Weaknesses in Meat and Poultry Inspection Pilot Should Be Addressed Before Implementation United States General Accounting Office (GAO) Report to the Committee on Agriculture, Nutrition, and Forestry, U.S. Senate GAO-02-59 17dec01 (released 7jan02)

The Honorable Tom Harkin
Chairman
The Honorable Richard G. Lugar
Ranking Minority Member
Committee on Agriculture, Nutrition, and Forestry
United States Senate

While the food supply in the United States is generally safe, foodborne bacteria, viruses, and other pathogens can cause a significant number of illnesses, a small fraction of which are life-threatening. According to estimates from the Centers for Disease Control and Prevention (CDC), foodborne diseases cause 76 million illnesses in the United States each year, including 325,000 hospitalizations and 5,000 deaths. Meat and poultry products contaminated with bacteria such as Salmonella, Campylobacter, Escherichia coli (E. coli), and Listeria cause most foodborne illnesses and deaths. To control the spread of foodborne illnesses through meat and poultry products, approximately 3,400 U.S. Department of Agriculture (USDA) inspectors at 1,300 slaughter plants are stationed along slaughter lines to provide continuous inspection and conduct organoleptic examinations—using sight, touch, and smell—of each and every carcass.

In 1997, USDA announced the need to modify its meat and poultry slaughter inspection program to make industry more responsible for identifying carcass defects. This approach is consistent with the agency’s previous adoption of the Pathogen Reduction: Hazard Analysis and Critical Control Point (HACCP) regulations. The HACCP approach is risk-based and makes industry, rather than federal inspectors, responsible for identifying steps in food production where food safety hazards are most likely to occur and for establishing controls that prevent or reduce it. USDA had not extended the HACCP principles to slaughter inspections because the agency has traditionally provided continuous inspection of each and every carcass. However, USDA believes that changing its traditional inspection system would also reduce inspectors’ reliance on organoleptic inspections, allow for a shift to prevention-oriented inspection systems based on risk, and permit redeployment of its resources to better protect the public from foodborne diseases.

Before making a permanent change to its slaughter inspections system, USDA developed a model to test whether such a change would continue to ensure the safety of meat and poultry products. This model is being tested at several volunteer chicken and hog plants. At plants participating in the project, plant personnel, instead of USDA inspectors, examine each carcass to make an initial determination whether it is unacceptable and should be removed from the slaughter line. At the participating plants, USDA first measures the performance of traditional inspection systems and then compares it with the performance of modified inspections. USDA will propose regulations to modify its inspection system nationwide if the project shows that the modified inspection system performs as well as or better than traditional inspections. A reduced number of USDA inspectors are still at each plant to verify that safety and quality standards are met. In addition, at least one USDA inspector is positioned near the end of each slaughter line in order to comply with a federal court ruling that USDA cannot delegate the inspection of carcasses to plant employees. As of July 2001, the federal cost associated with this project was approximately $5.7 million, and USDA estimates that it will spend another $1.2 million. Australia and Canada have tested similar modified inspection programs for meat and poultry and approved their use on a nationwide basis.

Concerned about the design, methodology, and reliability of the pilot project, you asked us to (1) describe the objectives, design, and scope of the project; (2) identify limitations, if any, in the project’s design and methodology; and (3) any design and methodology limitations notwithstanding, determine if the data generated by the project will allow USDA to reach valid conclusions on the relative effectiveness of modified and traditional inspection methods in ensuring food safety and quality. In addressing these objectives, you also asked us to obtain information from similar projects in Australia and Canada.

In response to your request, we reviewed USDA’s pilot project as implemented at chicken and hog plants. However, we were able only to analyze data from chicken plants because complete data from hog plants were not yet available. We also visited Australia and Canada to learn about similar pilot projects. Appendix I describes our methodology in detail.

USDA’s original objective in implementing the pilot project was to test whether a prevention-oriented inspection system that uses plant personnel to examine each carcass for safety and quality and USDA inspectors to verify that safety and quality standards are met can provide a level of product safety and quality equal to or better than traditional inspections. At 11 chicken and 3 hog plants that are voluntarily participating in the before and after phases of the project, plant personnel, instead of USDA inspectors, initially determine which carcasses and parts are unacceptable and should be removed from the slaughter line because they are diseased or unwholesome. USDA is, therefore, able to use fewer inspection personnel at these plants. Using safety and quality performance standards developed for the project, an independent contractor measured how well pilot plants’ inspection systems performed against these standards. The contractor measured how well USDA inspectors identified carcass defects under traditional inspections and then conducted similar measurements once plant personnel assumed those duties. Also, as a part of the pilot project, USDA inspectors at the project plants have examined carcasses to evaluate how well plant employees detect carcass defects. USDA will analyze these data to determine if at least the same level of safety and quality is maintained under the modified inspections. USDA has announced that preliminary results from the chicken pilot project show that plants’ inspection systems perform better under the modified inspections, and that it expects to propose regulations to modify its slaughter inspection system for all chicken plants early in 2002. Both Australia and Canada have recently tested and adopted modified inspection programs that resemble USDA’s pilot project in that they include the replacement of some government inspectors with plant personnel.

USDA’s pilot project for chickens has several design and methodology limitations that compromise the overall validity and reliability of its results. Hence, it is questionable whether the data generated by the project are indicative of how all of the chicken plants’ inspection systems would perform if modified inspections were adopted nationwide. First, the chicken pilot that USDA designed lacks a control group—a critical design flaw that precludes a comparison between the performance of the inspection systems at those plants that volunteered to participate in the pilot and that of plants that did not participate. Without a control group, USDA cannot determine whether changes in inspections systems are due to personnel changes or other possible explanations, such as the addition of chlorine rinses. Second, the chicken plants that volunteered to participate in the baseline measurement phase of the pilot were not randomly selected, and they did not include plants from all chickenproducing areas or plants of all sizes. Thus, the results cannot be generalized to the entire population of chicken slaughter plants in the United States. Third, the pilot project’s methodology did not take into account variables such as seasonal changes and plant modifications that could affect project results. For example, after the project began, many plants added antimicrobial rinses and washers, which usually reduce the levels of microbial contamination. USDA acknowledges some of the project’s limitations but maintains that the design is consistent with that of Page 4 GAO-02-59 Meat and Poultry Pilot Inspection Project other USDA studies and that it is difficult to conduct this type of experiment without relying on volunteer participants. Finally, USDA’s pilot project did not include features of the modified inspection systems in Australia and Canada that would be important considerations in ensuring the successful implementation of a modified inspection system nationwide. For example, during the pilot project, USDA did not require the training of plant employees. Such training would be important in ensuring that plant personnel are as competent as federal inspectors in identifying carcass defects that can affect product safety and quality. Similarly, the pilot project did not require the use of statistical process controls as required in Australia and Canada. Statistical process controls provide a means to determine whether the plant’s production processes are performing within established performance standards. This technique would enable USDA and the plants to better measure and control their performance daily and over time to ensure continuous improvement. Additionally, USDA allowed plants with repeated noncompliance records to continue participating in the pilot project. If USDA decides to permanently modify its inspection system, a phased-in approach such as Canada’s would help to ensure that plants continually improve their ability to ensure product safety and quality and that those that have difficulty return to traditional inspections.

Notwithstanding the project’s design problems, which we believe make the results unreliable, we found that, so far, the data themselves do not conclusively demonstrate that modified inspections are at least equal to traditional inspections. Part of the difficulty is that the two data sets collected for this pilot show somewhat different results. These data show whether the pilot plants met the seven organoleptic performance standards—two food safety standards and five other consumer protection (quality) standards—that USDA developed to compare, first the inspection systems’ performance at plants under traditional inspections and then under modified inspections. The modified inspection systems at plants in the pilot project were expected to perform at least as well as they did under traditional inspections. However, the contractor’s data show that while the majority of the chicken plants’ inspection systems met or exceeded four or more of the seven organoleptic standards adopted by USDA, none met all seven. Specifically, the contractor’s data show that most of the inspection systems at these plants did not meet the safety standard for the presence of fecal material, which could contain harmful bacteria such as E. coli. However, the traditional inspection system at these plants was also unable to meet this standard. Most inspection systems also did not meet quality standards for defects such as feathers and oil glands, but USDA officials pointed out that such defects are of a lesser concern because they are not associated with foodborne illness and they can generally be removed from the carcass after it leaves the slaughter line and moves to the processing area. USDA’s contractor also measured whether the modified inspection systems at these plants met two other standards that existed before the pilot project—the pathogen reduction performance standard for Salmonella and generic E. coli. The data show that 9 of the 11 inspection systems met the Salmonella standard and that 80 percent of the samples collected for generic E. coli were within the acceptable ranges. On the other hand, the data collected by USDA inspectors at these chicken plants show better results than those shown by the contractor’s data. According to USDA’s data, 7 of 10 plants’ inspection systems met 6 of the 7 standards. While the data generated by the pilot project yielded inconclusive results, the pilot project’s inspectors and veterinarians that we surveyed for the most part believe that a modified inspection system results in safer products. More than half of those surveyed said that modified inspections are equal to or somewhat better than traditional inspections in ensuring product safety and quality. Modified inspections allow for additional time to inspect slaughter line operations more thoroughly because inspectors, except for the carcass inspector, are no longer tied to a fixed location.

This report reiterates our previous recommendation for legislative revisions aimed at reducing the potential for further legal challenges by providing USDA with clear authority to modify its inspection system. In addition, this report makes several recommendations to the Secretary of Agriculture aimed at ensuring that, if USDA decides to implement a modified inspection system, such a system will effectively ensure product safety and quality. In commenting on a draft of this report, USDA said that the report’s recommendations for executive action are appropriate and that the issues noted in the report can be satisfactorily addressed. USDA also provided technical comments which we incorporated as appropriate.

At slaughter plants, FSIS inspectors are located at fixed stations and examine carcasses to identify and remove defects that impair product safety and quality. Defects that are likely to present a public health risk, such as the presence of potentially pathogenic fecal material, are considered food safety hazards. Product quality defects, such as feathers or bruises on the carcasses, are aesthetic flaws that rarely present a direct safety risk.

Inspectors perform different tasks, depending on the animal species being slaughtered. For example, in chicken plants, the inspectors are located at fixed stations immediately after the place on the slaughter line where carcasses are opened and the organs removed. From these positions, the inspectors examine the exterior, interior cavity, and organs of each carcass. The inspectors then determine which carcasses must be condemned, which have conditions that may be trimmed, and which may proceed to the next stage in the slaughter line. Carcasses are then trimmed, rinsed, and, as the final step of the slaughter line, placed into a chiller. From here, carcasses move to the processing area of the plant, where they are cut up and packaged. In contrast, in hog plants, the inspectors are located at three fixed stations on the slaughter line. At the head station, the inspectors excise and evaluate the head lymph nodes; at the viscera station, they evaluate most organs; and at the final carcass station, they examine each carcass for possible contamination, disease conditions, and other adulterants before it goes into the chiller or cooler.

For a variety of reasons, including responding to recommendations from the National Academy of Sciences and GAO that FSIS shift its resources to a prevention-oriented, risk-based inspection system, in 1996 USDA embarked on regulatory reform efforts to reduce foodborne hazards in meat and poultry products. The agency adopted a new science-based process control system, the Pathogen Reduction: Hazard Analysis and Critical Control Point system. HACCP is a risk-based approach that makes industry, not federal inspectors, responsible for identifying steps in food production where contamination is most likely to occur and for establishing controls that prevent or reduce contamination.

USDA anticipated that the HACCP framework would also require changes in the roles and responsibilities of its inspection workforce, as industry, not government, becomes increasingly responsible for the safety of its products. For slaughter line operations, however, making industry more directly responsible for the safety and quality of their products was problematic because of a legal requirement for continuous carcass-bycarcass inspection. USDA believed that, to more fully integrate the HACCP concept into meat and poultry slaughter operations, its traditional inspection system would have to be modified to transfer some of the carcass defect detection responsibilities to industry personnel. Before implementing such an approach, FSIS decided to conduct a pilot to test and evaluate whether product safety and quality could be maintained at a comparable level if plant employees perform the carcass defect detection duties and a reduced number of federal inspectors remain to verify product safety and quality through increased testing and observation.

The pilot project is known as the HACCP-Based Inspection Models Project (HIMP) and its original design was intended to measure the effect of completely removing FSIS inspectors from their fixed locations on the slaughter line while maintaining one oversight and one verification inspector. In April 1998, the inspectors’ union filed suit in the United States District Court for the District of Columbia to enjoin the USDA from proceeding with the pilot project, on the grounds that it violated the Federal Meat Inspection Act, 21 U.S.C. section 604, and the Poultry Products Inspection Act, 21 U.S.C. section 455, by not requiring federal government officials to perform carcass-by-carcass postmortem inspections. In June 2000, the United States Court of Appeals for the District of Columbia Circuit ruled that delegating the task of inspecting carcasses to plant employees violated the acts because both statutes require that federal inspectors, rather than private employees, determine whether a product is adulterated. As a result of this ruling, in September 2000, FSIS redesigned the pilot project for chickens and hogs and placed at least one FSIS inspector back at a fixed location on each slaughter line to inspect each carcass.

Following a district court determination in January 2001 that the redesigned pilot project does not violate the acts, the inspectors’ union appealed in February 2001. That appeal is still pending. A detailed description of the litigation surrounding the pilot project is contained in appendix II.

Figure 1 depicts inspection responsibilities under FSIS’ traditional and modified inspections on a chicken slaughter line.

Figure 1: Traditional and Modified Inspection Systems on a Chicken Slaughter Line

Traditional Inspection System

Modified Inspection System

Note: Under the traditional inspection system, the slaughter line’s speed at most chicken plants is 91 birds per minute. Each inspector is responsible for about 35 birds per minute. Therefore, three FSIS inspectors are required to inspect carcasses on a slaughter line that runs at a speed of 91 birds per minute. The figure above shows only one FSIS inspector’s station in detail. Under the pilot project, a plant employee (sorter) replaces the FSIS inspector who examined the carcasses to determine their condition.

As figure 1 shows, under traditional inspection systems, three FSIS inspectors are present at fixed locations on the slaughter line. In contrast, at chicken plants that participate in the pilot project, one FSIS inspector is present at a fixed location on each slaughter line, while a verification inspector monitors the entire line. The verification inspector is free to move along the slaughter line to continuously observe and evaluate the plant’s implementation of its HACCP system and process controls. This inspector is also responsible for randomly selecting and examining 80 carcasses per line per shift to verify that the plant is complying with the performance standards for food safety and quality. A slaughter line averaging 91 chickens per minute would process over 43,000 chicken carcasses during each shift. The carcass inspector is located at the end of the slaughter line, immediately before carcasses enter the chiller. This inspector visually examines each carcass to comply with a court mandate that each carcass should receive FSIS inspection. At participating hog plants, carcass inspectors are positioned at the head, viscera, and final carcass inspection stations. Most of the inspectors and some veterinarians at the pilot project plants received promotions.

Pilot Project Makes Industry More Responsible and Accountable for Product Safety and Quality, While Continuing Government Monitoring With Fewer Resources

The pilot project’s hypothesis was that a prevention-oriented inspection system that uses plant personnel to examine each carcass can provide a level of product safety and quality equal to or better than traditional inspections. At selected chicken and hog plants that voluntarily agreed to participate in the project, plant personnel, instead of FSIS inspectors, initially determine which carcasses and parts are unacceptable and should be removed from the slaughter line. Under this project, FSIS is able to use fewer inspection personnel at each plant. FSIS used an independent contractor to collect and analyze organoleptic and microbial data under the traditional inspection system first and then under the modified inspection system. FSIS developed organoleptic performance standards to measure how well the inspection system at participating plants performed after shifting from traditional to modified inspections. In addition, as part of their duties, FSIS inspectors continued to collect organoleptic performance data at these plants. Recently, both Australia and Canada tested and adopted modified inspection programs that resemble USDA’s pilot project.

Under Pilot, Industry Is More Responsible and Accountable for Product Safety and Quality, While FSIS Provides Inspection Using Fewer Government Resources

To initiate the pilot project, FSIS invited meat and poultry plants that slaughter young, healthy animals—market hogs, fed cattle, and young poultry (including young turkeys)—to volunteer to participate in the project.1 Sixteen chicken, 5 hog, and 5 turkey plants initially volunteered to participate.2 No cattle plants volunteered. In general, plant managers that volunteered to participate in the pilot said that they elected to do so because they believe they can enhance product safety and quality if they have better control of slaughter line operations and, at the same time, increase productivity.

FSIS contracted with the Research Triangle Institute3 (RTI) to collect and analyze organoleptic and microbial data at plants that volunteered to participate in the pilot project. The purpose of this data collection is to evaluate whether the inspection systems at plants operating under modified inspections perform at least as well as the traditional inspection systems. In August 2000, RTI completed its collection of organoleptic and microbial data at 16 chicken, 5 hog, and 5 turkey plants under the traditional inspection. These baseline data documented the accomplishments of the traditional inspections system, and FSIS decided to adopt the results as performance standards for measuring the accomplishments of the same plants’ inspection systems after they shifted to modified inspections. After a transition period during which plant personnel practiced their new roles, RTI again collected organoleptic and microbial data under the modified inspection system to provide a beforeand- after comparison.4

After RTI collected data on the performance of the traditional inspection system at the volunteer plants (baseline data), 5 of the 16 chicken plants and 2 of the 5 hog plants dropped out of the project.5 As a result, RTI has collected modified inspections data at 11 chicken and 3 hog plants. These data were collected over a 5- to 6-week period. Some managers from plants that dropped out cited uncertainty about the future of the pilot project because of the court’s action, which resulted in modifications to the pilot’s original design.

Ten additional chicken plants have since volunteered to participate in the pilot, bringing the total number of participating plants to 21, but RTI will collect data at only 5 of the additional 10 plants. According to FSIS, no traditional inspection performance data (baseline data) will be collected at these additional plants. Table 1 shows the number of slaughter plants that participated both in the baseline data collection phase (traditional inspections) and in data collection after they switched to modified inspections.

Table 1: Number and Types of Pilot Plants That Participated in Traditional and Modified Inspections

              Plants that completed baseline data   Plants that completed data
              collection under traditional          collection under modified
Species       inspections                           inspections               
Chickens      16                                    11 (10)a
Hogs           5                                           3
Turkeys        5                                           0

aOne plant dropped out after it completed data collection under the modified inspection system.
Source: USDA, FSIS.

Because turkey plants have not participated in modified inspections and because data from hog plants under modified inspections are not yet available, this report discusses pilot project results from chicken plants only. (See app. III for information on the development of performance standards for hogs.)

At the pilot project plants, FSIS has been able to use fewer inspection resources than would otherwise be used under traditional inspection systems, resulting in an overall 22 percent reduction (from 259 to 202 inspectors) in the number of FSIS inspectors on the slaughter floor. At most of the 14 chicken and hog plants, the reduction of inspectors ranged from 1 and 10 inspectors, depending on the number of slaughter lines and the speed of those lines. Under traditional inspections, there were three inspectors per line on average, compared with two inspectors per line under modified inspections. At three of the pilot project plants, the change in the number of inspectors was not entirely due to the modified inspection system.

The Pilot’s Scope Included the Development of Safety and Quality Standards to Measure Performance of Traditional Inspections as Compared With Modified Inspections

To collect and analyze organoleptic data on the defects that the FSIS inspectors had overlooked under traditional inspections at chicken plants, RTI veterinarians at each volunteer plant examined approximately 2,000 randomly selected carcasses that had passed FSIS’ slaughter inspection. RTI veterinarians examined carcasses to record the same type of defects that FSIS inspectors identify when they examine each carcass. These included safety defects, such as visible fecal material and evidence of septicemia and/or toxemia, and other consumer protection defects (quality defects) such as feathers, hairs, or bruises. FSIS grouped this information into two food safety and five other consumer protection (quality) categories:

FSIS then developed performance standards for each of these seven categories. The quality performance standards were set at position 12 (out of 16) on the basis of measurement of the traditional inspection system at the 16 chicken plants. That is, 25 percent of the plants’ inspection systems (4 different plants) would have to improve on their baseline results in each of the 7 performance categories. FSIS officials explained that they decided to adopt the 12th position of the baseline results as the standard to ensure that most participants would be able to meet them. The officials stated that this represents a reasonable tightening of the traditional system’s accomplishments.

To better protect consumers from defects that may be harmful, FSIS set the two food safety standards at zero rather than adopting the results of the 12th position. Thus, a performance standard of zero for fecal contamination means that FSIS expects all carcasses slaughtered during a shift to be free of fecal material. Conversely, an 80-percent performance standard for dressing defects such as feathers and oil glands means that if 80 of 100 carcasses have these defects, the performance standard is met. The performance standards represent the percentage of carcasses per shift with each defect that FSIS considers acceptable. Table 2 shows the traditional inspection system’s baseline results for the 16 chicken plants.

Table 2: Traditional Inspection System’s Results (Baseline Data) From 16 Chicken Pilot Plants: Defects per Shift in Percents

Note: Ranking numbers represent different plants for different food safety and OCP standards. Although the 12th position’s performance was 0.05 percent for Food Safety 1 and 1.5 percent for Food Safety 2, FSIS adopted a zero tolerance policy for these two standards. For Food Safety 2, the zero tolerance standard was used because it is already a regulatory requirement. Source: RTI.

RTI also measured the prevalence of generic E. coli and Salmonella under traditional inspections and modified inspections at the 11 chicken plants to determine if changes in the inspection system would have an effect on the microbial profile of carcasses. RTI collected and analyzed 300 samples for both microorganisms over a 6-week period. RTI analyzed the Salmonella samples to determine if they met FSIS’ already existing regulatory standard for Salmonella and the generic E. coli samples to determine if the results were in the acceptable, marginal, and unacceptable ranges established by FSIS.

FSIS is also collecting data on product safety and quality defects at chicken plants in the pilot project and is analyzing these data to compare how well traditional inspection systems perform versus modified inspection systems. FSIS inspectors collect the data as they conduct scheduled and unscheduled verification tests at each chicken plant. These tests evaluate, among other things, how well plant employees detect carcass defects. When the pilot project is concluded, FSIS plans to analyze these data to determine if at least the same level of safety and quality is maintained under the pilot program. FSIS has announced that preliminary results from the chicken pilot project show that inspection systems perform better after the plants shifted to modified inspections, and that it expects to propose regulations to modify its traditional slaughter inspection system for all chicken plants in 2002.

The Australian Quarantine and Inspection Service and the Canadian Food Inspection Agency have tested and adopted alternative inspection systems similar to that tested in the FSIS pilot project. (See apps. V and VI for additional details.)

Australia has tested and adopted a modified inspection system for all meat plants that produce products for domestic consumption. In addition, Australia is now in the process of implementing the Meat Safety Enhancement Program at plants that export meat. Under the program, plants are responsible for plant process controls, including sanitation, the microbial monitoring of pathogens, and detecting product defects. The program incorporates the use of government-licensed company employees, acting under the supervision and oversight of government veterinarians, to replace government inspectors. In order to be licensed by the government, the company’s employees are required to undertake about 600 hours of training and pass an examination. The Australian Quarantine and Inspection Service sets public health product standards and audits plant operations to ensure that food safety standards are met. A key feature of the program is the mandatory use of statistical process controls to track plants’ performance over time.

Canada has adopted a new poultry inspection system, known as the Modernized Poultry Inspection Program that allows plant personnel to assume defect detection duties and remove defective carcasses from the slaughter line under the supervision of government inspectors. The program is being implemented in a phased approach. Performance standards for Canadian poultry plants participating in this program are similar but not identical to FSIS’ standards for chicken plants in the U.S. pilot project. Like the Australian program, the Canadian program has stringent training requirements for industry personnel replacing inspectors. Industry personnel are required to demonstrate competency through testing and certification.

Design and Methodology Limitations Compromise Overall Validity and Reliability of Pilot Project Results

The pilot project has several design and methodology limitations that compromise the overall validity and reliability of its results. For example, the lack of a control group prevents valid comparisons between the inspection system of participating plants and that of nonparticipating plants. In addition, the participating plants were not randomly selected; therefore, results from these plants cannot be generalized to the entire population. Finally, the pilot project does not appropriately explain how variables, such as seasonal changes and plant modifications, could affect the project’s results. In addition, the pilot project does not include important features of similar programs in Australia and Canada.

FSIS officials characterized the design of the pilot project as a classic before-and-after experiment that is typical of other FSIS experiments. A classic before-and-after experiment involves both a control and an experimental group and thus provides a basis for comparing the results of the two groups. Ideally, the groups are as similar as possible, except for the variable that the experiment is attempting to measure. However, because this level of rigor is rarely achievable in applied settings, quasiexperimental designs are often used in which comparison groups are established to compare the impact of a new program with the status quo. According to our discussion with FSIS officials, the volunteer plants were not separated into a control group and an experimental group because the agency felt that it would have been unfair to exclude volunteer plants from taking part in the modified inspection aspect of the experiment. Nevertheless, under a quasi-experimental design, FSIS could have used as a control group the five chicken plants that decided to discontinue participation in the pilot project after their baseline performance under the traditional system was measured. The agency could have then compared the results from these plants with the results obtained from the plants that participated in the modified inspections. Lacking such a comparison, it is difficult to determine how the inspection systems at the plants that did not participate in the pilot project would have performed under the performance standards established for the pilot project.

The managers of two large chicken plants that dropped out of the pilot said that they proposed another approach to FSIS. Namely, these managers wanted to conduct what could have been a controlled experiment within their own plant. Both of these plants have multiple slaughter lines, and the managers wanted to operate half of their slaughter lines under traditional inspections and the other half under modified inspections. However, FSIS would not authorize such an arrangement because the agency wanted the entire plant to be under one inspection system. Officials said that it would have been too difficult to maintain two separate inspection systems in the same plant. As a result, both of these plants withdrew from the pilot project.

The Australian pilot project compared three plants that volunteered to participate in the pilot project with three plants operating under traditional inspections, so the pilot had both an experimental group and a control group. The plants in the experimental group removed the government inspectors from the slaughter line and replaced them with plant personnel, while the control group plants made no changes to their inspection system. The pilot project measured both microbial and organoleptic data at plants in both groups to determine how well they complied with standards. The results were then compared in order to determine how the modified inspections affected food safety and quality. The Canadian Food Inspection Agency did not use a control group.

Plants Participating Are Not Randomly Selected and Do Not Represent the Population

Because the volunteer chicken plants were not randomly selected, the results cannot be generalized to the population of plants that slaughter young chickens. FSIS is not authorized to compel plants to participate in any pilot project; therefore, it had to rely on volunteer plants instead of randomly selecting participants. However, FSIS did have the option to randomly select a subgroup to serve as a control group from the volunteer plants, but decided not to pursue that option. Australia and Canada also did not select plants on a random basis.

FSIS officials acknowledge that the number of plants in the pilot was not statistically derived. The officials believe, however, that the 16 chicken plants that originally volunteered constitute a sufficient number to estimate results for the population. However, only 11 chicken plants have participated in all phases of the pilot project. These plants represent about 6 percent of the population of young chicken plants nationwide, or 11 of 186 plants in 1999. We estimate that 125 randomly selected plants would need to participate in the project in order for FSIS to obtain results that can be generalized to the population of chicken plants.6

As a result of the pilot project’s reliance on volunteer plants, the results of the pilot do not provide sufficient geographic representation. That is, the 11 chicken plants that have participated in all phases of the pilot project are located in 7 of 35 states that produce young chickens, or about 20 percent of the chicken-producing states. These plants are concentrated in the southern region of the United States and account for about 57 percent of U.S. young chicken production. (See fig. 2.) Unlike FSIS, the Canadian Food Inspection Agency designed its pilot to include chicken plants from the western, eastern, and central provinces. As a result, the Canadian pilot project has better geographic representation.

In addition, there are eight large plants, three small plants, and no very small plants in the pilot project. According to FSIS, there are 112 large, 61 small, and 13 very small plants that slaughter young chickens in the United States. We discussed with officials at all participating plants in the pilot project whether the plants represent the population of chicken plants in the United States. Most of them said that, in their opinion, the plants are representative of the industry nationwide. FSIS officials are also of the opinion that all plants that volunteered to participate are similar to nonparticipants in all respects except for the fact that they volunteered. However, since only volunteer plants participated in the pilot, potential selection bias is introduced because such plants may be predisposed to better performance.

Pilot Project’s Methodology Does Not Take Into Account Variables That Could Affect Results

The pilot project does not adequately control for the effect that seasonal variations and plant modifications can have on the its results. Because seasonal changes can cause animals to develop diseases and/or poor health, two poultry science experts that we consulted said that it is imperative to collect samples over sufficiently long periods of time to cover all seasons. Seasonal variations significantly affect microbial results, especially in the case of bacteria such as Salmonella, which was measured in the pilot project. However, the pilot project’s design called for 6 consecutive weeks of microbial data collection. Accordingly, the experts we consulted believe that microbial data collected during the pilot project are of limited value because the data cover less than 2 months at each plant.

The pilot project’s results may also be affected by another set of variables. Most of the plant managers at pilot project plants told us that they introduced changes to their slaughter lines to improve product safety and quality after the baseline performance of the traditional system was measured. These changes included the addition of new brushes and scrubbers, plain water washes, and antimicrobial washes. Some managers also modified their evisceration equipment and developed computerized systems to track carcass safety and quality defects. All the inspectors and veterinarians at pilot project plants that we surveyed indicated that at least one change had been made after modified inspections were implemented at the plant. In effect, it is not possible to discern whether the before-andafter results measured by the two food safety and five quality standards are to be solely attributed to the shift from traditional to modified inspections at these plants.

According to FSIS officials, the pilot project was designed to measure the accomplishments of the traditional inspection system and the accomplishments of the models inspection system allowing for measuring multiple process changes—not isolated comparisons between FSIS inspector performance and plant employee performance. If this is the case, then the pilot project should have included the measurement of all the process changes introduced, not just the only change introduced by the agency; namely, the transfer of carcass defect detection from FSIS inspectors to plant personnel. We understand that, at their initiative, plants decided to make changes to their process during this pilot. However, FSIS did not collect data on what changes were made or on what the effect of those changes may be on the overall pilot project results.

USDA officials disagree that the pilot project does not take into account variables that could affect its results. They stated that, if the results from all plants are viewed collectively, their sample extended over a sufficient period of time to consider the effect of seasonal variations, evisceration line effects, and other non-quantifiable variables. Even if all plants are considered collectively, FSIS is still missing microbial data for the entire year. According to officials, from January through April, no data were collected at any of the pilot plants. Moreover, according to the experts we consulted, FSIS should have a year’s worth of microbial data from each plant in the pilot to be able to account for seasonal variations.

In discussing these limitations with USDA officials, they explained that the focus of the project is studying an entire system, including all changes made by the plants, and not simply comparing the work of company employees with that of government inspectors. However, we note that the project’s stated objective was to determine whether a modified inspection system is at least as good as the traditional system. In order for such a comparison to be valid, USDA would have to determine that other variables introduced into the slaughter line, such as the addition of rinses or brushes, are not the cause of changes noted after the shift in inspection systems took place.

The pilot project did not address three important features of similar programs in Australia and Canada. Those features would help to ensure that plant personnel are able to discharge their carcass defect detection duties as well as federal inspectors and that plants are capable of continuing to ensure the safety and quality of their products under a modified inspection system. First, the pilot project did not require that, prior to assuming their new duties, plant personnel receive any training and pass competency tests. Second, the pilot project did not require the use of statistical process controls as a means to ensure that plants continually improve their processes under the modified inspection system. Third, the plants continued in the pilot project irrespective of repetitive noncompliance with regulatory requirements.

FSIS did not require that plant employees complete training before assuming the carcass defect detection duties that FSIS inspectors’ performed. Furthermore, FSIS did not establish a way to measure plant employee’s knowledge and competence. FSIS officials told us that training for plant personnel will not be required if the system is adopted nationwide. Plant personnel and FSIS inspectors at each pilot project plant did participate, however, in a transition phase during which they practiced their new roles. FSIS officials told us that the transition phase was also used to explain how and when data were to be collected during the project.

If FSIS implements modified inspections nationwide, plant personnel will be assuming carcass defect detection duties previously performed by FSIS inspectors who have received formal training on how to identify carcass defects and determine whether they are suitable for human consumption. However, unless FSIS institutes training requirements, plant personnel will not have received comparable training. We asked plant managers participating in the pilot project about the training that they have provided for plant employees. They replied that they provided various levels of training. Individual plants have sole discretion for determining the type of training provided for their employees as FSIS does not approve or monitor plants’ training programs. According to plant officials, FSIS is supportive but does not endorse a training program that the International HACCP Alliance developed for plant personnel participating in this pilot project.7 In some instances, the Alliance, not FSIS, has accredited plants’ training programs but it does not certify trainees who complete these courses.

Seven of the 14 pilot project managers whom we interviewed support a training program for plant personnel. About two-thirds of the managers also were in favor of developing a joint training program between industry and FSIS. At nine pilot project plants, plant employees were required to pass a test before being assigned to their new jobs, and seven plants adopted some kind of continuous on-the-job training. The managers told us that classroom training for plant employees ranged from 4 to 80 hours. The duration of on-the-job training ranged from 8 to 160 hours.

In contrast, the Australian modified inspection program requires plant personnel to complete about 600 hours of classroom training as a prerequisite to participation in the modified inspection program. Plant personnel must also pass a competency test and be certified. Upon completion of the on-the-job training, state authorities license the plant personnel. The Australian government’s veterinarian-in-charge at each plant is authorized to withdraw or deny license renewal of plant personnel on the basis of performance.

Similarly, under Canada’s modified inspection program, government inspectors train plant personnel. The required training includes standardized classroom training, in-plant training, a final examination, and a demonstration of competency on verification tests and finished product standards. At the completion of the training, the Canadian Food Inspection Agency accredits plant employees. Plant employees must be reaccredited every 6 months, and the agency’s inspectors reserve the right to conduct additional reaccreditation of plant employees at any time. Sixty-seven percent of the FSIS inspectors and veterinarians whom we surveyed believe FSIS should develop a training program that includes mandatory testing and certification of plant employees.

Although FSIS has encouraged plants in the pilot project to use statistical process controls, it is not a requirement of the pilot. However, threequarters of the inspectors and veterinarians responding to our survey indicated that their plants use statistical process controls. About a third of these respondents believe that the use of statistical process controls have the effect of making food safety and quality better, while a quarter of them believe that it does not make a difference. At the pilot project plants, FSIS inspectors do not review or analyze plant data using statistical process controls. However, by applying process controls, FSIS could obtain a more comprehensive assessment of the effectiveness of the modified inspection system because it could review the performance of an individual plant’s data to determine whether the plant’s systems are continually improving over time.

In contrast, the Australian and Canadian food inspection agencies use statistical process control methods to audit a plant’s performance over time. This method entails verification by the plant’s management that its production process is stable, that it is capable of producing products that meet performance standards, that it takes appropriate actions if changes occur in the process before it results in unacceptable products, and that it takes actions to continuously improve its process.

Project Does Not Fully Take Into Account Plants’ Repeated Instances of Noncompliance With Regulatory Requirements

FSIS allowed all plants to continue participating in the pilot project even though several had multiple instances of noncompliance with regulatory requirements for fecal contamination.8 Our analysis of noncompliance records issued by FSIS inspectors at 11 chicken pilot plants showed an increased number of such records after the plants shifted to modified inspections for fecal noncompliance. In two instances, however, FSIS took additional action and notified plants of problems with repeated noncompliance. FSIS officials told us that they are now considering how many repeated instances of noncompliance will result in further regulatory action.

Figure 3: Fecal Noncompliance Records Under Traditional and Modified Inspections

Note: Data are provided for a 12-month period before and after the pilot project’s implementation. However, at some plants, data reported after the implementation of the pilot project covered periods of less than 12 months, data for plant 7 cover 8 months, data for plants 1 and 10 cover 10 months, and data for plant 9 cover 11 months. It is likely that a full twelve months of data for these plants would result in an even higher number of noncompliance records.

Source: USDA, FSIS.

As figure 3 shows, the number of noncompliance records resulting from fecal material increased significantly at several plants after they shifted from traditional inspections to modified inspections. At one plant, FSIS inspectors issued 469 fecal material noncompliance records during a 10-month period after the plant shifted to modified inspections. Many noncompliance records reflect multiple violations recorded within the same record. At one plant, FSIS inspectors reported that over 90 percent of the fecal material noncompliance records reflected repetitive failures.

FSIS officials acknowledge the rise in the number of noncompliance records and attribute it to the fact that at pilot project plants, FSIS inspectors have more time to verify compliance with regulatory requirements and to document instances of noncompliance when they occur. The officials also point out that, as part of the their verification duties, FSIS inspectors now examine 80 carcasses per line per shift compared with 20 carcasses under the traditional inspection system.9 In addition, unscheduled verification tests of additional carcasses are conducted under each inspection system. Officials also told us that staffing shortages existed under traditional inspections and that in some instances, it was not possible to complete verification activities on 20 carcasses for each line. Even accounting for the fourfold increase in the number of verification checks taking place under the modified inspections system, one plant experienced a more than twenty-fold increase in the number of noncompliance records for fecal contamination. At other plants, however, the number of noncompliance records did not even double.

Another factor may also account for the increases in noncompliance records issued for fecal failures. Although inspectors and veterinarians working at pilot project plants who responded to our survey were not comparing line speeds to those of the traditional inspection system, about 60 percent of them think that line speeds are too fast under modified inspections to ensure product safety. For example, one inspector stated that fast line speeds cause inexperienced employees to miss carcasses with visible fecal contamination. FSIS does not regulate line speeds under this pilot. However, under traditional inspections, each inspector is responsible for about 35 carcasses per inspector per minute. Canadian inspectors told us that line speeds need to be regulated under modified inspection systems.

In contrast to FSIS, the Canadian Food Inspection Agency decided to proceed with a phased approach to nationwide implementation of Canada’s modified inspection program, the Modernized Poultry Inspection Program. Under this approach, poultry plants must pass a preliminary assessment before being accepted into the program and are required to pass additional audits as they move from one phase of the program to the next. Producers that choose not to participate in or do not qualify for a partially modified or fully modified inspection system are inspected under the traditional system.

Pilot Project Results Are Inconclusive, but FSIS Inspection Personnel Generally Support Modified Inspections

Notwithstanding the design and methodology limitations, we found that data from the chicken pilot that RTI and FSIS inspectors collected show somewhat different results and do not provide a conclusive basis for FSIS to make a decision regarding the merits of one inspection system compared with another. However, 71 percent of the FSIS inspectors and veterinarians that we surveyed believe that product safety is equal to or somewhat better under modified inspections. Some of them commented that under the modified system, they are able to oversee the entire slaughter line and have more time to collect carcass samples for detailed examination. However, they also raised concerns about the modified inspection system.

Contractor’s Data Do Not Conclusively Show That Modified Inspection Systems at Chicken Pilot Plants Performed “As Well As or Better Than” Traditional Inspection Systems

RTI’s data do not conclusively show that inspection systems at chicken plants perform as well as or better than they did after they switched to modified inspections. Under modified inspections, none of the inspection system at these plants met all of the performance standards that FSIS developed to measure their performance. Table 3 shows the performance standards that the inspection systems at these plants were expected to meet after shifting to modified inspections.10

Table 3: Performance Standards for Inspection Systems at Chicken Plants in the Pilot Project

Data from the inspection systems at the 11 chicken plants whose performance RTI measured under both inspection systems show that, while the majority of the systems met four or more of the performance standards for food safety and quality under modified inspections, none met all seven of the standards.11 The data show that 10 of the 11 plants’ inspection systems met the food safety standard for septicemia and/or toxemia, but only 1 of the 11 systems met the zero-tolerance food safety standard for visible fecal material, which could contain harmful bacteria such as E. coli. Most plants’ inspection systems, however, were also unable to meet the fecal material standard under the traditional inspection system. All 11 systems at these plants met the quality standard for bruises and sores (OCP2), but only 3 met the quality standard relating to the presence of feathers (OCP4). According to FSIS officials, carcasses with product quality defects are of a lesser concern than those with food safety defects because they are not associated with foodborne illness and quality defects can generally be removed from the carcass after it leaves the slaughter line and moves to the cut-up and processing area.

Three of the 11 plants’ inspection systems met six of the seven standards, 3 plants met five of the standards, 4 plants met four standards, and 1 plant met two standards. Table 4 shows which systems met the performance standards that FSIS set for the pilot project.

Table 4: Eleven Chicken Plants’ Inspection Systems Compliance With Performance Standards (RTI’s Data)

As discussed earlier, under its contract with FSIS, RTI collected microbial data as well. Specifically, it analyzed samples for the presence of Salmonella and generic E. coli. The results show that 9 of the 11 chicken plants’ inspection systems met the performance standard for Salmonella. Also, about 80 percent of the samples collected at the11 plants were within the acceptable ranges that FSIS specifies for generic E. coli.

Irrespective of whether plants’ inspection systems met the performance standards, we analyzed these data to determine if, under modified inspections, the systems performed at least as well as they did under traditional inspections. Table 5 summarizes the results of this analysis.

Table 5: Eleven Chicken Plants’ Inspection Systems Performance After Shifting to Modified Inspections (RTI’s Data)

The data show that some of the plants’ inspection systems improved or remained unchanged in certain categories such as the food safety standard for septicemia and/or toxemia. On the other hand, the systems’ ability to meet the quality performance standards deteriorated over time. In general, most measures of performance showed deterioration in one or more plants. In fact, every inspection system showed deterioration for at least one of the nine standards measured. At 5 of the 11 plants, results for Salmonella were worse under modified inspections than they were under the traditional inspection system—only 2 improved their performance in this category. FSIS has stated that any new inspection system should perform at least as well as the current system of inspections.

FSIS’ Data Do Not Conclusively Show That Plants’ Inspection Systems Perform “As Well As or Better” Under Modified Inspections

FSIS’ data for 10 chicken plants in the pilot project also show that none of their inspection systems met all seven standards.12 In contrast to the RTI data, however, the FSIS data show that 7 of the 10 systems met six of seven standards, and the remaining 3 plants met five of the seven standards. FSIS officials stated that the differences in results might be caused, at least in part, by differences in data collection. As part of FSIS inspectors’ verification duties at the chicken plants participating in the pilot project, the inspectors sample and analyze carcasses to determine whether they meet performance standards. Inspectors are not required to collect and analyze microbial samples, as RTI was required to do for the pilot project. The data collected cover a longer period than RTI’s and include much larger samples per plant. The FSIS officials acknowledge, however, that disparities in the FSIS and RTI data for some performance standards, such as dressing defects (e.g., feathers and oil glands), suggest a need for further investigation.

The FSIS data show that 7 of the 10 chicken plants’ inspection systems met six of the seven performance standards, but none of the plants met the zero-tolerance standard for fecal material over time. It should be noted, however, that the actual performance of the 12th position for this standard was 1.5 percent, not zero as required by FSIS’ performance standard. So noncompliance with this standard may not provide a basis for comparing performance of the modified with traditional inspection systems. After shifting to modified inspections, each of the 10 plants’ systems had reached a performance that was better than 1.5 percent for this standard. We analyzed data on individual inspection systems’ performance from the time that each plant implemented modified inspections through June of 2001. Table 6 shows a summary of the inspection systems’ performance at individual plants.

Table 6: Ten Chicken Plants’ Inspection Systems Compliance With Performance Standards (FSIS’ Data)

Table 6 shows which plants’ inspection systems did not meet specific performance standards. That is, plant 6 did not meet the food safety standard for infectious conditions (e.g., septicemia and/or toxemia); plant 8 did not meet the food quality standard for animal diseases such as airsaculitis; and plant 2 did not meet the food quality standard for digestive content contamination (ingesta). At the pilot plants, FSIS inspectors are issuing noncompliance records when plants do not meet the performance standards; however, FSIS officials have not yet decided how many instances of noncompliance will be tolerated before the agency can decide to take further action to ensure regulatory compliance.

We reviewed data collected by FSIS inspectors after the plants shifted to modified inspections during two different time periods. In comparing these two periods, we found that in several categories, the inspection system’s performance at each plant deteriorated over time. Table 7 summarizes the changes for these 10 plants.

Table 7: Changes in Chicken Plants’ Inspection Systems Performance Over Two Time Periods After Shifting to Modified Inspections (FSIS’ Data)

On the basis of this information, in addition to providing plant managers with feedback on a daily basis, FSIS inspectors could identify plants whose inspection systems need to improve. In addition, FSIS could aggregate data from these plants and use this type of analysis to evaluate whether a specific performance standard needs to be revised. For example, we made two observations by analyzing the data as follows:

• As shown in table 7, 9 of the 10 plants’ performance under the OCP4 standard deteriorated over time. While this standard permits 80 percent of carcasses per shift to have this defect and still meet the standard, this trend may suggest the need for FSIS to investigate why the inspection system’s performance at an individual plant is deteriorating.

• As shown in table 7, 9 of the 10 plants maintained their performance and one improved under the FS2 standard.

Australian officials told us that under their modified inspection system, the Australian Quarantine and Inspection Service requires plants to use process controls to demonstrate continuous improvement in plant sanitation, microbial pathogen reduction, and quality defect reduction. The agency uses these data to monitor plant performance and enforce compliance with standards.

Most FSIS Inspectors and Veterinarians Believe That Modified Inspections Are Equal to or Better Than Traditional Inspections for Ensuring Product Safety

About 70 percent of the respondents to our survey of FSIS inspectors and veterinarians believe that modified inspections are equal to or somewhat better than traditional inspections for product safety while more than half believe that modified inspections are the same as or better than traditional inspections for product quality. Table 8 summarizes inspectors’ and veterinarians’ responses to major topics covered by our survey. (See app. IV for further detail on the survey responses.)

Table 8: Summary of Responses to GAO’s Survey of USDA Inspectors and Veterinarians

aThis change was experienced by 75 percent of respondents at their plants after modified inspections were implemented.
bSome inspectors cited more time and freedom to find defects and write noncompliance records and more carcasses are sampled under modified inspections.
Note: Because of rounding, numbers may not add up to 100 percent.

Nearly all of the inspectors and veterinarians provided additional written comments on several issues of concern and made suggestions for improvement. The inspectors commented that under the modified system, they are able to oversee the entire slaughter line and have more time to collect carcass samples for detailed examination. In addition, 27 of the 210 inspectors and veterinarians commented that they were concerned about regulatory noncompliance and enforcement issues. Particularly noteworthy are comments that indicate that stronger measures should be in place to reduce repetitive instances of fecal material noncompliance and that inspectors should have more authority to hold a plant accountable for multiple and/or repetitive problems. In addition, four inspectors would like each failure of the zero-tolerance standard for fecal material to be documented in separate noncompliance records rather than recording multiple instances in a single record, which may obscure the frequency of the problem. Thirty-nine inspectors and veterinarians also raised concerns regarding product quality. In addition, seven inspectors stated that the quality performance standards need to be tightened so that fewer defects are overlooked.

In addition, 30 inspectors and veterinarians commented that they are concerned that the zero-tolerance standard for fecal material is not adequately enforced, especially since the carcass inspector examines only the back of the carcass and does not look inside the cavity. Seven suggested that FSIS place a mirror or provide some other mechanism to facilitate observation by the carcass inspector.

Finally, 62 inspectors commented that working conditions under the modified inspection system are better because the inspectors no longer need to make repetitive motions, as they did when they continuously examined each carcass on the slaughter line by touch. Sixteen inspectors also said that they like the ability to physically move from their location at least once every hour. Twenty-five of them mentioned, however, that the carcass inspectors’ location might expose them to agents, such as chlorine or trisodium phosphate that could adversely affect their health.

We believe that a risk-based inspection system—such as the one that USDA is pilot-testing at chicken plants and is starting at hog plants—has merit in concept and is consistent with the existing risk-based framework for HACCP. However, while we support this approach, we believe that the design of this pilot will not permit USDA to reach conclusions about whether the new system of modified inspections performs as well as the traditional system.

If, as planned, USDA undertakes to modify its current inspection system for chicken plants, it will need to consider several factors that we discuss in this report and that are also key features of the Australian and Canadian modified inspection systems. First, if USDA issues regulations that require all chicken plants to modify their inspection systems concurrently without first determining whether individual plants are able to meet standards, it may risk including those plants with repeated records of noncompliance. Such plants may have difficulty meeting their new responsibilities. Continued participation in a modified inspection system should depend on the plants’ ability to maintain good performance. Second, adding a requirement that plants use statistical process control systems to identify variations in performance will allow the plants to better manage and control their production processes and will also allow USDA to appropriately monitor and verify inspection systems’ performance at each plant over time. Third, without requiring that plant personnel receive adequate training to undertake carcass defect detection responsibilities, USDA may jeopardize product safety and quality. At the very least, USDA should ensure that personnel conducting carcass defect detection duties have knowledge and training comparable to that of USDA inspectors who were previously responsible for these duties. Last, inspectors and veterinarians responding to our survey provided several useful comments on how to enhance a modified inspection system. For example, they believe that USDA needs to address multiple instances of plant noncompliance with regulatory requirements. As discussed in this report, many plants had repeated instances of noncompliance with a critical food safety performance standard.

In addition to the current legal challenges, USDA faces future potential legal challenges unless its statutory inspection authorities are revised. These challenges will likely hinder the Department’s objective of modifying its inspection system at meat and poultry slaughter plants. We continue to believe that, as we have recommended before, the Congress should consider revising the Meat and Poultry Acts, 21 U.S.C. sections 604 and 455, to provide FSIS with the flexibility and discretion to target its inspection resources for the most serious food safety risks. Such revisions would eliminate the requirements that USDA has traditionally implemented through continuous carcass-by-carcass government inspection and replace them with a risk-based inspection system that includes government oversight and verification.

If USDA decides to implement modifications to its inspection system, we recommend that the Secretary of Agriculture direct FSIS to

We further recommend that, if in addition to the current pilot project for chickens, USDA decides to conduct similar pilots for other species—hogs, turkeys, or cattle—the Department take steps to ensure that the pilot’s design and methodology are sufficiently rigorous to allow more valid conclusions than those in this chicken pilot.

We provided USDA with a draft of this report for review and comment. In written comments, USDA said that the report’s recommendations were appropriate and that the agency will address them when it seeks public comment prior to making any regulatory changes. USDA provided technical comments, which we incorporated in the report as appropriate. USDA’s comments and our responses are contained in appendix VII.

We performed our review from December 2000 through October 2001, in accordance with generally accepted government auditing standards.

As agreed with your offices, unless you publicly announce its contents earlier, we plan no further distribution of this report until 30 days from the date of this letter. At that time, we will send copies to congressional committees with jurisdiction over food safety issues; the embassies of Australia and Canada; the Secretary of Agriculture; the Director, Office of Management and Budget; and other interested parties. We will make copies available to others on request.

If you have any questions about this report, please contact me at (202) 512-3841. Major contributors to this report are listed in appendix VIII.