Scale-dependent approaches to modeling spatial epidemiology of chronic wasting disease.
This e-book is the product of a second workshop that was funded and promoted by the United States Geological Survey to enhance cooperation between states for the management of chronic wasting disease (CWD). The first workshop addressed issues surrounding the statistical design and collection of surveillance data for CWD. The second workshop, from which this document arose, followed logically from the first workshop and focused on appropriate methods for analysis, interpretation, and use of CWD surveillance and related epidemiology data. Consequently, the emphasis of this e-book is on modeling approaches to describe and gain insight of the spatial epidemiology of CWD. We designed this e-book for wildlife managers and biologists who are responsible for the surveillance of CWD in their state or agency. We chose spatial methods that are popular or common in the spatial epidemiology literature and evaluated them for their relevance to modeling CWD. Our opinion of the usefulness and relevance of each method was based on the type of field data commonly collected as part of CWD surveillance programs and what we know about CWD biology, ecology, and epidemiology. Specifically, we expected the field data to consist primarily of the infection status of a harvested or culled sample along with its date of collection (not date of infection), location, and demographic status. We evaluated methods in light of the fact that CWD does not appear to spread rapidly through wild populations, relative to more highly contagious viruses, and can be spread directly from animal to animal or indirectly through environmental contamination. We discovered that many of the wellpublished methods were developed for fast-spreading human diseases, such as influenza and measles. While these methods are applicable to fast spreading wildlife diseases, such as foot-and-mouth disease or West Nile virus, many are not likely to work well for CWD. Only limited data exist to evaluate geographic and spatial spread because many locations where we find CWD tend to be locations where samples have just been taken or sample sizes have just become large enough to have a high probability of detecting a low prevalence. Consequently, methods that work well to describe or predict the spread of foot-and-mouth disease throughout England, which occurred within a year, do not work well for describing or predicting CWD spread. We did not exclude methods that we regarded as inappropriate; rather, we included methods that are commonly used for disease epidemiology and then discussed their applicability for modeling the spatial epidemiology of CWD. We hope including inappropriate methods with an explanation of why they are ill-suited for CWD will make it easier to drop them from consideration and explain to others why they were not recommended for spatial modeling of CWD. We organized the three chapters by scale and extent for which each method was developed or best suited. The first chapter covers methods appropriate to multi-jurisdictional or multi-state modeling, which we call “regional” scale. The second chapter covers methods appropriate for within state areas such as wildlife management units or metapopulations, which we call “landscape” scale. The third chapter covers methods appropriate for population or individual-based modeling, which we call “fine” scale. We know this rubric is somewhat artificial because many methods work at multiple scales. We hope, however, that this structure addresses some of the challenges faced by managers that work at local, regional, state, and national scales. Further, the resolution of empirical data often changes with spatial scale, which affects the utility of different modeling approaches. For example, individual-based models work best at modeling spread within populations, while risk analysis is most useful for summarizing data over larger scales such as a region. Because some methods are applicable at several scales, however, we included a graphic at the beginning of each method that indicates the range of scales for which it applies. For example, the graphic to the right indicates that the method is most applicable for regional-scale modeling. There is also a question of resolution as well as scale and extent for each method. CWD surveillance data have been collected over large areas, such as a wildlife management unit or state, but the resolution of the data may be fine scale with GPS locations for many samples. For each method, we described the required resolution of the data and describe the type of data required, as well as what questions the method could answer and how useful the method is, given typical CWD data. For each scale, we presented a focal approach that would be useful for understanding the spatial pattern and epidemiology of CWD, as well as being a useful tool for CWD management. The focal approaches include risk analysis and micromaps for the regional scale, cluster analysis for the landscape scale, and individual based modeling for the fine scale of within population. For each of these methods, we used simulated data and walked through the method step by step to fully illustrate the “how to”, with specifics about what is input and output, as well as what questions the method addresses. We also provided a summary table to, at a glance, describe the scale, questions that can be addressed, and general data required for each method described in this e-book. We hope that this review will be helpful to biologists and managers by increasing the utility of their surveillance data, and ultimately be useful for increasing our understanding of CWD and allowing wildlife biologists and managers to move beyond retroactive fire-fighting to proactive preventative action.