heat waves are known to affect mortality rates. Severe heat waves effects
in the United States have been described for the cities of St Louis (1966)
, St Louis and Kansas City (1980) , Philadelphia (1993) [4) and
Chicago (1995) .
The effects of heat waves appear to be important in Portugal. Portuguese
heat waves episodes and their consequences are currently well documented
Severe heat waves effects on mortality in June 1981 were acknowledged, initially
in the concelho (small administrative unit) of Cascais within the district
of Lisbon, and an estimation of excess of deaths prepared for the entire
district of Lisbon was presented later (1988) . In 1998, a study based
on national mortality data, carried out by the Observatório Nacional
de Saúde (National Observatory of Health, ONSA), estimated the number
of heat waves related deaths nationwide at about 1900 .
In July 1991, Portugal was struck by another heat waves. Again, its effects
on mortality were studied and a nationwide estimate of 1000 excess deaths
was made .
Given the impact of the two previous heat waves, in 1999 ONSA created ÍCARO
(standing for ‘Importância do CAlor: Repercussão sobre
os Óbitos’, which means ‘the importance of heat and
its repercussions on mortality’), a system, which sought to conceive
and operate an alert system for heat waves that influence mortality; and
to study the characteristics and effects of heat waves . The result of
a joint action between ONSA and of the Centro de Vigilância, Previsão
e Informação (the Insitute of Meteorology’s Surveillance,
Forecast and Information Service), this system generates the ÍCARO
index, calculated and reported to other institutions, daily between 15
May and 30 September each year. This index indicates the possibility of
occurrence of heat waves, with probable influence on mortality for the region
of Lisbon, with an anticipation of 3 days (further details are available
in “Fontes de informação”, “ÍCARO” at www.onsa.pt ).
In the summer of 2003, between 29 July and 13 August. all districts in
Portugal experienced unusually high temperatures. At least 8 of the 18
mainland Portuguese districts had daily maximum temperatures above 32ºC
during all this period. Four districts, corresponding to the non-coastal
interior of Portugal, had daily maximum temperatures above 35ºC during
the entire period.
When records going back to 1980 were consulted, it was seen that, for the
first time during this period, 15 out of the 16 days between 29 July and
13 August had maximum temperatures above 32ºC in the district of Lisbon,
including a noteworthy consecutive run of 10 days of such high temperatures.
A 5 day run of temperatures above 35ºC was also recorded for the first
time since 1980.
On 12 August 2003, while the heat waves was still happening, ONSA designed
a preliminary study that aimed to assess the influence of the heat waves
on mortality in the general population. Results were preliminary, because
it was not yet possible to consider the full effect of the heat waves at
Mortality data were obtained from the 31 national civil registrars covering
all the district capitals of mainland (continental Portugal), and representing
approximately 41.5% of overall mortality. The daily number of deaths registered
from 1 June to 12 August was requested on 12 August, and obtained by 19
Period of time studied
The reference period used for the death toll was 14 days, comprising data
from 30 July (the first day of the heat waves +1 day for the death registration
delay) and 12 August.
Expected number of deaths
The expected number of deaths (E) if the heat waves had not occurred was
calculated using three periods within July 2003 (for comparison purposes):
15-28 July; 1-14 July; and 1-28 July, all excluding heat waves-influenced
The product of 14 days multiplied by the average daily number of deaths
registered for each of the three reference periods was used for the calculation
of E, generating three expected death estimates.
Comparison of the expected and observed number of deaths
The number of deaths registered in each one of the national civil registrars
was summed, constituting the total number of deaths observed (O).
The excess of deaths caused by the heat waves was calculated by the difference
O–E, for each of the 3 different E values. These differences represent
the number of heat waves-related deaths, for each of the comparison periods.
p = (O–E) / E = 1 – O/E = 1 – r represents the proportion
of the excess of deaths in relation to the expected deaths.
Estimation of the total excess of deaths related to the heat waves
in mainland Portugal
The total number of deaths related to the heat waves was estimated by
p x ECont in which
ECont = 3486, the number of expected deaths, in the reference period, in
the mainland and was calculated by the product of 14 days and 249 deaths.
This last number is the daily average number of deaths in the same period
as the heat waves in mainland Portugal in 2001 (the most recent mortality
data available from ONSA).
Confidence interval estimates
Excess mortality 95% confidence limits ( ELow ; EUpp )
for each comparison period were obtained from the 95% confidence for
the O/E ratio ( rLow ; rUpp ) calculated by the ‘exact
method’ described by Silcocks that uses the relation between the
Beta and Binomial distributions .
ELow = ( rLow – 1 ) x 3486 and EUpp = (
rUpp – 1) x 3486.
The total number of deaths registered between 30 July and 12 August was
1966. The daily average number was calculated as 140.4 [TABLE 1].
The excess deaths estimates varied slightly for the three comparison
periods used [TABLE 2].
The estimation of excess of deaths within the selected concelhos and
during the period of the heat waves studied, amounted in 539 deaths,
when the last 14 days prior to the heat waves are used as a comparison
period. When compared with number of deaths registered during the first
two weeks of July, the same estimate results in an excess death toll
The estimated excess of deaths, using the period 15-28 July for comparison,
represents a heat-related death toll raise of about 37.8%.
Estimates of the total number of heat waves related deaths within the
Using for comparison the period of two weeks that preceded the heat waves
(15-28 July), an excess of about 1316 deaths was estimated [TABLE 3].Although
the 3 reference periods provide similar estimates, as presented in table
3, the estimate using the period 15-28 July should be the most accurate,
given its proximity to the first day of the heat waves.
The best estimate of the effect of the heat waves of August 2003 on the
mortality of the Portuguese mainland population was 1316 deaths. For
comparison periods further away from the start date of the heat waves,
the estimates produced values of 1271 and 1227 deaths. These results
were consistent among themselves, conveying additional confidence to
the estimates, although for the reasons mentioned above, the value
1316 deaths, calculated based on the period 15-28 July, should be the
Rural versus non-rural population
The exclusive use of civil registrars in district capitals only was chosen
because of time and organisational restraints. This option may have
introduce some systematic bias, forexample, there may be under
-representation of rural areas where
1. the ability of the population to resist heat might be different from
other areas – contributing to some inaccuracy in estimates;
2. age distribution might be different – having more elderly citizens
might translate in a greater effect on mortality.
In line with past experience which has found higher mortality in older
age groups, this could represent an underestimation of mortality.
This rapid method is not meant to estimate rural and non-rural effects
of heat waves. As long as the interest lies in identifying and assessing
a heat waves effect, the use of mainly non-rural areas can be an additional
factor of difficulty. The use of a balanced or proportional sample of
civil registrar’s offices in future estimates is advised to prevent
this potential bias.
Preliminary nature and limitations of estimates
The urgency to find out the dimension of the heat waves’s effects
on mortality imposed the following limitations:
1. the period studied ended on 12 August. It is expected that the heat waves’s
influence continued beyond this date;
2. the use of a sample of national civil registrars of mainland Portugal
made up of all district capitals instead of all concelhos.
The first limitation necessarily induces an underestimation of heat waves
excess related deaths
The second limitation may produce either an under- or an overestimation,
since the populations of the concelhos not represented in the sample
may have experienced different effects of the heat waves.
The average number of deaths registered annually in the participating
31 national civil registrars corresponds to about 41.5% of the total
deaths registered in mainland Portugal (data from 2001). Therefore, the
likelihood that including deaths registered in the remaining national
civil registrars would change the estimates profoundly is low.
Alternative explanations for the excess of deaths
The excess number of deaths could have originated, totally or partly,
from simultaneous phenomena besides the heat waves:
1. The presence of a high number of tourists increases the population
of the Portuguese mainland during the month of August. However, their
presence is equally high during the month of July, especially in the
last two weeks, which was the time period used as the main comparison
period. Also, most tourists visiting Portugal during summer do not belong
to the oldest age groups, and so are unlikely to have made much contribution
to the heat waves related excess of deaths;
2. Visits by Portuguese who have emigrated abroad, and their descendents
are more frequent in August than in July, and this could contribute to
the increase of mortality, independently of the influence of the heat waves.
However, such visitors tend to bein good health and belonging to younger
age groups, and therefore more resistant to the harmful effects of heat waves
than the elderly.
3. Road accidents are more frequent during the month of August, which
could influence excess deaths. However, this should not influence the
current estimates, since the number of fatal victims of road accidents
in August is very similar to that of July (157 deaths were reported during
July 2002 and 130 during August 2002 - data from Direcção
Geral de Viação (Directorate-General for Transport) 
4. A series of forest fires happened during the same time period as the
heat waves. According to the general media, there were 18 fatalities. These
deaths may only indirectly be attributed to the heat waves. While they
influence the total number of deaths to 12 August 2003, only some of
these deaths (and presumably a small proportion) will have been registered
in the national civil registrars participating in this study.
The 2003 heat waves influenced the excess of mortality less than the 1981
There may be several reasons for this:
1. Access to and quality of healthcare is better in 2003 than in 1981;
2. Contrary to the events of 1981 and 1991, there was an alert for the
2003 heat waves and intervention deployed by the Serviço Nacional
de Bombeiros e Protecção Civil (National Service of Firemen
and Civil Protection), in order to diminish the effect of the heat waves;
3. Athough there is no scientifically sustained confirmation yet, it
seems natural that heat waves striking early in the year should have more
influence on mortality than those which occur later. The adaptation of
the individual to progressively rising temperatures should explain a
higher resistance to heat.
4. The nature of the July/August 2003 heat waves was unusual, having three
different temperature peaks. The daily maximum air temperatures dropped
considerably for one day between the first and second peak (3-4 August)
in the coastal districts of Portugal, where most of the Portuguese population
lives. Therefore it can be argued that the majority of the Portuguese
population was not exposed to a consecutive period of heat stress longer
that that of 1981, and this could be a major explanation for the observed
reduced heat waves impact.
Although all these reasons may be behind the reduction of the effects
of the 2003 heat waves, it does not appear possible to determine the relative
influence of each.
In conclusion, at the date of issue of the report, the heat waves of 2003
was estimated as having caused about 1316 deaths up to 12 August, mainly
in the elderly.
Definitive 2003 heat waves effect
The work and methodology presented here gave sound evidence of severe
impacts on the health and mortality of the population, and this stimulated
the responsible institutions to exert unusual efforts to gather complete
information and knowledge abut what really happened. The Portuguese
Direcção-Geral da Saúde (General Directorate of
Health) was able to obtainthe death certificates for summer 2003 earlier
than usual to create a database. A joint report on the complete and
final effects of the heat waves effects , prepared by the Direcção-Geral
da Saúde and ONSA, was finished in April 2004, and replaced
all previous plans for future studies .
This report showed that the heat waves’s impact on mortality occurred
between 30 July and 15 August. For the estimation of the expected number
of deaths without the heat waves effects, mortality data by district, sex
and age group for the years of 2000 and 2001 were used. A total excess
mortality of 1953 deaths (1866-2039:95%CI) was estimated. These deaths
were observed in the older age groups, mainly 75 years old and above.
The number of excess deaths estimated for women were more than twice
the number estimated for men. Mortality effects were observed in all
Portuguese districts, and the non-coastal districts had higher relative
increases in mortality (Guarda, Castelo Branco, Portalegre and Évora).
Causes of deaths most strongly associated with the heat waves were ‘heatstroke’ (O/E=70.0)
and ‘Other disorders of fluid, electrolyte, and acid-base balance’ (O/E
= 8.65). Other important associated causes of death accounting for higher
mortality were ‘diseases of the circulatory system’ (758
estimated excess deaths, of which about 370 were ‘cerebrovascular
disease’, about 145 were ‘ischaemic heart disease’ and
118 were ‘heart failure’), ‘diseases of the respiratory
system’ (about 255 excess deaths) and ‘all malignant neoplasms’ (about
131 excess deaths).
These full effect estimations clearly show that the 2003 heat waves was
different from the 1981 heat waves in several ways: most importantly, the
mortality impact differed in age groups, with children being spared,
and it was more intense for women.
Direct comparisons between final full heat waves effect and preliminary
estimate is not possible, for two main reasons: the time periods in the
two methodologies are different; and the underlying starting data are
While the final full effect estimate is based on the date of death, basic
data of the preliminary estimation methodology is solely the number of
deaths registered by the civil registrar’s offices on each given
working day, which does not account for locally displaced deaths, holidays
and other similar phenomena.
This limitation could be overcome by having a wider period of civil registrar
office notifications, allowing for all deaths during the intended period
of study to be accounted for, but such a solution is against the intended
nature of the methodology meant to give a timely estimate of possible
effects of the heat waves. This is a valuable solution when definitive
mortality data is not available quickly enough.
The sooner this rapid method is applied to estimate heat waves effects,
either during or after the heat waves occurrence, the more likely it is
that mortality impact estimates will be biased towards the lower limits,
but when significant impact is shown, the method’s objectives are
In the summer of 2004, a system of daily mortality surveillance was established
that will henceforth operate annually in tandem with the ÍCARO
surveillance system. This new daily mortality surveillance system was
created based on the experience and methodology described in this paper.
This system consists mainly of collecting very simple data (the total
number of registered deaths) on a daily basis from a sample of 67 civil
registrar’s offices distributed throughout the districts of mainland
Portugal. Of these 67 offices, 31 are from district capitals and participated
in the work presented here.
We are most grateful to the General Director of the Direcção
Geral dos Registos e do Notariado and the Conservatórias do Resgisto
Civil who promptly collaborated with ONSA and without whom this work
would not have been possible.
We are particularly grateful to Dr Teresa Abrantes and Dr Fátima
Coelho of the Portuguese Meteorology Institute, for helping us reviewing
air temperatures for recent years, particularly the 2003 summer data.
We are grateful to the Fundação de Ciência e Tecnologia
(FCT) that partially funded this work (Projecto POCTI/ESP/39679/2001).