity have been demonstrated to be associated with mortality and hospital admissions (Davis et al., 2016; Rockl¨ov and Forsberg 2010;

Zhang et al., 2014). Morbidity and mortality might increase with higher temperature partly due to the combination with high humidity, stressing the thermoregulatory system (McGregor and Vanos 2018; Vanos et al., 2010). Studies on the effect of humidity on hospitalizations have shown contradictory results (Lin et al., 2009; Schwartz et al., 2004). Other

* Corresponding author. Department of Epidemiology and Global Health, Umeå University, 901 85, Umeå, Sweden.

E-mail addresses: osvaldo.fonseca@umu.se (O. Fonseca-Rodríguez), ssherid1@kent.edu (S.C. Sheridan), erling.lundevaller@umu.se (E.H. Lundevaller), barbara.

schumann@umu.se (B. Schumann).

Contents lists available at ScienceDirect

Environmental Research

journal homepage: www.elsevier.com/locate/envres

https://doi.org/10.1016/j.envres.2020.110535

Received 19 September 2020; Received in revised form 18 November 2020; Accepted 23 November 2020

(2)

studies (Barnett et al., 2010; Kim et al., 2011) did not show differences between several temperature metrics (minimum, maximum, and mean) and indexes, such as apparent temperature (Steadman 1984) and the humidex (CCOHS 2019), in predicting mortality. Despite the strong physiological evidence, there is a lack of epidemiologic support for the worsening of heat stress due to humidity (Goldie et al., 2015). Davis et al. (2016) reported a large number of articles showing contradicting results regarding the effect of humidity on all-cause mortality and morbidity due to cardiovascular and respiratory diseases. Moreover, these health outcomes could be influenced by other factors, such as wind speed, solar radiation, and air pressure (Ferrari et al., 2012; Hajat et al., 2010), which have been shown to produce a collective effect on human health (Driscoll 1990; McGregor and Vanos 2018).

Considering that several meteorological variables can contribute to weather vulnerability, the estimation of their synergetic effects on health is particularly useful (Hondula et al., 2014). One approach, the spatial synoptic classification (SSC), has been used in biometeorological applications (Davis and Kalkstein 1990; Sheridan 2002) to estimate the

effect of the entire suite of weather conditions on human morbidity and mortality (Hondula et al., 2014; Sheridan 2002). The SSC classifies daily weather conditions into one of seven weather types: dry polar (DP), dry moderate (DM), dry tropical (DT), moist polar (MP), moist moderate (MM), moist tropical (MT), and transition (TR). This classification is based on air temperature, dew-point temperature, sea-level pressure, wind speed, and cloud opacity, which are measured every 6 h (Sheridan 2002). Characteristics of each weather type differ according to season (i.

e., DP in winter is colder than in summer) and geographic region (i.e., MT is hotter in more tropical regions than more poleward locations).

The SSC has been used in a wide variety of health-related studies (Fonseca-Rodríguez et al. 2019, 2020; Hondula et al., 2014; Kysely and Huth 2010; Sheridan and Lin 2014; Urban and Kysely 2018; Vanos et al., 2014), and it has been employed in several heat-health warning systems in the USA and other countries (Kalkstein and Sheridan 2007; Sheridan and Kalkstein 2004).

Certain weather types are more oppressive than others because they increase the mortality above the specific seasonal baseline (Sheridan

Fig. 1. Study locations. Black crosses represent the main weather stations in each study area (Malm¨o, MMX; Bromma, BMA; ¨Ostersund, OSD; and Umeå, UME).

(3)

et al., 2009). DT and MT in summer and DP and MP in winter are associated with increased all-cause and cause-specific mortality in Sweden (Fonseca-Rodríguez et al. 2019, 2020). We expected that the extremely hot and extremely cold fraction of those weather types could also affect the hospitalization due to cardiovascular and respiratory diseases. Hence, this study aimed to assess the relationships between the extreme subsets of the hot (dry tropical plus [DT+] and moist tropical plus [MT+]) and cold (dry polar plus [DP+] and moist polar plus [MP+]) weather types and cardiovascular and respiratory hospitalizations according to age and sex in summer and winter, respectively, in four Swedish locations.

The present study is the first to analyze the effect of extreme weather types on cause-specific hospitalizations in Sweden, and adds to the body of literature on the weather-morbidity relationship. Findings will help the Swedish health care system to be prepared for the future when the number of vulnerable people will likely increase, and climate change might pose a challenge for population health and the health system.

2. Materials and methods 2.1. Study locations and data

The study was conducted in four Swedish locations, two in the

south—southwest Skåne (Scania) county, surrounding Malmö, with 21 municipalities, and Stockholm county with 26 municipalities—and two in the north—Jämtland county, centered on Östersund, with 12 municipalities, and eastern Västerbotten county where 5 municipalities in the Umeå region were included (Fig. 1). Selected locations represent different climatic and environmental conditions and can thus give a good picture of regional variations. The population in these locations is relatively large compared to other southern and northern locations in Sweden, which increases the sample size.

Meteorological data consisting of hourly and daily temperature, dew point, wind speed and direction, air pressure at sea level, and cloud cover were used to categorize the daily weather. These data were obtained from weather stations having no or very few missing values (Table 1), which allows us to categorize the daily weather based on the SSC methodology. In order to improve the completeness of data, meteorological observations from neighboring weather stations inside the corresponding study location were used. Missing hourly data were completed by interpolation from adjacent hours. We replaced under 2%

of observations at each station; hence, the spatial and temporal interpolation had a negligible impact on the daily weather’s characterization.

The same method was previously described (Fonseca-Rodríguez et al.

2019, 2020).

According to the K¨oppen-Geiger climate classification system, Table 1

Sample sizes of weather types and health outcomes for the four study locations in Sweden, 1991–2014.

Season Location Weather

Type Days Cardiovascular disease hospitalizations Respiratory disease hospitalizations

All Men Women <65

years

≥65

years All Men Women <65

years

≥65 years

Summer SKÅNE Other

Weather Type

2,884 160,400 90,279 70,116 42,558 117,837 62,305 32,667 29,637 27,426 34,878

DT+ 21 1,147 647 500 296 851 410 227 183 177 233

MT+ 47 2,374 1,349 1,025 608 1,766 1,045 541 504 419 626

STOCKHOLM Other Weather Type

2,789 533,604 294,753 238,832 143,091 390,494 204,822 104,852 99,960 90,912 113,900

DT+ 47 8,378 4,606 3,771 2,194 6,183 3,303 1,691 1,612 1,385 1,918

MT+ 111 19,832 10,917 8,912 5,200 14,629 7,733 3,869 3,864 3,015 4,718

J¨AMTLAND Other Weather Type

2,881 26,673 15,368 11,305 6,800 19,873 11,465 6,052 5,413 5,271 6,194

DT+ 32 247 144 103 60 187 102 58 44 40 62

MT+ 36 286 164 122 71 215 132 57 75 53 79

V¨ASTERBOTTEN Other Weather Type

2,898 41,196 23,087 18,109 10,717 30,479 14,175 7,460 6,715 6,742 7,433

DT+ 13 156 88 68 26 130 49 24 25 18 31

MT+ 32 387 215 172 94 293 111 52 59 37 74

Winter SKÅNE Other

Weather Type

3,399 209,352 117,895 91,445 56,391 152,949 105,555 54,159 51,390 51,049 54,500

DP+ 88 5,201 2,914 2,287 1,383 3,818 2,934 1,459 1,475 1,410 1,524

MP+ 143 8,360 4,631 3,728 2,273 6,086 4,398 2,188 2,210 1,970 2,428

STOCKHOLM Other Weather Type

2,770 581,825 321,733 260,066 158,461 423,338 289,471 146,339 143,118 141,077 148,380

DP+ 260 51,659 28,567 23,092 13,811 37,848 27,321 13,627 13,689 13,170 14,146

MP+ 599 124,014 68,670 55,334 34,221 89,783 59,112 30,270 28,839 29,162 29,947

J¨AMTLAND Other Weather Type

3,073 31,317 17,884 13,433 7,968 23,349 16,360 8,678 7,682 8,218 8,142

DP+ 107 1,049 610 439 266 783 535 295 240 275 260

MP+ 448 4,574 2,624 1,950 1,179 3,395 2,197 1,140 1,057 1,046 1,151

V¨ASTERBOTTEN Other Weather Type

2,844 45,822 25,813 20,009 12,078 33,744 20,747 10,660 10,086 10,486 10,260

DP+ 418 6,458 3,689 2,769 1,730 4,728 3,056 1,559 1,497 1,545 1,511

MP+ 356 5,406 3,071 2,335 1,462 3,944 2,346 1,238 1,108 1,146 1,200

Weather types: DT+ = Dry Tropical plus, MT+ = Moist Tropical plus, DP+ = Dry Polar plus, MP + Moist Polar plus.

Missing days in summer: Skåne = 0, Stockholm = 5, Jämtland = 3, Västerbotten = 9; Missing days in winter: Skåne = 0, Stockholm = 1, Jämtland = 2, Västerbotten = 12.

(4)

Sweden has three different climate zones and is dominated by the warm humid continental climate (Dfb), which includes our two southern study locations, Skåne and Stockholm. The two northern study locations, J¨amtland and V¨asterbotten, are in the subarctic climate (Dfc) region.

The third climate present in Sweden is the oceanic climate (Cfb), and it covers areas in the south of the country, including part of Skåne (Climate Data.org, 2020).

The study period was from January 1, 1991, to December 31, 2014.

Daily hospitalizations due to cardiovascular causes (ICD-9: 390–459, ICD-10: I00–I99) and respiratory causes (ICD-9: 460–519, ICD-10:

J00–J99), based on the International Statistical Classification of Dis- eases and Related Health Problems (ICD) (WHO 1978; WHO 2016), were obtained from the Linnaeus database at the Centre for De- mographic and Ageing Research (CEDAR), Umeå University, Sweden.

The Linnaeus database has data on the entire Swedish population; the Swedish personal identification numbers are used to link individual records from Swedish register data on death date and causes, hospitalization, and various socioeconomic conditions (Malmberg et al., 2010).

We stratified the number of daily cardiovascular and respiratory hos- pitalizations in groups by sex and age (<65 and ≥ 65 years) for each location. The hospitalization database is of very high quality, the level of completeness is assumed 100%, and there are no missing days in the hospitalization database.

Data from hourly and daily measurements for temperature, wind speed, precipitation, cloud cover, dew point, and air pressure were obtained from the Swedish Meteorological and Hydrological Institute (SMHI), recorded at the weather stations located in each of the study areas. The weather stations Malmö A (MMX), Bromma Airport (BMA), Ostersund (OSD), and Umeå Airport (UME) are located in Skåne, ¨ Stockholm, Jämtland, and Västerbotten, respectively. The geographic locations of the four study areas and weather stations are shown in Fig. 1.

The daily weather at each location was classified into one of the seven SSC weather types—DP, DM, DT, MP, MM, MT, and TR—according to the SSC (Sheridan 2002). The meteorological data consisted of temperature, dew point, wind speed and direction, air pressure at sea level, and cloud cover, measured every 6 h. The process has been described in more detail in our previous studies (Fonseca-Ro- dríguez et al. 2019, 2020). In this study, we assessed the effect of extreme hot and cold weather types. The hottest fraction of the tropical weathers (DT and MT) was classified as DT+ and MT+, respectively;

similarly, the coldest fraction of the polar weathers (DP and MP) was classified as DP+ and MP+, respectively.

DT+ and MT + subsets occur when the morning and afternoon apparent temperature values are both more than one standard deviation above weather-type means for the specific location; DP+ and MP + occur on days when the morning and afternoon apparent-temperature values are both more than one standard deviation below the weather- type means at the specific location (Sheridan and Kalkstein 2004).

The characteristics of the weather types for each location and month are shown in Table S1.

2.2. Statistical analysis

The study period was divided into two seasons, extended summer, from May to August, and extended winter, from November to March, which were assessed separately. We estimated the cumulative effect over 14 days of the two extremely hot weather types (DT+ and MT+) on hospitalizations due to cardiovascular and respiratory diseases in summer. Likewise, in winter, we assessed the cumulative effect over 28 days of DP+ and MP+ (the coldest weather types) on hospitalizations due to cardiovascular and respiratory diseases. We used a shorter maximum lag in summer than in winter because it has been demonstrated that heat has a more immediate effect on morbidity than cold (Ye et al., 2012). Hence, we created binary variables for each of the weather types considered in summer (DT+ and MT+) and in winter (DP+ and MP+), respectively,

and days with different classifications (DP, DM, DT, MP, MM, MT) served as the reference category. The presence and frequency of the weather types depend on the season (e.g., MT and DT did not occur in winter).

The total cases for each location, cause of hospitalization by season, and weather type, including the frequency of weather type, were computed (Table 1). The effect of the weather types on cause-specific hospitalizations was analyzed by time series analysis with quasi- Poisson regression. The quasi-Poisson model accounts for over- dispersion by fitting an extra dispersion parameter. We applied a distributed lag non-linear model (DLNM) to estimate the exposur- e–response function over the study locations. The dlnm package was used to create a cross-basis function to represent the delayed non-linear exposure–outcome relationships (Gasparrini 2011a; Gasparrini et al., 2010). The logknots function defines knots for lag space at equally spaced log-values and describes the lag–response function (Gasparrini et al., 2019).

The model used in our study was as follows:

Yt̃quasi − Poisson(μ_t,θ) Var (Yt) =θμ

log(μ_t) =α+β crossbasis(weather types) + γ ns(year) + δ ns(DOY) +εDOW +Offset(log ofPopulation)

The model outcome was daily hospitalizations (Yt) due to cardiovascular and respiratory diseases in each region. The cross-basis matrix of binary variables (weather types) was constructed for each of the weather categories. A natural spline (ns) was used to control for long- term trends (year), with three degrees of freedom. We accounted for the seasonal effect using a natural spline fitted to the days of the year (DOY), with three degrees of freedom. In addition, we controlled for the day of the week by including the categorical variable DOW (day of the week). The daily population under risk was used as the offset variable. α is the intercept, and β, γ, δ, and ε are the other coefficients.

The total population at risk and the population stratified by sex and age were linearly interpolated to the daily level by annual population counts for each location. Sub-analyses were conducted, stratifying the population by sex and age (<65, ≥65 years), in order to assess possible differences in weather vulnerability. In addition, splitting analyses by age group (<65 and ≥ 65 years), we make our results easier to compare with other studies that use a similar age stratification. We computed the relative risk (RR) of cause-specific hospitalization with a 95% confidence interval (CI) for each weather type.

We tested different combinations of maximum lags from 7 to 14 days in summer and from 21 to 28 days in winter using equally spaced knots in the cross-basis matrix, the number of degrees of freedom (df) for the lag-response function from 3 to 6, and df for seasonal and long-term trends. Parameters were chosen based on our previous studies conducted in these locations. Apart from lagged weather effects on disease occurrence, we also considered that people could delay the visit to health facilities for different reasons. One standard model was selected in order to facilitate the comparison among weather types, seasons, and locations. The final model was selected based on the quasi-Akaike information criterion (QAIC). In the final model, we used a maximum lag of 14 days in summer and 28 days in winter and 3 df for the lag-response function in both seasons. In sensitivity analyses, we removed the seasonality parameter from the model, and the month was used as a categorical variable. These, however, did not improve the result, and the changes were not substantial.

Finally, in order to have collective effect estimates for the northern region (V¨asterbotten and J¨amtland) and southern region (Skåne and Stockholm), the RRs were pooled using a multivariate meta-analysis with the mvmeta package (Gasparrini 2011b). All statistical analyses were performed using R statistical software, version 3.6.0 (R

(5)

Development Core Team 2019).

3. Results

Table 1 shows the number of days classified as other weather types and oppressive weather types (summer, DT+, MT+; winter, DP+ and MP+) and the number of hospitalizations due to cardiovascular and respiratory diseases, stratified by sex and age group.

Our data show a strong seasonality of hospitalizations, with a substantial dip in July, which differs from the mortality patterns that showed a smoothed reduction of deaths in summer. In the supplementary file, the time series of the daily cases of cardiovascular (Supple- mentary Figure 1) and respiratory hospitalizations (Supplementary Figure 2) by location is also shown. The monthly rate of cardiovascular and respiratory hospitalizations by location are shown in Supplementary Figure 3 and Supplementary Figure 4, respectively. In this study, we were interested in estimating the lag-distributed relative risk. Hence, the cumulative effect estimates of DT+ and MT + over 14 days in summer and DP+ and MP + over 28 days in winter were part of the analysis results and are presented in Supplementary Table 2.

3.1. Summer: cardiovascular diseases

In the southern locations, Skåne and Stockholm, the dry and very hot DT + weather type, appeared to reduce the risk of hospitalization due to cardiovascular diseases in the total population and all subgroups during the entire lag period of 14 days. MT+ (humid and very hot weather) was associated with a reduced risk of hospitalization between lags 3–11 among the entire population, men, and elderly people (≥65 years old) and after day 8 for women and young people (<65 years old).

In the northern locations of J¨amtland and V¨asterbotten, DT + seemed to produce a protective effect in the total population and all subgroups at shorter lags; however, the effect was less clear, showing very wide confidence intervals. MT + showed a delayed reduction of risk, with higher precision at medium lag, around day 10, in the total population, women, and elderly people (Fig. 2).

3.2. Summer: respiratory diseases

Hospitalizations due to respiratory diseases tended to increase in southern locations due to DT + among men and elderly people in the first few days after exposure. The effect of MT + on respiratory hospitalizations in all groups was immediate, larger, and lasted for approxi- mately 5 days, followed by a delayed risk reduction until lag 14 (Fig. 3).

In northern locations, DT + had an immediate decreasing effect on hospitalizations due to respiratory diseases in the total population, men, and people under 65 years of age. Likewise, a risk reduction was observed among elderly people between lags 4 and 7. The risk of hospitalization because of respiratory causes due to MT + weather rose at shorter lags in the general population, women, and elderly people;

nevertheless, it showed a risk reduction from around lag 10 onward.

However, we should take into account the low precision, mainly in the northern locations, because of the wide confidence intervals (Fig. 3).

3.3. Winter: cardiovascular diseases

In the south, DP + did not produce any effect on cardiovascular diseases. On the other hand, MP + slightly reduced the risk at shorter lags (0–2 days), but the risk increased after that in all subgroups (Fig. 4).

In the north, the risk of cardiovascular diseases due to DP + weather was reduced immediately and during the first 5 days in the total population and all population strata, except for people under 65 years of age.

Also, MP + increased the effect of cardiovascular diseases at longer lags for women and elderly people; on the other hand, among people under 65 years of age, the risk increased at shorter lags (Fig. 4).

3.4. Winter: respiratory diseases

In southern locations, DP+ was associated with a considerable increase in hospitalizations due to respiratory diseases between days 9 and 20 after exposure for all groups except people under 65 years of age. MP +reduced the risk at shorter lags in all groups, with a delayed increase, except for elderly people (Fig. 5).

In the north, DP + showed a protective effect at shorter lags, while

Fig. 2. Relative risk (RR) with 95% confidence interval of cardiovascular disease hospitalizations due to extreme hot weather types (dry tropical—DT+ and moist tropical—MT+) in two southern and two northern locations in summer (May through August) for the total population (TOTAL), stratified by sex (MEN and WOMEN) and age (<65 and ≥ 65 years).

(6)

the risk increased at longer lags in the total population, men, and elderly groups and at medium lags in women. MP + appeared to produce a protective effect at shorter lags in the total population, men, and elderly.

On the other hand, this weather type increased the risk of hospitalization after lag 20 in women and young people.

4. Discussion

This is one of the few studies analyzing the effect of weather variables on hospitalizations in Sweden and the first to estimate morbidity effects of weather types based on the SSC in this country. Specifically, we studied the effect of extreme heat (DT+ and MT+) and extreme cold (DP+ and MP+) weather types on hospitalizations due to cardiovascular Fig. 3. Relative risk (RR) with 95% confidence interval of respiratory disease hospitalizations due to extreme hot weather types (dry tropical—DT+ and moist tropical—MT+) in two southern and two northern locations in summer (May to August) for the total population (TOTAL), stratified by sex (MEN and WOMEN) and age (<65 and ≥ 65 years).

Fig. 4. Relative risk (RR) with 95% confidence interval of cardiovascular disease hospitalizations due to extreme cold weather types (dry polar—DP+ and moist polar—MP+) in two southern and two northern locations in winter (November to March) for the total population (TOTAL), stratified by sex (MEN and WOMEN) and age (<65 and ≥ 65 years).

(7)

and respiratory diseases.

4.1. Summer

Based on studies of cardiovascular mortality, it would be expected that hot weather types would increase hospitalizations due to cardiovascular diseases. However, the results of our study pointed in the opposite direction. The very hot weather types (DT+ and MT+) reduced the risk of hospitalization due to cardiovascular diseases in the total population and the subgroups according to sex and age. Interestingly, in our previous study conducted in the same study areas, we found a clearly increased risk due to the hot weather types DT and MT on cardiovascular mortality in southern locations, with MT being the most oppressive weather type (Fonseca-Rodríguez et al., 2020). However, the extreme hot fraction of those weather types reduced the risk of hospitalization, with MT+ (hot and humid) showing the most evident risk reduction.

Similar results were found by Urban and Kysely (2018) in Prague, Czech Republic, on the opposing effects of hot weather types DT and MT, with both increasing mortality and reducing morbidity due to cardiovascular diseases. Likewise, Urban et al. (2014) found no relevant short-term effect of heat on cardiovascular morbidity in either sex. In addition, cardiovascular morbidity showed a reduction due to heat, while mortality risk increased, similar to our findings in the present and previous studies (Fonseca-Rodríguez et al. 2019, 2020). Also, a study conducted in New York, USA, found no effect of heat (DT or MT) on cardiovascular hospitalizations, while heat did increase mortality (Sheridan and Lin 2014).

In general, higher temperatures are not associated with an increase in cardiovascular hospitalizations (Kovats and Hajat 2008). Even protective effects have been reported in locations of northern Europe, including Stockholm, supporting our findings (Michelozzi et al., 2009).

A study from Gothenburg, Sweden, found a risk reduction of hospitalizations by myocardial infarctions in the total population for both sexes and age groups (≤75 and > 75 years) (Wichmann et al., 2013). In En- gland and Wales, however, no evidence of any adverse effect of heat on myocardial infarctions was found (Bhaskaran et al., 2010). Other authors have shown increased mortality and morbidity due to

cardiovascular diseases associated with heat (Schwartz et al., 2004;

Tong et al., 2010), while no effect of humidity has been found (Schwartz et al., 2004).

Heat exposure has a negative effect on cardiovascular diseases by causing dysfunction of thermoregulatory mechanisms (Bouchama and Knochel 2002; Steadman 1984; Tian et al., 2012). The heat causes a body response, producing vasodilation and leading to sweating, loss of fluids and salt, and an increase of the workload, which can have fatal consequences for people with congestive heart failure (N¨ayh¨a 2005). It can also cause hemoconcentration and increase the risk of thrombosis (Keatinge et al., 1986).

We observed a decrease in cardiovascular hospitalizations, the opposite effect on mortality (Fonseca-Rodríguez et al., 2020). We as- sume that the different effects of hot weather types on cardiovascular disease hospitalizations and mortality are due to the rather immediate occurrence of heat-related deaths: a considerable number of people die from an acute episode of cardiovascular disease before being hospitalized, as previously suggested by other authors (Bouchama 2004;

Michelozzi et al., 2009; Rockl¨ov 2010). In this regard, our results are also coincident with other studies (Kovats et al., 2004; Linares and Diaz 2008; Urban et al., 2014; Urban and Kysely 2018; Zacharias et al., 2014). In particular elderly people who live alone could be more vulnerable; they may die at home before being able to reach the closest hospital (Zacharias et al., 2014).

The effect of the hot weather types (DT+ and MT+) on respiratory diseases was immediate, increasing the risk of hospitalizations, particularly in the south. The effect of MT+ was larger than the effect of DT+, despite the higher temperatures reached during DT+ in contrast with MT+. However, MT + weather is also characterized by humid conditions with high overnight temperatures (Supplementary Table 1). This characteristic of the MT weather type is highly relevant, considering that heat-related illnesses tend to be more sensitive to minimum rather than maximum temperatures (Hayhoe et al., 2010). Additionally, we must consider that air conditioning is not common in Swedish households, increasing the exposure of the population to high minimum temperatures. In our previous study (Fonseca-Rodríguez et al., 2020), MT had a larger impact on respiratory mortality than DT. In New York, Sheridan Fig. 5. Relative risk (RR) with 95% confidence interval of respiratory disease hospitalizations due to extreme cold weather types (dry polar—DP+ and moist polar—MP+) in two southern and two northern locations in winter (November to March) for the total population (TOTAL), stratified by sex (MEN and WOMEN) and age (<65 and ≥ 65 years).

(8)

and Lin (2014) found a non-significant increase in hospitalizations due to respiratory diseases related to DT and MT weather types, while respiratory mortality increased due to both hot weather types.

Lin et al. (2009) demonstrated the effect of extremely high apparent temperatures on the immediate risk of hospital admissions for respiratory diseases. They showed that the interaction of high relative humidity and high temperature intensifies the impact on respiratory hospitalizations, while low relative humidity moderated the effect of high temperature. They also revealed a similar effect on males and females; on the other hand, the effect of high temperatures increased with age, particularly in the elderly population. Likewise, an increasing daily temperature has also been associated with a higher risk of respiratory hospitalizations in the total population, but particularly in the elderly population (Kovats et al., 2004). Other studies have also reported an increase in emergency room visits for respiratory diseases in relation to extremely hot temperatures (Lavigne et al., 2014). In addition, our results are also supported by a meta-analysis that found a significant increase in morbidity due to respiratory diseases in elderly people, attributed to heat (Bunker et al., 2016).

Specifically, in our southern study locations, the risk of respiratory hospitalizations was associated with the very humid and hot MT + weather type in all groups, while the dry and hot DT + affected more men and the elderly population. Roy´e et al. (2016) observed that respiratory diseases were linked to a hotter and humid weather type in Galicia, Spain. In Stockholm, a significantly high risk of respiratory hospital admissions was reportedly associated with a high apparent temperature; the increase in risk was evident in all age categories, but it was particularly high among the elderly population (Michelozzi et al., 2009).

The results obtained by Lin et al. (2009) suggest an interactive effect of relative humidity and temperature on health, with a larger impact of heat at higher humidity levels. Heat might cause inflammation of the airways, leading to dyspnea. This can affect people with a vulnerable health condition because of a pre-existing chronic respiratory disease (Lavigne et al., 2014; Michelozzi et al., 2009). Thus, temperature-related inflammation could explain the increasing number of respiratory hospitalizations associated with heat.

Heat can also affect respiratory health in general through thermo- regulation responses and the inhalation of hot air (Bunker et al., 2016; Li et al., 2019). In humans, the primary mechanism of body heat regulation is by sweating, and excessive heat increases ventilation, causing thermal hyperpnea (White 2006). Hyperventilation may trigger dynamic hy- perinflation and dyspnoea, particularly in people with pre-existing respiratory diseases such as chronic obstructive pulmonary disease (Anderson et al., 2013), due to the persistent pulmonary and systemic inflammation and ventilatory impairment in these patients (Mannino et al., 2006). Furthermore, inhaling hot air, mainly if the air is humid, may lead to acute bronchoconstriction of the airway, impairing respiration in asthmatic patients (Hayes et al., 2012).

In contrast to cardiovascular diseases, the risk of respiratory hospitalization due to MT + increased at shorter lags. Respiratory disease symptoms rarely lead to sudden death, unlike cardiovascular diseases;

thus, respiratory disease patients may be more likely to reach the closest hospital compared to those who suffer a stroke or myocardial infarction.

During the study period, the case fatality rate of cardiovascular diseases was 2.57 (95% CI 2.54–2.59) times higher than respiratory diseases, supporting this assumption.

4.2. Winter

In the southern locations, the moist and very cold weather (MP+) was associated with an increased risk of hospitalization due to cardiovascular diseases at longer lags in all groups. In the north, this weather type increased the risk of hospitalization only in women and elderly people. In the United Kingdom, the risk of myocardial infarction (specific cardiovascular diseases) increased between days 8 and 14 with

lower temperatures. Additionally, an increased risk (though non- significant) was observed when the relative humidity was low or high (Bhaskaran et al., 2010). In the US, Schwartz et al. (2004) also showed that hospital admissions for heart disease increased at high and low temperatures, but no effect of relative humidity was detected. However, in our study, the humid and cold MP + increased cardiovascular hospitalizations, but not the dry and cold DP+, despite its lower temperatures. This could indicate that humidity amplifies the negative effect of low temperature on cardiovascular health.

The physiological response in the organism to reduce heat loss consists of peripheral vasoconstriction, it increases of hemoconcentration and blood pressure, causing an increase of the workload of the heart. Also, coronary circulation is affected, causing heart problems (Keatinge et al., 1984; N¨ayh¨a 2005). Additionally, it has been observed that exposure to cold can elevate fibrinogen levels in plasma, red cell and platelet counts, blood viscosity, and arterial pressure, increasing the risk of thrombosis (Keatinge et al., 1984; Woodhouse et al., 1994).

In our previously research, we showed an immediate increase in cardiovascular disease mortality related to MP (Fonseca-Rodríguez et al., 2020). In the present study, on the other hand, we found a small protective effect of extreme cold (MP+) at shorter lags and increased risk at longer lags. In this line, Urban et al. (2014) did not find any significant effect of cold on mortality and hospital admissions due to cardiovascular causes for either sex. In Gothenburg, Sweden, cold pro- duced a non-significant reduction of hospitalizations due to myocardial infarction at short lags (Wichmann et al., 2013), which is similar to our results regarding cold weather and cardiovascular hospitalizations mainly at shorter lags in southern locations, associated with MP+, and in northern locations, related to DP+. The protective effect or the lack of association between cold and cardiovascular hospitalizations could be explained by the housing conditions with good heating systems and insulation, which protect residents from cold weather in Sweden (Hanninen et al., 2011; Wichmann et al., 2013). However, MP+ is related to a delayed increase in cardiovascular hospitalizations in southern locations. MP weather is warmer than DP and warmer in the south than in the north. This suggests that the behavior of people might change depending on the registered temperatures. During the coldest days (DT+), people are likely to spend more time at home, while on milder days, they might be more willing to do outdoor activities and, consequently, be more exposed to low ambient temperatures (Martí- nez-Solanas and Basaga˜na, 2019).

Moreover, we found that extreme cold (MP+) was associated with a reduction of hospital admissions due to cardiovascular diseases at shorter lags. In general, extreme cold could be a hindrance and could postpone the hospital visit. This might explain the immediate reduction of hospitalizations due to cardiovascular diseases and the delayed increase in hospitalizations (Schwartz et al., 2004). Therefore, vulnerable people could die from cardiovascular diseases during extremely cold days with inclement weather conditions before going to the hospital.

Some studies have found that cardiovascular mortality increases during days with extreme weather conditions with very low temperatures and heavy snowfall (Baker-Blocker 1982; Gorjanc et al., 1999; Hopstock et al., 2012). DP + days are extremely cold, but for MP + days, snowfall could be abundant. Therefore, both cold-weather types might affect hospital visits.

Hospitalizations due to respiratory diseases increased at longer lags after exposure to cold weather during winter, with this effect being stronger in the south. In Galicia, Spain, the dry and cold weather was associated with an increase in respiratory hospital admissions (Roy´e et al., 2016). This is in line with findings from our southern and northern locations. DP has also been associated with influenza transmission in New York, US (Davis et al., 2012). Several studies have shown that exposure to cold air increases the number of viral respiratory infections, mainly in the upper respiratory tract (Kalkstein and DeFelice 2014;

M¨akinen et al., 2009).

The delayed effect of cold weather on people seeking medical

(9)

attention could be due to the incubation period of typical winter infections. In our study population, respiratory infections were some of the most frequent causes of hospital admission. Infectious diseases such as influenza A (H1N1), with median incubation periods of 1.4–1.6 days (Nishiura and Inaba 2011), or influenza A (H7N9), with a mean incubation period of 3.4 days (Virlogeux et al., 2015), might be the cause of cold-related hospitalizations, with a delay of several days. This might be the case for many other infectious respiratory diseases with high incidence in winter. People might also delay their visit to the hospital while the symptoms are not severe.

Exposure to cold weather has been related to respiratory diseases through different ways (Zhang et al., 2020). Low temperatures might reduce the lung function (Donaldson et al., 1999), or cause inflammation and bronchoconstriction affecting respiration (Koskela et al., 1996), and increased secretions in the airways may lead to an acute exacer- bation of chronic obstructive pulmonary disease (Li et al., 2011). Cold can also reduce the resistance to infections of the respiratory system (Eccles 2002; The Eurowinter Group 1997). Also, behavioral factors contribute to respiratory risk: During winter, the transmission of infections is favored by indoor crowding in rooms with low levels of humidity (Tellier 2009), among other factors.

The differences in the effect of dry and cold DP+ and the moist and cold MP + on respiratory hospitalizations could be related to the type of respiratory disease (e.g., infectious vs. non-infectious diseases).

Different infectious diseases, such as influenzas A and B, and influenza- like illnesses have peaked at different temperatures (Dai et al., 2018).

The humidity affects the transmission and survival of the influenza virus; low absolute humidity could favor both aspects, while high absolute humidity reduces the spread of the diseases (Shaman and Kohn 2009). In general, respiratory tract infections increase due to cold temperatures and low humidity. Also, a significant decrease in humidity has been associated with the occurrence of respiratory tract infections two weeks later (M¨akinen et al., 2009).

However, it must be considered that there are other seasonal effects on influenza and other infectious respiratory diseases (Davis et al., 2012). For example, in winter, in particular, during extreme cold or snowy and stormy weather, people tend to spend more time indoors, sometimes in crowded spaces, enhancing disease transmission (Lofgren et al., 2007).

Other respiratory problems could increase due to exposure to cold and high humidity (Roy´e et al., 2016). The cold could reduce the ability of the respiratory tract to expel secretions and foreign bodies. At the same time, high humidity facilitates an increase in microorganisms, such as fungi and bacteria, which can cause pneumonia, bronchitis, and asthma (Shaman and Kohn 2009). Nevertheless, cold weather is not only related to infectious diseases, but also to other respiratory diseases. In our study population, asthma was one of the top five respiratory causes for hospital admissions; the other respiratory causes were pneumonia, chronic diseases of the tonsils and adenoids, chronic obstructive pulmonary disease, and respiratory failure. People with asthma are more susceptible to functional disabilities or health problems because of exposure to cold weather compared to individuals without pre-existing respiratory diseases (Hyrk¨as-Palmu et al., 2018). The results obtained by Lee et al. (2012) indicated that the DP weather type in autumn and winter corresponds to increased hospital admissions and increased cases of asthma in New York City (NYC) for young people. Similar results have also been reported by other authors for New York and Greece (Jamason et al., 1997; Nastos et al., 2006).

In general, the prevalence of respiratory diseases increases during colder months and decreases during July in the northern hemisphere.

This seasonal effect is particularly pronounced in respiratory pathol- ogies (Roy´e et al., 2016). As mentioned, cold conditions favor the transmission of certain viruses (Shaman and Kohn 2009; World Health Organization (WHO) 2018). Also, COPD, pneumonia, asthma, and other respiratory diseases increase in cold months, showing a clear seasonal pattern (Hicks et al., 2018; Lee et al., 2012; Williams et al., 2017).

4.3. Strengths and limitations

Previous studies about the effect of weather variables on health outcomes worldwide and in Sweden focused on mortality rather than on morbidity. This is the first study to assess the effect of extreme hot and cold weather types based on the SSC on cause-specific hospitalizations in different Swedish locations in summer and winter over a 24-year period.

Thus, the combination of several aspects, such as the long study period, the southern and northern locations, and the climates, provided us a broad view of the relationship between weather and cardiovascular and respiratory diseases in the country. In addition, by stratifying the population by sex and age, we could show the specific effect of heat and cold on those groups.

On the other hand, some limitations must be addressed. The small number of daily hospitalizations due to cardiovascular and respiratory diseases in the northern locations, J¨amtland and V¨asterbotten, affected the precision of effect estimates of the very hot weather types during summer. In addition, the frequency of extremely hot weather types was low in the north, which also increased the uncertainty around the estimations in the northern study locations. Particular weather effects among subgroups were, therefore, difficult to assess.

The exposure variable has intrinsic limitations as well. While the concept of a weather type makes intuitive sense as a holistic measure of the atmospheric state, nevertheless, some information is lost when all weather data are categorized into one of several aggregates. There is, of course, within-weather type variability that cannot be accounted for in our method.

Thus, we emphasize that the SSC is a relative rather than an absolute classification system of the weather (Hondula et al., 2014; Sheridan 2002), and the same weather type also does have a different range of temperature, humidity, etc. in different seasons and geographic regions.

Further the adaptation to heat and cold varyvaries among populations, as physiological, behavioral, and technological adaptation may sub- stantially modify climate–health relationships (McMichael et al., 2006).

Thus, we acknowledge that the results of this study, showing the effect of extreme hot (DT+ and MT+) and cold (DP+ and MP+) weather types, cannot be directly extrapolated to other geographic areas with different climates. However, results can be generalized to countries with similar climates and socioeconomic structures, but it has to be done with caution.

5. Conclusions

In this study, we assessed the effect of extremely hot and cold weather types on hospitalizations due to cardiovascular and respiratory diseases, stratified by sex and age, during summer and winter in southern and northern Sweden. The present study offers important information about the relationship between extremely hot and cold weather types and cardiovascular and respiratory hospitalizations, complementing preceding studies on all-cause (Fonseca-Rodríguez et al., 2019) and cause-specific (Fonseca-Rodríguez et al., 2020) mortality linked to oppressive weather types based on the SSC. However, we must consider that the conclusions of this study are mainly based on the results from southern locations because of the more precise effect estimations.

We observed some small overall risk increases in cardiovascular and respiratory diseases due to extreme hot and cold weather types, mainly in southern locations where the population at risk is large. Even though RRs are comparably small, given the large number of people exposed, the absolute number of hospitalized cases might be substantial. Future research should investigate the opposing effects of heat on cardiovascular deaths and hospitalizations, especially in summer, which could also give insights into possible failures in protecting vulnerable people from the heat. Partly protective effects for hospitalizations might be due to heat/cold-related mortality at short lags, indicating a need for faster health care access, especially for cardiovascular events.

(10)

Mainly the hot and humid (MT+), but also the dry and hot weather (DT+), immediately increased the risk of respiratory diseases among the elderly population in southern locations (less clear in the north) during summer. Patients with pre-existing respiratory conditions and their health care providers need to be aware of short-term risks during these days. They should take some protective measures, such as avoiding sun exposure, staying inside, and drinking abundant water. Hospitals and elderly care centers should have cool rooms for such high-risk groups, even in colder climate zones. In winter, the staff of health care centers must be aware that exposure to cold weather can trigger cardiovascular and respiratory diseases. In particular, vulnerable people must avoid unnecessary exposure to cold and should wear proper clothing and stay in warm and well-insulated indoor environments. Risk communication to health care providers and people with chronic diseases can be improved to avoid health risks related to heat and cold.

Credit author statement

Osvaldo Fonseca-Rodríguez: did the data curation, carried out the statistical analyses and visualization of results, and wrote the original draft. Scott C. Sheridan: reviewed and edited the manuscript, contributed to the methodology and conceptualization, and contributed to funding acquisition. Erling H¨aggstr¨om Lundevaller: reviewed and edited the manuscript. Barbara Schumann: reviewed and edited the manuscript, contributed to the study conceptualization, and was in charge of funding acquisition and project administration.

Funding

This research was funded by the Swedish Research Council Formas, grant number FR-2017/0009.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

This research was funded by the Swedish Research Council Formas, grant number FR-2017/0009. Open access funding provided by Umeå University.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.

org/10.1016/j.envres.2020.110535.

References

Anderson, G.B., Dominici, F., Wang, Y., McCormack, M.C., Bell, M.L., Peng, R.D., 2013.

Heat-related emergency hospitalizations for respiratory diseases in the Medicare population. Am. J. Respir. Crit. Care Med. 187, 1098–1103. https://doi.org/

10.1164/rccm.201211-1969OC.

Astrom, D.O., Edvinsson, S., Hondula, D., Rocklov, J., Schumann, B., 2016. On the association between weather variability and total and cause-specific mortality before and during industrialization in Sweden. Demogr. Res. 35, 991–1009.

Baker-Blocker, A., 1982. Winter weather and cardiovascular mortality in Minneapolis-St.

Paul. Am. J. Publ. Health 72, 261–265. https://doi.org/10.2105/ajph.72.3.261.

Barnett, A.G., Tong, S., Clements, A.C.A., 2010. What measure of temperature is the best predictor of mortality? Environ. Res. 110, 604–611. https://doi.org/10.1016/j.

envres.2010.05.006.

Martínez-Solanas `E, Basaga˜na, X., 2019. Temporal changes in the effects of ambient temperatures on hospital admissions in Spain. PloS One 14, e0218262. https://doi.

org/10.1371/journal.pone.0218262.

Bhaskaran, K., Hajat, S., Haines, A., Herrett, E., Wilkinson, P., Smeeth, L., 2010. Short term effects of temperature on risk of myocardial infarction in England and Wales:

time series regression analysis of the Myocardial Ischaemia National Audit Project (MINAP) registry. BMJ 341, c3823. https://doi.org/10.1136/bmj.c3823.

Bouchama, A., 2004. The 2003 European heat wave. Intensive Care Med. 30, 1–3.

https://doi.org/10.1007/s00134-003-2062-y.

Bouchama, A., Knochel, J.P., 2002. Heat stroke. N. Engl. J. Med. 346, 1978–1988.

https://doi.org/10.1056/NEJMra011089.

Bunker, A., Wildenhain, J., Vandenbergh, A., Henschke, N., Rockl¨ov, J., Hajat, S., Sauerborn, R., 2016. Effects of air temperature on climate-sensitive mortality and morbidity outcomes in the elderly; a systematic review and meta-analysis of epidemiological evidence. EBioMedicine 6, 258–268. https://doi.org/10.1016/j.

ebiom.2016.02.034.

CCOHS, 2019. Canadian Centre for Occupational Health and Safety). Humidex Rating and Work. http://www.ccohs.ca/oshanswers/phys_agents/humidex.html.

November 2019.

Chen, T.L., Li, X., Zhao, J., Zhang, K., 2017. Impacts of cold weather on all-cause and cause-specific mortality in Texas, 1990-2011. Environ. Pollut. 225, 244–251.

https://doi.org/10.1016/j.envpol.2017.03.022.

Climate Dataorg, 2020. Sweden Climate. Accessed June 15 2020. https://en.climate-dat a.org/europe/sweden-105/.

Dai, Q., et al., 2018. The effect of ambient temperature on the activity of influenza and influenza like illness in Jiangsu Province, China. ScTEn 645, 684–691. https://doi.

org/10.1016/j.scitotenv.2018.07.065.

Davis, R.E., Kalkstein, L.S., 1990. Development of an automated spatial synoptic climatological classification. IJCli 10, 769–794. https://doi.org/10.1002/

joc.3370100802.

Davis, R.E., Rossier, C.E., Enfield, K.B., 2012. The impact of weather on influenza and pneumonia mortality in New York City, 1975-2002: a retrospective study. PloS One 7, e34091. https://doi.org/10.1371/journal.pone.0034091.

Davis, R.E., McGregor, G.R., Enfield, K.B., 2016. Humidity: a review and primer on atmospheric moisture and human health. Environ. Res. 144, 106–116. https://doi.

org/10.1016/j.envres.2015.10.014.

Donaldson, G., Seemungal, T., Jeffries, D., Wedzicha, J., 1999. Effect of temperature on lung function and symptoms in chronic obstructive pulmonary disease. Eur. Respir.

J. 13, 844–849.

Driscoll, D.M., 1990. A perspective on weather-human response relationships. Int. J.

Environ. Stud. 36, 19–25. https://doi.org/10.1080/00207239008710580.

Eccles, R., 2002. An explanation for the seasonality of acute upper respiratory tract viral infections. Acta Otolaryngol. 122, 183–191. https://doi.org/10.1080/

00016480252814207.

Ferrari, U., et al., 2012. Influence of air pressure, humidity, solar radiation, temperature, and wind speed on ambulatory visits due to chronic obstructive pulmonary disease in Bavaria, Germany. Int. J. Biometeorol. 56, 137–143. https://doi.org/10.1007/

s00484-011-0405-x.

Fonseca-Rodríguez, O., Lundevaller, E.H., Sheridan, S.C., Schumann, B., 2019.

Association between weather types based on the spatial synoptic classification and all-cause mortality in Sweden, 1991–2014. Int. J. Environ. Res. Publ. Health 16, 1696. https://doi.org/10.3390/ijerph16101696.

Fonseca-Rodríguez, O., Sheridan, S.C., Lundevaller, E.H., Schumann, B., 2020. Hot and cold weather based on the spatial synoptic classification and cause-specific mortality in Sweden: a time-stratified case-crossover study. IJBm 64, 1435–1449. https://doi.

org/10.1007/s00484-020-01921-0.

Gasparrini, A., 2011a. Distributed lag linear and non-linear models in R: the package dlnm. J. Stat. Software 43, 1–20.

Gasparrini, A., 2011b. Package’mvMeta’. Multivariate Meta-Analysis and Meta- Regression.

Gasparrini, A., Armstrong, B., Kenward, M.G., 2010. Distributed lag non-linear models.

Stat. Med. 29, 2224–2234. https://doi.org/10.1002/sim.3940.

Gasparrini, A., Armstrong, B., Scheipl, F., 2019. Package ‘dlnm’, vol. 2, pp. 3–9.

Goldie, J., Sherwood, S.C., Green, D., Alexander, L., 2015. Temperature and humidity effects on hospital morbidity in Darwin, Australia. Annals of Global Health 81, 333–341. https://doi.org/10.1016/j.aogh.2015.07.003.

Gorjanc, M.L., Flanders, W.D., VanDerslice, J., Hersh, J., Malilay, J., 1999. Effects of temperature and snowfall on mortality in Pennsylvania. Am. J. Epidemiol. 149, 1152–1160. https://doi.org/10.1093/oxfordjournals.aje.a009770.

Hajat, S., et al., 2010. Heat-health warning systems: a comparison of the predictive capacity of different approaches to identifying dangerously hot days. Am. J. Publ.

Health 100, 1137–1144. https://doi.org/10.2105/AJPH.2009.169748.

Hanninen, O., et al., 2011. Seasonal patterns of outdoor PM infiltration into indoor environments: review and meta-analysis of available studies from different climatological zones in Europe. Air Quality Atmosphere and Health 4, 221–233.

https://doi.org/10.1007/s11869-010-0076-5.

Hayes, D., Collins, P.B., Khosravi, M., Lin, R.L., Lee, L.Y., 2012. Bronchoconstriction triggered by breathing hot humid air in patients with asthma role of cholinergic reflex. Am. J. Respir. Crit. Care Med. 185, 1190–1196. https://doi.org/10.1164/

rccm.201201-0088OC.

Hayhoe, K., Sheridan, S., Kalkstein, L., Greene, S., 2010. Climate change, heat waves, and mortality projections for Chicago. J. Great Lake. Res. 36, 65–73. https://doi.org/

10.1016/j.jglr.2009.12.009.

Hicks, A., Healy, E., Sandeman, N., Feelisch, M., Wilkinson, T., 2018. A time for everything and everything in its time - exploring the mechanisms underlying seasonality of COPD exacerbations. Int. J. Chronic Obstr. Pulm. Dis. 13, 2739–2749.

https://doi.org/10.2147/COPD.S146015.

Hondula, D.M., Vanos, J.K., Gosling, S.N., 2014. The SSC: a decade of climate-health research and future directions. Int. J. Biometeorol. 58, 109–120. https://doi.org/

10.1007/s00484-012-0619-6.

Hopstock, L.A., Fors, A.S., Bonaa, K.H., Mannsverk, J., Njolstad, I., Wilsgaard, T., 2012.

The effect of daily weather conditions on myocardial infarction incidence in a