Lyme Disease remains the most prevalent vector borne disease in North America. The increasing human incidence of Lyme disease in the United States is attributed to the increase in the geographic range of its vector tick Ixodes scapularis. The density of nymphal ticks infected with Borrelia burgdorferi, the bacterial agent of human Lyme disease provides the most accurate assessment of human risk for both peri-domestic and occupational/recreational settings. Unfortunately collection of such data is prohibitively costly. Inconsistencies in diagnoses and reporting of human cases of Lyme disease result in unreliable estimates of risk, overestimating in areas of misdiagnoses and underestimating in areas of poor access to health facilities. In Wisconsin there is poor spatial correlation (<50 %) between past human case data and the location of future human cases of Lyme disease.
I am currently using Multi-layer Feed Forward artificial neural network capability in Veritas/Hampson-Russell ISMap software, to combine data layers for Wisconsin with high spatial correlation as measured by Moran’s I : remotely sensed Enhanced Vegetation Index (EVI) images from NASA’s newly available Moderate Resolution Imaging Spectroradiometer (MODIS) as well as atmospheric products for key dates, recent and archival canine sero-prevalence to Borrelia burgdorferi, and a previous model of tick habitat suitability we had developed for the upper mid-west. Co-kriging of past human case data with the combined habitat layers increased spatial correlation between 1998 and year 2000 human case data to 91%. Low co-kriging error terms suggested that 2001-2003 human case data can be used to identify spreading areas of infection and human risk of Lyme disease for the years 2004-2006. For that part of the Upper Mid-West the spread of infected ticks is estimated at about 20 kilometers per year in a predominately southeasterly direction. Additional MODIS images are needed to increase confidence intervals.
