While previous studies focused on single cities, the present project investigated the health impact of weather on a large scale through a variety of climatic conditions and of socio-economic and demographic characteristics, applying the same methodology, thus allowing for comparison between cities and the pooling of results.
Previous studies showed that the temperature level corresponding to the minimum mortality level varies from city to city and across different latitudes according to the local climate and probably reflecting adaptation by the local population to the temperature range in both the hot and cold season [11, 12, 22, 32–35]. The analysis of the heterogeneity of the effect in European areas was accounted for in the present project, describing city-specific "change points" of the dose-response relationship and the specific shape of the dose-response curves. In the pooled analysis, heterogeneity was reduced grouping the cities according to a priori geographical and meteorological characteristics.
Few studies have examined the effect of heat on outcomes other than mortality. In Chicago, during the July 1995 heat wave an 11% increase in hospital admissions was observed, with 35% of the increase among patients over 65 years . More recently, studies performed in London and 12 US cities reported an increase of admissions for specific causes in the elderly and evidence of a harvesting effect [8, 9]. To date, the present study is the largest one to investigate weather and hospital admissions in Europe.
In the present study, the role of meteorological variables other than temperature was investigated, assuming that they may contribute to the negative health effects. Therefore, an exposure indicator including dew point temperature was chosen for the time series analysis [23, 25], and the excess mortality/morbidity associated with specific air masses was explored using a climatologic classification based on synoptic indexes [37–39]. The choice of maximum apparent temperature as the exposure variable for the time series analysis was driven by the fact that it comprises temperature and dew point temperature in a single parameter. The simplified formula for this indicator was considered the most suitable one on the basis of data availability and quality. Several other indexes of thermal discomfort have been proposed in meteorological literature [40–43]. Among these, apparent temperature has been used as exposure measure in recent studies assessing the heat effect [20, 25, 44–48]. The importance of using heat stress indicators that combine temperature and a measure of humidity is due to the fact that on hot days the degree of humidity influences the body's ability to cool itself by evaporation and perspiration. From a modeling point of view, given that apparent temperature is obtained as linear combination of temperature and square dew point temperature, there is a strong correspondence between a regression model where a linear term for apparent temperature is included and a model where a linear term for air temperature plus a quadratic term for dew point temperature are introduced. Therefore, the percent variation in mortality/morbidity associated to 1°C increase in apparent temperature is expected to be similar to that of a model adjusting for dew point temperature.
When comparing the results between cities it is important to consider that the use of mean apparent temperature for Barcelona led to lower threshold values and an underestimation of the effect. Furthermore, regarding the pooled estimates, the combination of results obtained using different exposure indicators increases heterogeneity and the risk of bias in the meta-analytic estimates, in particular when thresholds or exposure-response curves are combined. Therefore, sensitivity analyses was performed excluding Barcelona from the meta-analysis and no significant differences were found.
This project focused on time series rather than on heat wave episodes, using an approach, that has been successfully used in the analyses of the effects of air pollution. Such methods have the advantage that the population under study serves as its own control, and covariates that vary between subjects, but not over time, are not potential confounders [49, 50].
Most time series studies performed so far showed an immediate effect of heat on mortality, with the maximum impact within two or three days (lag 0–3) [5, 10–13, 34], while for the cold season, an effect of up to fifteen days has been observed [10, 12, 13]. Knowledge on the lag time between exposure to extreme weather conditions and negative health outcomes is important for the development and planning of prevention plans by public health authorities and health care providers.
Evidence that the increase in mortality is followed by a deficit that, partly compensates the negative effect (harvesting) is contradictory . In the present study the heterogeneity of mortality/morbidity displacement patterns between cities was systematically investigated.
There is much disagreement in literature concerning human acclimatization to changing weather [37, 52, 53]. While the issue was examined by comparing the threshold temperature in different geographic locations, the possibility of a short-term acclimatization was also evaluated by comparing the dose-response function in the first period of the summer with the effects modeled in the later part of the season. This allowed also for comparison of the impact of the first heat wave in one summer with the following ones.
Given the small number of events (mortality and admissions), a unique definition of winter and summer season was chosen in order to reach a reasonable statistical power, and sensitivity analysis was performed, focusing on the three central summer months (June–August).
The relationship between increase in air pollution levels and acute health effects has been well described in the USA and in Europe [50, 16]. The levels of some of the pollutants associated with increase in mortality and hospital admissions are higher during the summer period in many European areas. A synergistic effect of warm temperature and air pollution on mortality has been suggested from time series analysis conducted in Athens, whereas no effect modification detected in a study in Philadelphia, USA [54, 55]. The present study investigated the independent effect of meteorological variables of that of ambient air pollution, and explored whether there is synergy between the two factors.
The results of the mortality analysis were also used for the development of experimental Heat/Health Watch Warning Systems (HHWWS) in five cities (Rome, Barcelona, London, Paris, Budapest). An air-mass-based climatologic index was developed, applying a synoptic approach. Using meteorological forecast data, these models are able to predict the arrival of an oppressive air mass 72 hours in advance. Such warning systems have been successfully implemented in the United States , whereas in Europe before the 2003 heat-wave, only in Rome a HHWWS was experimented and implemented. Therefore, the pilot study in five PHEWE cities represents an important innovation in the field of heat health prevention. The development of a standardized protocol allows for knowledge and technology transfer to other European cities in the future.
The usefulness of early warning systems is closely linked to public health strategies aiming at the prevention of negative health effects of heat. In the present project an overview of existing prevention programs in the participating cities was obtained through a questionnaire. Moreover, physiological and behavioral adaptation measures, experiences with different HHWWS, urban planning, housing standards, and socio-economic determinants of vulnerability were summarized in a comprehensive literature review. The quantification of the effect of heat/air masses exposures in the different populations was addressed through a health impact assessment (years-of-life-lost approach). These results will contribute to policy development, public health decision-making, and will be an important input for cost-benefit analysis and risk communication. Guidelines for preventive strategies and health care actions taken to lessen morbidity and mortality effects can then be based on evidences arising from this project, namely the literature review, the investigation of the state-of-the-art in the participating cities (feasibility), and the identification of susceptible populations.