Medea

Acción Estratégica Medea

Spatio-temporal modelling

Background

In the MEDEA I project, the geographical distribution of mortality was studied for a group of causes in various Spanish cities. As a consequence of the findings of that project, we now know in detail the geographical variation of mortality risks in these cities, and the relationships between these risks and other factors of social or environmental origin have been determined.

In MEDEA I mortality risks were considered to be quantities that did not vary with time and they were estimated on the basis of this assumption. This hypothesis is valid when the study period is short since there is consequently less opportunity for the risks to vary. However, the present MEDEA project proposed to extend the study period to cover from 1996 to 2007, i.e. 4 years more, and hence the hypothesis of risks remaining static over the study period is more difficult to uphold in the new project.

Objectives

  • Visualisation, in dynamic form, of the temporal evolution of mortality risks over the study period in the cities considered by MEDEA.
  • Adaptation of the methodology employed in the setting of the MEDEA project.
  • Writing of a protocol for the spatio-temporal modelling of risks in the MEDEA project setting.
  • Coordination of the work of the various nodes involved in order to ensure compatibility of their results.

Methodology

The methodology used will be that described in Martínez Beneito et al. (2008), which considers jointly the spatial and temporal dependence of the risks to be estimated. Moreover, these two components will be weighted as functions of the dependence of the two types exhibited by the data. Finally, the size of some cities considered in the study will mean that adaptation of the methodology developed previously in the MEDEA project will represent a computational challenge which will have to be faced.

References

Martínez-Beneito MA, López Quílez A, Botella Rocamora P. An autoregressive approach to spatio-temporal disease mapping. Statistics in Medicine. 2008, 27: 2874-2889