809 PROBABILISTIC RISK ASSESSMENT AND SLAUGHTERHOUSE PRACTICES: MODELLING CONTAMINATION PROCESS CONTROL IN BEEF DESTINED FOR HAMBURGER Tanya Roberts Scott A. Malcolm Clare A. Narrod USDA/Economic Research Service 1800 M St. NW, Room 3077N Dept. of Geography and Environmental Engineering AAAS Fellow USDA/FSIS/OPHS/ERAD Washington, DC 20036-5831 Johns Hopkins University 1400 Independence Ave (202) 694-5464 Baltimore, MD 21218 Washington, DC 20250 (301) 962-0227 (202) 501-7400 ABSTRACT The art and science of risk assessment as applied to foodborne pathogens is still in its infancy and limited to what can be measured and quantified. 1 Many important process components are omitted from models, including this one, because of lack of data. Still the models may yield insights into process control and evaluation and/or into data collection priorities and need to be proposed for scientific evaluation and further refinement. This paper models four beef slaughterhouse events with two levels of process control of generic E. coli. Monte Carlo simulation is used to characterize the distribution of contamination of average raw burgers. For slaughter plants with each process at “level 1” control, <1.5% of raw hamburgers have >4 log 10 colony forming units (CFU) of generic E. coli per quarter-pound hamburger. In contrast, plants with all “level 2” processes produced >93% of raw hamburgers with this level of generic E. coli. Sixteen scenarios are used to measure the sensitivity of the output distribution to changes in process control regimes at each step. I. INTRODUCTION Forty foodborne pathogens are estimated to cause 6.5 million to 33 million human illnesses annually with up to 9,000 deaths and unquantified chronic illnesses. 2 Medical costs and productivity losses for seven foodborne pathogens are estimated at $6.6 billion to $37.1 billion annually. 3 In response to increased public health concerns about foodborne pathogens, both the USDA and the FDA have instituted mandatory and voluntary pathogen reduction programs, such as Hazard Analysis and Critical Control Point (HACCP) systems, for foods under their jurisdiction. In the summer of 1998, the National Academy of Sciences studied how to improve the U.S. food safety system. Their most important conclusion 4 was that the food safety system: “…be science-based, with a strong emphasis on risk analysis, thus allowing the greatest priority in terms of resources and activity to be placed on the risks deemed to have the greatest potential impact (p. 5).” This paper develops a process model that includes the major functions of beef slaughter plants. Average carcass contamination is modeled as the sum of random variables that represent either contamination or decontamination of the carcass surface. The probability distributions of the component variables are roughly estimated from the available literature, with an emphasis on identifying the range of values. For each step in the process, two different levels of practice indicating level of process control are defined. The data available on slaughter operations is scarce, and subject to both variability and uncertainty in each step. Since the focus of this work is on identifying opportunities for process control, no attempt is made to separate variability and uncertainty. The output of the model is a distribution of contamination of an average raw hamburger. Monte Carlo Simulation is used to create this distribution and assess the control of the process. Output process control is defined to mean the probability that an average burger is below some specified level of contamination. By comparing the levels of output process control under combinations of process control for the component steps, the contribution of changes in process control regimes at each step can be evaluated.