Development of the questionnaires
The development of the psychosocial questionnaires for food safety began with an examination of published studies and behavior change theories to identify the psychosocial factors associated with a wide array of personal health choices, including food safety practices.
A panel of experts (n= 8) in food safety and health behavior change, tests and measurements, and psychology identified 3 psychosocial factors (that is, beliefs, locus of control, and self-efficacy) as being key to understanding current food safety practices and creating interventions to change practices.
Food safety beliefs Health beliefs appear to influence health practices ranging from smoking, contraceptive use, exercising, compliance with medical regimens, and healthy eating (Conner and Norman 2005). Only a few studies investigating food safety beliefs have been published (Unklesbay and others 1998; Medeiros and others 2004), so the food safety belief questionnaire needed to be almost completely developed de novo. A review of key cognitive-behavioral theories and previously reported health behavior studies that employed belief scales served as the starting point for development of the beliefs questionnaire (Fein and others 1995; Byrd-Bredbenner 2000, 2004; Finckenor and Byrd-Bredbenner 2000).
Guidelines for the development of belief scales were carefully followed (Linn and Gronlund 2000; Torabi and others 2001; DeVellis 2003). Development of the scales began with identification of the belief constructs needed to provide insight into how food safety behavior change programs should be framed to evoke improved behaviors. The 3 constructs initially identified were personal interest in learning about food safety, importance of cleanliness/sanitation to the participant, and perceptions of the prevalence (threat) of food poisoning and one's susceptibility to food poisoning.
Belief scales typically take the form of semantic differential, Guttman, Thurstone, and Likert scales (Hopkins 1998; DeVellis 2003; Thorndike 2005). A Likert-type format with 5-response choices (strongly agree, agree, uncertain, disagree, strongly disagree) was selected because this format provides flexibility in item construction, has a familiar format, provides a sufficient degree of accuracy and discrimination ability, can be completed rapidly, was most suitable to the constructs being measured, and is relatively easy to construct and score (Labovitz 1970; Tiku 1971; Cox 1980; Hopkins 1998).
Next, a bank of 10 to 24 belief statements for each construct was constructed by revising items from the literature and/or creating items de novo. To prevent a response set, some statements were worded positively and others were worded negatively. All items were reviewed by the panel of experts for clarity and being representative of the construct measured (that is, content validity), then refined.
The statements were pretested with a group of young adults from a variety of majors that were enrolled in freshman/sophomore level courses at the same northeastern university (n= 180) during spring semester 2004. Participants responded to each belief statement by indicating their agreement/disagreement with, or uncertainty about, the statement. The statements reflecting each construct were mixed throughout the pretest to prevent a response set. For each statement, a score of 5, 4, 3, 2, or 1 was assigned to responses of strongly agree, agree, uncertain, disagree, or strongly disagree, respectively, for positively worded statements. The scoring was reversed for negatively worded statements. An overall mean score was computed for each scale by summing the score of the items in the scale and dividing by the number of items in the scale. Thus, mean scale scores ranged from 5 (strongly positive) to 1 (strongly negative).
Pretest data were subjected to exploratory principal components factor analysis using an orthotran/varimax rotation method to refine and shorten the scales, establish the unidimensionality of the statements reflecting each construct, and compute a preliminary internal consistency measure for each scale (that is, Cronbach alpha coefficient) (Norman and Streiner 2000). At the beginning of factor analysis procedures, all 56 belief statements were included. Oblique solution factor loadings for all statements were examined and 1 statement at a time was removed from the analysis in an iterative fashion. A statement was selected for elimination based on the strength/weakness of its commonality and factor loading, whether it was the lone item loading on a factor, and/or whether its strongest factor loading made contextual sense in relation to the content of the other items loading strongly on that factor. This process was systematically repeated until the remaining statements (n= 29; 5 to 11 per construct) made contextual sense with the other scale items and generated acceptable Cronbach alpha coefficients (range was 0.72 to 0.87).
The exploratory factor analysis indicated that the importance of cleanliness/sanitation to the participant and personal interest in learning about food safety constructs were indeed individual, unique constructs. However, the perceptions of the prevalence (threat) of food poisoning and one's susceptibility to food poisoning construct appeared to be 3 rather than 1, that is, personal susceptibility to food poisoning, threat of food poisoning in the United States, and personal threat of food poisoning. As a result of the division of this single construct into three, 4 additional statements were written to supplement items in these scales. The panel of experts reviewed the 33 statements in the revised scales to confirm their content and construct validity.
To further refine the belief scales, a pilot test was conducted with a 2nd sample of young adults from a variety of majors that were enrolled in freshman/sophomore level courses at a northeastern university (n= 77) during fall semester 2004, none of whom participated in the pretest. The pilot test results were subjected to the same type of factor analysis procedures used with the pretest and items were eliminated in a similar fashion. The food safety beliefs instrument had 28 items (4 to 10 per scale) with acceptable Cronbach alpha coefficients (range was 0.69 to 0.92). The belief scale items corresponded to an eighth-grade reading level according to the Flesch-Kincaid Grade Level score. The panel of experts once again reviewed the items and confirmed the content validity of the scales.
Food safety locus of control Locus of control refers to the extent to which people view the attainment of a particular outcome as being either within their control, where their action determines the outcome (internal), or outside their control, where outcome is controlled by forces of powerful others (external) or by luck or chance (AbuSabha and Achterberg 1997; Brown 1999). Locus of control is a construct that may assist health educators in understanding learners' beliefs regarding their personal responsibility and ability to influence their own health (Brown 1999). For instance, some studies indicate that a strong internal locus of control is associated with greater long-term weight management success (Senekal and others 1999) and dietary healthfulness in pregnant women (Saturnino-Springer and others 1994). In addition, those with hypertension and type 2 diabetes mellitus have been found to have a higher external locus of control than those with no history of these diseases (Plescia and Groblewski 2004). Other studies, however, have found no difference in weight loss or dietary healthfulness between individuals with an internal compared with external locus of control (Nir and Neumann 1991; Murphy and others 2001).
Although there are various domains within locus of control, such as the health domain (Wallston and Wallston 1978), social domain, nutrition domain, and weight domain (Stotland and Zuroff 1990), a review of the literature did not reveal that a food safety domain had been elucidated. Thus, because of the potential importance of the locus of control construct on food safety behaviors, this domain was defined and scales created.
Food safety locus of control was defined as the degree to which an individual believes food safety (avoidance of food poisoning) is controlled by internal factors (that is, largely under a person's own control) or external factors (that is, largely under the control of powerful others or determined by luck or chance factors) (AbuSabha and Achterberg 1997; Brown 1999). The Health Locus of Control Questionnaire (Wallston and others 1976; Wallston and Wallston 1978) was modified for use in this study (see Table 1). There were 6 Likert-type items in each of 3 scales (internal, external: powerful others, external: chance). Each of the 18 items in this questionnaire was scored from 1 to 6 (strongly disagree, disagree, slightly disagree, slightly agree, agree, and strongly agree). A total score for each scale was computed by summing the score of each item and dividing by the number of items on the scale, so scale scores could range from 1 to 6.
Table 1—. Food safety locus of controla final factor loadings and item convergence values
|Internal locus of control scale item (Cronbach alpha = 0.63)|
| If I only eat food prepared in a sanitary manner, I can keep from getting sick.||c||c|
| If I am careful about the food I eat, I can avoid food poisoning.||c||c|
| My physical well-being depends on how well I take care of myself.b||c||c|
| It is my own behavior that determines whether I get food poisoning.||0.421||0.737|
| If I get food poisoning I am to blame.||0.456||0.792|
| I am directly responsible for protecting myself from food poisoning.||0.440||0.745|
|External locus of control scale items (Cronbach alpha = 0.63)|
| I can only maintain my health by consulting health professionals.b||0.697||0.325|
| If I see an excellent doctor regularly, I am less like to have health problems.b||0.768||0.708|
| Other people play a big part in whether I stay healthy or get food poisoning.||c||c|
| Health professionals keep me healthy.b||0.746||0.774|
| The type of care I get from others determines how well I recover from an illness like food poisoning.||c||c|
| Following doctor's orders exactly is the best way for me to stay healthy.||0.723||0.722|
|Chance locus of control scale items (Cronbach alpha = 0.59)|
| When I stay healthy, I'm just plain lucky.b||0.400||0.706|
| No matter what I do, if I am going to get food poisoning I will get food poisoning.||0.681||0.688|
| Most things that affect my health, like food poisoning, happen to me by accident.||0.575||0.427|
| Luck plays a big part in determining how soon I will recover from an illness.b||0.639||0.669|
| Even when I take care of myself, it is easy to get food poisoning.||c||c|
| If it 's meant to be, I will stay healthy.b||0.310||0.559|
The modified questionnaire was reviewed by the panel of experts for content validity. The questionnaire was pretested with the same undergraduate students who participated in the pretest of the beliefs questionnaire (n= 180). The statements from each scale were mixed throughout the pretest to prevent a response set. The pretest findings revealed that the locus of control scales generated Cronbach alpha coefficients of 0.69, 0.53, and 0.59 for the internal, external: powerful others, and external: chance scales, respectively. While the Cronbach alphas are not as high as generally desired, they are in keeping with typical administrations of this instrument as reported by its developer (Wallston 1993). Minor editorial changes were made to increase the clarity of items. The Flesch-Kincaid reading level for the locus of control scales was determined to be at the ninth-grade level.
Food safety self-efficacy Self-efficacy, a psychosocial construct originally developed by Bandura (1977), appears to be a powerful predictor of behavior change (McAuley and Blissmer 2000; Rejeski and others 2003; Hallam and Petosa 2004) in some cases. Self-efficacy is an individual's confidence in his or her ability to perform a particular recommended health behavior (for example, exercising) or abstain from an unhealthy behavior such as smoking (Healey and Thombs 1997). Perceived self-efficacy is thought to influence which health behaviors will be initiated, the degree of effort expended, and the persistence of the behavior (Lev and Owen 1996; Lev and others 1999; Etter and others 2000). Self-efficacy is often measured in specific environmental contexts or conditions, such as when one is at a party, in a hurry, or has had a stressful day (Ounpuu and others 1999).
Many studies have documented clear associations between level of self-efficacy and other psychosocial constructs such as beliefs, attitudes, stage of change, and social norms for certain health behaviors (Campbell and others 1998). High self-efficacy scores frequently are associated with the more advanced stage of change status and greater readiness to change. For example, significant increases in self-efficacy scores were observed as stage of change increased for exercise, sun protection, smoking, dietary fat intake, and fruit and vegetable intake (Herrick and others 1997; Campbell and others 1998; Ounpuu and others 1999). (Stage of change is a construct of the Transtheoretical Model that describes behavior change as a process in which an individual moves through a series of 5 distinct stages of change: precontemplation, contemplation, preparation, action, and maintenance [Prochaska and DiClemente 1984; Prochaska and others 2002]).
No questionnaire measuring self-efficacy related to safe food handling could be located, so food safety self-efficacy was defined (that is, an individual's confidence in his or her ability to perform specific recommended food handling [food poisoning prevention] behaviors) and some environmental contexts identified (that is, time pressure, hunger status). Then, an item pool was constructed using self-efficacy measures from previous nutrition and health studies as a guide (Lev and Owen 1996; Lorig and others 1996; Healey and Thombs 1997; Lev and others 1999; Etter and others 2000; Byrd-Bredbenner 2004).
The structure of self-efficacy items generally takes 1 of 2 Likert-type forms (Lev and Owen 1996; Sallis and others 1988). In the 1st form, participants are asked to indicate their agreement/disagreement/uncertainty with statements like “I have confidence in my ability to prepare food in a sanitary way” and “I have confidence in my ability to keep my food safe to eat.” In the 2nd form, participants are asked to rate how confident they are (that is, I am sure I could do it, I could do it, I do not know if I could do it, I could not do it, and I am sure I could not do it) that they could motivate themselves consistently for at least 6 mo to perform tasks like “prepare food in a sanitary way” and “store food at an appropriate temperature so it is safe to eat.” The panel of experts selected the 2nd form because it was more succinct (that is, it had a single stem “Please rate how confident you are that you could really motivate yourself to do things like these consistently for at least 6 mo” followed by a list of behaviors), easily incorporated a specific time frame (that is, 6 mo) within the stem, and permitted answer choices that were more action related.
The pool of self-efficacy items was reviewed by the panel of experts for clarity, comprehensiveness, and contextual value, and to establish content validity. A total of 33 self-efficacy items was pretested with the same undergraduate students who participated in the pretest of the beliefs and locus of control questionnaires (n= 180). The items were scored 5, 4, 3, 2, or 1 for the responses noted above (that is, I am sure I could do it, and so on). Total self-efficacy score was computed by summing the score of each item and dividing by the total number of items on the scale. Thus, higher scores reflect greater self-efficacy.
The pretest findings revealed that the self-efficacy questionnaire generated a very strong Cronbach alpha coefficient (that is, 0.91). To reduce participant burden and minimize item duplication, the expert panel identified 9 items to eliminate. In addition, minor editorial changes were made to increase clarity.
To further explore validity of this measure, mean self-efficacy score was compared using analysis of variance with participants' self-identified stage of change related to the way they prepare food in terms of food safety. No studies could be located that used the concept of stage of change for food safety behaviors; thus, instruments used in previously reported health behavior research were modified to assess food safety stage of change (Glanz and others 1994; Greene and others 1994; Prochaska and others 1994; Read and others 1996; Auld and others 1997; Browne 2003). Participants were asked to identify which of these 5 statements best described them: (a) I have no intention of changing the way I prepare food to make it safer to eat in the next 6 mo; (b) I am aware that I may need to change the way I prepare food to make it safer to eat and I am seriously thinking about changing my food preparation methods in the next 6 mo; (c) I am aware that I may need to change the way I prepare food to make it safer to eat and I am seriously thinking about changing my food preparation methods in the next 30 d; (d) I have changed the way I prepare food to make it safer to eat, but I have been doing so for less than the past 6 mo; and (e) I have changed the way I prepare food to make it safer to eat and I have been doing so for more than the past 6 mo. As reported by other health behavior change researchers (Herrick and others 1997; Campbell and others 1998; Ounpuu and others 1999), self-efficacy score tended to increase significantly as stage increased.
The shortened questionnaire (24 items) was pilot tested with the same group of undergraduates (n= 77) that pilot tested the belief questionnaire described above. The Cronbach alpha coefficient for the shortened self-efficacy questionnaire was 0.90. The Flesch-Kincaid Reading Grade Level score was grade 10.
College and university instructors from across the United States were invited via email to recruit students in their introductory-level general education courses to complete an online food safety survey. Most participating instructors awarded extra credit points to students who completed the survey. From January to October 2005, a total of 4548 young adults enrolled at 21 colleges and universities completed the knowledge questionnaire as part of a larger nationwide survey of young adults (participating colleges and universities were located in these states: Arizona, California (3 schools), Florida (2 schools), Georgia, Indiana, Kansas, Montana, Louisiana, Minnesota, Missouri, New Jersey (2 schools), North Carolina, North Dakota, Oklahoma, Pennsylvania, South Carolina, and Texas). The colleges and universities were located throughout the United States and included those ranging from community college to research institutions. Participants who were older than the age criterion set in this study for young adults (that is, 17 to 26 y) were eliminated from the analysis; the final sample size was 4343.