The Office for National Statistics (ONS) and its predecessors have produced weekly national mortality data since the 1850s. The purpose of our weekly deaths system is to provide a quick estimate of any increase in deaths related to events such as an influenza outbreak, or period of exceptional weather.

This paper will examine the impact of the 2003 heat wave on mortality in England and Wales by age group and region. Air pollution and temperature data will also be presented. The impact on mortality will be compared with the rapid estimates produced by the weekly deaths reporting system, which produced the first estimates of excess mortality in England and Wales in August 2003.

Method
The specific heat wave episode in August was defined as starting when maximum daily Central England Temperature (CET) first exceeded average values (1971 to 2000) by 8°C and ending when temperatures returned to average levels. Excess mortality by age group and region was calculated for the 10 day heat wave period (4 to 13 August 2003). In addition, a value for excess mortality over the whole of July and August was also calculated; no further calculations were made for this time period.

The mortality data, based on date of death, were adjusted on a regional basis for the small proportion of deaths still unregistered at the time of the analysis (approximately 1%). Deaths were assigned to Government Office Regions (GORs) of residence.

Excess mortality was calculated as observed deaths minus the expected mortality (average of deaths in the same years 1998 to 2002). The number of observed deaths was treated as a Poisson variable. The 95% confidence limits for this value were then subtracted from expected mortality to give confidence limits for excess mortality. These limits were then calculated as a percentage of expected mortality to give percentage confidence limits for excess mortality.

The Met Office (the government agency that supplies meteorological data on the weather and climate) supplied temperatures for each GOR during the episode. Within each GOR, the maximum and minimum of the daily maxima were then identified. Daily values were generated for a national 5 km grid by interpolation of data from approximately 560 stations. Daily data for the London Weather Centre weather station, and Central England Temperature (CET) were downloaded from the British Atmospheric Data Centre. CET is representative of a roughly triangular area in central England enclosed by the cities of Preston, London and Bristol.

Measurements of the ambient air concentrations of ground level ozone (daily maximum of a running 8 hour mean) and PM₁₀ (particulate matter of diameter less than 10 µm) (24 hour mean) were obtained from the UK National Air Quality Archive. Air pollution data was collected to allow a description only of a potential confounder of the association with temperature.
Excess pollutant exposure was calculated as the difference between the daily regional concentrations averaged over the episode and the equivalent values observed during same period in 2002, when mean concentrations were considered to be typical mean values for August.

Our current rapid weekly response system takes deaths which have been registered in England and Wales in the previous week and uses this to make an estimate of what the final registered numbers will be [1]. Estimates of excess were made by comparing the registered deaths for each heat wave week in 2003 with the average number of deaths registered for the same week in the previous five years.

Subsequent analysis based on the day when deaths occurred will include deaths which were registered immediately as well as deaths which were registered in the weeks or months following the death.

Results: England and Wales
In England and Wales, there were 2139 excess deaths (16% increase, CI: 14% to 18%) during the August heat wave. The Central England Temperature (CET) peaked at 31.5°C (88.7°F) on 9 August [FIGURE 1]. This coincided with the peaks in the concentrations of ozone and PM₁₀ in England. The peak in daily deaths in England and Wales occurred two days later on 11 August. In England and Wales, mortality in people over the age of 75 increased by 22% (CI: 20% to 25%), more than the increase seen for other age groups: (11% (CI: 6% to 15%) for the 0-64 age group and 3% (CI: -1% to 6%) for the 65-74 age group).

Overall in July there were fewer deaths than expected (-1%), despite a slight increase in temperatures and mortality during mid-July (FIGURE 1). Overall in August there was an excess of 5%. The increase in mortality over the heat wave episode (4-13 August) was followed by a decrease in deaths (-4%) in the period 24 to 29 August 2003.

Results: Regions
The impact was greatest in the southern half of England, particularly in London, where deaths for all ages increased by 42% (CI: 36% to 48%) [TABLE 1].

All regions had an excess for people over the age of 75 years. However the greatest excess in the over 75 age group was in the London region with a 59% (CI: 51% to 67%) increase. London experienced night time temperatures of 26-27°C (79 to 81°C) during the heat wave, and a maximum of 37.9°C (100.2°F) was recorded in London on 10 August ([FIGURE 2].

Concentrations of ozone and PM₁₀ peaked in London on 6 and 8 August respectively [FIGURE 3]. Excess concentrations of PM₁₀ and ozone were highest in London and South East regions respectively [TABLE 2].

Results: Comparison with initial estimates from the weekly deaths reporting system
Data presented above [FIGURE 1] of the number of deaths occurring each day show that deaths started to increase on 4 August 2003. This was included in the week of 2 to 8 August (registration week 32). An estimate was made of deaths registered in this week on Thursday 14 August. Our estimation method takes into account short delays in data being sent to ONS [1].

The first clear indication of the upturn in deaths registered following the hot temperatures was given by ONS on 21 August, when an estimate of 907 excess deaths was made for registration week 33, 8 to 15 August 2003. This estimate was of a similar order of magnitude to the final excess mortality for that week.

However, this method did not indicate well the mortality excess in the week following the heat wave. The initial estimate of excess mortality was -75, but the final excess was 309 deaths for that week. The excess in registrations in this week are likely to be deaths that occurred in the previous two weeks but that were not registered until week 34.

Table 3 shows initial and final information on death registrations in England and Wales for the period 2 to 22 August 2003. Registrations for the weeks before and after this period were around average levels. Final number of deaths registered in weeks 32 to 34 (2 to 22 August) were a total of 1828 deaths above average (average of the same week over the previous five years). This final estimate of death registrations is much higher than the initial estimated excess over the same period of 910 deaths.

Subsequent work was based on the days when deaths actually occurred. As already described, there were 2139 excess deaths over the episode of 4 to 13 August, when temperatures were at their highest. As expected, this was more than the final excess of 1828 deaths registered over the three weeks, as deaths registered much later than week 34 are included in the number of deaths by date of occurrence.

Discussion
The heat wave had a major effect on mortality in England and Wales, but not to the extent of that observed in France where hot temperatures were maintained for much longer [2].

Excess mortality was much greater than that observed with previous heat waves in the UK. In Greater London it was estimated that the 2003 heat wave was associated with a 40% increase in mortality (all ages) compared with an excess of 16% in 1995 and 15% in 1976 [3,4]. Excess mortality in England and Wales was 10% in 1976, compared with 16% in 2003. Temperatures during the 1976 event were of a comparable magnitude, but the increasing ageing population in England and Wales [5] may have contributed to the increased 2003 excess.

The 2003 heat wave in the United Kingdom occurred relatively late in the summer. Deaths in July overall were slightly below expected levels. Temperatures and mortality did increase in mid-July, and mortality was above what would normally be expected for the time of year. However, high temperatures were not sustained, and did not reach the levels seen in August. There was a small dip in mortality following the heat wave in August, indicating possible displacement of a proportion of deaths by the heat wave.

High ozone concentrations are an important co-exposure during heat waves in England. High ozone concentrations were reported during the 1976 heat wave [6]. Excess exposure to ozone and PM₁₀ were recorded for all regions in England, most notably in London and the South East. Between 21% and 38% of the excess deaths (where excess deaths were predictions based on previous time series studies of air pollution and mortality) in the 2003 heat wave were estimated to be attributable to ozone and PM₁₀, although that study assumed no interaction between high temperatures and high pollutant exposures [7]. This study has not attempted to separate out the effects of pollutants and temperature.

Cities are usually more affected by increasing temperatures than surrounding areas where building density is lower [8]. The nocturnal urban heat island in London is greatest in the summer months, and has increased since the 1960s [9].

The elderly (over 75 years) are most vulnerable to heat related mortality, as has been shown in other heat wave studies in the UK [3] and in other countries [10]. When older people live alone, they may not receive the care they need during a heat wave (for example, adequate hydration) and they are also unlikely to call for medical attention, and therefore may die at home without being admitted to hospital [11].

Smaller increases were seen in many regions in the 0-64 year age group, which may reflect an increase in mortality in children and infants who are also at risk from heat-related deaths [8], or an increase in mortality in sick adults (e.g., those with chronic cardiorespiratory disease).

The 2003 data used for mortality is provisional. There is some uncertainty about the number of deaths that have still not been registered, but a reliable estimate of final values can be made.

The weekly deaths reporting system provided a useful indication of the impact of the hot weather in England and Wales, although initial figures were an underestimate. The peak in the number of deaths occurring in England and Wales was on 11 August 2003; the registration week which included this date showed a clear excess (the system picked up the biggest impact week). Any heat warning system would also need to make use of additional information such as temperature data.

When deaths are registered the information is passed to ONS. When a death is registered in a particular week it does not mean that the death necessarily occurred in that week; some deaths are registered in the weeks following the death. Data by date of occurrences is not therefore available as rapidly as deaths by date of registration. Estimates based on the numbers of deaths registered for the week therefore provide a more timely indicator. Initial figures by date of death were published in October 2003 [12].

As climate change continues, heat waves are very likely to increase in frequency and intensity [13] and are likely to exacerbate London’s urban heat island [9]. The weekly deaths reporting system can play a useful role as a quick i ndication of the impact on mortality of an event such as a heat wave. The prevention of deaths has been addressed in a heat wave plan that has recently been published by the Department of Health [14].