Visualization and exploratory analysis of epidemiologic data using a novel space time information system

Abstract

BACKGROUND: Recent years have seen an expansion in the use of Geographic Information Systems (GIS) in environmental health research. In this field GIS can be used to detect disease clustering, to analyze access to hospital emergency care, to predict environmental outbreaks, and to estimate exposure to toxic compounds. Despite these advances the inability of GIS to properly handle temporal information is increasingly recognised as a significant constraint. The effective representation and visualization of both spatial and temporal dimensions therefore is expected to significantly enhance our ability to undertake environmental health research using time-referenced geospatial data. Especially for diseases with long latency periods (such as cancer) the ability to represent, quantify and model individual exposure through time is a critical component of risk estimation. In response to this need a STIS - a Space Time Information System has been developed to visualize and analyze objects simultaneously through space and time. RESULTS: In this paper we present a "first use" of a STIS in a case-control study of the relationship between arsenic exposure and bladder cancer in south eastern Michigan. Individual arsenic exposure is reconstructed by incorporating spatiotemporal data including residential mobility and drinking water habits. The unique contribution of the STIS is its ability to visualize and analyze residential histories over different temporal scales. Participant information is viewed and statistically analyzed using dynamic views in which values of an attribute change through time. These views include tables, graphs (such as histograms and scatterplots), and maps. In addition, these views can be linked and synchronized for complex data exploration using cartographic brushing, statistical brushing, and animation. CONCLUSION: The STIS provides new and powerful ways to visualize and analyze how individual exposure and associated environmental variables change through time. We expect to see innovative space-time methods being utilized in future environmental health research now that the successful "first use" of a STIS in exposure reconstruction has been accomplished.

Figure 1

Change in water supply systems over 50 years (1935, 1965, 1995) Over the years many towns in Oakland County and Genesee County begin to purchase surface water (from Detroit).

Figure 2

Participant movement over 20 years Cases (circles) and controls (squares) continue to move in, out, and around the study area. In 1960 there were two cases and one control. By 1982 four more cases and two more controls moved into the study area and in 2001 the same number of cases and controls remain in the area however one case and one control have moved addresses.

Figure 3

Box plot of arsenic exposure in 1988 for cases (left) and controls (right) The median is the black line that bisects the box. The upper and lower quartiles, the medians of the upper and lower halves of the data, are the edges of the black box. The "whiskers" on the box, the bars at the top and bottom, are 1.5X the interquartile range.

Figure 4

Time graph of arsenic exposure in cases (top) and controls (bottom) Notice the increase in arsenic for both sets after 1951. The increase in arsenic is much larger for controls and remains high for at least two individuals.

Figure 5

Arsenic in drinking water (2003/2004) Each point represents an arsenic value taken from the kitchen tap at the present residence of each participant.

Figure 6

Local Moran analysis at two spatial scales Local Moran analysis with five nearest neighbours is on the left, and with ten nearest neighbours is on the right. Notice the appearance of the high-high cluster to the north, and the increase in size of the low-low cluster to the west as the size of the local neighbourhood is increased