Effects of apparent temperature on daily mortality in Lisbon and Oporto, Portugal

Climate change impacts on human health are a global concern [1]. Recent extreme weather events, such the European heatwave in 2003, have attracted renewed interest on weather-related health effects [2, 3]. Since climate change will likely increase the mean temperature, as well as the frequency of heat events [4] there is an urgent need to evaluate the links between climate and human health, to better identify vulnerable populations and take preventive measures.

Previous studies have reported that days of usually low and high ambient temperatures are associated with increases in mortality and morbidity [5–9]. Some groups adopted for a time series approach, while others focus on extreme events such as heat waves and cold spells using episode analysis. In Europe, populations living in urban environments as well as the elderly have a high risk of mortality from ambient heat exposure [10]. If future populations become more urbanized and the number of elderly continues to increase, the issue of heat-related mortality will probably become more severe.

Being part of the Iberian Peninsula, Portugal has a mild Mediterranean climate. Observed climate trends in Portugal show that between 1970-2000 there was an increase in the average temperature at a rate of 0.5°C/decade, which is more than twice than what was observed for the mean world temperature [11]. Quantitative studies looking at the effects of temperature on human health in Portugal are surprisingly few [12, 13]. In this paper we investigate the association between mean apparent temperature and daily mortality in Lisbon and Oporto during the warm season using modern statistical methods accounting for confounding effects of air pollution.

Figure 1 shows the associations between apparent temperature and mortality in both cities during the warm season, controlling for day of the week and long-trend trends in mortality. For both cities the mortality risk increases with increasing apparent temperature.

In the exploratory analyses, we found that lag-01 apparent temperature had the best model fit in both study areas, compared with the moving averages of multiple days. We therefore report here the results analysing the effect of apparent temperature only during warmer months and for lag-01.

Following the 2003 heatwave in Europe, various studies have assessed the effects of temperature extremes in various cities [2, 3, 19, 20]. During this heatwave period, in Lisbon and Oporto excess mortalities of about 400 and 183 were reported respectively [21]. Although Lisbon was one of the few cities in Europe that had a heatwave early warning system already in place and alerts were sent out (via the media) to warn the population [22], the final analyzes of the 2003 heatwave health effect showed that the population in Lisbon (OR of 1.41) was more vulnerable than the population in Oporto (OR of 1.3) [21]. Improvements to the early warning system are currently being investigated to allow for better models and a more active response action so as to ensure more health and life gains in the future during extreme weather events [23].

Since climate change is also likely to change the average temperature, in the current study, we restricted our data to the warmer months and to a five-year period so that the estimate for mortality represents the average ambient temperature that is commonly experienced, not the worst scenario that may occur in years with unusual weather patterns, such extremes temperatures and heat waves. To our knowledge, this is the first study to analyse the effects of mean apparent temperature on mortality in any Portuguese city.

Our results show an association between mean apparent temperature and daily mortality in Lisbon and Oporto, during the warmer months. These associations persist even when controlling for the effects of air pollution. Overall, the population in Lisbon was more vulnerable than the population in Oporto. It is interesting to note that this result is in agreement with observation reported during the 2003 heatwave were Lisbon was more vulnerable to acute heatstress [21].

When analyzing by cause of death, we observed a strong temperature effect on cardiovascular mortality, in agreement with other studies [24–28]. From a public health point of view, this finding is important since cardiovascular diseases are the leading cause of death for both men and woman in Portugal, accounting for ~40% of all deaths during the study period. This increase in mortality is likely to be related to the stress placed on the respiratory and circulatory systems to increase heat loss though skin surface blood circulation [29].

Direct comparisons with previous investigations is difficult because earlier studies have used different climate variables, seasons in the year, modelling approaches, temperature cut-offs and percentiles [3, 5, 15]. Nevertheless, a previous study that analyzed the effect of temperature on all-cause mortality in New Delhi, Sao Paulo, and London found a 2.2%, 1.6% and 1.4% increases per degree Celsius above 20°C [7]. Although this study used temperature (not apparent temperature), the cut-off temperature in this study was near to the mean apparent temperature as well the mean observed temperature in Lisbon, and the results are similar.

More recent studies using analysis method similar to ours allow for more suitable comparisons. For example, a study in California, found a 2.3% increase in all-cause mortality per 10° F (about 5.5°C) increase in mean apparent temperature [17], while in another study in nine US cities an increase of 2.74% in all-cause mortality was found [30]. It is difficult to directly compare our results with other epidemiologic studies from European cities because of the various definitions used to classify temperature exposure and the different methodologies applied [8]. Studies in European cities typically use maximum apparent temperature [8], maximum temperature [9] or mean temperature [24] to analyze the effects of heat exposure above specific heat thresholds. Nevertheless, most of these studies yield results similar to ours in that risks for deaths from cardiovascular and respiratory diseases increase with increasing summer ambient temperatures.

Our finding that the elderly are the group most at risk in the warmer periods is also consistent with previous studies [8, 10, 31, 32]. In our study the risk of death for people > 65 years was higher for cardiovascular diseases in Lisbon while in Oporto the risk was higher for respiratory diseases. A possible explanation for a higher risk for respiratory diseases in Oporto than in Lisbon is likely related to the higher pollution levels in Oporto. The latter city is surrounded by heavy industry, whereas Lisbon is a financial and administrative city surrounded by green areas and agriculture. The higher risk for cardiovascular disease in Lisbon can also be partly explained by the fact that this city is not only naturally warmer but it is also more prone to heat island effects since it is more urbanized and has a higher density of high rise buildings. Nevertheless, these differences in cause of death between these two locations should be explored in more detail in future investigations.

The reasons why confounding effect of air pollutants on heat related deaths differs between locations is an ongoing area of investigation. Thus, the inclusion of air pollutants to examine confounding and effect modification is an important contribution. We found a lower effect when adjusting for PM₁₀ in Lisbon, and a decrease in the effect with PM₁₀ and ozone in Oporto. Our results are consistent with those from O'Neill and colleagues [33] who found a small decrease in the association between mortality and apparent temperature when adding PM₁₀ and ozone singly or jointly. A recent analysis of the effect of temperature in nine US cities, with similar methods to ours, also found a lower effect when adjusting for ozone [29].

The results from our study support the importance of examining subgroups in specific locations. The fact that in both cities the elderly were identified as being more sensitive to heat exposure, together with the fact that the elderly population in Portugal is increasing makes this finding an important public health concern warranting special attention in current and future health programs. Since climate change is likely to result in more frequent, intense and longer duration episodes of heat waves the need for continued vigilance of heat exposure as well as the development of better heat warning systems is paramount. The fact that in this study a lag-01 was used also allows for better planning of intervention actions. In addition, our results also highlight which populations are more vulnerable thus allowing for targeted public awareness of ways to avoid heat exposure. Public education targeting homes for the aged, health care centres, pharmacies, and associations/societies with close contact with the aged will be key at reducing population vulnerability to heat exposure in these two cities.

In this study we used mean apparent temperature to assess heat exposure. Although this index is a mean daily value, it requires that first hourly apparent temperatures be calculated from which the daily mean is then derived. Consequently, hourly climate data is always needed. Obtaining hourly climate data at the city level from future climate models is currently not very practical, thus limiting the use of this index in studies assessing climate change health impacts due to heat exposures. Heat exposures in terms of maximum daily temperature (not apparent temperature) would be better suited for studies aiming to use the exposure-response relationship in climate change investigations.

In future analysis it would also be interesting to explore the effect of temperature in daily hospital admissions. Acclimatization and individual susceptibility should also be investigated as possible explanations for the differences between the cities. Finally, the effect of access to prevention and to emergency care [34] between cities is also an area warranting future research.

In conclusion, our study provides evidence of increased mortality due to mean apparent temperature exposure in Portugal, even when adjusting for air pollution. As changes in temperature due to climate change are expected to increase, the need for local epidemiologic studies which can help to analyse the effects on human health increases. Results of such studies can contribute towards decision making at national and local governmental agencies interested in developing better warning and health systems that allow them to be better equipped to prevent heat-related mortality. Since the response functions differ between locations, further research focusing on socio-economic, health systems management and geographical characteristics is needed.