Spatial Forecasting of Disease Risk and Uncertainty

Abstract

Because maps typically represent the value of a single variable over 2-dimensional space, cartographers must simplify the display of multiscale complexity, temporal dynamics, and underlying uncertainty. A choropleth disease risk map based on data for polygonal regions might depict incidence (cases per 100,000 people) within each polygon for a year but ignore the uncertainty that results from finer-scale variation, generalization, misreporting, small numbers, and future unknowns. In response to such limitations, this paper reports on the bivariate mapping of data "quantity" and "quality" of Lyme disease forecasts for states of the United States. Historical state data for 1990-2000 are used in an autoregressive model to forecast 2001-2010 disease incidence and a probability index of confidence, each of which is then kriged to provide two spatial grids representing continuous values over the nation. A single bivariate map is produced from the combination of the incidence grid (using a blue-to-red hue spectrum), and a probabilistic confidence grid (used to control the saturation of the hue at each grid cell). The resultant maps are easily interpretable, and the approach may be applied to such problems as detecting unusual disease occurrences, visualizing past and future incidence, and assembling a consistent regional disease atlas showing patterns of forecasted risks in light of probabilistic confidence.

Geography, Spatial Data Analysis, and Geostatistics: An Overview. 地理学、空间数据分析及地统计学：综述

“…The table shows that 40% of geostatistical references found in “mainstream” geography journals refer to the use of kriging in relation to temperature, rainfall erosivity, various atmospheric and groundwater pollutants, soil properties, vegetation and species distribution patterns, and DEM creation. By contrast, kriging has been little used in human geography, an exception being De Cola's (2002) mapping of Lyme disease.…”

The Visualisation of Uncertainty for Spatially Referenced Census Data Using Hierarchical Tessellations

“…TRUST has the ability to use data of varying scales whereby tessellations can be created from the active display or the full extent in the ArcMap module of ArcGIS. Therefore, TRUST is appropriate for data exploration such as fine-scale cluster detection and for the identification of broad patterns, something that symbology uncertainty information cannot do (De Cola 2002).…”

Comparative Study of Association Rule Mining and MiSTIC in Extracting Spatio-temporal Disease Occurrences Patterns

“…[31][32][33] Further, health experts are interested in understanding the consequent risks and build forecast models, either in the short or midterm as part of the disease management strategies. Currently, the latter are based on spatial, temporal and spatio-temporal co-occurrence of disease incidences, but in most of these cases the approaches consist of either a characteristic decoupling of spatial and temporal interdependencies [25][26][27][28][29][30], or an inherent dependency on various etiological and geographical factors. The challenge for data mining in health or any other monitoring application is the factor that needs to be monitored -either the outcome or the causative factors that lead to this outcome.…”