The Food and Drug Administration has completed testing a system that relies on data mining technology to screen imported foods for contamination, and plans to deploy it for agencywide use, according to an annual report FDA released on Monday.
In its Food Protection Plan report, FDA said it successfully tested the Predictive Risk-based Evaluation for Dynamic Import Compliance Targeting system at five seaports in the Los Angeles area and has established a steering committee to deploy the system agencywide.
The value of U.S. food imports total about$70 billion a year, but FDA inspected only 1 percent of imported food from 2002 through 2007, according to a Government Accountability Office report released in September. An outbreak of Salmonella Saint Paul this summer caused grocery stores nationwide to remove tomatoes from shelves before FDA identified jalapeno and serrano peppers grown in Mexico as the source of the illness.
PREDICT can better detect possible problems in food shipments than existing FDA systems, GAO reported. The system can use up-to-date data from FDA laboratory tests and other information, such as weather events that could affect water quality, to identify imports that might require closer inspection, the audit agency added.
The system also uses the agency's data from tests on imported fresh produce, information that current FDA electronic screening do not use, GAO said. FDA officials plan to test PREDICT further at an unidentified border crossing by March 2009.
New Mexico State University developed the system, and FDA said it plans to run a procurement this winter to build the production version of PREDICT. The contract is expected to have a value of more than $10 million, according to a post on a procurement Web site maintained by the Health and Human Services Department.
An FDA spokeswoman did not return calls requesting more information on the system. A spokesman at New Mexico State University said he could not provide details on PREDICT by deadline.