Quantitative microbiological risk assessment (QMRA), predictive modelling and HACCP may be used as tools to increase food safety and can be integrated fruitfully for many purposes. However, when QMRA is applied for public health issues like the evaluation of the status of public health, existing predictive models may not be suited to model bacterial growth. In this context, precise quantification of risks is more important than in the context of food manufacturing alone. In this paper, the modular process risk model (MPRM) is briefly introduced as a QMRA modelling framework. This framework can be used to model the transmission of pathogens through any food pathway, by assigning one of six basic processes (modules) to each of the processing steps. Bacterial growth is one of these basic processes. For QMRA, models of bacterial growth need to be expressed in terms of probability, for example to predict the probability that a critical concentration is reached within a certain amount of time. In contrast, available predictive models are developed and validated to produce point estimates of population sizes and therefore do not fit with this requirement. Recent experience from a European risk assessment project is discussed to illustrate some of the problems that may arise when predictive growth models are used in QMRA. It is suggested that a new type of predictive models needs to be developed that incorporates modelling of variability and uncertainty in growth.