Mortality as a function of temperature. A study in Valencia, Spain, 1991-1993

Increased mortality is associated with both very low and very high ambient temperatures. This study assesses the relationship between daily numbers of deaths and variations in ambient temperature within the city of Valencia.

The daily number of deaths from all causes (total deaths and only those occurring in people aged over 70), as well as those deaths from specific causes (e.g. cardiovascular and respiratory diseases, malignant tumours and all causes except external ones) occurring within the city of Valencia were related to the average daily temperature using autoregressive Poisson regression controlling for seasonality, day of the week, holidays, air pollution, influenza incidence, and humidity. Temperature was measured within the regression model as two complementary variables: 'Heat' and 'Cold'; also taken into account were their delayed effects up to 2 weeks after measurement.

Graphical analysis revealed a relationship between temperature and mortality according to the time of year. For the cooler months (November-April), the temperature at which mortality was lowest was the 'minimum' (i.e. around 15 degrees C), while for the warmer months (May-October), it occurred at around 24 degrees C. Because of this, a stratified analysis was undertaken with different values for the 'Heat' and 'Cold' variables according to which of the two seasons was involved. During the colder months of the year, higher temperatures tended to exert a rapid influence on mortality and the lower temperatures a more delayed relation. During the hot season it is the heat variable which more clearly manifests an effect, and this is prolonged over the two following weeks. Variations also occur according to age and cause of death. The effect of temperature is greater in persons aged over 70 years of age, and it is also greater in cases of circulatory and respiratory diseases.

A statistically significant association has been found between temperature and mortality. This relationship is not monotonic, but mortality increases in proportion to the variance in ambient temperature from a range of temperatures that varies from winter to summer.