DEP - Dissertações de mestrado
Permanent URI for this collection
Browse
Browsing DEP - Dissertações de mestrado by Author "Bulhosa, Carolina"
Now showing 1 - 1 of 1
Results Per Page
Sort Options
- The impact of heatwaves on mortality in the Lisbon district – ICARO system revisitedPublication . Bulhosa, Carolina; Antunes, Marília; Nunes, BaltazarTemperature is an environmental factor that influences human comfort and health, so much that both extreme heat and cold increase mortality. Studying the effect that extreme heat has on mortality is of utmost importance, in order to try to predict and mitigate the consequences of global warming, with special attention to the most vulnerable and least adaptive population. In 1991, a heat health warning system that monitors possible increases in mortality due to extreme heat was created - the ICARO system. It was initially developed based on a time series statistical model using dynamic regression techniques and a dynamic threshold, which were calibrated for Lisbon data concerning the 1981 and 1991 heatwaves. The purpose of this work is to formulate a new kind of model to study the heat-mortality relation, aiming to optimise/update the ICARO system. Since the effect of extreme heat on mortality is not limited to the time period when it occurs but is delayed in time, using a model from the family of distributed lag non-linear models (DLNM) seems to be appropriate. A DLNM is based on a bidimensional function, called “cross-basis” function, which describes the shape of the relationship simultaneously along the space of the predictor - temperature -, and along its lag dimension. Therefore, this type of functions allows to explain an exposure-response effect, considering both the intensity and timing of a combination of several past exposures, up to a determined maximum lag. The model proposed here, was calibrated with data from the district of Lisbon from 1980 to 2017, restricted to the months between May and September. The total counts of daily deaths were explained in function of the daily maximum temperatures, through a cross-basis function allowing for a maximum lag of 10 days. The model also accounted for two time functions to control for seasonality and trend. The day of the week and annual average population entered the final model, as well. The results revealed that heat has a sustained effect up to 4 days, causing an overall increase in the relative risk of death for temperatures above 30 oC. However, temperatures below 15 oC during summer confer some protection. The predictive performance of the DLNM and the ICARO model were assessed and compared, through a cross-validation method. It revealed that both models have a good capacity to predict the highest peaks of mortality, but the DLNM tends to underestimate the magnitude of the lower ones. Overall, the DLNM obtained is considered a good model, since it seems to capture at least the main features of the studied relationship. There are some possible future developments for this theme, such as simpler modelling choices for the cross-basis function and accounting for some of the known risk factors for the heat-related mortality.