A multilevel analysis to explain self-reported adverse health effects and adaptation to urban heat: a cross-sectional survey in the deprived areas of 9 Canadian cities
In this study, multilevel analyses made it possible to process building-level and
DA-level variance as a source of information 51], in contrast to other approaches used to analyze the same dataset 35], 36].
The results of random parts of the analyses show that a 3-level model supports the
identification of indicators associated with self-reported adverse health impacts
and the adaptation index when it is very hot and humid in summer. Therefore, they
emphasize the need to develop public health interventions aimed at both residents
of very disadvantaged DAs, property managers in these sectors and municipal authorities
in large urban centres already struggling with the urban heat island.
The importance of the role of buildings and the environment as risk (or prevention)
factors of self-reported health impacts and adaptation to heat is strengthened by
the contribution of four building-level and DA-level covariables. However, this number
of four variables remains well below the number of individual-level covariables, despite
important efforts to measure various aspects of the context (see Additional file 1: Table S1). This observation has been reported in the literature and some explanations
have been given on this subject 52]–54], including the fact that census data do not characterize neighbourhoods in detail
and that a DA is an administrative geographical unit and not a neighbourhood as perceived
by respondents.
The effect of buildings and the environment on self-reported adverse health impacts
and on the adaptation index could also be covered by some covariables. We cannot make
any conclusions regarding this aspect due to the cross-sectional design of our study.
Also in the absence of a proven theoretical framework, because although some interesting
avenues have been proposed 41], 55], 56], none have been validated. The establishment of such a framework would be very useful
for research and public health surveillance, in particular for clarifying the kind
of links between covariables; in our study, most are related only to the prevalence
of self-reported adverse health impacts or to the adaptation index.
Specifically, in this study, the prevalence of self-reported negative health impacts
when it is very hot and humid in summer is 46Â %; although high, this could correspond
in reality to the highly disadvantaged DAs of the most populous cities in Québec,
as discussed elsewhere 35], 36].
As already reported in the literature 41], 57]–59], poor physical or mental health is a powerful indicator of risk to heat. Similarly,
residential neighbourhoods considered somewhat or very polluted due to high urban
traffic density are in all likelihood more paved and therefore hotter. Such highly
paved and thus more impermeable neighbourhoods contribute to the negative health impacts
of heat 2], 3], 60], 61] and result in the deployment of various adaptations 9], 10].
In addition, hormonal differences may explain the difference in self-reported adverse
health impacts by gender 62], 63] for women aged 45 to 64 in particular (menopause period) 35]. Regardless of gender, the 45–64 age group, however, remain at higher odds of impacts
according to our results. This is contrary to what is generally reported in this area
of research 42] but plausible in very poor populations such as ours. In fact, the proportion of 45–64
years old having more chronic health problems increases according to a decrease in
family income 64], while seniors having survived the same conditions would in all likelihood have been
in better health. Another explanation for the increased odds of heat impacts for the
45–64 group would be greater exposure, mainly because of the requirement to leave
home for family or work reasons 35] compared to older people, who are more confined 10]. Clarifying this issue would be important for public health because those 65 and
over are generally the target group for national heatwave plans.
Physical inactivity, a risk factor for several non-communicable diseases 57], is associated with a higher prevalence of self-reported adverse health impacts.
The average level of physical activity declines with age; the result is a downward
trend in fitness 64], 65]. It is known that poor physical fitness leads to a low cardiovascular reserve and
a low tolerance to humid heat 66]. However, physical inactivity was not associated with self-reported adverse health
impacts in multivariate models in our previous articles 35], 36]. It is thus possible that this multilevel analysis highlights a potential contribution
of deprived neighbourhoods to physical inactivity 67], 68]. Further studies are needed on this.
In our study, physical inactivity, like functional limitations, is also associated
with a lower self-reported adaptation to heat, according to the index. More specifically,
these two conditions characterize very poor populations 64], 65], such as our sample drawn from DAs that are farther (in time and distance) from cool
greenspaces than more affluent DAs 69].
Improving the urban development of the poorest sectors of large urban centres would
then be more beneficial for countering the health impacts of heat and facilitating
adaptation. Over a third of respondents in our study also expressed this need. In
addition, this initiative would have many co-benefits 55], 57]. For example, according to our results, DAs that offered better walkability were
on average more adaptive according to the index, which could mean a reduction in the
use of air conditioning in the home 70], which contributes to the outside heat and indirectly by exacerbating the air pollution
in some cases 57]. Similarly, better equipped DAs could reduce residential mobility and improve levels
of well-being in neighbourhoods 71], 72] while encouraging adaptation by means other than air conditioning in the dwelling.
These recent hypotheses, however, remain to be confirmed.
As already published 7], 42], but not in our previous multivariate analyses 35], 36], social support is associated with health impacts from heat with this multilevel
approach. According to our results in the present article, the network that helps
prevent impacts would be less diverse and more likely to be family than the network
that promotes adaptation in a way other than air conditioning in the home. Further
studies are needed in this respect, as networks with strong links could potentially
exacerbate rather than reduce vulnerability to the effects of heat in the elderly,
according to some authors 73].
The relationship between satisfaction with the indoor temperature of the dwelling
in summer and the decrease in the prevalence of self-reported adverse health impacts,
as well as the relationship between satisfaction with insulation quality and reduced
adoption of behaviours measured by index of adaptation, clearly highlights the contribution
of the building occupied in reducing exposure to heat, beyond the individual sensitivities.
The fact that each of these relationships is observed for both the dwelling (individual-level)
and the building (building-level) reinforces the importance of the use of these perceptions
for public health surveillance. Especially in highly disadvantaged areas of major
urban centres, where there are old buildings whose insulation meets less efficient
standards than today’s, unless upgrading has been done since they were built 36].
Last, in this study the variables were self-reported which is generally considered
easy to implement and cost-effective in several situations. As reported elsewhere
35], the validity of self-reported versus medical-based diagnoses and behaviours has
been well established over time, several countries and data collection methods, especially
as a tool for predicting future risks and as an epidemiologic survey tool for prevention
and public health actions. Perceived health is thus a reliable and valid subjective
measurement of the overall state of health and is widely used by statistical and health
agencies throughout the world. This self-reporting approach for evaluating environmental
exposures is also documented if not as well assessed. Moreover, the objective measures
are far from perfect, as already mentioned 12], 17]. From a public health perspective, it remains important to validate these results
in other contexts.