Bayesian modelling as a tool for risk assessment

Introducing Docent Jukka Ranta, Ph.D., Senior Researcher in the Risk Assessment Research Unit at Evira.

The modelling of food safety risks is a relatively recent and minor research area in Finland. Senior Researcher Jukka Ranta of the Risk Assessment Unit has been able to use new approaches based on Bayesian methods in his research into food risk assessment by applying them to evidence-based risk calculation that combines several sets of data.

The potential of computational methods soared in the 1990s, opening up new approaches based on Bayesian methods to interesting application problems in epidemiology and biosciences, for example. This has interested Ranta since his master's thesis on stochastic epidemic models (1994), and the risk assessment of food is a natural continuation of this.

The Risk Assessment Research Unit was created, in part, to respond to the need for the production of processed data to support national decision-making.

“The combination of Bayesian methods and applications and the discovery of new information at their interface is interesting. There is always something new to find,” says Ranta. According to Ranta, it is inspiring and challenging to separate the general structure from the details and to shape it into a feasible mathematical form, whose dependencies can be estimated. Ranta is motivated by taking the modelling ideas into new applications to support decision-making, sometimes on a very large scale, both at a national and international level. The national results of different countries are compared and evaluated at institutions such as EFSA. Developing his own modelling skills and those of his colleagues in the research unit is also important for him to ensure the completeness and continuity of the research.

Sometimes, the research career of a mathematician and risk assessor can be a bit of a lonely wrestling match with theories. “Advice for the most difficult modelling problems can only be provided by other statistics experts, who are not readily available – if at all,” remarks Ranta. Ranta’s current key tasks include the completion of the ongoing projects and the formulation of new problems to be modelled.

Looking to the future, Ranta believes that evolving areas include the analysis and modelling of data and data pools, the synthesis of different data sets, and the refinement of data in these data sets into knowledge.

“As the amount of data grows and its interpretation and structures become more complex, it becomes necessary to analyse it in more detail and also to evaluate the uncertainty of the findings,” says Ranta.

Bayesian modelling is used in a growing number of areas. According to Ranta, there is still untapped potential in it that has not been fully understood. It is, therefore, wise to identify favourable applications. Not everything can be solved with models.

“This requires the purposeful development of resources for method research, networking, the training of researchers, and a long-term strategy which focuses on the quality of scientific research and monitors development”, explains Ranta.

© Finnish Food Safety Authority Evira, Mustialankatu 3, FI-00790 Helsinki Finland