Predicting the Risk of Lyme Disease: Habitat Suitability for Ixodes scapularis in the North Central United States

Marta Guerra, Edward Walker, Carl Jones, Susan Paskewitz, M. Roberto Cortinas, Ashley Stancil, Louisa Beck, Matthew Bobo, Uriel Kitron

Disclosures

Emerging Infectious Diseases. 2002;8(3) 

In This Article

Methods

In Wisconsin, a statewide survey of parks and forests was conducted to determine the presence of I. scapularis. Sites were selected to represent each region in the state, with 28 of 59 states parks and forests selected. In Michigan, three sites were selected in Menominee County, where I. scapularis had previously been identified[7]. In Illinois, paired positive and negative sites were sampled in Ogle and Rock Island counties, and additional sites were sampled along the Illinois River. Data are presented separately for each collection site.

Tick collection was conducted at a total of 138 sites in July and September-October 1996, June 1997, and May-June 1998. The most comprehensive trips were made June 14-26, 1997, and May 27 through June 3, 1998, in the southern part of the study region. In several natural areas, more than one site was dragged, and results for each site were considered separately.

Questing I. scapularis ticks were collected in two ways: 1) by dragging a 1-m2 white flannel cloth through vegetation for a total of at least two hours at each site (timed dragging), or 2) by dragging 1000 m on a grid (distance dragging). Timed dragging was conducted by teams of 4 or 5 persons, with each person dragging for 30 minutes. Distance dragging was also conducted by teams of 4 or 5 persons, which required an estimated 2 to 2.5 hours per grid. Thus, each site was dragged for a total of 2 or 2.5 hours per visit. All calculations of tick numbers are per 1 drag-hour.

Nymphs and adults were maintained alive in plastic vials with moistened cotton balls on ice for B. burgdorferi culture. Larvae were placed in vials containing 70% ethanol for later identification.

Small mammals were trapped overnight during July and October 1996, June 1997, and June 1998 at 13 selected sites in Wisconsin, and at all the Michigan and Illinois sites. Sherman live traps (H.B. Sherman Traps, Inc, Tallahassee, FL) were placed approximately 10 meters apart and baited with bread and peanut butter. Approximately 35 to 50 traps were placed per site, and 0 to 15 mice and 0 to 7 chipmunks were trapped at each site. White-footed mice and chipmunks were anesthetized with the inhalant anesthetic methoxyflourane (Shering-Plough, Inc., Madison, NJ), examined for ticks, and ear-tagged, and their sex and weight were recorded (LACAC animal use protocol # 99099). Ticks were removed and placed in vials containing 70% alcohol for identification.

For each site, the average number of each stage of the deer tick was calculated per hour of dragging. The number of ticks per dragging hour is based on an average of all drags. There was no situation where all or most ticks were found on one drag. The average number of larvae and nymphs was determined per small mammal captured. These data were not pooled with the dragging data because animals were not trapped at all sites.

A site was classified as negative (0) if I. scapularis was never found on vegetation or small mammal hosts. There was no case where ticks were found only on small mammals but not on drags. A site was rated 1 if only one stage of the tick was found, regardless of the quantity. A rating of 2 was given if all stages of the tick were found at low density (<10 larvae, <4 nymphs, <2 adults), and a rating of 3 indicated all stages were found at higher density.

We considered several types of classification, including calculating each stage separately and each collection trip separately. Although repeat visits increase the chance that a site will be classified as positive for the presence of ticks, there were no sites where more than one stage was found in only one visit. The finding of only one stage, however, may indicate accidental introduction without establishment. We selected a very conservative and coarse classification to account for the limitations of such an extensive field survey and to allow for differences in weather conditions, time of day, and other variables.

Soil Data. After removing the layer of leaf litter, soil samples were collected at each site from the uppermost 6 inches of topsoil. Data on predominant vegetation, leaf litter thickness, slope, and compass direction were also recorded at each site. Particle size analysis was performed on 10 gm samples of soil[37]. The pH and the percentages of sand, silt, and clay were measured for each sample, and the soil texture class was determined from a combination of these percentages. The percentages and classes were compared with site positivity using Spearman rank correlation.

Forest Moisture Index. The classification of forest type was derived from the predominant trees at each site. The number of mature trees (>4 inches in diameter) were counted within a 50-m2 grid at each site and identified according to species[38]. The most common species were used to classify the forests via a moisture index[38]. The sites were divided into five categories: dry, dry/mesic, mesic, wet/mesic, and wet.

Data Sources. Geographic coordinates of sites were determined by using a Trimble Geoexplorer (Trimble Navigation, Ltd., Sunnyvale, CA) global positioning system (GPS) and exported by using the Trimble Pathfinder software into ARC/INFO and ArcView GIS (ESRI, Redlands, CA). The generated georeferenced database was overlaid on digitized state coverages of environmental data. Land cover and elevation data for Wisconsin were obtained from WISCLAND/GAP (University of Wisconsin and Wisconsin Department of Natural Resources, Madison, WI) at a scale of 1:40,000. WISCLAND/Upper Midwest GAP analysis created land cover classifications based on Landsat Thematic Mapper (TM) data and stratification of the satellite imagery with a hierarchic classification system into wetlands, urban areas, and upland areas. For Illinois, land cover, elevation, and quaternary geologic data were obtained from the Illinois GIS (Department of Natural Resources, 1996, Springfield, IL) at a scale of 1:500,000. Bedrock geology data were obtained from the Digital Geologic Map and Mineral Deposits of Minnesota, Wisconsin, and Michigan (U.S. Geological Survey, Reston, VA) at a scale of 1:1,000,000 for Wisconsin and Michigan, and from the Illinois GIS at a scale of 1:500,000. Soil data, including order, texture, drainage, and quaternary geology, were obtained from STATSGO (U.S. Department of Agriculture, Washington, DC) with a resolution of 2.5 km2.

Environmental Variables. Land cover data were grouped into five ordinal categories: agriculture, grasslands, coniferous forest (in which ≥75% of trees maintain leaves all year), mixed forest (neither deciduous nor coniferous species make up >75% of land cover), and deciduous forest (at least 75% of trees shed foliage simultaneously in response to seasonal change).

Bedrock geology was classified as Precambrian, which consists of volcanic and metamorphic rocks, and sedimentary deposits from the Silurian, Ordovician, and Devonian eras[40]. Quaternary geology information was obtained from the USDA Forest Service North Central Research Station (General Technical Report NC-178). Categories were classified as outwash plains and pitted outwash, lake plain, till plain, ground moraine, loess, and plateau.

Soil orders are defined by amount of organic matter present, pH, and the type of vegetation growing on the soil[40]. In Wisconsin and northern Illinois, 8 of 12 soil orders are represented: mollisols (present under prairie), alfisols (deciduous forests), spodosols (coniferous forests), entisols and inceptisols (both of which are associated with poorly developed soils),histosols (peat and muck), and vertisols and paleosols (which represented <1% of the area). These orders were classified into ordinal categories based on increased fertility and decreased acidity: 0 = histosol and spodosol, 1 = entisol, 2 = inceptisol, 3 = mollisol, and 4 = alfisol.

Soil texture[40] was divided into seven groups in order of increasing particle size, ranging from clay (<2 mm) through silt (2 to 50 mm) to sand (0.05 to 2.0 mm). Drainage was divided accordingly into seven categories (STATSGO, Washington, DC), from very poorly drained to well drained. Excessively drained soils were ranked as 0 since they are too dry to support a biotic environment[40].

For each site, yearly and seasonal rainfall averages and average snowfall per year were obtained from the weather station (NOAA) nearest each site. Elevation ranged from 495 m in northern Wisconsin to 197 m in western Illinois. Precipitation, elevation, and remote sensing indices were treated as interval-level data.

Statistical Analysis. All analyses were performed by using SPSS software (SPSS, Chicago, IL). Soil texture classifications of samples from the sites were compared with those listed in STATSGO, the soils database (STATSGO, Washington, DC, and Spearman rank correlation was used to assess correlations between field data and data from the GIS. Univariate analysis was initially performed by using chi square contingency tables to determine significant associations between site positivity and environmental variables coded as previously described. Discriminant analysis was performed by using only the significant (p<0.25) environmental variables from the univariate analysis[41]. A linear discriminant function was obtained from the combination of variables that best characterized the differences between the groups. A stepwise approach was used to enter variables one at a time until the discriminating power between tick abundance categories ceased to improve. Analyses were performed by grouping the outcome variables into positive or negative sites and into the four abundance categories described previously.

As mentioned, since a site classified as category 1 (finding only one stage of the tick) could result from introduction into an unsuitable habitat, categories 0 and 1 were combined for additional analysis. Only 112 sites were used in the analysis, with no more than three sites included per natural area where multiple sites were sampled. The resulting classification functions were then used to predict tick abundance categories and assess how well the functions discriminated. Separate discriminant analyses were performed by using the seven indices obtained from the remote sensing data at three spatial scales and the precipitation data.

Logistic regression analysis was performed by using the primary environmental factors as independent variables and the positive and negative sites as outcome variables. Forest moisture index was excluded from the model because this variable was not available as digitized geographic coverage.

To develop a risk map for Lyme disease in the area studied, a grid was created encompassing the states of Wisconsin and Illinois with a resolution of 2.5 km2 per cell. The grid was overlaid with the selected coverages by using ARC/INFO and ArcView GIS(ESRI, Redlands, CA), and data values corresponding to each layer were assigned to each cell. The Summarize Zones procedure from the ArcView Analysis Menu was used to calculate summary attributes for features by using a grid scheme that divided the entire study area into 2.5-km2 cells. Each cell was assigned a value for each layer included in the logistic regression based on the most common category. The logistic equation was then used to generate the probability of the presence of I. scapularis within each 2.5-km2 cell of the grid map. The map was generated with probabilities divided into quartiles and deciles.

Comments

3090D553-9492-4563-8681-AD288FA52ACE
Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.

processing....