The European Journal of Public Health Advance Access originally published online on October 6, 2006
The European Journal of Public Health 2006 16(6):583-591; doi:10.1093/eurpub/ckl063
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Heat-related mortality |
August 2003 Heat Wave in France: Risk Factors for Death of Elderly People Living at Home
S. Vandentorren1, P. Bretin1, A Zeghnoun1, L. Mandereau-Bruno1, A. Croisier1, C. Cochet2, J. Ribéron2, I. Siberan1, B. Declercq1 and M. Ledrans1
1 Department of Environmental Health, French Institute of Public Health Surveillance, Cedex, France
2 Technical and Scientific Building Center (CSTB), Paris, France
Correspondence: Stéphanie Vandentorren, MD, Department of Environmental Health, Institute for Public Health Surveillance, 12 rue du val d'Osne, 94415 Saint Maurice Cedex, France, tel: +33 1 55 12 53 08, fax: +33 1 41 79 67 68, e-mail: s.vandentorren{at}invs.sante.fr
Received May 17, 2006, accepted March 28, 2006
 |
Abstract |
|---|
 Top Abstract MethodsResultsDiscussionReferences |
|---|
The August 2003 heat wave in France resulted in many thousands of excess deaths particularly of elderly people. Individual and environmental risk factors for death among the community-dwelling elderly were identified. We conducted a case–control survey and defined cases as people aged 65 years and older who lived at home and died from August 8 through August 13 from causes other than accident, suicide, or surgical complications. Controls were matched with cases for age, sex, and residential area. Interviewers used questionnaires to collect data. Satellite pictures provided profiles of the heat island characteristics around the homes. Lack of mobility was a major risk factor along with some pre-existing medical conditions. Housing characteristics associated with death were lack of thermal insulation and sleeping on the top floor, right under the roof. The temperature around the building was a major risk factor. Behaviour such as dressing lightly and use of cooling techniques and devices were protective factors. These findings suggest people with pre-existing medical conditions were likely to be vulnerable during heat waves and need information on how to adjust daily routines to heat waves. In the long term, building insulation and urban planning must be adapted to provide protection from possible heat waves.
Keywords: deaths, elderly, epidemiological survey, heat wave
In its minimum, maximum, and mean temperatures and its duration, the August 2003 heat wave was the worst in France since record-keeping began in 1873. Indeed, this heat wave struck France in early August after a warm month of June with temperatures 4–5°C above seasonal averages. It resulted in nearly 15 000 excess deaths during the period from August 1 through August 20.1 Studies of earlier heat waves report that risk factors for death include population vulnerability2 and environmental factors, such as living in urban areas.3,4 The first descriptive studies of this heat wave by the French Institute of Public Health Surveillance (InVS) found that most of the heat-related deaths occurred among the community-dwelling elderly, that is, those living in their own home (alone or with others).5 Identifying the risk factors for these deaths is an important public health priority to prevent a repetition of this toll in future heat waves.6 A case–control study was conducted to determine individual risk factors for death in this population and thereby help define effective public health strategies for population groups at high risk.
|
Methods |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
Study design and case definition
The matched case–control study was conducted in October 2003 in four different areas: Paris, some of its suburbs (Val de Marne), and two cities south of Paris (Tours and Orleans, respectively, 235 and 130 km away). These areas were selected because of their location in the regions most severely affected by the heat wave and their architectural variety. Eligible case subjects were persons aged 65 and over who died from August 8 through August 13, lived in the selected communities, and had been at home at least 24 h before death or hospital admission. All deaths related to accident, suicide, or surgical complications were excluded. Assuming that 10% of the control subjects were exposed to any risk factor and defining the significance level at 0.05, we set the sample size at 300 cases and 300 controls to provide a statistical power of 80% to detect an odds ratio (OR) of 2.0. Eligible cases were randomly selected from the eligible death certificates, and their names and addresses obtained from their municipal vital records office. Subject's next-of-kin were contacted. Eligible controls, that is, those who had lived at home between August 8 and August 13 inclusive and matched cases for age (±5 years), sex, and residential area (we subdivided each of the areas into zones of 100 000 residents according to census records, each zone relatively homogeneous from a socioeconomic standpoint7,8) were randomly selected from telephone records, including unlisted numbers.
Data collection
Investigators, specially trained in interviewing persons in mourning, questioned, face-to-face where possible or otherwise by telephone, the cases' next-of-kin, or a friend or neighbor in the absence of a relative. Controls or their next-of-kin were interviewed in the same way. Information on living conditions came from direct observation. Factors related to housing conditions were grouped in three categories: related to the type of building, the particular dwelling unit, and the principal rooms, especially the bedroom. Information about medical conditions was obtained by calling the family physicians of cases and controls to know whether the patient suffered from mental disorders, cardiovascular, neurological diseases, or other chronic disease. Despite the significant effort spent on completing the questionnaires, it was not possible to collect reliable data for some items such as quantity of liquids per day, bath or shower frequency and opened windows because respondents had to have been much closer to the cases to answer such detailed questions.
Satellite pictures (acquired on 9 August 2003 at 10:17 a.m. and 10:18 a.m. UTC by Landsat 5) provided environmental data for assessing the heat-island profile of the area around the homes during the heat wave. The surface temperature in the study area was calculated using 120m-pixel1 thermal infrared images. The vegetative cover, an important factor, which mitigates the heat island effect, was estimated by combining near-infrared and visible images and calculating the normalized difference vegetation index9 by 30m-pixels.2 After each address was geo-coded, the mean of the surface temperature within a 200 m radius area around each home was calculated using a GIS. The proportion of the 200 m radius area with an elevated normalized difference vegetation index (classes 4–10), corresponding to the proportion of the area with chorophyllous activity, was also computed. This proportion is later called, ‘the vegetation index’.
Statistical analysis
Three separate analyses of deaths were performed: deaths from all causes (one control per case), heat-related deaths (one or two controls per case), and cardiovascular-related deaths (one or two controls per case), the latter because they were known to account for many of the excess deaths.1 Heat-related deaths were defined as those whose death certificates mentioned ‘dehydration’, ‘hyperthermia’ or ‘heat stroke’ as the primary cause of death.
Matched data were analysed with a conditional logistical regression model to investigate the relation between each potential risk factor and mortality.10,11 The relationships between the quantitative risk factors and mortality were initially studied non-parametrically to explore any possible nonlinear effects. Nonlinear transformation was found to be not required; these factors were introduced linearly into the multivariate model. To avoid co-linearity, the model included only one of any group of highly correlated variables representative of the same phenomenon. The multivariate analysis used conditional logistical regression and included only the potential risk factors shown to be important in the univariate analysis (P < 0.20) and to have relatively low rates of missing values. The final multivariate regression model contained the factors that were significant predictors of death (P < 0.10). Despite the matching for age, cases remained older than controls. The univariate and multivariate analyses were adjusted on age by introducing a linear term age into the model. Data analysis used SAS software.12
|
Results |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
Sample description
The study included 315 case and 282 control subjects. Informants were the decedents' child (48%), spouse (19%), other relative (23%), or other friend or neighbour (7%). The rate of participation was 89% among cases and 59% among controls. This rate of participation among controls was quite acceptable. About 36% among controls refused to participate (the main reasons given were lack of time, not interested or fear of receiving someone at home) and 5% of the controls could not be analysed because their interviews were incomprehensible due to confusion. The case sample analysed was similar to the population of deaths from which it was randomly drawn in terms of sex (P = 0.40), age (P = 0.10), and residential area (P = 0.32).
Controls were more often widowed, divorced or single than the overall French population, according to the 1999 census.
The average age was 85.1 years (SD: 8.1) in cases and 82.1 years (SD: 7.5) in controls. Most of the people lived in Paris (205 cases and 184 controls) and in Val de Marne (89 cases and 79 controls), few people lived in Tours (12 cases and 9 controls) or Orleans (9 cases and 10 controls). Cases were similar to controls in terms of sex (P = 0.44) with 72.7% of females in cases and 69.8% in controls.
Causes of death
Cause of death was reported for 254 (81%) of the 315 cases and most often involved either a cardiovascular cause (37%) or the heat (35%). Other causes included cancer (7.5%), respiratory (6.3%), and neurological (4.3%) diseases.
Univariate analysis of all causes of death
Health
Lack of mobility was strongly associated with death (table 1), most especially for those confined to bed (OR = 7.52). As expected, pre-existing medical conditions were also important risk factors for death (table 1). In decreasing order of odds ratios, these included mental disorders, neurological disease, cardiovascular disease, cancer, obesity, and high blood pressure.
|
Social factors
Lower social status was identified as a risk factor (table 2), with the risk of death three times higher among manual workers than managers. A complete absence of social activities, that is, no social, religious, cultural, or leisure activities, multiplied the risk of death by a factor of six. Surprisingly, the use of home attendants (including for paramedical care, help moving, house cleaning, or even meal delivery) was also associated with a higher risk of death.
|
Housing conditions
Table 3 reports the housing factors for which there were not too many missing values (as there were, for example, for building materials). For the first category (building-related), two principal risk factors were obtained: (i) construction date: buildings built before 1975 were associated with a higher risk of death than more recent ones; (ii) comfort level (the percentage of dwelling units with private toilets): buildings with higher comfort levels seemed safer. The principal risk factors related to the individual dwelling units were, in decreasing order, lack of thermal insulation (assessed by building date and questions about insulation) and living on the top floor. In addition, the number of rooms decreased the risk and the number of windows per 50 m2 increased it. The gradient of risk according to the floor (upper floors excluded) was not statistically significant. Finally, two aspects of the bedroom's location were risk factors: its location directly under the roof and the duration of sunlight there.
|
Environmental factors
Of the environmental indicators that might reflect a heat-island effect around the subjects' homes, the temperature—calculated as the mean of the surface temperatures for a 200 m radius around the home—was a risk factor while the vegetation index9—also calculated for a 200 m radius around the home—was described as protective.
Behavioural factors
Finally, it is noteworthy that the behaviour of the case subjects during the heat wave directly influenced their likelihood of death (table 4). This includes behaviour such as visiting air-conditioned or cool places, dressing lightly, or using cooling techniques and devices (defined as any of: a mister, cool, or cold bath or shower, sponge bath, mechanical fan, or air conditioner). On the other hand, opening the windows during the afternoon (instead of opening them at any other time) was associated with an increased risk of death.
|
Multivariate analysis
In the multivariate regression model, the following factors were adjusted: occupation, mobility, behavioural adjustment to the heat wave (going out, dressing, using advice for staying cool), medical conditions (cardiovascular, neurological diseases, mental disorders and high blood pressure), living conditions such as thermal insulation and having a bedroom under the roof and the thermal index (200 m radius). The multivariate analysis enabled us to identify several statistically significant risk factors (table 5). Lack of mobility (bed confinement) remained a major risk factor, together with certain pre-existing medical conditions, including mental disorders, cardiovascular diseases, and neurological diseases. Of the social factors, only being a manual worker was still a significant risk factor. Housing characteristics that remained associated with death during the heat wave were thermal insulation of the unit or building and location of the bedroom under the roof. Temperature also remained a major risk factor. Finally, behaviour such as dressing lightly and using cooling devices and techniques remained protective against death.
|
Analysis of cardiovascular causes of death
Similar analyses were performed for 91 cases and 170 controls and adjusted for age because cases were older than controls (1.17 years). Results were similar to the analysis of all causes of death (table 5).
Analysis of heat-related causes of death
Finally, 87 cases to 164 controls were matched and adjusted for age because cases were older than controls (1.6 years). In the multivariate analysis (table 5), most of the significant factors found in the all-cause analysis were again found to be risk factors, including confinement to bed [OR = 7.12, 95% confidence interval (95% CI) 1.54–33.02] and cardiovascular diseases (OR = 4.73, 95% CI 1.23–18.24). This heat-related analysis also identified high blood pressure as a risk factor (OR = 4.05, 95% CI 1.50–10.90). Odds ratios in this analysis were estimated compared to the all-cause analysis for bedroom under the roof (OR = 5.43, 95% CI 1.33–22.21 compared with 4.06), dressing lightly (OR = 0.08, 95% CI 0.02–0.32 compared with 0.22), and using a cooling device or technique (OR = 0.19, 95% CI 0.04–0.85 compared with 0.32).
|
Discussion |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
This study showed that the elderly at greatest risk of death during this catastrophic heat wave were those confined to bed, with a cardiovascular or neurological disease or mental disorder, those living in old buildings without insulation or in the areas with the greatest heat island effects, and those with a bedroom located directly under the roof. Behaviour adapted to the heat was an important protective factor against death. The results of the three different analyses were consistent, with the strength of the relations of some factors greater in the analysis of heat-related causes of death. This result emphasizes the importance of these factors, which include a bedroom under the roof, dressing lightly, and using cooling devices and techniques.
The narrow case definition may have been too sensitive, but the study period covered only the peak of the heat wave, so that we can assume that we focused on subjects who died from heat. This definition probably prevented misclassification linked to information bias that might have occurred if cases had been recruited from the start of the heat wave. Moreover, a specific analysis was focused on a group of heat-related deaths. Finally, the cases included did not differ from the case pool in terms of age, sex, or residential area and were thus representatives of all the potential case-subjects. The design of the study may have introduced some bias. Thus, telephone recruitment of controls may have introduced a selection bias. Indeed, to respond to the phone, people had to be able to reach the phone and were consequently less confined to bed. To minimize this bias, we also conducted phone interviews during lunchtime to reach home attendants. This selection bias must have affected few other variables: the variable ‘mental disorders’may have been affected by this bias since mentally sick elderly were not able to participate in phone interviews. However, we noted that dialogue was only incomprehensible for 5% of control. Conversely, the variables ‘matrimonial status’ and ‘lived alone’ may have been affected by a selection bias. This may be due to a residual confounding. The information brought by the variables ‘living alone’ and ‘mobility’ are not strictly independent. Indeed the people ‘living alone’ are most likely to be not confined to bed, able to dress and to wash themselves whereas people experiencing mobility difficulties are less susceptible to live alone. All considered, the effect of this bias was minimal as it concerned mainly this variable and did not affect other factors studied, in particular environmental factors, which are important to consider in terms of prevention. They were probably associated, however, with less information bias and fewer inaccuracies. Controls may have responded more precisely than the case informants about habits, although close relatives or friends responded precisely to detailed questions. As the nature of the study did not allow blinding of the interviewers, it might have introduced a bias that is the reason why we checked all medical information by calling the family physician of cases and controls. Recall bias was probably limited because interviews with controls were conducted immediately after those of cases. The supplemental adjustment for age should have reduced the confounding bias: age was an important risk factor even after matching.
When we take these points into account, our results are consistent with those described in previous studies. Some of the risk factors we identified have already been pointed out, including lack of mobility and social isolation,13,14 history of disease,15–17 and lower social status.18,19 We also identified high blood pressure as a risk factor, although it may indirectly reflect diuretic intake. This study focused on environmental factors and showed that housing conditions, such as a bedroom under the roof or the absence of any or good insulation, are risk factors for death, although it is generally difficult to estimate their weight. Other factors, such as the floor (height), the duration of sunlight in the home, the number of windows, and the ability to create drafts were significantly associated with death only in the univariate analysis, perhaps because of a lack of statistical power. Some of these factors have previously been associated with increased death rates.14,20,21 Air conditioning is frequently described as a protective factor in North American studies,22 but was not found to be significantly associated with death in our study, probably because its use in France remains relatively rare. Earlier studies also describe the role of environmental factors, such as heat-island effects.23–27 Our study emphasizes the importance of the outdoor temperature near the dwelling unit and shows that levels of risk can vary for inhabitants in the same district. This is an important finding using an original use of satellite data. The vegetation index was also found to be a protective factor, as previously reported.25 Several earlier studies document variations in the risk of death within a city, even when there is no clear relation with environmental factors.26–28 Finally, our study showed that the subjects who were able to adapt their behaviour to the heat were protected by their actions. These behavioural factors provided the most significant protection in our analysis, and their importance for prevention is primordial.
Several types of recommendations for prevention plans can be deduced from these results. In the short term, special attention during heat waves should be paid to the elderly, those confined to bed or suffering from cardiovascular, neurological, or mental disorders, all of whom are especially vulnerable during heat waves. Secondarily, the population as a whole and especially these vulnerable groups require information on adjusting their daily routines to a heat wave. Avoiding alcoholic beverages,21 increasing the frequency of showers and baths, dressing lightly, using cooling devices or techniques, opening windows when the outdoors temperature is cooler than indoors, and visiting cooler or airconditioned places are simple preventive measures that are widely available and could reduce excess mortality during hot weather. In the long run, improving insulation in existing buildings and urban planning are needed. Specifically, increasing urban vegetative cover and use of paving and roofing materials that are more reflective and less absorbent could reduce heat-related deaths.29 These steps are costly and time-consuming but their implementation is necessary, given the likelihood that the frequency of heat waves will only increase in France and elsewhere.30
Key points
|
|
Footnotes |
|---|
1A pixel correspond to a 120 meters square
2A pixel correspond to a 30 meters square
|
References |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
1 Hémon D, Jougla E The heat wave in France in August 2003. Rev Epidemiol Sante Publique 2004;52:3–5.[Medline]
2 Basu R, Samet JM. Relation between elevated ambient temperature and mortality: a review of the epidemiologic evidence. Epidemiol Rev 2002;24:190–202.
3 Jones TS, Liang AP, Kilbourne EM, et al. Morbidity and mortality associated with the July 1980 heat wave in St Louis and Kansas City, Mo. JAMA 1982;247:3327–31.[Abstract]
4 Besancenot JP. Vagues de chaleur et mortalité dans les grandes agglomérations urbaines. Environnement Risques et Santé 2002;1:229–40.
5 Impact sanitaire de la vague de chaleur d'août 2003 en France. Bilan et perspectives. Institut de Veille Sanitaire, 2003. Available at http://www.invs.sante.fr/publications/2003/bilan_chaleur_1103/vf_invs_canicule.pdf (accessed on November 25, 2003).
6 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;92:830–3.
7 IAURIF, INSEE. Atlas des franciliens. Tome 2001;2:81
8 IAURIF, INSEE. Atlas des franciliens. Tome 2000;1:97
9 Dousset B, Gourmelon F. Satellite multi-sensor data analysis of urban surface temperatures and landcover. ISPRS J Photogrammetry Remote Sensing 2003;58:43–54.[CrossRef]
10 Breslow NE, Day NE. Statistical Methods in Cancer Research. Volume I—The Analysis of Case–Control Studies. IARC Scientific Publications No. 32. Lyon: IARC, 1980.
11 Schlesselman JJ. Case–Control Studies: Design, Conduct, Analysis. New York: Oxford University Press, 1982: xv + 354.
12 SAS Institute Inc. SAS/STAT Software: Changes and Enhancements Through Release 8.02. Cary, NC: SAS Institute Inc., 1999.
13 Naughton MP, Henderson A, Mirabelli MC, et al. Heat related mortality during a 1999 heat wave in Chicago. Am J Prev Med 2002;22:221–7.[CrossRef][ISI][Medline]
14 Semenza JC, Rubin CH, Falter KH, et al. Heat related deaths during the July 1995 heat wave in Chicago. N Engl J Med 1996;335:84–90.
15 Buffat J, Brinquin L Le coup de chaleur: de l'épidémiologie à la prévention. Climat et Santé 1996;15:5–24.
16 Kaiser R, Rubin CH, Henderson AK, et al. Heat related death and mental illness during the 1999 Cincinnati heat wave. Am J Forensic Med Pathol 2001;22:303–7.[CrossRef][ISI][Medline]
17 Kunst AE Temperature et mortalité aux Pays Bas. Essai d'analyse chronologique. Climat et Santé 1996;15:43–64.
18 Curriero FC, Heiner KS, Samet JM, et al. Temperature and mortality in 11 cities of the eastern United States. Am J Epidemiol 2002;155:80–7.
19 Klinenberg E. Review of heat wave: social autopsy of disaster in Chicago. N Engl J Med 2003;348:666–7.
20 Whitman S, Good G, Donoghue ER, et al. Mortality in Chicago attributed to the July 1995 heat wave. Am J Public Health 1997;87:1515–18.
21 Kilbourne EM, Choi K, Jones TS, Thacker SB Risk factors for heatstroke. A case–control study. JAMA 1982;247:3332–6.[Abstract]
22 Kilbourne EM. Heat waves and hot environments. In: Noji K,(editor). The Public Health Consequences of Disasters. New York, NY: Oxford University Press, 1997:245–69
23 Jones TS, Liang AP, Kilbourne EM, et al. Morbidity and mortality associated with the July 1980 heat wave in St Louis and Kansas City, Mo. JAMA 1982;247:3327–31.[Abstract]
24 Rooney C, McMichael AJ, Kovats RS, Coleman MP Excess mortality in England and Wales, and in Greater London, during the 1995 heatwave. J Epidemiol Community Health 1998;52:482–6.[Abstract]
25 Dousset B, Gourmelon F Satellite multi-sensor data analysis of urban surface temperatures and landcover. ISPRS J Photogrammetry Remote Sensing 2003;58:43–54.
26 Schuman SH Patterns of urban heat-wave deaths and implications for prevention: data from New York and St Louis during July, 1966. Environ Res 1972;5:59–75.[Medline]
27 Buechley RW, Van Bruggen J, Truppi LE Heat island equals death island?. Environ Res 1972;52:482–6.
28 Palecki MA, Changnon SA, Kunkel KE The nature and impacts of the July 1999 heat wave in the Midwestern United States: learning from the lessons of 1995. Bull Am Meteorol Soc 2001;82:1353–62.[CrossRef]
29 Etude des facteurs de décès des personnes âgées résidant à domicile durant la vague de chaleur d'août 2003. Available on http://www.invs.sante.fr/publications/default.htm.
30 Schar C, Vidale PL, Luthi D, et al. The role of increasing temperature variability in European summer heatwaves. Nature 2004;427:332–6.[CrossRef][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  A Fouillet, G Rey, V Wagner, K Laaidi, P Empereur-Bissonnet, A Le Tertre, P Frayssinet, P Bessemoulin, F Laurent, P De Crouy-Chanel, et al. 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 Int. J. Epidemiol., April 1, 2008; 37(2): 309 - 317. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Bouchama, M. Dehbi, G. Mohamed, F. Matthies, M. Shoukri, and B. Menne Prognostic Factors in Heat Wave Related Deaths: A Meta-analysis Arch Intern Med, November 12, 2007; 167(20): 2170 - 2176. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||