Why posting set results is bad policy

Policymakers at the U.S. Department of Agriculture's Food Safety and Inspection Service (USDA-FSIS) say that consumers have a right to know information concerning the safety of America's food supply and experience has shown it is best for them to give consumers such information fully, quickly and before they hear it from others. As a consumer, I agree with that general approach to addressing matters of food safety, but I wonder what information, specifically and directly related to the safety of America's food supply, FSIS intends to convey by publishing the salmonella performance categories of certain broiler slaughter establishments.

On March 28, FSIS began posting on its website the results of completed sample sets derived from its Salmonella Verification Sampling Program for young chicken (broiler) slaughter establishments with Category 2 or Category 3 performance status. To date, no sound science has been offered in support of the decision to implement these postings. There are numerous reasons why it is bad policy.

Data Is Generic, Not Timely

A review of the analysis completion dates of sample sets listed on the FSIS website shows many of the posted Salmonella Verification Sampling sets are eight or more months old. When these timeframes are added to the two-plus months needed to complete an average Salmonella Verification Sampling set, it is difficult to understand how the USDA can consider such information timely. Additionally, the salmonella performance category of each establishment is formulated using data that is qualitative (i.e. positive or negative) not quantitative, and the salmonella serotypes isolated from the positive samples have no bearing on the category. It is hard, therefore, to see how the USDA expects consumers to glean any relevant food safety information from these postings.

The FSIS founded this program on the notion that any serotype of salmonella, if present in/on food at a high enough concentration, can cause human illness. This is why, prior to 2006, the USDA did not consistently isolate and report serotypes from salmonella-positive samples, and why now, despite the fact that certain salmonella serotypes are rarely if ever attributed to human food borne illness, FSIS counts all salmonella-positive samples equally when determining the salmonella performance category of a broiler establishment.

Faulty Assumption

The FSIS adopted a similar, non-scientific, approach when it developed the salmonella performance categories in 2006. The development of Category 1 was based on an assumption that a 50 percent reduction in the prevalence of salmonella on carcasses tested at the post-chill location would result in a measurable reduction in human illnesses attributed to salmonella. Setting aside, for the moment, the problems associated with using Salmonella Verification Sampling data to inform decision making in this way, such an assumption marginalizes all the other routes (several of which have proven to be significant) by which transmission of salmonella to humans can occur. The overly simplistic approaches that the USDA has taken to address salmonella associated with raw young chicken products conflict with basic Hazard Analysis and Critical Control Point (HACCP) principles, and are not adequately supported by proper food safety risk assessments.

In April of 2006, it appeared as if the USDA Under Secretary for Food Safety, Dr. Richard Raymond, had a clear understanding of this when, speaking at a national food-borne epidemiologists meeting, he said the Agency's salmonella performance measure was, "scientifically unsound . . . for generic salmonella, including those serotypes that are not, or are rarely attributed to food-borne illness" and "not consistent with FSIS' goal of transitioning to a more robust risk-based inspection system." Why then, almost two years later, and without instituting any substantive changes to address these problems, would FSIS begin publishing on its website salmonella performance categories formulated using data from the Salmonella Verification Sampling Program?

Data Doesn't Track With Human Illnesses

This question becomes even more relevant when one compares data published in the Salmonella Verification Sampling Program Quarterly Progress Reports (also on the website) to the incidence of human salmonella infections reported in the April 11 (2008) issue of the Morbidity and Mortality Weekly Report (MMWR) from the Centers for Disease Control and Prevention (CDC). On a national basis, the prevalence of salmonella on broiler carcasses, tested by the FSIS at the post-chill location, declined from a high of 19.7 percent for the second quarter of 2005 to 8.5 percent for the fourth quarter of 2007. The percentage of all broiler establishments achieving a Category 1 performance status also increased from 35 percent for the first quarter of 2006 to 74 percent for the fourth quarter of 2007. According to the MMWR, however, when 2007 was compared with 2004-06, "the estimated incidence of infections caused by salmonella did not decline significantly." And, despite the fact that 2007 FSIS Salmonella prevalence data only reached a height of 10.6 percent (for the third quarter), the CDC reported that "the incidence of salmonella infections in 2007 (14.92 cases per 100,000) was the furthest from the national target for 2010 (6.80 cases)." These facts raise serious questions about the significance of Salmonella Verification Sampling data, and show how ill-conceived FSIS attempts to use it, such as the creation and posting of salmonella performance categories, lack scientific validity.

No Consumer Benefits, But A Downside

While it is clear that publishing completed salmonella sample set results for Category 2 and Category 3 broiler establishments will not translate into any tangible food safety benefits for consumers, posting such information without explaining its relevance or intended purpose (as the FSIS has done), could result in profound negative consequences. Consumers will believe these results are significant and may assume the Salmonella Verification Sampling Program is the single-most important measure of process control by which to judge broiler establishments. Doing so will incorrectly lead consumers to conclude that product from Category 1 plants is somehow safer than that from Category 2 plants, establishing Salmonella Verification Sampling data as the means by which to determine product acceptability, and Category 1 plants as the only acceptable ones from which to purchase product. Adoption of such an approach by our foreign trading partners would compound matters by making future U.S. trade negotiations more difficult. The realization of any of these complications as a result of the implementation of this non-scientific policy would be truly unfortunate.

Given that posting completed salmonella sample set results for Category 2 and Category 3 broiler establishments will not improve food safety and may very well lead to consumer confusion and/or restriction of domestic and international trade of U.S. broiler products, the FSIS should reconsider its decision to publish such information, and cease to do so. And, the USDA should begin to remove politics from the current regulatory food safety inspection system and work more closely with the poultry industry to develop scientifically-based approaches that will help address our common goal of reducing the number of human illnesses attributed to salmonella and the consumption of U.S. poultry products.