Indices for performance evaluation of predictive models in food microbiology

TLDR

The authors propose two complementary, simple indices to evaluate predictive food‑microbiology models. The indices quantify prediction confidence and detect bias that could lead to fail‑dangerous outcomes. Using literature data, the indices were shown to be objective and interpretable, revealing limitations of growth‑rate‑based evaluation and highlighting that while they may miss some systematic deviations, they offer a useful first step toward defining validated models in food microbiology.

Abstract

Two complementary measures are proposed as simple indices of the performance of models in predictive food microbiology. The indices assess the level of confidence one can have in the predictions of the model and whether the model displays any bias which could lead to ‘fail‐dangerous’predictions. The use of the indices is demonstrated using data collated from independent and published literature. This analysis supports previous reports that evaluation of predictive models by comparison to published microbial growth rate data may be inappropriate because of limitations in that data. The indices may fail to reveal some forms of systematic deviation between observed and predicted behaviour. It is concluded, however, that the indices provide an objective and readily interpreted summary of model performance and may serve as a first step towards the development of an objective and useful definition of the term ‘validated model’in predictive food microbiology.