Sampling Effort

  Monitoring Goal

  Test Characteristics

Current Settings


Entered Data

  Upload Data

  Model Settings

Chronic Wasting Disease Weighted Surveillance

This application provides tools for planning weighted surveillance sampling for chronic wasting disease (CWD), and estimating underlying prevelance after sampling has been completed. Models are available for white-tailed deer (Odocoileus virginianus), mule deer (Odocoileus hemionus), and elk (Cervus canadensis) populations, and an additional model allows combining samples from elk and mule deer in a single analysis.

Weighted surveillance is based on the simple principle that within a population there exists heterogeneity among individuals with regard to disease risk. To maximize efficiency and potentially increase the likelihood of detecting new disease foci, weighted surveillance programs exploit this heterogeneity by focusing disease detection efforts in groups most at risk. The underlying methods were orginially developed by Walsh et al. (2010) using a Frequentist statistical approach. These results were then refined and recast into a Bayesian statistical framework by Heisey et al. (2014). It is imporant to note that the use of weighted surveillance techniques requires that prior information is available to estimate heterogeneity in individual risk. A general discussion of CWD surveillance is given in Walsh et al. (2012).

The first tool called Design, is used for planning weighted surveillance activities. The user specifies how much confidence they need, and what minimum prevalence they would like to detect in the reference class (e.g., I would like a 95% confidence of detecting at least one case if the prevalence is at least 1%). This information then provides the total number of points required to meet the specified confidence for the chosen minimum prevalence. The user can then select from the potential sources of surveillance samples, and vary the number of samples arising from each source. This provides a means of setting sampling objectives for each source to ensure the requisite number of points is reached. This tool can also be used to evaluate in real-time how close a program is to achieving its goal given the number of samples collected from each source to date.

The second tool called Estimation is for use after sampling for CWD detection has occurred and no positive cases were found. It provides the means to estimate the potential underlying prevalence rate of CWD given the number of negative samples collected during surveillance. Of particular interest is the upper bound of the credible interval, which means there is, for example, a 95% probability that true prevalence is at or below the reported level given your sampling effort and lack of cases detected.

It's important to note that these tools are entirely independent - you can use either one however you wish but there's no need to use both, and changes made in one tool won't have any affect in the other.

The techniques and theory underpinning them used in this application are described in Heisey et al. (2014). This application takes advantage of the R statistical software for estimation. Heterogeneity of risk classes are based on the chronic wasting disease information described in Jennelle et al. (2018) and Walsh et al. (2012). The former provides a description of the Wisconsin estimates and a recent case study of the application of weighted surveillance to CWD detection. The elk and mule deer weights are based on results reported in Walsh et al. (2012).


The toolbar at the top-right of your screen has links to help pages, bug tracking, contact information and more.


This tool was developed in collaboration with Dan Walsh, USGS and is hosted and maintained by SpeedGoat.