Introduction

Climate change is the biggest global health threat of the twenty-first century, and it will affect directly or indirectly all populations (Haines et al. 2009). The primary health impacts are mediated through changes in weather, in particular by heat waves and changes in ambient temperatures. The direct effects of temperature such as heat-related mortality are likely to increase as temperatures rise towards the end of the twenty-first century, although the effects on overall mortality would be offset by adaptation of population (Donaldson et al. 2003a; Christidis et al. 2010) and lowering of winter mortality (Keatinge et al. 2000) even though this is contested (Staddon et al. 2014). Heat-related health hazards are particularly likely in northern areas, where temperature is predicted to rise most (IPCC Fifth Assessment Report 2013).

Deaths from cardiovascular, respiratory, and all causes increase not only during identified heat waves (Näyhä 1981; Hajat et al. 2006; Martiello and Giacchi 2010; Kollanus and Lanki 2014; Lee et al. 2016; Zhang et al. 2016) but also during a normal summer if temperature exceeds the optimal temperature for the area (Näyhä 2007; Hajat and Kosatsky 2010; Guo et al. 2013). Comparable increases are seen in hospital admissions due to cardiovascular (Semenza et al. 1999; Morabito et al. 2005) and respiratory diseases (Michelozzi et al. 2009). The vulnerable population groups include the elderly, people suffering from debilitating medical conditions, women, and the socially deprived and disadvantaged ones (Basu 2009; Hajat et al. 2010; Stafoggia et al. 2006). People living alone (Semenza et al. 1996), in isolation (Hajat et al. 2010), or in institutions and those unable to care for themselves (Bouchama et al. 2007) are adversely affected by summer heat. Physical exertion, either occupational (Fleischer et al. 2013) or leisure-time related (Centers for Disease Control and Prevention 2011), conveys an extra risk, as does inactivity due to confinement to bed (Semenza et al. 1996; Bouchama et al. 2007; Martiello and Giacchi 2010). People living in colder parts of the world may be more susceptible to the effects of heat stress than those living in warmer areas because they are less physiologically adjusted and have less environmental protection (Keatinge et al. 2000; Guo et al. 2013). Finally, also other co-morbid conditions such as obesity (Vandentorren et al. 2006) and alcoholism (Kilbourne et al. 1982) will further increase the hazards of heat exposure.

Most studies on heat-related adverse effects among the general population are limited to mortality and hospital admissions. However, deaths and acute episodes of disease only constitute a fraction of the entire spectrum of heat-related harms, and a comprehensive assessment should include subjective symptoms which may forecast more severe events (Josseran et al. 2010). Unlike other environmental hazards, heat-related symptoms are easily perceived and enable people to avoid the heat exposure, before any disease attacks occur (Gronlund 2014). However, only one small study has reported heat-related symptoms (e.g., shortness of breath) in the general population during a severe heat wave in Australia (Nitschke et al. 2013). In fact, the vast majority of people in the northern climate suffer from heat-related complaints even during a normal summer, with most of them being the elderly or women (Näyhä et al. 2014). As summer temperatures are predicted to rise and heat waves to occur more frequently and with higher intensity (IPCC Fifth Assessment Report 2014), a higher occurrence of heat-related adverse effects can be expected. It is therefore prudent to (1) identify the groups most at risk and (2) estimate the prevalence of complaints within these groups. Our previous study asking about heat-related symptoms in the National FINRISK 2007 Study focused on sex and age effects in 28 symptoms and complaints among the Finnish population but did not look for other vulnerable groups (Näyhä et al. 2014). The present paper using the same database determines the prevalence of heat-related cardiorespiratory symptoms in population subgroups classified according to a number of social, occupational, and health characteristics. We selected cardiorespiratory symptoms for study, because they can be meaningfully linked with cardiovascular and respiratory events which constitute the major part of heat-related mortality (Kilbourne 1999). Identification of subgroups at special risk could reduce the costs by targeting advice and intervention.

Material and methods

The area studied

Finland is a subarctic Northern European country with cold winters and cool, short summers. The study areas were south (the cities Turku and Loimaa with nine adjoining rural municipalities and the cities of Helsinki and Vantaa), east (the counties of North Savo and North Karelia), and north (the province of Oulu), as shown in Fig. 1. In the south, the population is mainly urban and the climate maritime, while in other areas, the population is more rural and the climate more continental. In 1981–2010, the mean temperature in July (the warmest month) ranged from +18 °C in the south (Helsinki) to +16 °C in the north (Oulu). In summer 2006, i.e., the summer preceding the present survey, the temperatures exceeded the long-term average by 1–2 °C in the areas studied, and the number of hot days (highest daily temperature ≥25 °C) compared with that in the reference period 1981–2010 was greater in the south (13 vs 1) but not in the north (8 vs 9).

Fig. 1
figure 1

The areas of the FINRISK 2007 cold-heat substudy. Isotherms are mean July temperatures, 1981–2010

Full size image

Study population

The National FINRISK Study is a series of nationwide surveys conducted in Finland since 1972 at 5-year intervals with the aim of monitoring the risk factors for chronic diseases. The participants of the present study consisted of a subsample of the FINRISK 2007 study conducted in January–March. A random sample of 2000 people aged 25–74 years, stratified by sex and 10-year age groups, was drawn in each of the five study areas. The entire sample comprised 9957 people, of which two thirds were invited to participate in a more detailed study on temperature-related symptoms and one third underwent a dietary survey (they did not participate in the cold-heat study). Out of all 9957 subjects, 6733 (68 %) participated in the main study, and 4193 subjects belonging to the temperature subsample attended at the clinic and were given the Oulu Cold and Heat Questionnaire (OCHQ 2007), which they were asked to return in a pre-paid envelope. The response rates in the temperature subsample and dietary survey were similar (63 and 62 %, respectively). The questionnaire was returned by 4007 subjects and linked with the FINRISK study data. The details of the study protocol are reported elsewhere (Vartiainen et al. 2010). The study protocol was reviewed and approved by Coordinating Ethics Committee of the Helsinki and Uusimaa Hospital District. All participants gave a written informed consent.

The questionnaires

The OCHQ 2007 asked about symptoms and complaints experienced in warm and hot weather (the English translation of the questionnaire is given elsewhere (Näyhä et al. 2014)). The questionnaire was designed by the study team at Kastelli Research Center, Oulu, which comprised specialists in thermophysiology, cardiology, chest medicine, physiatrics, epidemiology, occupational medicine, and public health science. The questions were designed based on the experience gained in previous FINRISK studies (Raatikka et al. 2007; Ikäheimo et al. 2014). The heat-related cardiorespiratory symptoms were elicited by asking the respondents to tick “yes” or “no” to whether “Does warm weather cause you to have any of the following symptoms?”: shortness of breath, prolonged cough or coughing bouts, wheezing of breath, increased excretion of mucus from the lungs, chest pain, and cardiac arrhythmias. As many of the symptoms may overlap, a composite variable denoted as cardiorespiratory symptoms was created as follows: if the respondent had experienced at least one of these six individual symptoms, the outcome was coded as 1, otherwise 0.

In the main FINRISK questionnaire, the participants were asked about diagnosed medical conditions; education; professional field; marital status; physical workload; leisure-time physical activity (Hu et al. 2003); smoking; and the usual quantity and frequency of beer, wine, and spirits consumed during the past 12 months (Sundell et al. 2008). Body height and weight were measured at the survey site (to an accuracy of 1 mm and 100 g, respectively) and converted to body mass index (BMI; kg/m2). Information on sex, age, and place of residence were available from the Finnish National Population Register Centre. The definitions and classification of the variables are shown in Table 1.

Table 1 Subjects classified according to demographic and personal characteristics
Full size table

Data analysis

The proportion of respondents who reported having experienced cardiorespiratory symptoms during warm weather was treated as the prevalence of individuals having such a tendency. Logistic regression was used to assess the relationship between heat-related cardiorespiratory symptoms (yes/no) and demographic and individual factors, one factor at a time, but adjusting for sex and age. A fully adjusted model was then fitted including all the variables. To allow for curvilinear age trends, age (in 1-year classes) was smoothed by natural cubic splines with 3 degrees of freedom. The results were first expressed as odds ratios (ORs) with their 95 % confidence intervals (CIs). The ORs express the relative odds for having cardiorespiratory symptoms in each class compared with a reference class. To have the actual prevalence for all classes of the explanatory factors, we calculated marginal predictions from the adjusted logistic regressions conditioned at mean values of all other factors in the model (Lane and Nelder 1982; Graubard and Korn 1999) (details are in Appendix). The adjusted prevalence in a given category then expresses the model-predicted prevalence of cardiorespiratory symptoms in a stereotypic individual having average values of all factors in the model. Adjusted prevalences were also calculated according to age, letting age vary from 25 to 74 years, and also at fixed values of professional and educational groups and pre-existing diseases. Compared with ORs, the model-adjusted predictions illustrate better the scale of the group differences. The calculations were performed using the R software, release 3.01 (R Development Core Team 2012).

Results

Characteristics of subjects

The mean age of the subjects was 51.1 years (men 51.8, women 50.5). Of the subjects, 46 % were men; 72 % were married or cohabiting; 15 % had academic education; 45 % were engaged in office work or services, 11 % in industry, and 3 % in agriculture or related work; and 41 % were economically inactive (students, housewives, pensioners, or unemployed). A physician-diagnosed cardiovascular disease was reported by 1111 subjects (28 %), and of these, 953 (86 %) had arterial hypertension and 201 (18 %) coronary heart disease. A diagnosed lung disease was reported by 417 subjects (11 %), 374 (90 %) of these having bronchial asthma and 71 (17 %) chronic bronchitis or emphysema. Further details on subject characteristics are in Table 1.

Crude prevalence of heat-related cardiorespiratory symptoms

Out of all 3811 subjects who answered any of the 6 questions on heat-related cardiorespiratory symptoms, 469 (12 %) reported at least 1 symptom (9 % of men, 15 % of women). Cardiac arrhythmia was reported by 6 % (4 % of men vs 9 % of women), chest pain by 2 % (2 vs 2 %), dyspnoea by 5 % (3 vs 7 %), cough by 2 % (1 vs 2 %), wheezing by 2 % (2 vs 2 %), and mucus production by 3 % (3 vs 3 %).

Figure 2 shows the age trend in the prevalence of all heat-related cardiorespiratory symptoms in the form of smoothed splines. Between the ages of 25 and 74 years, the prevalence increased from 3 to 28 % (2 to 24 % in men, 4 to 32 % in women), but among subjects having a pre-existing cardiovascular disease, the prevalence was higher throughout the age range (from 8 to 31 %) and still higher among those having a lung disease (12 to 56 %). Women had a higher prevalence than men, especially at older ages.

Fig. 2
figure 2

Prevalence of heat-related cardiorespiratory symptoms among all subjects and those having a diagnosed cardiovascular or lung disease. The circles are the empirical prevalences in each 1-year age interval. The line shows the prevalence smoothed by natural cubic spline with 3 degrees of freedom, and the shaded area is its 95 % confidence band

Full size image

The crude prevalence of cardiorespiratory symptoms also varied depending on a number of other factors, from 6 to 34 % (Table 2). The prevalence was particularly high among pensioners, the unemployed, and agricultural workers, with 3.8-, 3.0-, and 3.0-fold excesses, respectively, compared with participants engaged in industry. Having only a basic education was associated with a 3.1-fold prevalence compared with having an academic education. The subjects who were physically inactive during leisure time showed a 2.8-fold higher prevalence than the active ones, but physical workload was not associated with the symptoms. The prevalence of symptoms increased with increasing body mass index, reaching a 2-fold excess among the severely obese (BMI ≥35 kg/m2) compared with normal weight individuals (BMI <25 kg/m2), and the prevalence was relatively high among women, heavy alcohol consumers, current smokers, those residing in the north or east, and the divorced or widowed ones.

Table 2 Prevalence of heat-related cardiorespiratory symptomsa and odds ratios (ORs) from adjusted logistic regressions, together with their 95 % confidence intervals (CIs)
Full size table

A pre-existing cardiovascular disease was associated with a 2.8-fold prevalence compared with those having no such disease (23 vs 8 %), and a subgroup analysis showed a higher prevalence ratio separately for coronary heart disease (4.0; 44 vs 11 %) and a lower one (2.0; 20 vs 10 %) for arterial hypertension. The subjects having a pre-existing lung disease showed a 3.4-fold prevalence compared with those having no such disease (34 vs 10 %), with somewhat different prevalence ratios separately for bronchial asthma (3.2; 32 vs 10 %) and chronic bronchitis (4.2; 50 vs 12 %). These specific conditions were grouped into cardiovascular and respiratory disease groups for further analyses.

Adjusted analyses

Most variations in the crude prevalence were repeated in the sex- and age-adjusted analyses (Table 2), and further adjustments for all other variables caused only minor changes. The ORs for region of residence and smoking reduced to insignificance as did those for BMI, mainly due to inclusion of cardiovascular and lung diseases to the model. However, being engaged in heavy physical work now showed elevated odds for having the symptoms.

The right-hand column of Table 2 translates the full adjusted model to adjusted prevalence figures in each subgroup assuming average values of all other explanatory factors. The adjustment reduced some of the high-prevalence figures, mostly to less than 20 %, notably among pensioners, the divorced/widowed ones, severely obese subjects and those having only basic education.

However, the estimated combined effects of two or more factors were much larger. This is illustrated in Fig. 3 in form of model-predicted age patterns in the prevalence of cardiorespiratory symptoms, stratifying by sex, cardiovascular and lung diseases, professional field, and educational level. First, the rising overall age trend was greatly reduced compared with the unadjusted trends. At the age of 74 years, for example, the adjusted prevalences of symptoms among all participants were only 7 and 12 % among men and women, respectively, while the unadjusted prevalences in Fig. 2 were 24 and 32 %, respectively. Thus, only about a third of the symptom prevalence (7/24 % in men; 12/32 % in women) could be attributed to age alone.

Fig. 3
figure 3

Model-adjusted prevalence of heat-related cardiorespiratory symptoms by age, separately for all subjects (ALL), those having pre-existing cardiovascular (CVD) and lung disease (LD), and those having no such diseases (NoCVD, NoLD), separately for four professional fields and educational classes

Full size image

Wide variations in prevalence still existed when stratifications were made by disease and professional field (Fig. 3a). Among men and women aged 50 years, for example, who had cardiovascular and lung diseases, the prevalence of symptoms reached 34 and 50 %, respectively, and varied from 21 to 61 % depending on professional field. Figure 3b shows the respective prevalence estimates by educational level, with variations from 27 to 59 % at the age of 50 years, depending on educational class. Figure 3 also illustrates the female excess in the estimated prevalence. Thus, at the age of 50 years, for example, the prevalences of cardiorespiratory symptoms among women and men who worked in industry and had no cardiovascular or lung disease were 6 and 3 %, respectively, i.e., a difference of 3 %, while the respective figures among pensioners who had a cardiovascular and lung disease were 61 and 45 %—a difference of 16 %. The respective sex differences were similar between respondents with academic vs basic education.

Discussion

Summary of findings

The present survey is the first one to identify and describe quantitatively the large spectrum of vulnerability factors which underlie heat-related cardiorespiratory symptoms. During a normal summer, only a moderate proportion (12 %) of people aged 25 to 74 years in this northern population suffer from heat-related cardiorespiratory symptoms, but the prevalence can exceed 60 % in specific groups with either one or multiple risk factors such as pre-existing cardiorespiratory diseases or poor social status. Identification of these groups will help us to target pre-emptive measures appropriately not only during heat wave proper (IPCC Fifth Assessment Report 2014) but also during a normal summer (Basu 2009; Hajat et al. 2006).

Factors underlying heat-related cardiorespiratory symptoms

Heat-related mortality and morbidity increase by age (e.g., Basu 2009; Michelozzi et al. 2009). We have previously shown that a wide range of heat-related symptoms and complaints also increase with age (Näyhä et al. 2014). However, the present results show that the high prevalence of such symptoms among the aged is mainly attributable to factors other than age. At the age of 74 years, for example, only about a third of the symptom prevalence could be attributed to age alone, the rest being due to cardiorespiratory morbidity and unfavorable personal characteristics such as overweight, excessive alcohol consumption, or being divorced or widowed. The effect of age is explained by age-related deterioration of thermoregulation, involving both physiological and behavioral changes (Stapleton et al. 2014), declining sweat output, diminished vasodilatation, lower skin blood flow and reduced cardiac output, diminished awareness of heat, lowered sensation of thirst (Kenny et al. 2010), and reduced mobility and cognition (Hansen et al. 2011).

Patients suffering from cardiovascular or lung diseases have an elevated risk for dying or being admitted to hospital during hot periods (Hajat et al. 2010), and we noted a high prevalence of heat-related cardiorespiratory symptoms among participants having such conditions. The body’s response to heat stress involves peripheral vasodilatation to shunt blood to the skin where heat can be lost by conduction. The resulting decrease in blood pressure due to the vasodilatation needs to be compensated with increased cardiac output, and people with any disease that impairs the ability to increase heart rate or stroke volume would be under increased cardiovascular strain. Furthermore, sweating to lose heat by evaporation imposes further stress on the heart through dehydration (Kenny et al. 2010). Dehydration decreases plasma volume and increases red blood cell concentration and blood viscosity, making it harder to circulate. Heat stress also causes the release of additional platelets into the circulation. These changes in blood properties may lead to an increased risk for coronary events (Donaldson et al. 2003b). Finally, an additional risk is conveyed by the medications some of these patients use, e.g., diuretics, beta-blockers, and anticholinergics (Stöllberger et al. 2009), which impair thermoregulation. Although the present study focused on the disease groups accounting for a majority of heat-related morbidity and mortality, we recognize that also other conditions, such as metabolic (e.g., diabetes), renal, neural, or psychiatric diseases, may be aggravated with heat exposure (Hajat et al. 2010).

We observed a substantially higher prevalence of cardiorespiratory symptoms among women than men. This is understandable in terms of women having a higher surface-to-mass ratio, greater subcutaneous fat thickness (Seidell et al. 1988), and lower sweat production rate (Dehghan et al. 2013). Women also have higher heat mortality (Basu 2009; IPCC Fifth Assessment Report 2014; Stafoggia et al. 2006). While consistent, the finding must be taken with caution, since women tend to report more health-related symptoms than men (Barsky et al. 2001).

The prevalence of heat-related cardiorespiratory symptoms increased consistently with lowering education and was high among pensioners, the unemployed, and participants engaged in agriculture. This is in line with studies reporting high heat mortality in low educational and poor socioeconomic groups (Vandentorren et al. 2006; Gronlund 2014). Agricultural workers frequently suffer from heat-related symptoms (Mirabelli et al. 2010; Kravchenko et al. 2013) and have high heat mortality (Gronlund 2014; Xiang et al. 2014). In our sample, hard physical work was associated with heat-related cardiorespiratory symptoms but leisure-time physical activity was not, inactive persons showing the highest prevalence. This could result from an over-representation in the inactive group of sick individuals who describe themselves as immobile because of ambulatory or motivational problems.

Earlier studies have shown that people living alone have higher heat mortality than others (Semenza et al. 1996; Bouchama et al. 2007), as have single, widowed, and divorced individuals (Stafoggia et al. 2006). We did observe a relatively high prevalence of heat-related cardiorespiratory symptoms among the divorced or widowed but not among the single ones. One might speculate that people who have been left alone after marriage are less able to care for themselves.

There are several comorbid or behavioral conditions that may account for increased heat-related health risks. Our findings are in line with studies reporting an elevated risk of heat hazards among obese individuals (Vandentorren et al. 2006). As we failed to confirm an independent association of obesity with heat-related cardiorespiratory symptoms, the finding could have been confounded by the cardiorespiratory diseases that obese individuals are more likely to suffer. Heavy consumption of alcohol diminishes the contractibility of the heart, lowers blood pressure, and may lead to dehydration (Hajat et al. 2010). Our finding of a high prevalence of heat-related cardiorespiratory symptoms among the heavy alcohol consumers suggests that the risk may be limited to a small-population segment. Smoking could be entertained as a risk factor for heat hazards as it worsens endothelial function and reduces the capacity of the skin vessels to dilate (Avery et al. 2009), but we did not observe any independent association of smoking with the symptoms with any certainty.

We also noted a higher prevalence of heat-related cardiorespiratory symptoms in northern than in southern Finland, which could be expected from studies reporting that heat-related mortality is higher (Keatinge et al. 2000) and the threshold temperature for heat mortality lower (Hajat and Kosatsky 2010; Guo et al. 2013) in northern than southern areas. However, our adjusted analyses could not confirm the initial finding.

Strengths

The strength of our study is the large, representative population living in a cold climate, where people are known to be more vulnerable to heat hazards than those living in a warmer climate (Keatinge et al. 2000; Guo et al. 2013). We had information on most personal and demographic characteristics increasing individual’s vulnerability, such as low education, which is a known risk factor for poor health (IPCC Fifth Assessment Report 2014). We focused on heat-related cardiorespiratory symptoms, which can be meaningfully interpreted as antecedents of severe cardiorespiratory events. The use of 1-year age classes and the flexible smoothing method allowing for curvilinear trends allowed us to determine the prevalence of symptoms at the highest ages, where they most likely occur. We also presented adjusted prevalence figures, which are more informative at the population scale than ORs alone. Since our findings pertain primarily to normal summer heat rather than heat waves proper, they have particular relevance in the northern climate, where extremely hot periods are rare. Only an estimated 20 to 50 % of all heat-related mortality is attributable to identified heat waves (Hajat et al. 2006).

Limitations

While the validity of the questions on heat-related cardiorespiratory symptoms would seem adequate, individual differences in threshold temperature at which symptoms become manifest may have caused an unknown bias to the prevalence figures. It is also possible that the respondent’s expectations of what (s)he should answer may have affected the results. We recognize that we focused on symptoms, not on actual health effects, and it remains unclear how well heat-related symptoms predict future morbidity and mortality. However, the outcome variable was composed of six separate questions on individual symptoms, which can be meaningfully linked with future cardiac and respiratory events. A composite variable was regarded as better than single questions, since in patients’ mind, the symptoms partly overlap. One limitation is that medications some subjects may have used were not taken into account. Use of pro re nata prescriptions such as nitroglycerin would lead to under-estimation of the effects we describe. The response rate was satisfactory, but some unknown bias due to selective participation remains a possibility.

Practical inmplications

The results of our study aids in early recognition of vulnerable groups for heat exposure. This is important as the most obvious strategy to prevent any heat harms is to target preventive measures to those at risk (Bouchama et al. 2007). Since these symptoms often occur before any actual disease attacks, the mass media should warn high-risk individuals some days before any significant heat waves (Diaz et al. 2006), and according to British experience, giving personal warnings by telephone may be useful (Bhaskaran et al. 2011). Pre-emptive measures include seeking shelter in cool premises, adequate fluid intake, light clothing, taking cool showers, and avoiding excessive physical exercise in work and leisure time and avoiding excessive alcohol consumption. People with cardiovascular or lung conditions should ask their doctors if their medication was to be adjusted.

Conclusions

We have identified a number of population subgroups with a high prevalence of heat-related cardiorespiratory symptoms. These are commonly perceived among people with pre-existing lung or cardiovascular disease, agricultural workers, unemployed, pensioners, and people having only basic education. Timely public health and individual measures taking into account individual susceptibility may improve adaptation to higher temperatures and prevent large loss of life during the warm season.