My EPIC research involves using methods from statistics in combination with mathematical models to learn and predict how epidemiological systems behave to inform the design of better disease control. With colleagues I am developing statistical methods that enable such predictions from reported disease cases, even when the at-risk population e.g. of farms or wildlife, is unknown. This will allow analysis of a much wider range of disease threats than is possible with currently available methods, including African Swine Fever in wild boar.
Disease systems are made up of multiple elements e.g. individual animals on different farms and associated wildlife populations, that interact dynamically through a range of processes such as movement of livestock between farms, transmission of disease between individuals and progression of disease within individuals. The potentially complex nature of these interactions can result in less favourable than anticipated outcomes from actions aimed to control disease and, in some cases, totally unintended consequences e.g. culling of badgers may increase TB risks to cattle.
Mathematical modelling provides a framework which can be used to integrate key aspects of disease systems in a logically consistent way. Modern computational methods now enable exploration of the behaviour of potentially very complex models under a wide range of different scenarios which, for example, can be used to assess the impact of proposed disease control strategies. Unfortunately, it is often too costly or impractical to measure the values of critical parameters needed for such models, for example representing disease transmission or duration of infectivity. In addition, the inclusion, or nature, of particular processes in the model may be disputed. However, modern statistical methods enable estimates of an increasing range of model parameters to be obtained from incomplete and noisy field data e.g. observations of disease outbreaks. With scientists in EPIC and beyond I am developing statistical computing methods to enable information relevant to increasingly complex models to be reliably extracted from a widening variety of data. In addition, these tools are being extended to enable assessment of model structure to be informed by the available data e.g. to determine which processes should be included and how.
Review: Models of natural pest control: towards predictions across agricultural landscapes. Nikolaos Alexandridis Glenn Marion*, Rebecca Chaplin-Kramer, Matteo Dainese, Johan Ekroos, Heather Grab, Mattias Jonsson, Daniel S.Karp, Carsten Meyer, Megan E.O'Rourke, Mikael Pontarp, Katja Poveda, Ralf Seppelt, Henrik G. Smith, Emily A. Martin and Yann Clough
Using Combined Diagnostic Test Results to Hindcast Trends of Infection from Cross-Sectional Data. G Rydevik, GT Innocent, G Marion, RS Davidson, PCL White, C Billinis, P Barrow, PPC Mertens, D Gavier-Widen, MR Hutchings
Website design by Innovation Digital Limited