Food Protection Trends - April 2010 - 215

TABLE 2. Risk scoring grid Severity Frequency Low High Low 1 2 High 3 4

RESULTS
Descriptive analysis
Experts identified 15 food safety hazards that pose microbiological, chemical, and/or physical safety hazards to food during transportation and warehousing or storage (Table 3). The panel also identified 11 food product sectors to be considered when assessing the frequency and severity of these food safety hazards. In addition, the experts collectively identified the following modes of transport, as well as storage/warehousing, as having distinct risk rankings for food safety hazards: truck; rail; water; air; and intermodal. Different types of water transportation (e.g., deep sea freight versus inland water freight) were not considered separately, because of experts’ suggestions that food safety hazards are not related specifically to the type of water transportation. The total number of food safety hazards scored by panel members across food product sectors and transport modes substantially increased the respondent burden in the third round. An average of 13 out of 15 experts provided risk scores for each of the 990 individual risk rankings requested. Only 7 percent of all problem-sector-mode combinations resulted in an average risk score of 4 (high frequency, high severity) (61%). The majority of problem-sector-mode combinations resulted in average risk scores of 1 (low frequency, low severity) and 3 (low frequency, high severity) (28%). An analysis of the risk score data leads to the following observations: • The top 5 food safety hazards that were the greatest concern across all modes of transportation were: – Lack of security for transportation units or storage facilities, – Improper holding practices for food products awaiting shipment or inspection, – Improper refrigeration or temperature control of food products, – Improper management of transportation units or storage facilities to preclude cross-contamination, and – Improper loading practices, conditions, or equipment.

discussions with industry experts and asked each expert to expand the various lists and provide additional comments on these issues. The objective of Round 2 was to assess the risk posed by each of the fifteen food safety hazards by food sector and transport mode identified in Round 1. Experts were asked to assign a risk score from 1 to 4 based on the hazard’s frequency and severity (Table 2). Thus, each expert first had to assess whether the problem occurred at a high or low frequency for the specified food sector and mode of transport (i.e., how widespread the problem is) and then to evaluate whether the probability that the problem could render the food unsafe was high or low (i.e., assess the severity of potential consequences, such as mortality, morbidity, and economic impacts, of the problem). Panel members were directed to skip questions for which they lacked sufficient knowledge for an informed assessment. The objective of Round 3 was to obtain background information on preventive controls that may eliminate or mitigate the risk of food safety hazards in food transportation and warehousing/ storage from our expert panel. Again, we presented our own findings on preventive controls to the panel and asked them to expand the list. In Round 4, experts identified the set of preventive controls necessary to eliminate or mitigate the risk posed by each of the fifteen food safety hazards. Experts were asked to ensure that the controls had the broadest applicability across all food product sectors and modes of transport. As noted above, we used the Delphi technique to reach consensus for each round of questioning. Iteration rounds helped to stabilize results and increase agreement among participants. At the completion of all rounds, including iteration rounds, we sent each participant a summary of his responses for review

and final confirmation. Because full consensus was not attainable, we relied on accepted aggregation procedures to pool expert estimates, where applicable.

Data analysis
We used Stata (14) to perform descriptive univariate analysis as well as factor analysis on the data collected. For factor analysis, we used Stata’s factor, rotate, and score functions. Factor analysis is a data reduction technique that reduces the number of variables used in an analysis by creating new variables (called factors) that combine redundancy in the data. A factor analysis looks for correlations among the variables and the first step is to determine the number of relevant factors. While Stata’s algorithms used to solve factor analyses include methods of determining an appropriate number of factors, it is also possible to specify (fix) the number of factors in the analysis. For this study, we allowed the algorithms to determine the number of factors and also used judgment in determining the appropriate number of factors. The output from the factor analysis generates a table that relates each variable to each factor and assigns a numerical value between -1 and 1 to each relationship. The numerical values are referred to as factor loadings and reflect the strength of relationship between the factors and the variables. Variables that are closely related to one another should all load highly on the same factor. Stata’s score command produces estimates of these factors, which we used to develop indices of riskiness by food sector. Specifically, the method allowed us to generate an overall risk score for each food sector by mode of transport that combines the information in all of the fifteen food safety hazards, as well as multi-factor risk scores separately by food sector and mode of transport.

APRIL 2010 | FOOD PROTECTION TRENDS

215

Food Protection Trends - April 2010

Table of Contents for the Digital Edition of Food Protection Trends - April 2010