CHARACTERIZATION OF THE ASSOCIATION BETWEEN SHORT TERM VARIATIONS IN DAILY MORTALITY AND ADVERSE ENVIRONMENTAL CONDITIONS USING TIME SERIES METHODOLOGY.

摘要

Available from UMI in association with The British Library.; Most of the literature review reports the use of regression models to investigate mortality-environmental relationships. The goal of this research work has been to devise improved methods of statistical and computational analyses to study mortality associated with daily weather and pollution concentrations making allowance for influenza epidemics. Daily data from Greater London from 1959 to 1976, from November to February have been analysed.; The strategy for analysing the association between mortality and environmental variables has been (1) To describe the main characteristics of mortality, pollution and weather series with univariate ARIMA models. (2) To distinguish between sporadic sudden changes and step level changes increases in mortality. (3) To assess excess mortality associated with influenza. (4) To prewhiten all mortality, weather and pollution series to allow for influenza epidemics in the mortality-environmental modelling relationship. (5) To build up transfer function models to capture the variability of both the mortality and the environmental variables. (6) To develop a methodology for building a common model to describe a cause-effect time series relationship when multiple time series are available. This method is essentially a combination of several identification procedures for transfer function models. (7) To develop a computational system on SAS, to facilitate the analysis of time series. For this, an interactive system called the TODAY system, and a set of specific-purpose programs were created. The TODAY system was fundamental for building up univariate ARIMA time series models and for detecting sporadic and level increases in mortality. The specific-purpose programs, complement the statistics computed by SAS such as the IMPULSE program, which computes impulse and step response functions and the CORNER program which produces corner tables utilized to identify parsimonious transfer function models.; The transfer function models which related temperature and smoke with mortality, indicates that deaths are related to the cumulative effect of adverse temperature that was reached one to five days earlier. There is a contemporaneous impact of pollution on mortality. Pollution influences cardiovascular more than respiratory mortality. There is a general decrease in excess mortality due to pollution from 1959 to 1974. Rainfall, mean wind speed and relative humidity did not shown any association with mortality. (Abstract shortened by UMI.)