The variability introduced by partial sample analysis to numbers of Cryptosporidium oocysts and Giardia cysts reported under the information collection rule

Subsampling of the concentrated sample pellet resulting from the information collection rule (ICR) method for detection of Cryptosporidium oocysts and Giardia cysts is often required to facilitate the terminal microscopy step. Variability in counts of (oo)cysts in subsamples suggests that sampling error may occur as a consequence of subsampling and subsequent partial analysis of this pellet. Variability is modeled here as a Poisson distribution and is incorporated into a model that accounts for other uncertainties in the detection procedure. The result is to greatly expand confidence intervals (CI) around reported concentrations. For example, the 90% CI around the concentration reported based on detection of one (oo)cyst in a pellet subsample is increased nearly threefold when sampling error arising from partial sample analysis is added to other potential sources of variability. Consideration of a distribution in which (oo)cysts are clumped, such as a negative binomial distribution, further increases uncertainty, producing confidence intervals ranging over several orders of magnitude. The relevance of the uncertainty arising from partial sample analysis to risk assessment is demonstrated. For example, to produce water with an acceptable 1 in 10,000 risk of infection, the level of water treatment prescribed may increase from 3- to 4.5- (or greater) log removal of (oo)cysts. Implications of this additional uncertainty in the ICR method on reported (oo)cyst concentrations argue for expanded reporting procedures to include an indication of the uncertainties associated with reported concentrations and/or improvements in the detection method.