A qualitative probabilistic model of microbial outbursts in foods

Sequences of counts of potentially harmful organisms in foods usually exhib it an irregular fluctuating pattern. The counts are determined by the inter play of numerous random factors that tend to promote or inhibit the organis ms' growth. The counts can be recorded as zero, indicating that either the organism is not present or is below a minimum detectable level, or they can fluctuate randomly within characteristic bounds. An outburst is said to oc cur when the population surpasses a specified threshold determined by safet y or quality considerations. The growth pattern in this 'explosive' mode is also governed by a combination of random mechanisms that determine the gro wth rate and eventual decline of the population. This paper presents a prob abilistic model for such scenarios. The model parameters represent the unde rlying distribution of the fluctuations, the detection and explosion thresh olds and the probability of continued growth after an outburst has begun. A simplified version of the model was used to simulate examples of microbial histories that resemble those of sensitive foods. It is also used to eluci date how the frequency, intensity and duration of outbursts are affected by the parameters of the model. In addition, we demonstrate how to estimate t he model's parameters from actual records and illustrate the efficacy of th e estimation method with simulated data. The utility of such models for ris k assessment will depend on the availability of long records of microbial c ounts that include outbursts in order to test their predictive ability. Bec ause the presence of a harmful agent is not always sufficient to cause food poisoning, models of this kind can only estimate the expected frequency of outbursts but not the frequency of actual food-poisoning outbreaks. (C) 20 01 Society of Chemical Industry.