Food Protection Trends - December 2008 - 919

FIGURE 1. Types of products produced by meat and poultry processorsa and evaluated the structure and effectiveness of the questionnaire form. We also tested the questionnaire with personnel at processing plants and with industry trade associations; we subsequently revised the questionnaire based on their suggestions. The survey instrument and study design were approved by the Office of Management and Budget’s (OMBs) information collection clearance process. Survey administration We implemented a variety of procedures aimed at maximizing the response rate to the survey, including many of the procedures recommended by Dillman (11). Various industry trade associations sent e-mails and posted information in their newsletters to encourage their membership to participate in the survey. We administered the survey over a period of approximately 18-weeks from July to November 2005. We contacted plants by telephone to screen for eligibility and to identify the target respondent for the survey, mailed the questionnaire via Federal Express to target respondents, sent a thank-you/reminder postcard, and made a series of follow-up telephone calls to nonrespondents to encourage participation. We received 944 completed surveys; 423 plants were eligible but did not complete the survey (i.e., non-respondents); 183 plants were ineligible (e.g., they conducted slaughter activities or were out of business); and, for 102 plants, we were unable to determine their eligibility for the survey. We calculated weighted response rates (respondents / [nonrespondents + respondents]) by stratum, using the initial sampling weights adjusted for unknown eligibility so that cases with unknown eligibility were distributed between those eligible (non-respondents) and those ineligible in the same proportions that existed among cases with known eligibility. Ineligible plants were excluded from the response rate calculation. The overall weighted response rate was 66%. Response rates were higher for federally inspected plants than for state-inspected plants, and response rates were higher for large and small plants than for very small plants. Analysis procedures Before tabulating the survey data, we conducted data editing and coding. The edited and coded questionnaires were double-keyed for quality control purposes. To prepare the analysis dataset, we systematically reviewed the keyed data to identify and address any inconsistencies and outlying values. We weighted the survey data to reflect the selection probabilities of sampled establishments and to compensate for differential non-response. We performed χ2 tests to determine if differences in responses between HACCP plant sizes were statistically significant (i.e., large versus very small and small versus very small). Additionally, for some analyses, we performed χ2 tests to determine if differences in responses between plant types were statistically significant (i.e., plants producing RTE product versus NRTE product and plants producing RTE product versus inputs to further processing). All analyses were conducted using Stata, a statistical analysis software tool that takes the stratified sample design into consideration when computing variances (16). DECEMBER 2008 | FOOD PROTECTION TRENDS 919

Food Protection Trends - December 2008

Table of Contents for the Digital Edition of Food Protection Trends - December 2008