Daniela De Angelis
Statistical methods in epidemic modelling
This research programme concerns the development and application of statistical methodology to estimate characteristics and evolution of epidemics. The work, carried out through the link with the Health Protection Agency, is motivated by the need to provide evidence-based input to public health policies on infectious diseases in England and Wales. This is required to: plan interventions to reduce disease transmission; plan health care resources; and evaluate past interventions. Examples of policy areas include the Department of Health (DH) Strategy for Sexual Health; the DH Action Plan for Hepatitis C, and the DH Pandemic Influenza Plan.
For any epidemic, direct data on disease incidence and prevalence as well as their likely future evolution are typically not available. The problem is then to infer these unobserved characteristics from fragmented data from a variety of sources (surveillance systems, ad hoc studies, observational studies), possibly available at diverse levels of aggregation (individuals, local authorities, geographical regions), that are typically incomplete and often affected by biases. In these circumstances, standard statistical approaches are rarely appropriate.
The aim of this programme is to develop and apply statistical methods to characterise specific epidemics fully and correctly, exploiting the complex body of available information on different aspects of the disease. The driving motivation is that of providing sound statistical approaches to specific problems, while developing conceptual inferential frameworks that are more generally applicable. Key methodological elements of this work are:
- an emphasis on the need to combine information from a variety of sources and/or at diverse level of aggregation;
- the importance of carrying out critical appraisal of resulting models to explore the role of each piece of information in the estimation process, and to identify and resolve conflicts between data sources;
- the need for a comprehensive modelling of the phenomenon of interest in order to correct for potential biases.
Programme Members: Daniela De Angelis, Paul Birrell, Anne Presanis, Teresa Prevost and Yang Xia
For more details about this programme please click here.