Quantifying uncertainty associated with microbial count data: a Bayesian approach

We consider the problem of estimating bacterial concentration in a substance, given microbial count data. A Bayesian approach is proposed which naturally allows the incorporation of both plate-count data and extra information from confirmatory tests such as genotyping by polymerase chain reaction (PCR). The estimation methods yield posterior credible regions for bacterial concentration, in contrast to the previous methods, which generally only produce point estimates. The approach is illustrated with specific reference to the enumeration of the food-borne pathogen Escherichia coli O157 by spiral plating, although the methodology can be applied to any bacterium or counting method of interest. The results obtained provide guidance to the experimenter as to the number of confirmatory tests which should be performed, and also suggest that in the initial plate count one should err on the side of including rather than excluding colonies whose genotype seems unclear.