Introduction

Heat waves, meteorological events characterized by unusually high temperatures sustained for prolonged periods, can affect human health and survival, and are among the most dangerous environmental hazards [1]. Epidemiological evidence reveals the health impacts of heat waves on numerous mortality and morbidity outcomes, including hospital admissions, emergency department visits, ambulance attendances, occupational injuries, all-cause and cause-specific mortality [2–6]. The biological mechanisms that underlie the effects of heat waves on human health are not fully understood, although many physiological responses to extreme temperatures may be related to cardiorespiratory events [2]. Cardiorespiratory diseases are the leading cause of morbidity and death worldwide [7], and numerous epidemiological studies suggest an increase in cardiorespiratory events during heat waves, although the magnitude of these effects varies across countries and population groups [2].

The global mean temperature has been increasing since the pre-industrial era, and extreme heat events are becoming more intense and frequent [8]. As a consequence of climate change, the temperature will continue to rise, and in the future heat waves will be even more frequent and severe [9]. According to the latest report from The Lancet Countdown, vulnerability to the extremes of heat and exposure to heat waves have risen worldwide [10]. In the context of both climate change and ageing populations, the impacts of heat waves on health are expected to intensify [10, 11].

Although the health effects of heat waves are well documented, most study locations are concentrated in high income countries, principally in Europe, North America and Australia; few studies have been conducted elsewhere, except in China [3, 12]. There is a scarcity of studies in tropical regions characterized by high population density, low income and low health resources, which are susceptible to the impacts of extreme heat on health [3]. Investigations to improve the description of the health risks arising from heat waves in these regions are urgently required, to promote public policies to mitigate and adapt to the impacts of the climate crisis.

To the best of our knowledge, few epidemiological studies of heat waves-related mortality have been conducted in Brazil [12, 13]. One multi-city study examined the impacts of heat waves on all-cause mortality and reported the results at national level [14]. Another study conducted in 32 municipalities in the Brazilian Amazon, examined the effects of various heat wave definitions according to sex, age group, and cause of death [15]. Two studies were conducted in the city of São Paulo, estimating the effects of heat waves on all-cause [16] and elderly mortality [17], while two studies, comparing observed and expected deaths during heat waves, were performed for the Metropolitan Region of Rio de Janeiro [18, 19].

The aim of this paper is to estimate the effects of heat waves on cardiovascular and respiratory mortality in the city of Rio de Janeiro, Brazil between 2012 and 2017. We stratified our analysis by sex, age, and a combination of both strata, and assessed how these effects change according to heat wave intensity. Different to most previous study conducted in Brazil that used a single city exposure measure for all participants, we used an estimate of individual-level exposure, in order to account for the spatial variability of weather conditions and reduce exposure misclassification error.

Methods

Results

We recorded a total of 132,863 deaths from cardiorespiratory outcomes, which occurred in Rio de Janeiro between 2012 and 2017. We excluded deaths with non-geocoded addresses and those with lagged exposure in relation to the index or a referent period prior to 2012. We also restricted our analysis to the hot season. This resulted in 42,926 participants, 29,442 of whom died from cardiovascular and 13,484 from respiratory disease. Table 1 shows the summary statistics for the study population and the weather variables over this period.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Study population summary statistics and weather variables in Rio de Janeiro, Brazil, 2012–2017.

The data were restricted to the hot season (November to March).

https://doi.org/10.1371/journal.pone.0283899.t001

Table 2 shows the odds ratios (OR), and their 95% confidence intervals for the cumulative effects of heat waves of varying intensity on cardiovascular and respiratory mortality, for the entire study population, and stratified by age and sex. In general, the effect estimates rise as the heat wave intensifies. However, in some strata, when we use the most stringent heat wave criteria, the effect estimates are not statistically significant at 95%. Although the results were not statistically different, in general we found greater effect estimates for respiratory than cardiovascular mortality, but for the most stringent definition, the central OR decreased and become very imprecise.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Odds ratios (OR) and 95% confidence intervals (95% CI) for the cumulative effect of heat waves on cardiovascular and respiratory mortality, over 5 and 10-day lags respectively, in the city of Rio de Janeiro, Brazil, stratified by age and sex.

https://doi.org/10.1371/journal.pone.0283899.t002

Results stratified by sex and age were not statistically different but suggest that the older people and the women were more vulnerable to the effects of heat waves, although for some heat wave definitions, the odds ratios were higher among men and the younger group. There was also an increase in the mortality odds ratio for men during heat waves, but this was lower than for women, and all values demonstrated uncertainty above 5%. When we stratified our analysis of the older group by sex, for cardiovascular mortality, the results suggest that older women are the most vulnerable to heat wave effects, with higher central estimates than older men, except during the most intense heat waves (using the 99^th percentile). For respiratory disease, older women were more vulnerable during the less intense heat waves (using the 90^th and 92.5^th percentiles), while older men were more vulnerable to heat waves of greater intensity (using the 95^th, 97.5^th and 99^th percentiles).

The results of our sensitivity analyses are shown in the S4 Table. We observed very similar results when we changed the cross-basis parameters for the heat wave adjustment (using a quadratic b-spline for lags, and changing the degrees of freedom of the natural cubic spline to 3 and 5). Although not statistically different, the odds ratios were slightly lower when using 3- and 10- day lags for cardiovascular, and 3- and 5- day lags for respiratory mortality, and were slightly higher when absolute humidity was removed from the model. When we controlled for temperature, the added effect estimates of heat waves were observed for almost all heat wave definitions, except in the most intense heat waves (defined at the 99^th percentile of daily mean temperature) for respiratory mortality, although most were not statistically significant at the 95% level.

Discussion

In this study, we used a case-crossover design based on individual-level exposure estimates to investigate the effect of heat waves of varying intensity on cardiovascular and respiratory mortality, stratified by age and sex subgroups, in the city of Rio de Janeiro, Brazil. Our results show an increase in the risk of cardiovascular and respiratory mortality on heat wave days compared to non-heat wave ones, and suggest that the effects were greater for respiratory than cardiovascular mortality, consistent with the conclusion of a systematic review and meta-analysis [2]. In general, the effects increase as heat waves intensify, which is also in accordance with the literature [4]. The evidence suggests that women and elderly are more vulnerable to heat waves, as indicated by previous systematic reviews [12, 30]. Furthermore, when we stratified our analysis of the older group by sex, older women were more vulnerable to cardiovascular mortality, while the results for respiratory mortality were inconclusive, suggesting that, depending on heat wave intensity, both groups were vulnerable.

Our findings are in agreement with evidence from previous studies conducted in Brazil, most of which use single-city exposure measures and different study designs [16, 17, 31]. These studies observed that heat waves have a greater effect on respiratory than cardiovascular mortality, with greater risk from higher intensity heat waves. Son et al. [16] analyzed the modification of heat wave effects in the city of São Paulo on total non-external mortality and observed that estimates increase with age and are greatest for those aged 75 or above. Moraes et al. [31], in a study restricted to older people (> = 65 years) and conducted in the city of São Paulo, found that mortality risk during heat waves was greater for total cardiovascular and respiratory diseases among women. The study of Diniz [17], also restricted to people aged 65 or above and conducted in the Metropolitan Region of São Paulo, found that heat waves have a greater effect on women for both cardiovascular and respiratory outcomes. A study in the Metropolitan Region of Rio de Janeiro [18] observed high numbers of excess deaths during four heat wave events in 2010 and 2012, and the observed-expected death ratios were higher for women than men and for the older group. In a recent paper conducted in the Brazilian Amazon [15], we also found that heat wave effects increase with heat wave intensity and were greater for the elderly and women, however, as opposed to this study, there, the effects were greater for cardiovascular mortality than for total non-external and respiratory mortality.

In previous research conducted in Rio de Janeiro [32], we found that the cardiovascular mortality risk increased with ambient temperature, and estimates of the heat effect were greater for older people and women. In the present study, our sensitivity analyzes suggest that, beyond the temperature effect, an additional effect arises from the duration of heat sustained over several consecutive days, in line with Gasparrini and Armstrong [33].

The heterogeneity effect of heat waves across population subgroups can be explained by several factors. In general, older people may have other comorbidities, which can make it harder for the body to respond to thermal exposures. But even in healthy older people, the changes in cardiovascular function that occur with aging can compromise the body’s ability to perform thermoregulation during exposure to heat waves [34]. Gender or sex differences can be explained by differences in age structure, occupational and socioeconomic status, mobility and exposure patterns, in addition to physiological differences in coping with heat [12, 35]. Further, in our study, the mean age of women (82.4 years) in the older group is higher than that of men (78.9 years). Women may have greater difficulty in dissipating heat due to their fitness levels, the higher percentage of body fat, and possible differences in skin conductance. Furthermore, differences in hormone levels, especially the changes in post-menopausal estrogen levels, may be associated with inflammatory responses during thermoregulation [35].

We estimated residential temperature and humidity levels at individual resident addresses, based on the inverse distance weighting method, in order to account for the spatial variability of the weather factors, and reduce the exposure misclassification error. Nonetheless, this estimation method has limitations, since it is based on the values of the observed variable at the weather stations, rather than taking other determinants of the exposure variable into account, such as altitude, land cover, proximity to water, solar radiation, wind speed, etc. The Guo et al. [36] study of ordinary kriging, which took account of other environmental variables, provided better spatial predictions of temperature than inverse distance weighting. Although we used an individual-level exposure estimation, we did not consider other determinants of individual exposure related to daily mobility or an environment with warmer or milder temperatures, for example an air-conditioned one.

Another study limitation is the loss of deaths that could not be geocoded, which may have influenced our effect estimates. Losses during geocoding tend to be concentrated in places with poorer urban infrastructure, where cartographic databases tend to be of low quality, coinciding with areas of lower income and greater segregation, which are essential health determinants. There was a higher proportion of losses among Black and mixed race people and those with lower educational levels compared to the study population, 54% vs 37%, and 74% vs 61%, respectively. However, we believe that, if present, any such bias would have distorted our results towards the null. Further, the geocoding of residence does not account for differences in how travel and indoor/outdoor activity patterns impact exposure. However, this type of measurement error is likely to be non-differential [37], since this kind of mobility pattern may have been controlled by self-matching (the same week day for a month) in the time-stratified case-crossover design.

No standard definition exists for heat waves, either in the scientific literature or in policy, but heat waves are typically defined based on intensity, referring to level of temperature, and duration, i.e., consecutive days of high temperature. We analyzed five definitions with varying thresholds of temperature intensity, and a duration threshold of two or more days. Future studies could consider the impacts of alternative heat wave definitions with higher duration thresholds, as well the timing of heat waves within seasons, to examine whether, for example, the impacts of the first heat wave of the hot season are different from those of later ones. Additional work is also needed to examine the effect modification of certain environmental factors, such as humidity and air pollution, and individual- and contextual-level socioeconomic characteristics.

Despite our research limitations, few studies of heat waves have been based on individual-level temperature estimates, allowing exposure estimates for a given day to vary within the same city, and, to our knowledge, this is the first such study carried out in Brazil. Individual-level estimates of exposure are an interesting approach to estimating thermal exposure in large cities since temperature levels can be heterogeneous within cities, mainly because of the occurrence of heat islands. Future studies should explore more robust methods for estimating individual-level meteorological exposure. Another important study contribution is that our findings suggest that women, older people, and particularly older women, constitute the groups most vulnerable to the effects of heat waves, which may support the design and implementation of warning systems and heat-health action plans, enabling health and related sectors to adapt to climate change.

Supporting information