Spatial epidemiology

Question 1

GIS - definition, types of data (2), tools (3)

Answer 1

A GIS provides an integrated set of tools that allow input, storage, manipulation, analysis and visual representation of spatial data.Spatial data: measurements taken at a specific location or within specific regions (e.g. census tract, health district).

Deal with different types of data:

1. Vector (series of pairs of coordinates) including points, lines, polygons; and
2. Raster (arrays of pixels - grids/cells).

Tools:

1. Data manipulation tools - buffers
2. Spatial analytical tools - interpolation, measuring distance, i.e. characterizing spatial properties of the layer (derived from geography)
3. Spatial statistical tools - assessing autocorrelation/identifying clusters

Question 2

GIS - applications (3)

Answer 2

1. Disease mapping (depict spatial variation in risk e.g. disease atlas)
2. Ecological analysis (distribution of disease in relation to explanatory covariates e.g. etiological questions)
3. Cluster detection (assess whether clusters exist and where they are located e.g. surveillance)

Question 3

Mapping disease - methods (2), representation (3)

Answer 3

Methods:

1. Standardized rates - e.g. estimates of incidence or prevalence
2. Significant deviations from the area average e.g. observed vs expected counts

Representation:

1. Dot density maps
2. Proportional circles
3. Cloropleth (grey-scale)

Smoothing

Question 4

Cluster detection - methods (2)

Answer 4

Clustering - test null hypothesis that cases are randomly distributed

1. Global statistics: indicate whether clustering (spatial autocorrelation) is somewhere in a study area (e.g. Moran’s I - compares the differences between neighboring pixels and the mean to provide a measure of local homogeneity. The value range is between +1 and -1, where +1 = strong positive spatial autocorrelation (clustering), 0 = spatially uncorrelated data (random dispersion), and -1 = strong negative spatial autocorrelation (perfect dispersion - think chess board)
2. Local statistics: identify location of clusters (e.g. Scan statistics - superimposes a number of circular or elliptical windows of varying sizes to determine the group of contiguous areas with the most significant risk)