Exploratory analysis of seasonal mortality in England and Wales, 1998 to 2007.

BACKGROUND This article reports research carried out to inform possible methods of describing seasonal mortality in relation to extremes of temperature. In particular, since different methods are currently used to assess excess winter mortality and heatwave related mortality, we aimed to find out whether a single method could be used to measure all seasonal mortality in relation to temperature. In order to do this the project investigated whether there are temperatures above or below which excess deaths occur, and explored whether it is possible to predict reliably how many deaths would occur at extreme temperatures. METHODS Daily and monthly Central England Temperatures for 1998 to 2007 were supplied by the Met Office Hadley Centre and daily death occurrence data between 1993 and 2007 was extracted from the death registrations database held by the Office for National Statistics (ONS). Least squares regression, based on the previous five years of data, was used to predict expected mortality, and excess mortality was calculated as the difference between the expected mortality and the observed mortality on any given day. Statistically significant increases in both daily deaths and temperatures were investigated with the probability of excess mortality assessed on those days. Two regression models were calculated, one for deaths and temperature and one for excess deaths and temperature. RESULTS Five days with statistically significant excess mortality were identified over the period 1 January 1998 to 31 December 2007, the largest being on 31 December 1999. Three of the five days identified coincided with extremely hot weather occurring in August 2003 and July 2006. However, more extreme temperatures were seen on some days with no excess mortality, so predicting mortality using extreme temperatures alone would cause frequent false positive results. Regression models based on daily death and temperature explained only 8 per cent of the variance in summer mortality and 7 per cent of the variance in winter mortality. The models based on excess deaths and temperature explained 20 per cent of the variance in excess mortality in summer, but only 1 per cent of the variance in excess mortality in winter. CONCLUSION There is a weak but significant relationship between temperature and mortality in both the summer and winter months. While in winter mortality does increase as it gets colder, winter mortality is variable and high mortality can occur on relatively mild days. Similarly, in the summer high temperatures are often associated with relatively increased mortality, but a single hot day does not always lead to excess deaths. Daily mortality cannot be predicted from temperature alone: the prevalence of influenza in winter and factors such as air pollution in summer should also be considered.