Contributors employed a variety of methods to estimate the number of excess deaths during and just after the 2003 heat episodes and to relate daily death counts to weather, to concentrations of air pollutants, and to demographic and social characteristics. Indeed, the varying emphases and methods demonstrated by the six national contributors provide complementary evidence of what happened in 2003, and to whom. While the absence of uniform methods does limit between-country comparisons of the health impact of the 2003 heat waves, these reports taken together suggest that weather alone does not explain the varying tolls of excess death within and between countries.

All six reports demonstrate that the mortality impact of the 2003 heatwave was greatest on the very old: for example, excess mortality in France was estimated at 20% for those aged 45-74 years, at 70% for the 75-94 year age group, and at 120% for people over 94 years [3]. There was no evidence of excess mortality in infants and children in any of the six reports. Among the elderly in France, Portugal and Italy, the three countries which stratified deaths by sex, rates were higher in females [3,4,7]. The strength of the age effect and the direction of the sex effect differ from those described for the 1995 heatwave in Chicago, United States [12], and indeed from those of the 1981 heatwave in Portugal [13]. Also rather unexpected is the observation from Spain that mortality impacts were more pronounced in rural villages than in the provincial capitals [5]; heatwave deaths are generally assumed to be an urban phenomenon, related in part to endogenous production of heat by city-based buildings, traffic, and factories, and to heat retention by inner city asphalt and concrete [14].

Investigators in Rome and Turin calculated rates of excess death as a function of socioeconomic level. They report that the greatest excess was in people living in areas of the lowest socioeconomic level, and suggest that finding may be upwardly biased, due in part to the phenomenon that those who have the means to do so leave Italian cities in summer, leading to an overestimate of the denominator for economically advantaged elderly people resident in the city in summer [7].

It has been observed that few deaths during heat waves are declared to be due to hyperthermia, heatstroke and other classic heat illnesses [15]. Reports from both France and Portugal observe that in 2003, deaths certified as caused by ambient heat constituted an important proportion of the death excess [3,4]; in France, 2852 of 11 891 (24%) excess deaths among people over 74 years were medically certified as directly heat-related.

Assessment of deaths per day offers speculative evidence for the effect of heatwave severity and duration on excess mortality. In the Netherlands [6], deaths per week vary with average weekly maximum temperature over the entire temperature range. In Milan [7] and in Spain [5], heat events of shorter duration earlier in the summer show a moderate death excess, compared with the deep and prolonged mortality spikes associated with the 10 day August heat event. In Paris, the daily death record shows that excess deaths are apparent within 1-2 days of rising August temperatures, crescendo during the unrelenting heat of 2-12 August, drop as temperatures fall, and reach baseline levels by 19 August [3]. Taken together, these graphs of deaths per day suggest that the lag time from extreme heat to excess death is around 1-2 days, and that cumulative mortality effects occur when hot weather is prolonged.

Between-city comparisons offer insight into the influence on deaths during heat waves of air pollution, population adaptation, and community preparedness. In Portugal, August forest fires led to a reported 18 accidental casualties [4]: presumably the attendant air pollution may also have had an impact on short-term mortality. Coincident to the high temperatures, elevated concentrations of ground-level ozone and PM10 were recorded in London and the south east of England, those areas of England and Wales where excess mortality was most in evidence [2]. In France, varying concentrations of ground-level ozone in cities subject to differing meteorological circumstances has allowed investigators to assess the joint effects of ozone and heat: these appear to be additive; while the apportionment of deaths to heat versus ozone differed markedly between cities, it appears that for France overall, during the period 3-17 August 2003, heat had the preponderant impact on mortality [16]. French contributors suggest that Marseille’s experience of a heatwave in 1983 and the existence there of a risk management plan for hospitals and a public communication strategy may have led to a death excess in that city of only 26%, compared with 53% for Nice (although the proportion of the very old in Nice is somewhat larger than in Marseille) [3].

Are those who die during heat waves already near death, with extreme heat advancing the date of their demise by only a few days to weeks, and thus creating a compensatory deficit in expected deaths during the days following the heat event? US investigators have suggested so [17], and the very high rates of excess death in Europe during 2003 among the very old tends to support that concept. However, while deaths for all ages in England and Wales declined by 4% between 24-29 August when compared with expected numbers, there was no evidence of mortality displacement in France or Spain, neither during late August, nor during September, October, and November [2,3,5].

Related to this notion of mortality displacement by heat waves is the hypothesis advanced by contributors from Spain that excess 2003 summer mortality may relate in part to the lower than expected level of mortality there during the winter of 2002-2003, leading to a larger than expected pool of people in fragile health by summer 2003 [5]. Italian contributors suggest that the lower levels of excess mortality in Rome during the August heatwave, compared with two earlier episodes in June and July, may be attributable to a reduction by late summer of the pool of susceptible persons [7].

Taken together, the six articles presented here offer a quantitative overview of the short-term effects on mortality of prolonged extreme heat in Europe, a meteorological phenomenon likely to become more frequent towards the mid twenty-first century [18]. The experience of 2003 shows that those most likely to die of the heat are the old, the chronically ill, and the isolated. Both northern and southern Europe are at risk.

Still, there may be good news amidst these descriptions of catastrophe. The Netherlands has seen a decline in the influence of heat on mortality when the 1950s are compared with the 1970s or 1990s [6]. Recent community and/or institutional experience with extreme heat may lead, as in Marseille, to a dampening of the impact of subsequent heat waves [3]. And, in contrast to the situation in Portugal in 1981 and 1991, the official alerts and interventions deployed there in 2003 may in part explain a relative reduction in the impact of extreme heat on mortality [4]. Perhaps most important, national plans for hot weather preparedness and response have been implemented or refined in five of the six contributing countries since 2003.

The task ahead is to build coordinated, sustained national and local programmes to reduce the vulnerability of the population to extreme heat. This involves pre-event planning and effective during-event intervention. While research is needed to guide these programmes, surveillance, such as that described in this issue, can monitor the effectiveness of our efforts and pinpoint the evolution of vulnerabilities.


Acknowledgements
Thank you to France Labrèche, Institut national de Santé publique du Québec and Bettina Menne and Tanja Wolf, WHO Regional Office for Europe, European Centre for Environment and Health, Rome, who provided pertinent reports and publications.