Matthew (Matt) Ferrari

I use mathematical and statistical tools to understand patterns of disease incidence, and the effects of heterogeneity, in time and space.

Measles dynamics in developing countries

Measles still kills hundreds of thousands of children each year in developing countries. Attempts to eradicate the disease through mass vaccination are hampered by both logistical and epidemiological challenges; for instance, high birth rates can make it difficult to maintain the necessary 95% vaccine coverage.

In collaboration with Medecins Sans Frontiers we are investigating local and regional dynamics of annual measles epidemics in West African countries (Niger, Tchad, Democratic Republic of Congo), in order to recommend vaccination strategies to minimize mortality and morbidity due to measles. We are using time series analysis and epidemic models to investigate:

• The nature of the strong annual seasonality in incidence at the regional scale
• Local variation in the scale of measles outbreaks

Dynamics of directly transmitted pathogens on host networks

I use simulation and analytical techniques to investigate how the spread of disease in social networks of hosts is affected by heterogeneities in contacts and local restrictions on transmission. These have important implications for the scaling of transmission across networks of different size and geometries — and can even lead to structural evolution of the network itself (as hosts are removed by mortality or acquired immunity).

Vector behavior and spatial transmission

Disease vectors can transmit pathogens while foraging. Given a heterogeneous host population, choice of foraging locations by vectors will lead to differential host exposure to pathogens. Bacterial wilt — a pathogen of gourd species — is transmitted by a beetle, Acaymma vittata. Using field and lab experiments we are investigating how vectors respond to plant quality, and the implications for epidemic spread and pathogen mediated host selection.

Statistical methods for estimating transmission rates

Disease incidence data are often gathered at spatial and temporal scales that are coarse relative to scales considered by quantitative epidemiological models of host-pathogen systems (e.g. case counts are generally reported over discrete time intervals, while many classic epidemic models employ differential calculus, which makes predictions in continuous time). Furthermore, observed data often suffer from incomplete reporting, imperfect diagnosis, measurement error and other biases. One of the great challenges in quantitative epidemiology is to develop statistical models that provide a coherent link between theory and data. I am developing:

• Discrete time, stochastic models to develop statistical methods to estimate transmission rates for incidence data
• Computational methods (e.g. Markov chain Monte Carlo) to account for the uncertainty due to imperfect measurement