The application of multidimensional scaling methods to epidemiological data.

Abstract

This paper illustrates the use of multidimensional scaling methods (MDS) to examine space-time patterns in epidemic data. The paper begins by outlining the principles of MDS. The model is then formally specified and illustrated by application to two data sets. The first is partly a tutorial example. It uses monthly reported measles morbidity data for the 31-year period from January 1960 to December 1990, collected for the 50 states of the USA, plus New York City and the District of Columbia. These data are used to explore the various ways in which MDS may be used to identify changing spatial patterns in geographically-coded data. In addition to their tutorial use, the data are also employed to search for any substantive changes in the geographical structure of measles epidemics in the USA that may have followed the introduction of mass vaccination in 1965. New England appears to have developed an epidemic profile distinct from the rest of the USA, and there is tentative evidence of an urban-rural split in epidemic characteristics. The second data set takes annual reported measles mortality data for New Zealand and the states of Australia from 1860 to 1949. MDS is used to show how the spatial relationships among these geographical units have changed over time in response to changes in the sizes of local susceptible populations.