Among the 3485 respondents living in the very disadvantaged dissemination areas of the nine most populated cities within Québec, the prevalence of physical and mental health impacts during very hot and humid summer conditions was respectively 44.0% (CI: 42.2–45.7) and 17.8% (16.4–19.1). Combined together, the prevalence of health impacts was 46.0% (CI: 44.2–47.8). This latter prevalence was used in the analyses.
In this study, the prevalence of self-reported adverse health impacts during very hot and humid summer conditions was 46% among participants, mainly on physical health with some simultaneous mental health impacts. While high, this prevalence could correspond to the reality of very disadvantaged dissemination areas in Québec’s most populated cities, where several conditions strongly correlate with high temperatures and thermal discomfort indices [34
], as described in the introduction. The prevalence of impacts is, however, a subjective measurement, the reliability and validity of which is well established, as mentioned in the introduction, but heat exposure self-reporting should be further studied.
The risk indicators associated with the prevalence of self-reported health impacts identified subgroups more likely to feel the harmful effects of heat. Two of these subgroups were clearly associated with prevalence, independently of age. They consisted of people who described most of their days as being rather or extremely stressful and of people reporting at least two diagnoses of chronic diseases, particularly those who considered their state of health as being fair or poor. Women and people on long-term medical leave characterized only those under 65 years of age. Finally, households with very low income (<$15,000) and air conditioning at home are associated with both groups.
High daily stress helped to explain the prevalence of self-reported health impacts in a context of heat, for all ages. In this study, 25% of the respondents said they felt such stress, which is similar to the 26.3% documented in the general population of Québec (≥15 years of age) in 2007–2008 [43
]. Stress does not always lead to disease [46
]. When it is prolonged however it can become chronic and play a negative role in health, either by causing functional disorders (e.g., sleep disturbances, mood disorders), or by initiating and maintaining diseases (e.g., inflammatory diseases, depression), or by promoting the adoption of unsuitable behaviours (e.g., excessive alcohol consumption). In addition, chronic stress could be one of the vulnerability factors of people with age-related pathologies [47
]. The addition of thermal stress could therefore be sufficient for those individuals feeling high daily stress to perceive more health impacts in a context of heat.
The relationship between multimorbidity (≥2 diagnoses) and the prevalence of self-reported adverse health impacts clearly illustrates that the state of health determines physiological or biological susceptibility to heat, independently of age, corroborating what was found elsewhere [29
]. The bivariate analysis showed that multiple and various chronic disorders can place a heavy toll on the body in a context of heat and humidity (Table 4
). However, only multimorbidity (≥2 diagnoses) was associated with the prevalence of impacts in the multivariate analysis. In population surveys, this indicator is related to individual characteristics such as advancing age and polymedication, but also to organizational characteristics including an increase in medical consultations and the use of emergency departments [51
]. Thus, multimorbidity could be an indicator of choice for identifying people who must receive particular attention from public authorities in surveillance, as well as in active support and follow-up during episodes of very hot and humid summer conditions. To this end, the priority would be identifying chronic disorders which compromise thermoregulation [50
]. Individuals with such diseases would be particularly vulnerable to heat impacts, particularly if they are unable to seek refuge in a cool area. Validating the number of chronic disorders to use as the action threshold would be equally important to pursue.
The perception of being in poor health was strongly associated with the prevalence of health self-reported impacts in the bivariate analyses. However, in the multivariate analysis, its main contribution was its interaction with multimorbidity in the two age groups. The combination of these variables may take into account distinct characteristics of a person's state of health in a context of heat. As mentioned earlier, these perceptions can reflect certain health aspects difficult to identify clinically and are useful for predicting help-seeking behaviours and health service use [10
]. Multimorbidity refers to diagnoses of chronic diseases and therefore to a clinical measurement, although self-reported here. In the study, 27% of the respondents described their state of health as fair or poor, which was much higher than the 9.8% documented in the general population of Québec (≥12 years of age) in 2007–2008 [43
]. This difference could be due to the fact that our sample (≥18 years of age) was, on average, relatively old (53 years of age) and from the most disadvantaged neighbourhoods of the most populated cities, but a more detailed explanation eludes us at this time.
The prevalence of self-reported health impacts was higher for women than for men, but only for those under 65 years of age. The women may have had a tendency to consider a wider range of factors when they evaluated their impacts given their tendency to do so for perceived health status [10
]. On the other hand, this sex-related difference may be biological [53
]. For example, female sex hormones substantially affect certain bodily responses to heat [54
], as is the case for pain [55
]. From 45–64 years of age in particular, women go through menopause, a period characterized by a sudden and dramatic decrease in hormone levels, contrary to men, whose drop in hormone levels is more gradual throughout andropause [56
]. Improving the understanding of the mechanisms underlying the differences in sex-based health impacts in a context of heat could therefore be beneficial in ensuring that clinical settings and public health can offer more personalized adaptation strategies.
In this study, air conditioning was associated with the prevalence of self-reported health impacts during hot and humid summer conditions. The transversal design of our study did not allow verifying the beneficial effects of air conditioning to reduce or stabilize health impacts, or, on the other hand, to reduce normal physiological adaptation and potentially increase heat-related impacts [57
]. A recent survey in five Canadian cities [58
] similarly reported that heat-related illnesses was higher among those with cardiovascular and respiratory illnesses, higher among younger respondents and bore no relationship with the availability of air conditioning at home. On the basis of scientific literature though, it remains very likely that air conditioning during heat waves in cities brings positive effects on heat-related morbidity and mortality when diagnosed in a medical setting [57
]. However, no epidemiologic study so far seems to have included indoor dwelling temperature as an exposure variable, thus precluding any firm conclusion on this topic, nor allowing population- and evidence-based recommended thresholds for setting home air conditioning optimal temperatures [59
]. This could be of great interest as air conditioning use increases steadily in several countries [60
], and is not innocuous at the community level, given its contribution to greenhouse gases and air pollutants, system-wide blackouts or other similar and costly maladaptations [62
The final two indicators of the prevalence of self-reported health impacts were long-term medical leave and very low income. While the first indicator involved a particular situation for those under 65 years of age, and the second characterized also those 65 years of age and older, both spoke to poverty. In fact, 75% of those under 65 years on long-term medical leave reported an annual income below $15,000. According to several authors, people very economically disadvantaged are one of the main groups at high heat-related risk for adverse health impacts [22
In fact, poverty brings its share of consequences which have an impact on health (e.g., the lack of resources necessary for adaptation to heat) [64
]. Consequently, one would expect that the low-rental housing context, which is aimed at supporting households in a difficult economic situation, would explain the prevalence of impacts not only in bivariate analysis, but also in multivariate analysis. This was not the case, suggesting that it is not living in low-rental housing per se
that contributes to health impacts during very hot and humid summer conditions, but rather that certain characteristics of its residents, such as the proportion of people under 65 years of age on long-term medical leave, which explains the prevalence of impacts.
In the present study, advancing age was associated with an increase in the prevalence of self-reported health impacts, but only up to 64 years of age, the threshold where the prevalence decreased (Table 1
). Therefore, those 65 years and older were less likely to report impacts than younger participants. A priori
, this observation seems contrary to what has been reported in the scientific literature, where advanced age is a risk factor for morbidity and mortality in a context of heat [22
]. This difference may be due to the subjective nature of our impact measurement. Thus, older people could have a less accurate perception of the consequences of heat on their health, particularly because of physiological changes (e.g., reduced thermoregulation) or perceptual changes (e.g., reduced perception of heat and thirst) associated with their aging [33
]. On the other hand, many older people could in fact be less at risk than younger age groups, particularly if they have the possibility of adapting well to heat, such as by not having to leave their dwelling for work or the family, contrary to younger working people. Older people do not comprise a homogeneous population simply because they belong to the same age group. A similar observation could be applied to people living alone, a characteristic generally considered as being a risk factor during very hot and humid summer conditions [67
], while in this study, its relationship with the prevalence of impacts was low in the bivariate analysis and not statistically significant in the multivariate analysis. Refining these indicators would therefore be most useful from the standpoint of public health surveillance and emergency preparedness.
3.4. Limitations of the Study
The study’s response rate was low, but the response rate by question (considered as being another measurement of the survey’s response rate) was very good. No response rate from Canadian surveys can be used for comparison purposes, particularly because none of these surveys specifically targeted the very disadvantaged DAs studied. Based on a previous qualitative study conducted in two of the sampled cities [68
] and due to the characteristics of the communities studied (large urban centres, multiethnic environments, etc
.), it seems, however, that this rate could in fact represent what can realistically be obtained in the very disadvantaged DAs of Québec’s large cities.
For ethical considerations, no information was collected from people deciding not to participate in the study. In order to make up for this limitation, certain statistics were compared to census data available by DA at the Institut national de santé publique du Québec
. On this basis, it is possible to propose that the data from the study are comparable (from the standpoint of response rates) to those of the 2006 Census regarding the percentage of:
non-family households (census = 56.1%; study = 61%);
non-family households of only one person (87.6% and 90%);
non-family households of at least two people (11.8% and 10%);
single-parent families among the family households (34.4% and 36.2%);
couples with or without children among the family households (65.4% and 63.8%);
couples with children among the family households (46.8% and 42.4%);
couples without children among the family households (53.5% and 57.6%);
people born elsewhere than in Canada (20.9% and 20%);
immigrants speaking neither French nor English (2.0% and 0, because speaking one of the two official languages was a selection criterion in the study);
dwellings requiring major repairs (11.1% and 14.2%);
single-family homes (individual, semi-detached, row housing) (5.7% and 6.1%).
The study, however, seemed to underestimate the average income per household (census = $29,779; study, estimate using the midpoint of the $5000 strata = $23,679), and overestimate the average age (census, ≥18 years = 47 years; study = 53 years). These differences could be due to the fact that half of the sample comes from low-rental housing, considering the design of our study. Low-rental housing is mainly intended for low-income households and those experiencing a difficult economic situation. Also, half of the low-rental housing clientele consisted of elderly people. Furthermore, the average income ($31,724) and age (49.1 years) of the group not living in low-rent housing in the study were comparable to the census data.
Based on this information and despite the low response rate, it is therefore possible to state that the samples in the study generally represented well the populations living in the visited DAs, and also all the very disadvantaged DAs in Québec’s nine largest cities, due to the sampling plan that was adopted in the study (and that was taken into account in the weighting of the data). The percentages of elderly people and people with low income (which reduced the average income) seem, however, to be overestimated due to the study's design.
Moreover, given the use of two distinct methods of recruitment and the change in procedure during data collection, the people recruited by telephone and those recruited door-to-door were compared. The results of these comparisons (presented in a methodological report, see Bélanger et al
]) indicate that door-to-door collection made it possible to increase the representation of the very disadvantaged DAs and to include people who would not have otherwise participated (e.g., incorrect contact information). Thus, the pairing of the two types of recruitment reduced the risk of selecting only those who were at home and who had only one stationary telephone. The sampling plan adopted in the study therefore minimized potential selection biases.