Food Protection Trends - January/February 2023 - 25

each type of training. Each plant was randomly assigned
one training topic as the focus of the evaluations in an effort
reduce the extra workload assigned to each facility and the
time taken during the training process.
Employees at each plant were offered a pre- and posttest
survey, which they were instructed to place in a sealed " ballot
box " created by the human resources department. The pretest
survey contained the behavioral intentions instrument, and
the posttest survey contained both the behavioral intentions
instrument and demographic questions. After training at each
location had been completed, the human resources manager
shipped the box to researchers at North Carolina State
University without breaking the seal on the box. All data
collection methods and data collection tools were approved
by the North Carolina State University Institutional Review
Board before the data were collected (protocol 9494).
Pilot testing
The survey instrument was tested during a pilot study
before it was used to collect data. The pilot test was
performed at a single processing plant. Data were collected
from 20 employees using an online form of the survey.
Only the allergen control programmed instruction training
delivery was evaluated.
Population and sample validation
Because we could not collect from every processing
location or every employee from responding processing
locations, a regression analysis was performed with response
rate and third party (Safe Quality Foods Institute) audit
scores to identify any differences in audit scores between
plants that responded and those that did not respond. The
presence of such differences could indicate that the data
from responding plants were skewed and not representative.
A logistic regression analysis was performed to identify any
differences between the third party audit scores of plants that
completed and returned supervisor audits and those that did
not. No significant difference was found between the audit
scores of plants that submitted supervisor audits and those
that did not (P = 0.372) (Table 1).
Data analysis
All pretest IBM construct questions were tested for
reliability and validity by calculating the Cronbach's alpha
scores and the principal components scores.
The trainings were implemented at all 66 of the dairy
company's processing plants. Evaluations were assigned and
sent to each location. Only 33% (N = 22) of all plants mailed
in completed employee surveys (Table 1). When considering
the number of workers employed at responding plants (N
= 3,620), the total response rate of completed surveys was
21.9% (N = 793). Also, 21.2% (N = 14) of all plants responded
with completed supervisor audit data (Table 1).
Pre- and posttest knowledge quizzes were sorted to
eliminate those that were < 50% complete. Scores of
quizzes with ≤ 50% completion were eliminated to control
for inaccurate data due to skipped questions while also
allowing for the possibility that some questions may be
skipped due to lack of knowledge. For each responding
plant, posttest noncompliance results were subtracted
from pretest noncompliance results to determine the total
change in noncompliance.
TABLE 1. Response rate for training evaluation surveys and audit forms
Training topic
Introduction
to good
manufacturing
practices and
material controls
Personal
responsibility and
communicable
diseases
Allergen control
Food defense
Total
Number
of plants
contacted
Number
of plants
responded
20
9
Plant
response
rate (%)
Estimated
employees
contacted
Employee
surveys
received
Estimated
response
rate (%)
Supervisor
audits
received
Supervisor
audit
response
rate (%)
45.0
1,742
380
21.8
5
25.0
19
13
14
66
6
3
4
22
31.6
23.1
28.6
33.3
791
224
863
3,620
165
99
224
793
20.9
44.2
17.3
21.9
3
4
2
14
15.8
30.8
14.3
21.2
January/February Food Protection Trends
25
Food Protection Trends - January/February 2023

Table of Contents for the Digital Edition of Food Protection Trends - January/February 2023
