You are here:
ESTIMATING PATHOGEN EXPOSURES - THE CRITICAL CHALLENGE FOR QMRA TO SUPPORT REGULATION AND MANAGEMENT OF WATERS
Citation:
ASHBOLT, N. ESTIMATING PATHOGEN EXPOSURES - THE CRITICAL CHALLENGE FOR QMRA TO SUPPORT REGULATION AND MANAGEMENT OF WATERS. Presented at Toxicology and Risk Assessment Conference, West Chester, OH, April 23 - 26, 2007.
Impact/Purpose:
To provide broad-based support for the Microbiological and Chemical Exposure Assessment Research Division and to facilitate outreach and communications with peers, stakeholders, clients, and the public.
Description:
Pathogen and indicator concentrations normally vary by several orders of magnitude in raw waters, and to an even greater extent during hazardous event periods. This variation in concentration typically dominate the estimate of infection generated in a quantitative microbial risk assessment (QMRA), particularly if results are not averaged over a year. In addition to this variation, numerous uncertainties result from our attempts to assay pathogens and model their behaviour through water systems.
Raw water [oo]cyst concentrations are typically presented with little or no reporting of specific recovery data representative of the site sampled. The uncertainty resulting from limited recovery data was estimated for the MicroRisk Project systems using available data to improve the quality of Cryptosporidium and Giardia raw water concentration estimates. Recovery datasets ranging from 3-99 data points were examined by Bayesian statistics representing three Approaches: I - no recovery data, II - limited, unpaired recovery data from samples, and III - paired recovery data. No useful relationships were seen between water turbidity or type and recoveries reported. Critically, Approach I underestimated [oo]cyst concentrations by about 100%, with little difference between Approaches II & III. Using the smallest (n=3) recovery dataset, the upper band of uncertainty were on average more than 10-times (and on occasion up to 100-times) greater than when using the fullest (n=99) dataset; however, limited reduction in uncertainty occurred beyond n = 20. Nonetheless, for QMRA purposes, recovery data should be collected as a pair with count data for an initial period at least, so that any relationships (priors in Bayesian statistics) may be ascertained without the need to rely on and apply trends from elsewhere. When conveying QMRA results for risk management, output uncertainty should be incorporated into the results via either a two-dimensional (variability and uncertainty) risk assessment or a sensitivity analysis that includes recovery uncertainties.