Statistical Support for Lyme Disease Landscape Modeling and Decision Tool Development
|OTIE performed statistical and GIS analysis for determining the generalized applicability of a model to predict Lyme disease incidence in Maryland and Pennsylvania using landscape variables such as population and land use (e.g., forest cover). We investigated the performance of a combined Maryland/Pennsylvania logistic regression using predictor-variables such as a blocking variable (state), additional explanatory variables (number of jobs in farming and/or forestry occupations from SF4 US Census Bureau), climatic variables, classifying percent forest from a continuous variable to a binary variable (e.g., >35% and < 65% forest), and spatial variables (latitude and longitude).
The final logistic regression model was ported to the Internet as an Exploratory Data Analysis tool which allows the user to interact with the data and see how model performance changes during What-If analysis. The user has the ability to change the threshold by which Lyme disease is considered a public health risk (e.g., 1/10,000 or 1/25,000) and the cut-off by which the predicted value is considered an excedence (e.g., 30%, 50%). OTIE programmed the Lyme Disease Interactive Viewer (LDIV) interface to include an interactive information view, accordion organizer, map transparency sliders, integration with the USEPA website, and interactive map controls.
OTIE employed efficient states programming which allowed for the addition of key LDIV components such as an interactive error matrix, map modeling sliders, and layer selections. OTIE included ESRI ArcGIS map queries which dynamically poll the map service as statistical modeling parameters are changed to provide a rich, interactive user experience.
This innovative cutting edge on-line decision support tool is being featured in a USEPA exhibit at the Smithsonian Museum.
|Client Name: USEPA NERL
Contract Value: $140K