Disease outbreak data typically consist only of cases with times/locations. However, most outbreak data analysis methods require knowledge of the host spatial distribution. There is often much uncertainty in the host’s distribution because; data simply do not exist e.g. wildlife populations, or are not publicly available for privacy reasons e.g. farm locations in the U.S.A.

BioSS colleagues have developed methods that allow analysis of outbreak data even when the host distribution is unknown. To do this, they assume that the host’s spatial distribution is associated with known spatial covariates e.g. land use data, climate data. The actual nature of this association is informed by the outbreak data. The methods allow for risk assessment and prediction of both future spread and control. Their proposed approach, iPAR: inference for populations at risk. is generic and can in principle be applied to a range of infectious diseases. In practice, covariates and case data are aggregated to a grid of cells covering the region of interest, and analysis can be conducted at the scale most appropriate to the problem at hand.

Applying iPAR to data on the spread of African Swine Fever (ASF) in Estonian wild boar revealed good agreement between model predictions of future spread, based on early-stage outbreak data, and the held-out outbreak data from the later stages of the outbreak. Susceptibility estimates were higher for broadleaf woodland than other land uses.

iPAR can be applied in many other contexts. It has been used for similar models for outbreaks of ASF in wild boar in Italy, and it is being tested on Avian Influenza case reports in both wild birds and domestic poultry. Given knowledge of the host’s distribution is often uncertain for many diseases, there are likely to be many other suitable applications.

Risk map for spread of ASF in wild boar in Estonia. Grey patches are recorded as infected at the end of the early stage of the outbreak. Red colours indicate probability of infection four months beyond the end of the early stage, based on a model fitted to early-stage data. The black dots show patches that did in fact become infected within this time horizon.

This work is funded by EPIC, Scottish Government's Centre of Expertise in Animal Disease Outbreaks, in collaboration with Thibaud Porphyre from the Laboratoire de Biométrie et Biologie Évolutive (CNRS) in France. You can read the associated preprint here.