IJE Advance Access originally published online on January 13, 2008
International Journal of Epidemiology 2008 37(2):309-317; doi:10.1093/ije/dym253
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Has the impact of heat waves on mortality changed in France since the European heat wave of summer 2003? A study of the 2006 heat wave
1 INSERM, U754, IFR69, Université Paris Sud 11, Epidémiologie Environnementale des Cancers, Villejuif, France.
2 Institut de Veille Sanitaire (InVS)/French Institute for Health Surveillance, France.
3 Météo-France, Toulouse, France.
4 INSERM, CépiDc, IFR69, Université Paris Sud 11, Centre d'Epidémiologie sur les Causes Médicales de Décès, Le Vésinet, France.
* Corresponding author. INSERM-U754, 16, avenue Paul Vaillant-Couturier, 94800 Villejuif Cedex, France. E-mail: denis.hemon{at}vjf.inserm.fr
 |
Abstract |
|---|
 Top Abstract IntroductionMaterial and methodsResultsDiscussionAcknowledgementsReferences |
|---|
Background In July 2006, a lasting and severe heat wave occurred in Western Europe. Since the 2003 heat wave, several preventive measures and an alert system aiming at reducing the risks related to high temperatures have been set up in France by the health authorities and institutions. In order to evaluate the effectiveness of those measures, the observed excess mortality during the 2006 heat wave was compared to the expected excess mortality.
Methods A Poisson regression model relating the daily fluctuations in summer temperature and mortality in France from 1975 to 2003 was used to estimate the daily expected number of deaths over the period 2004–2006 as a function of the observed temperatures.
Results During the 2006 heat wave (from 11 to 28 July), about 2065 excess deaths occurred in France. Considering the observed temperatures and with the hypothesis that heat-related mortality had not changed since 2003, 6452 excess deaths were predicted for the period. The observed mortality during the 2006 heat wave was thus markedly less than the expected mortality (
4400 less deaths).
Conclusions The excess mortality during the 2006 heat wave, which was markedly lower than that predicted by the model, may be interpreted as a decrease in the population's vulnerability to heat, together with, since 2003, increased awareness of the risk related to extreme temperatures, preventive measures and the set-up of the warning system.
Keywords Heat waves, mortality, excess mortality, preventive measures
Accepted 28 November 2007
|
Introduction |
|---|
|
TopAbstractIntroductionMaterial and methodsResultsDiscussionAcknowledgementsReferences |
|---|
In August 2003, Western Europe experienced a heat wave that was exceptional in terms of duration, intensity, geographic extent and health impact.1–5 In France, this event revealed an overall lack of reactivity by people (insufficient protective reflexes and help for the elderly), health professionals (poorly known diseases related to heat), the health system (difficulty perceiving the signals, lack of prevention recommendations) and the media (alarmist messages but no information on prevention).
In order to prevent the consequences of a new heat wave, a National Heat Wave Plan was set up by the Directorate General for Health to prevent the risks related to extreme temperatures.6 The plan includes several measures: set-up of a system for real-time surveillance of health data, compilation of scientific recommendations on the prevention and treatment of heat-related diseases, air-conditioning equipment for hospitals and retirement homes, drawing up of emergency plans for retirement homes, city-scale censuses of the isolated and vulnerable, visits to those people during the alert periods, and set-up of a warning system. All the Directorates General of the Ministry of Health were involved together with the Directorate General for Civil Defence and Security.
During alert periods, other specific measures are to be implemented: intensification of care offer, visits to isolated and vulnerable people, and repeated preventive message broadcasting by the media.
Since summer 2004, the French Institute for Health Surveillance (Institut de Veille Sanitaire: InVS), in close collaboration with the national weather service (Météo-France), defined and implemented a heat health watch warning system on the basis of biometeorological indicators7 and other criteria.8 The warning system operates from 1 June to 31 August (seasonal surveillance period). When the alert criteria are fulfilled, an awareness and action level is declared by the Prefect who manages the département (French administrative unit with an average population of 600 000). A third level (maximum mobilization) is implemented if the impacts of the heat wave overwhelm the health field: power cuts, drought, management problems in the funeral centres and heavy air pollution.
During an alert period and after it, the health impacts of the heat wave are estimated by comparing the observed and expected numbers of deaths. This comparison does not take into account the temporal pattern of temperatures. The national heat plan and the warning system are evaluated and improved each year (feedback from the various players, outside evaluation, exercises).
The alert system aims to give the public authorities 3 days’ prior warning that a heat wave may occur, in order for the National Heat Wave Plan measures to be put into operation. The preventive measures and the new alert system are aimed at modifying the behaviour of people, health institutions and health authorities with regard to high summer temperatures.
In July 2006, a severe heat wave occurred over a large part of France and lasted 18 days (11–28 July). It also affected other European countries.9 According to Météo-France, this event was the second most severe heat wave recorded since 1950; the 2003 heat wave was the most severe. The intensity and duration of the 2006 heat wave were similar to those of the 1976 heat wave in France, during which 5000 excess deaths occurred.10
The objective of the study was to predict the national daily mortality for the summers of 2004–2006 and to compare it to the observed mortality, using a model based on a combination of temperature indicators for the same day and the 10 preceding days. The model was built using the data from a 29-year period, from 1975 to 2003.11 The number of excess deaths observed during the 2006 heat wave was thus estimated and compared with the expected number of excess deaths, assuming that the relationship between temperature and mortality during summer 2006 was similar to that observed during the summers of 1975–2003. This comparison was intended to contribute to assessing the health impact of the prevention and alert measures, and the population's general awareness due to the 2003 heat wave.
|
Material and methods |
|---|
|
TopAbstractIntroductionMaterial and methodsResultsDiscussionAcknowledgementsReferences |
|---|
The study analysed the mortality in association with temperature over the 122 ‘summer’ days, from 1 June to 30 September of each year, from 1975 to 2006 (3904 summer days in all), for the whole of France.
Mortality data
The daily counts of all-cause mortality (Ot) by age and gender were provided by the French National Institute for Medical Research (Inserm). The yearly population estimates were supplied by the French National Institute of Statistics and Economic Studies (INSEE).
Temperature data
The daily minimum and maximum temperatures (Tmin and Tmax) were obtained from 97 weather stations representative of the climate affecting the populations of the 96 French départements by the national weather service. The national daily values of minimum/maximum temperatures were the average of those 97 values, weighted by the populations of each département.
In order to obtain information on the usual temperatures in summer, the reference minimum/maximum temperatures were obtained by dividing each summer period into 12 10-day periods and averaging the temperatures for each 10-day period over the 28 years (1975–2002).
A 10-day moving average of the mean temperature (average of the daily minimum and maximum temperatures) was also calculated.
For each day, a cumulative maximum temperature variable (CTmax) was defined as the sum of the number of degrees above a cut-off point (27°C) from the current day t to either day t – 10 or the last day with a temperature higher than the cut-off point. The variable was equal to zero if the temperature was below the cut-off point on the day considered:
|
|
is equal to 1 if Tmaxt – d is higher than 27°C and 0 otherwise.
Modelling the daily fluctuations in mortality as a function of observed temperatures
As detailed in a previous paper,11 the daily mortality rates were modelled using a generalized estimating equations (GEE) approach, with a Poisson distribution. This model enables both specification of an over-dispersion and a first-order autoregressive structure. Seasonal variations in mortality during summer were controlled using a quadratic function of the day (Season). In order to allow for the overall trend of death rates declining regularly, although not log-linearly, from year to year, the mortality in summer was adjusted by the mortality rate observed over a reference period (MRref). This approach takes into account the long-term mortality trend and turned out to be more robust than a linear term. Four months, October and November for year N – 1, April and May for year N, were chosen to constitute the reference period because they immediately preceded the summer for year N and were not subject to high yearly variations in mortality.
The model was thus:
|
|
In which PopJ is the population estimate for the year considered. The temperature factors (Temperature factork,t) are a combination of minimum/maximum temperatures, the moving average and the cumulative variable of the maximum temperatures, recorded on the day considered and the preceding 2 days. Interactions between minimum/maximum temperatures and the cumulative indicator were also added. This combination was considered to be the most predictive of daily mortality in summer, both for the days with usual temperatures and for the days during intense and lasting heat episodes. The model explained 79% of the extra-Poisson variability in the daily number of deaths observed during a series of 25 summers (1975–99) used to fit the model and yielded an excellent temperature based prediction of the daily death counts observed for a series of four consecutive summers (2000–03), totally disconnected from the period used to fit the model.
Evaluating the predictive performance of the model (1975–99) for the period 2000–03
In order to evaluate the predictive performance of the model, the parameters of the model were first estimated for the 25-year period 1975–99. The daily numbers of deaths for the four summers 2000–03 were then predicted from the observed daily temperatures and compared to the observed numbers of deaths.
Prediction for the period 2004–06
The parameters of the model were then estimated on the full 29-year period, 1975–2003.
The resulting model was used to predict the daily expected number of deaths (
) for the summers 2004–06 from the observed daily temperatures. The daily baseline number of deaths 
was the number of deaths predicted by the model for the summer days, 2004–2006, using the reference temperatures.
The observed excess mortality (Ot –) was defined as the difference between the observed number of deaths and the baseline number of deaths. The expected excess mortality (
) was defined as the difference between the expected number of deaths and the baseline number of deaths.
The heat wave period in summer 2006 was defined as the period with consecutive days of alert in at least one département. The first day of this period was that on which the biometeorological indicators were greater than or equal to the cut-off in at least one département (11 July) and the last day was that on which the minimum and maximum biometeorological indicators were still greater than or equal to the threshold in at least one département plus 2 days (28 July). In each département, a dual cut-off was defined for minimum and maximum temperatures, using a mortality-temperature study of 14 pilot cities, extended to cover the whole of mainland France.7 Other criteria such as forecast reliability, wind, humidity, air pollution and holiday departures were also taken into consideration in the alert proposal.8
The observed excess mortality and the expected excess mortality during the 2006 heat wave were studied by gender for the whole population and for the two age groups (55–74 years, 
75 years).
The confidence intervals of the difference between the predicted and observed deaths included the variance of the model estimates, the over-dispersed Poisson component and the residual autocorrelation for successive days.
|
Results |
|---|
|
TopAbstractIntroductionMaterial and methodsResultsDiscussionAcknowledgementsReferences |
|---|
The daily fluctuations in the mortality rate and the minimum/maximum temperatures were closely correlated during the summers 2004–06 (Figure 1). Several moderate heat episodes were observed in summer 2005. In 2006, the temperatures began increasing as of the first days of July, with daytime temperatures reaching 35°C, and night-time temperatures above 20°C. The decrease in temperature began on July 28 in the west. Temperatures were still very high at the end of the month in the south, where six départements were still on alert during the first days of August. Seventy departments (76% of the French population) experienced at least two days of alert from 29 June to 3 August 2006. In July, only one day (6 July) was free from alert, and the Paris area and the south-east of France were the most impacted regions (Figure 2). During the period of continuous alert, from 11 to 28 July, 69 départements experienced at least 2 days of alert.
|
|
Goodness-of-fit (1975–99) and predictive performance (2000–03)
In order to evaluate the predictive performance of the model, the daily numbers of deaths were predicted with the model for the four summers, 2000–03, after adjusting on the calibration period 1975–99. The observed mortality rates were very close to the expected mortality rates for the days with usual temperatures and for the 2003 heat wave (Figure 3).
|
When the four summers, 2000–03, were included in the calibration period, the estimate of the daily mortality rates during the periods with extreme temperatures was further improved. As a result, for the 29 summers from 1975 to 2003, the model explained 76% of the total extra-Poisson variability of the daily death count.
Temperature based prediction of mortality for summers 2004–06
During summers 2004 and 2005, the observed numbers of deaths were 2–8% lower than the temperature-based predicted deaths, independently of the daily temperatures (Table 1).
|
During the 3 months, June, August and September 2006, the difference between the observed and the expected numbers of deaths was very small for people aged
55 years, with ± 15 deaths per day (1.3% of the daily mortality in summer) (Table 1). The daily fluctuations in the expected mortality rates were closely correlated with the fluctuations in the observed mortality rates during those 3 months. A marked increase in mortality was observed for July 2006, when the heat wave occurred.
Heat Wave 2006
During the 2006 heat wave, from 11 to 28 July, 2065 excess deaths were observed (CI95% = 1630–2499), equivalent to an excess mortality of +9%. The expected number of excess deaths predicted from the observed temperatures was 6452 deaths (CI95% = 6178–6726) and the expected mortality ratio was +27% (Table 2). Consequently, the estimated mortality deficit was very substantial: –4388 deaths (CI95% = –4920 to –3855).
|
For subjects aged 75 years and more, 1254 excess deaths (CI95% = 907–1601) occurred during the heat wave, i.e. an 8% increase in mortality. The expected number of excess deaths was 5080 deaths (CI95% = 4824–5334), equivalent to an excess mortality of 34%. Thus, the excess mortality deficit was –3825 deaths (CI95% = –4263 to –3386).
For subjects aged 55–74 years, the observed and expected numbers of excess deaths were moderate, with 399 observed excess deaths (CI95% = 226–572) and 1141 expected excess deaths (CI95% = 1061–1221) (Table 2). The excess mortality deficit was –742 deaths (CI95% = –941 to –543).
Excess mortality occurred in both men and women during the 2006 heat wave (Table 2). The excess mortality and excess mortality deficit were greater for women, even for the elderly.
|
Discussion |
|---|
|
TopAbstractIntroductionMaterial and methodsResultsDiscussionAcknowledgementsReferences |
|---|
The comparison of the observed and expected mortalities during the 2006 heat wave has shown that 2065 excess deaths occurred in the total French population. In the absence of preventive measures, given the temperature levels, 6452 excess deaths were expected during that period. Consequently, the excess mortality was much lower than expected on the basis of the temperature-based model (–4400 deaths, approximately). These findings may be related to the establishment of preventive measures in the context of the National Heat Plan.
For summers 2004 and 2005, the model generated a moderate overestimate of the daily number of deaths of 2–8%, independently of the temperature level. Excluding the month of July, the daily expected and observed numbers of deaths were close in 2006. During the 2006 heat wave, the excess mortality was marked for the oldest people and for women.
Although the results were derived from a statistical model, the good predictive performance of the model for the 2000–03 period strengthens confidence in the results.
A long-term downward general mortality trend in France from 1975 to 2003 was observed. This trend would probably have been occurring, with or without interventions, from 2004 to 2006. This long-term time course of mortality from 1975 to 2006 is taken into account by the adjustment on the mortality observed in the reference period. The time course of the daily mortality in summer is thus relative to the evolution of mortality on the reference period.
The difference between mortality during the summer and mortality during the reference period seemed to have changed in 2004 and 2005 compared with the difference observed from 1975 to 2003. The mortality observed during the months used as the reference period (October, November, April and May) have a regular time course from year to year and is probably not responsible for the differences with respect to the estimates for the summers of 2004 and 2005.
It does not seem easy to choose between the different hypotheses that may be advanced to explain that result. After the 2003 heat wave, several preventive measures were taken by the population and by health institutions and authorities to limit the risks associated with high temperatures. Those measures may have modified the relationship between the reference period and summer mortalities during summers 2004 and 2005. The results may thus reflect a marked awareness of vulnerable people (mainly aged people) during these two summers following the 2003 experience, even though the weather was not extremely hot. This behaviour would have been specific to the summer period, compared with the reference period
A long-term harvesting effect of the excess deaths observed in 2003 has also been proposed as an explanation for the mortality deficit in 2004.12 However, the excess mortality in 2003 (+15 000 deaths) was smaller than the mortality deficit observed in 2004 (–25 000 deaths).13 The hypothesis therefore seems to be an improbable explanation for the overestimates for the summers of 2004 and 2005. The mortality during those two summers may also have been influenced by other factors that were not identified.
The pattern of the observed mortality deficit for summer 2006, compared to the expected deaths, was quite different since the deficit was only observed during the 18 days of the heat wave in July 2006.
The marked excess mortality deficit during the 2006 heat wave in France suggested a decrease in the population's vulnerability to high temperatures, probably associated with:
- — the French population's increased awareness of the risks related to extreme summer temperatures after the 2003 heat wave;
- — the set-up of preventive measures with regard to the effects of high temperatures by the health authorities and institutions after the 2003 heat wave;
- — the set-up and implementation of the heat health watch warning system (HHWWS) by the InVS and Méteo-France as of summer 2004.
- — the set-up of preventive measures with regard to the effects of high temperatures by the health authorities and institutions after the 2003 heat wave;
Other countries have established systems for preventing the effects of heat waves. Most were developed and implemented at a local level; rarely on a national scale.14–17 In some cases, the system simply consists of prevention messages included in weather forecasts. Some countries have assessed their warning systems and found evidence that higher levels of heat awareness and implementation of a warning system were responsible for lower levels of mortality when another heat wave occurred.18–20
The present study compared the daily mortality of the whole French population. Although temperatures are heterogeneous in different places in France, the daily population-weighted average of temperatures on the national scale turned out to be highly correlated with the daily number of deaths in summer. Considering the whole population markedly reduced the variability related to small frequencies in studies of specific urban areas. However, from an operational point of view, the alert criteria which determine warning system level should be determined for each département. A finer geographic analysis of the mortality–temperature relationship might thus be used to investigate whether a decrease in population vulnerability occurred in various regions of France.
In conclusion, while about 2100 excess deaths occurred during the 18-day period of the 2006 heat wave, the observed excess mortality was much lower than that expected on the basis of the model (4400 deaths less). The deficit may be related to a change in summer mortality response, which may be due to public awareness of the dangers of high temperatures, the preventive measures set up by the health institutions and authorities, and the set up of a heat warning system after the 2003 heat wave, even though the respective contributions of those factors cannot be distinguished. However, given the magnitude of the persisting excess mortality—2100 deaths—ongoing prevention remains necessary.
|
Acknowledgements |
|---|
|
TopAbstractIntroductionMaterial and methodsResultsDiscussionAcknowledgementsReferences |
|---|
This work was financed by Inserm (French National Institute for Health and Medical Research), InVS (French Institute for Health Surveillance) and DGS (Directorate General for Health). We would like to express our gratitude to the institutions that afforded us various forms of assistance in this study: Inserm, Météo-France, INSEE and the Directorate General for Health (DGS) and to the members of those institutions with whom we worked: F. Laurent, G. Pavillon for Inserm-CépiDc-IFR69; G. Desplanques for INSEE, G. Gayraud and JM. Veysseire for Météo-France, and Philippe Magne for the DGS. We also are very grateful to A. Mullarky for his skilful assistance in the preparation of the English version of this manuscript.
Conflict of interest: None declared.
KEY MESSAGES
|
|
References |
|---|
|
TopAbstractIntroductionMaterial and methodsResultsDiscussionAcknowledgementsReferences |
|---|
1 Fouillet A, Rey G, Laurent F, et al. Excess mortality related to the August 2003 heat wave in France. Int Arch Occup Environ Health (2006) 80:16–24.[CrossRef][Web of Science][Medline]
2 Kosatsky T. The 2003 European heat waves. Euro Surveill (2005) 10:148–49.[Medline]
3 InVS. Impact sanitaire de la vague de chaleur d'août 2003: premiers résultats et travaux à mener. Bull Epidemiol Hebdo (2003) 45–46:1–12.
4 Grynszpan D. Lessons from the French heatwave. Lancet (2003) 362:1169–70.[CrossRef][Web of Science][Medline]
5 Vandentorren S, Bretin P, Zeghnoun A, et al. August 2003 heat wave in France: risk factors for death of elderly people living at home. Eur J Public Health (2006) 16:583–91.
6 Direction Générale de la Santé. Plan national canicule. (2006) http://www.sante.gouv.fr/canicule/doc/plan_canicule_2006.pdf.
7 Pascal M, Laaidi K, Ledrans M, et al. France's heat health watch warning system. Int J Biometeorol (2006) 50:144–53.[CrossRef][Web of Science][Medline]
8 Laaidi K, Pascal M, Berat B, Strauss B, Ledrans M, Empereur-Bissonnet P. Système d'alerte canicule et santé 2006 (sacs 2006). In: Rapport opérationnel. (2006) . Saint-Maurice: Institut de Veille Sanitaire (InVS). 46.
9 Estimated daily mortality during July 2006 in England and Wales. Health Stat Q (2006) 32:107–11.[Medline]
10 Rey G, Jougla E, Fouillet A, et al. The impact of major heat waves on all-cause and cause-specific mortality in France from 1971 to 2003. Int Arch Occup Environ Health (2007) 80:615–26.[CrossRef][Web of Science][Medline]
11 Fouillet A, Rey G, Jougla E, Frayssinet P, Bessemoulin P, Hémon D. A predictive model relating daily fluctuations in summer temperatures and mortality rates. BMC Public Health (2007) 7:114.[CrossRef][Medline]
12 Valleron A, Boumendil A. Epidemiology and heat waves: analysis of the 2003 episode in France. C R Biol (2004) 327:1125–41.[CrossRef][Web of Science][Medline]
13 Toulemon L, Barbieri M. The Mortality Impact of the August 2003 Heat Wave in France. 25th International Population Conference IUSSP, Tours, 18–23 July 2005.
14 Tan J, Kalkstein LS, Huang J, Lin S, Yin H, Shao D. An operational heat/health warning system in Shanghai. Int J Biometeorol (2004) 48:157–62.[CrossRef][Web of Science][Medline]
15 WHO Regional Committee for Europe. Heatwaves: impacts and responses. WHO 1-11. (2003).
16 Michelozzi P, de’ Donato F, Accetta G, et al. Impact of heat waves on mortality - Rome, Italy, June-August 2003. MMWR Morb Mortal Wkly Rep (2004) 53:369–71.[Medline]
17 Kalkstein L, Jamason P, Greene J, Libby J, Robinson L. The Philadelphia hot weather-health watch/warning system: development and application, summer 1995. Bull Am Meteorol Society (1996) 77:1519–28.[CrossRef]
18 Weisskopf MG, Anderson HA, Foldy S, et al. Heat wave morbidity and mortality, Milwaukee, Wis, 1999 vs 1995: an improved response? Am J Public Health (2002) 92:830–33.
19 Tan J, Zheng Y, Song G, Kalkstein L, Kalkstein A, Tang X. Heat wave impacts on mortality in Shanghai, 1998 and 2003. Int J Biometeorol (2007) 51:193–200.[CrossRef][Web of Science][Medline]
20 Smoyer KE. A comparative analysis of heat waves and associated mortality in St. Louis, Missouri–1980 and 1995. Int J Biometeorol (1998) 42:44–50.[CrossRef][Web of Science][Medline]
21 Naughton MP, Henderson A, Mirabelli MC, et al. Heat-related mortality during a 1999 heat wave in Chicago. Am J Prev Med (2002) 22:221–27.[CrossRef][Web of Science][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  J. Rocklov, B. Forsberg, and K. Meister Winter mortality modifies the heat-mortality association the following summer Eur. Respir. J., February 1, 2009; 33(2): 245 - 251. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M Pascal Commentary: Our next challenge in heatwave prevention Int. J. Epidemiol., December 1, 2008; 37(6): 1365 - 1366. [Full Text] [PDF]
|
|
|
|
|
|
J. Sunyer Commentary: Evaluating response to heat waves Int. J. Epidemiol., April 1, 2008; 37(2): 317 - 318. [Full Text] [PDF]
|
|
|
|
|
|
J. Tan Commentary: People's vulnerability to heat wave Int. J. Epidemiol., April 1, 2008; 37(2): 318 - 320. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||