Conference is now cancelled
Using animal activity and abundance to estimate ‘rattiness’ and predict Leptospira infection risk: A multivariate geostatistical framework for combining multiple indices of abundance and activity
Max Eyre (CHICAS, Lancaster University)
The abundance and activity of animal populations are important drivers of the risk of spillover of zoonotic infections into other animal and human populations. Due to difficulties in measuring abundance and activity in many animal species, multiple imperfect indices are often used. However, current approaches for combining multiple indices of abundance and activity do not exploit the inferential benefits that might accrue from the joint spatial modelling of the different indices.
In this study, we have developed a class of multivariate generalized linear geostatistical models for modelling the cross-correlation in space among multiple imperfect indices of animal abundance and activity. We illustrate the development and application of the novel methodology in the context of a case study on Rattus norvegicus, a reservoir for Leptospira in vulnerable urban communities in Salvador, Brazil. More specifically, we consider three indices of R. norvegicus abundance: rat signs, live traps and tracking boards. We use the three outcomes in order to draw predictive inferences on a spatially continuous latent process which relates to R. norvegicus abundance and activity and which we refer to as "rattiness". We then use this framework to fit our model to human Leptospira infection.
We show how this new methodology can be used 1) to explore the association between the latent process, "rattiness" and environmental factors, 2) to test for residual spatial correlation, 3) to evaluate the relative importance of each of the three outcomes in estimating "rattiness" and 4) to identify "rattiness" hotspots. We then show how ‘rattiness’ can be used to predict human Leptospira infection risk.
This is a practical new methodology for pooling information for animal species which are difficult to detect and will be useful in predicting the risk of zoonotic spillover and monitoring reservoir populations.