Development of Food Literacy Assessment Tool (FLitT) targeting low-income adults

Abstract

Objective:

Develop and test validity and reliability of the Food Literacy Assessment Tool (FLitT) in low-income adults.

Design:

Face validity was tested using cognitive interviews, content validity using expert review, and internal consistency reliability and test-retest reliability based on 2 administrations of survey.

Setting:

Urban choice food pantry in Cincinnati, OH.

Participants:

10 and 98 low-income adults for the cognitive interview and survey, respectively

Variables Measured:

Knowledge, self-efficacy, and behavior required to plan and manage, select, prepare, and eat in relation to food.

Analysis:

Cronbach α and Kuder Richardson 20 (KR-20) for internal consistency reliability and Intraclass Correlation Coefficient (ICC) for test-retest reliability.

Results:

Cognitive interviews and expert feedback suggested modifications to improve the FLitT’s clarity and offered more response options. Testing shows acceptable internal consistency in self-efficacy (Cronbach α=.92) and behavior (Cronbach α=.90) but not in knowledge (KR 20=.51). The FLitT shows acceptable test-retest reliability for knowledge (ICC =.84), self-efficacy (ICC=.70), and behavior (ICC =.93).

Conclusions and Implications:

The FLitT was developed and tested for face and content validity and internal and test-retest reliability in low-income adults. Additional research is needed to conduct a second round of face validity and test construct validity using factor analysis with larger size.

INTRODUCTION

Low-income adults are at an increased risk for developing diet-related chronic diseases, including obesity,¹ diabetes,² hypertension,³ and heart disease.⁴ Increased disease risk stems from health disparities, including limited access to healthy foods,⁵ poor dietary quality, physical inactivity,⁶ and inadequate access to healthcare.⁷ Low income is associated with increased consumption of added sugar⁸ and added fat⁹ and decreased consumption of fruits and vegetables, whole grains, low-fat dairy,¹⁰ dietary fiber,¹¹ and lean meats,⁹ which can all contribute to disease risk. Although a nutritious diet is an essential component of health, healthy eating is complex and influenced by a combination of knowledge, skills, and decision-making capabilities as well as personal preference, availability, and affordability.¹²

Food literacy encompasses these components and processes. Among several definitions developed over the last decade, the most widely accepted definition consists of the contextually specific food related knowledge, skills, and behaviors required to plan and manage, select, prepare, and eat food in order to meet needs and support health.^{13, 14} The similar, yet separate term, nutrition literacy, is defined as the ability to find, make use of, and apply nutrition information.¹⁵ While nutrition literacy measurement tools exist, and have been used to identify needs and develop interventions,^16–21 these tools do not address the interrelatedness between knowledge and skill, an intricacy unique to food literacy.¹⁵ The interrelated quality of components and the context specific nature make measuring food literacy challenging.^{13, 14} However, a valid and reliable food literacy assessment tool can help nutrition educators develop tailored nutrition education targeting population specific needs.

Currently, few instruments measure food literacy in its totality. Those that do exist measure food literacy in particular countries, such as Australia, Iran, Italy, Switzerland, and the Netherlands,^22–25 measure food literacy specifically in children,^25–27 or are intervention specific.^{24, 28} Although instruments to assess food literacy are available, the only tool to measure food literacy in the US is designed for children.²⁶ Furthermore, to the authors’ knowledge, there is no tool to measure food literacy in low-income adult populations in the United States. Therefore, the purpose of this study was to describe the development of and revisions to the Food Literacy Assessment Tool (FLitT), targeted toward low-income adults, and assess its test-retest reliability and internal consistency reliability.

METHODS

Study Design

This cross-sectional study was designed to develop and test the validity and reliability of the FLitT. This study was approved by the University of Cincinnati Institutional Review Board.

Development of the FLitT

The conceptual framework of food literacy by Vidgen and Gallegos (2014), which includes “a collection of inter-related knowledge, skills and behaviours required to plan, manage, select, prepare and eat food to meet needs and determine intake,” was used by nutrition faculty and graduate students to develop the FLitT. The FLitT is comprised of 4 subscales that are theoretically designed based on the framework of Vidgen and Gallegos’s food literacy¹³: nutrition knowledge, self-efficacy of food related skills and behavior, food related behavior, and food consumption frequency. The nutrition knowledge subscale includes items on food groups based on the 2015 Dietary Guidelines for Americans, nutrition label reading, and unit price comprehension. The self-efficacy subscale includes items measuring confidence in ability, broken down by the four domains of food literacy: plan and manage, select, prepare, or eat. Items in the behavior subscale are also categorized into one of the four domains of food literacy. Finally, the food consumption frequency subscale assesses consumption of major food groups, sugar sweetened beverages, and processed foods, as well as frequency of behaviors, such as dining out, skipping meals, and consuming breakfast. The initial version of the FLitT was developed in 2018 and used for face validity testing. This version contained 64 items, including 18 items in the knowledge subscale, 21 items in the self-efficacy subscale, 16 items in the behavior subscale, and 9 items in the food consumption frequency subscale. The instrument had a 4.5 grade reading level according to the Flesch-Kincaid system in Microsoft Office. (FLitT available upon request.)

To assess face and response process validity, a convenience sample of 10 participants was recruited at St. Vincent de Paul Liz Carter Outreach Center, a non-profit organization in Cincinnati, Ohio. The center offers emergency assistance to local residents, including access to a client choice food pantry, where patrons have the opportunity to choose the types of foods which they receive within categories based on an allotment determined by household size. Participants were eligible for the study if they were at least 18 years of age, able to read, write, and speak in English, a client of the food pantry, and had a place to prepare food. Each participant was compensated with a $25 grocery store gift card. All cognitive interviews were conducted individually and audio-recorded in a private room by the same researcher. During the interviews, the FLitT was sectioned off into four sections, and the participants were asked to complete one section at a time followed by a discussion on the clarity, difficulty, vocabulary, and challenges associated with each section or specific item. The following semi-structured open-ended questions were used to assess interpretation of questions²⁹ such as “How did you feel about answering the questions for this whole section?” and “Did you have any difficulty with any specific question?” Some participants were further probed to determine if they had trouble answering specific questions: “What did you think this question is asking you?”, “Can you try to ask this question in your own words”, or “You spent a great deal of time on this question. What were you thinking about during this time?” After completing the initial probes above, the interviewer asked more specific questions depending on each respondent, including, “What does the term XXX (e.g., processed food) mean to you?”, “How sure are you of your answer to this question?”, “How did you arrive at this answer?”, and/or “What caused you to change your answer to this question?”. The interviewer observed also non-verbal cues during the interviews. The interviews lasted an average of 44 minutes 25 seconds (SD=9.54 minutes). All interviews were audio-recorded. Content analysis was conducted via open coding by entering perception of difficulty, time spent, and additional comments or notes in spreadsheet for each item of the FLitT, then compared across participants. Based on findings, revisions were made prior to content validity testing.

To assess content validity, a convenience sample of 10 registered dietitians (RDs) with experience working with underserved populations was invited to provide feedback on the FLitT via e-mail, and 4 RDs and 1 dietetic intern working under one of the RDs reviewed it. The FLitT, along with a form assessing length, relevance, clarity, wording, content, and format, were sent electronically to each potential reviewer. The research team met to review and discuss the comments and suggestions, such as reordering items, adding more examples for serving size equivalents, adding appropriate response options, and rewording items. Specifically, reviewers suggested including more relatable examples of serving size equivalents, such as “size of poker chip” to describe “1 Tablespoon peanut butter.” The FLitT was revised based on comments consistently provided by reviewers in order to improve its clarity prior to testing reliability.

Revised FLitT

The revised FLitT contained 75 items divided into 4 subscales: knowledge (15 items), behavior (24 items), self-efficacy (22 items), and food consumption frequency (14 items). Items within the self-efficacy and behavior subscales are further categorized into one of four domains: plan and manage, select, prepare, or eat. ¹³

Test-retest Reliability and Internal Consistency Reliability

A total of 98 participants were recruited at the same food pantry through distribution of flyers and individual approach in the waiting area to test the reliability of the revised FLitT. For this study, questions on demographics, health status, and other characteristics were included at the end of the FLitT survey. In order to be eligible, participants had to be at least 18 years of age, able to read, write, and speak in English, a client of the food pantry, have a place to prepare food, and were not participants of the face validity and response process validity research. A total of 101 individuals were screened, but 3 were unable to participate based on inclusion/exclusion criteria. An average of 10.30 days after the first survey (SD=4.23 days), 75 participants returned to complete the same survey a second time for the test-retest reliability. Both times, the surveys were self-administered in a quiet waiting area or a private multi-purpose room. Participants received a $10 grocery store gift card each time they completed the survey.

Data were analyzed using IBM SPSS V25.0. Descriptive statistics (mean ± SD, frequency) were conducted to summarize participant characteristics. Data from the first survey (n=98) was used to calculate internal consistency reliability. Cronbach α was used for subscales and domains with Likert Scale response options while Kuder-Richardson Formula 20 (KR-20), a special case of Cronbach α for dichotomous response options, was used for the knowledge subscale. Values ≥ 0.7 were considered acceptable.³⁰ The corrected item-total correlation and Cronbach α if item deleted of domains within the subscale or subscale were calculated using the Reliability procedure of the Scale. Test-retest reliability was calculated for each domain and subscale using intraclass correlation coefficient (ICC). Values between 0.50–0.75 were considered to have moderate reliability, 0.75–0.9 were considered to have good reliability, and values >0.9 were considered to have excellent reliability.³¹ Floor and ceiling effect were used to examine the distribution of participants responses at upper and lower limits, with 15% or more indicating floor or ceiling effect.³² Additionally, items in the knowledge subscale were assessed for item difficulty, indicating the number of participants who selected the correct response. Item difficulty closer to 1 indicates and easier question. For this subscale, items with item difficulty less than or equal to 0.74 are considered ideal.³³

RESULTS

Test-retest Reliability and Internal Consistency Reliability

Of the 98 participants, 48 were male and 48 were female, while 2 preferred not to answer. Sixty-seven participants were between 45–64 years old and 83 were Black or African American. Sixty-five participants had a high school degree or above, 77 were not employed, and 58 used food assistance (Table 1). Eighty participants reported receiving nutrition education and seventy-seven received counseling about food/diet from a health care professional. Nearly all participants were the primary person responsible for grocery shopping (n=89) and food preparation in their households (n=89). Almost all participants regularly prepared food, with 62 participants reporting preparing food every day and 20 preparing food 4–6 days per week. Some participants skipped items, as noted in the footnotes of Tables 4–6. Cronbach α ranged 0.76–0.92 for the behavior and self-efficacy subscales and KR-20 was 0.51 for the knowledge subscale. ICC of the subscales and domains ranged 0.50–0.93 (P<0.01) (Table 2).

Table 1.

Demographics of Participants (n=98)

Characteristic	n (%)
Gender
Male	48 (49.0%)
Female	48 (49.0%)
Prefer not to answer	2 (2.0%)
Age
18–24	3 (3.1%)
25–34	3 (3.1%)
25–44	14 (14.3%)
45–54	34 (34.7%)
55–64	33 (33.7%)
65+	11 (11.2%)
Racial/ethnic background
Black or African American	83 (84.7%)
White	9 (9.2%)
Other	4 (4.1%)
Prefer not to answer	1 (1.0%)
Highest level of education
Less than 9th grade	5 (5.1%)
9th-12th grade (no diploma)	28 (28.6%)
High school graduate, GED, or equivalent	30 (30.6%)
Vocational, trade school, some college	20 (20.4%)
Associate degree	10 (10.2%)
College graduate or above	5 (5.1%)
Job status
Employed – full time	9 (9.2%)
Employed – part time or seasonal	12 (12.2%)
Not employed – retired	9 (9.2%)
Not employed – disability	43 (43.9%)
Not employed – other	25 (25.5%)
Current food assistance use
Food stamps/SNAP	58 (60.2%)
WIC	0 (0.0%)
Other	2 (2.0%)
None	40 (40.8%)
Marital status
Single, never married	63 (64.3%)
Married or domestic partner	12 (12.2%)
Separated, divorced, or widowed	23 (23.5%)

Table 4.

Corrected Item-Total Correlation of Each Domain, Item Test-Retest Reliability, and % of Lowest and Highest Scores in Self-Efficacy Subscale of the FLitT in Low-Income Adults

Item (22 items)	Corrected Item-total Correlationa (n=91)	Cronbach Alpha if Item Deleted	Item Test-Retest Reliability (ICC) (n=75)	% of lowest scoreb (floor) (n=98)	% of highest scorec(Ceiling) (n=98)
Plan & Manage (5 items)
Plan a food budget to last for a week or month	0.60***	0.69	0.56***	3.1	30.6
Stick with my food budget to last for a full week or month.	0.69***	0.65	0.74***	4.1	22.4
Plan healthy meals within my food budget d	0.48***	0.73	0.44 **	2.1	28.9
Choose less expensive foods based on unit price (price per volume or weight of food)	0.51***	0.72	0.44**	1.0	25.5
Use other food items when the one I need is not available	0.36***	0.77	0.38*	0.0	18.6
Select (6 items)
Know when a fruit or vegetable is spoiled	0.36***	0.85	0.30	1.0	57.7
Read the Nutrition Facts labels	0.62***	0.81	0.59***	3.1	25.5
Select seasonal fruits	0.67***	0.80	0.53**	1.0	28.6
Select seasonal vegetables	0.69***	0.80	0.57***	1.0	22.4
Select food that is low in fat	0.66***	0.80	0.53**	3.1	17.3
Select food that is low in salt	0.68***	0.79	0.52**	2.0	29.6
Prepare (7 items)
Follow instructions on a food package or recipe d	0.51***	0.78	0.21	2.1	32.0
Use whole grain products in place of refined grain products (example: whole wheat bread instead of white bread) d	0.60***	0.77	0.47 **	1.0	27.1
Prepare tasty meals without adding too much salt d	0.65***	0.76	0.68***	1.0	27.8
Cook foods I have never cooked before if I have a recipe e	0.50***	0.79	0.61***	5.2	13.5
Prepare meals using tools I have (examples: stove, oven, knife, pan or pot) d	0.54***	0.78	0.25	2.1	30.9
Use healthier cooking methods (examples: panfry or bake instead of deep fry) d	0.54***	0.78	0.27	2.1	22.7
Cook meats to a safe temperature d	0.43***	0.80	0.64***	3.1	41.7
Eat (4 items)
Eat fruits every day d	0.69***	0.70	0.67***	4.1	29.9
Eat vegetables every day f	0.78***	0.65	0.69***	2.1	29.2
Eat whole grains every day (example: brown rice, 100% whole wheat bread, 100% whole wheat pasta) e	0.53***	0.78	0.59***	3.1	25.0
Stop eating when I am fulle	0.45***	0.82	0.76***	4.2	32.3

^aPearson’s correlation coefficient for corrected item-total correlation by domain;

^b“strongly disagree”;

^c“strongly agree”;

^dn=74;

^en=73;

^fn=72;

^*P <0.05;

^**P <0.01;

^***P <0.001.

Table 6.

Corrected Item-Total Correlation, Item Test-Retest Reliability, and Floor/Ceiling Responses in the Food Consumption Frequency Subscale of the FLitT in Low-Income Adults

Item (14 items)	Corrected Item-Total Correlationa (n=97)	Item Test-Retest Reliability (ICC) (n=75)	% of lowest score b (floor) (n=98)	% of highest scorec(ceiling) (n=98)
Eat 1 ½ - 2 ½ cups of fruit per day	0.28**	0.76***	1.0	31.6
Eat 2–4 cups of vegetables per day	0.34***	0.61***	1.0	40.8
Eat 2 ½- 5 ounces of whole grains per day	0.34***	0.61***	3.1	32.7
Eat 5–7 ounces of protein per day	0.38***	0.59***	0.0	40.8
Eat or drink 3 cups of low-fat dairy or dairy alternative per day	0.39***	0.70***	6.1	26.5
Drink regular soda or pop	0.49***	0.78***	18.4	18.4
Drink any of the following: Fruit drink, sweet tea, sports drink, energy drink, sweetened coffeed	0.58***	0.78***	9.2	31.6
Eat fast food	0.42***	0.73***	13.3	5.1
Eat processed food	0.47***	0.75***	10.2	11.2
Eat at a sit-down restaurant or buffet	0.34***	0.59***	31.6	4.1
Eat breakfast	0.25*	0.87***	3.1	53.1
Skip meals	−0.05	0.66***	40.8	5.1
Have a snack or dessert of any of the following: chips, candy, crackers, cookies, sugary cereal, pastries, ice cream	0.50***	0.65***	4.1	30.6
Snack between mealsd	0.31**	0.64***	10.2	30.6

Pearson’s correlation coefficient for corrected item-total correlation;

^b“never”;

^c“every day,”

^dn=74;

^***P < 0.001.

Table 2.

Internal Consistency Reliability (n=98) and Test-Retest Reliability (n=75) for Food Literacy Assessment Tool (FLitT) Subscales in Low-Income Adults

Subscale/Domains	Cronbach’s α (n=98)	KR-20 (n=98)	ICC (n=75)
Food Consumption Frequency	-	-	0.81***
Knowledge	-	0.51	0.86***
Behavior
Total	0.90	-	0.93***
Plan/Manage a	-	-	0.82***
Select	0.82	-	0.84***
Prepare	0.81	-	0.79***
Eata	-	-	-
Self-Efficacy
Total	0.92	-	0.70***
Plan/Manage	0.76	-	0.68***
Select	0.84	-	0.52**
Prepare	0.80	-	0.50**
Eat a	-	-	0.74***

^-indicates not calculated

^aCronbach’s α not calculated for subscales with < 5 items

^*P <0.05;

^**P <0.01;

^***P <0.001.

Knowledge Subscale.

As shown in Table 3, most items showed high or moderate item test-retest reliability, except those items measuring knowledge of plant-based protein and unit-price labels, and dairy (ICC=0.46, 0.49, P<0.01, 0.01, respectively) while the item assessing knowledge related to dairy was not statistically significant. The average item difficulty index was 0.62 and the majority of items showing item difficulty have values between 0.2 and 0.8, except items on fruit (0.89), leafy green vegetables (0.96), starch vegetables (0.86) and Nutrition Facts labeling reading (0.11).

Table 3.

Item Test-Retest Reliability and Item Difficulty Index in the Knowledge Subscale of the FLitT in Low-Income Adults

Item (15 items)	Item Test-Retest Reliability (ICC) (n=75)	Item Difficulty Index (n=98)
Which food belongs to the fruit group?	0.83***	0.89
Which food belongs to the leafy green vegetable group?	0.72***	0.96
Which food belongs to the starchy vegetable group?	0.67***	0.86
Which food belongs to the protein group?	0.65***	0.74
Which food is a plant-based protein?	0.46**	0.65
Which food belongs to the dairy group?	0.31	0.72
Which food is the whole grain?	0.60***	0.77
Which food is a good source of fiber?	0.69***	0.75
Which food is a good source of iron?	0.67***	0.30
What food is a good source of calcium?	0.64***	0.77
Which cow’s milk has the least fat?	0.70***	0.62
Which meat has the least fat?	0.70***	0.32
How many calories are in the whole container of soda?	0.50**	0.11
How many grams of sugar are in 1 serving of soda?	0.72***	0.55
Based on the unit price, which one is less expensive?	0.49**	0.32

^*P <0.05;

^**P <0.01;

^***P <0.001

Self-efficacy Subscale.

In the self-efficacy subscale, most items showed moderate corrected item-total correlation within domain (r=0.43–0.69, P<0.001). Only two items, measuring the confidence to use alternative food items when an item is unavailable and know when a fruit or vegetable is soiled, showed weak correlation with the subscale total (r=0.36, P<0.001). Only one item on eating vegetables every day, showed a strong corrected item-total correlation (r=0.78, P<0.001). Most items showed moderate test-retest reliability (ICC=0.52–0.74, P<0.01 or P<0.001); however, several items either showed poor test-retest reliability or non-significant ICC as shown in Table 4. Floor effect was not found in this subscale. In other words, there are no items in which more than 15% of participants selected “strongly disagree.” However, all items, except the item on cooking new foods with a recipe (13.5%), showed ceilings effect with more than 15% participants selected “strongly agree.” Particularly, 57.7% and 41.7% of participants strongly agreed with their confidence in knowing when a fruit or vegetable is spoiled and cooking meats to a safe temperate, respectively. Highly correlated items (r>0.7) were identified: “I am confident I can eat vegetables every day” and “I am confident I can eat fruits every day” (r=0.82, P<0.01); “I am confident I can plan a food budget to last for a week or a month” and “I am confident I can stick with my food budget to last for a full week or month” (r=0.70, P<0.01); and “I am confident I can select seasonal fruits” and “I am confident I can select seasonal vegetables” (r=0.85, P<0.01).

Behavior Subscale.

Nearly all corrected item-total correlation within each domain in the behavior subscale showed moderate correlation (r= 0.35–0.69, P<0.001), except one weak correlation (r =0.25, P<0.05) and one no significant correlation (measuring how often participants deep fry food) as shown in Table 5. Test-retest reliability for items of the behavior subscale generally indicated moderate reliability (ICC=0.52–0.75, P<0.01). Items measuring behavior related to use of a grocery shopping list, use of unit prices, and rinsing canned goods showed good reliability (ICC=0.76, 0.76, and 0.80 respectively, P<0.001); however, the item assessing use of measuring utensils showed poor test-retest reliability (ICC=0.49, P<0.01). Seventeen out of 24 items did not show floor effects (over 15% of participants responding “never” to these items); however, three items under the preparation, use of thermometer (45.9%), and trimming skin off (22.4%) and fat off (17.3%) chicken and other poultry, did. In addition, two items under selection, using nutrition information at restaurants (30.6%) and reading Nutrition Facts labels (25.5%) show floor effects. All but 5 items showed ceiling effect with more than 15% of participants responding “always.” Those without ceiling effects include reading Nutrition Facts label (14.3%), using nutrition information at restaurants (12.2%), cooking new foods (11.2%), using a thermometer (10.2%), and deep frying (9.2%). High correlation was seen between items assessing frequency of trimming fat off of meat and frequency of trimming fat off of chicken and poultry (r=0.76, P<0.01)

Table 5.

Corrected Item-Total Correlation of Each Domain, Item Test-Retest Reliability, and % of lowest and highest scores in the Behavior Subscale of the FLitT in Low-Income Adults

Item (24 items)	Corrected Item-total Correlationa (n=91)	Cronbach Alpha if Item Deleted	Item Test-Retest Reliability (ICC) (n=75)	% of lowest scoreb (floor) (n=98)	% of highest scorec(Ceiling) (n=98)
Plan & Manage (4 items)
Plan meals before grocery shoppingd	0.50***	0.68	0.71 ***	5.1	23.7
Have transportation to and from shopping for food	0.43***	0.72	0.71***	9.2	36.7
Make and bring a grocery list when you shopf	0.60***	0.62	0.76***	16.3	17.3
Plan time to cook meals	0.56***	0.64	0.72***	4.1	25.5
Select (8 items)
Read the Nutrition Facts label before selecting a new food	0.54***	0.80	0.67***	25.5	14.3
Look at nutrition information before ordering at a restaurant when it is available	0.48***	0.81	0.64***	30.6	12.2
Compare the prices of similar food items to select oned	0.35***	0.82	0.59***	7.1	28.6
Select foods that you believe are healthyd	0.68***	0.78	0.74***	4.1	34.7
Select seasonal fruits when grocery shopping	0.63***	0.78	0.75***	11.2	22.4
Select seasonal vegetables when grocery shoppingd	0.69***	0.78	0.72***	8.2	23.5
Make choices based on what is on sale when shopping for foodd	0.50***	0.80	0.55***	0.0	29.6
Use unit prices to select what you buy?e	0.46***	0.81	0.76***	14.3	21.4
Prepare (11 items)
Wash your hands before preparing foods	0.25*	0.82	0.60***	0.0	80.6
Prepare a meal from basic or raw ingredients (from scratch)	0.43***	0.80	0.68***	10.2	22.4
Measure ingredients using measuring cups and measuring spoonsd	0.42***	0.80	0.49**	15.3	22.4
Rinse canned vegetables before cooking	0.51***	0.80	0.80***	14.3	42.9
Cook foods you have never cooked befored	0.56***	0.79	0.67***	12.2	11.2
Use healthy cooking methods (cooking with no fat or salt, baking, steaming, boiling, and so on)	0.56***	0.79	0.73***	4.1	27.6
Deep fry food when cooking	0.17	0.82	0.50**	11.2	9.2
Use a thermometer when cooking meatd	0.41***	0.81	0.70***	45.9	10.2
Trim skin off chicken and other poultry before cooking it	0.58***	0.79	0.66***	22.4	22.5
Trim fat off chicken and other poultry before cooking it	0.68***	0.78	0.52**	17.3	37.8
Trim fat off meat (beef, pork, etc.) before cooking it	0.63***	0.78	0.63***	14.3	34.7
Eat (1 item)
Eat with family or friends	-	-	0.60***	3.1	22.4

^aPearson’s correlation coefficient for corrected item-total correlation by domain;

^b“never”;

^c“always”;

^dn=74;

^en=73;

^fn=72;

^**P <0.01;

^***P <0.001

Food Consumption Frequency Subscale.

Most item-total correlations in the subscale were weak or moderate (r=0.25–0.58, P<0.05, P<0.01 or P<0.001) but corrected item-total correlation of skipping meals was not significant, as shown in Table 6. Four items assessing the consumption of fruit intake, soda, other sugar sweetened beverages, processed foods, and breakfast showed good test-retest reliability (ICC=0.75–0.87, P<0.001) and remaining items were found to have moderate test-retest reliability (ICC=0.59–0.73, P<0.001). Floor effects were not observed for most items, with the exception of items measuring soda consumption (18.4%), eating at restaurants (31.6%), and skipping meals (40.8%) where greater than 15% of participants responded “never.” In contrast, more than 15 % of participants answered “every day” for all items, except for those measuring fast food consumption, professed food consumption, eating at a sit-down restaurant, and skipping meals.

DISCUSSION

The revised FLitT showed internal consistency reliability, and test-retest reliability, with the exception of the internal consistency reliability of the knowledge subscale. Based on face validity and response process validity testing, 4 items were revised to improve the instructions, 14 items were revised to minimize ambiguity, 14 items were revised to remove assumptions about the population, 3 items were revised to better assess knowledge, 1 item was revised to prevent the influence of participant bias, and 14 items were revised to more accurately assess consumption patterns. Some participants indicated that they answered items in the behavior subscale based on the habits of their household, despite the fact that “you” was used. Thus, “you” was underlined for emphasis. For the items assessing self-efficacy, “I can” was replaced with “I am confident that I can” to ensure clarity and improve face validity and response process validity. Scales measuring self-efficacy targeting adults were found to use either “I am confident that I can” ³⁴ or just “I can.”³⁵ Additionally, the original response options for the food consumption frequency subscale were replaced with response options based on NHANES surveys, which included an option to reflect consumption less frequent than once per week. For knowledge and food consumption items related to dairy, non-dairy examples and response options such as “I don’t know because I drink milk” were added because some participants mentioned that they had lactose intolerance and avoided dairy items. All participants of the cognitive interviews were African American, and it is estimated that 24% of African Americans have lactose intolerance.³⁶ Items related to meat consumption were similarly amended to include an option that states “I don’t know because I don’t eat meat” where appropriate.

The expert reviewers for the content validity testing noted that the arithmetic involved in comparing Nutrition Facts labels may present a challenge. Research shows that lower levels of nutrition label numeracy are associated with Black and Hispanic race, unemployment, older age, low educational achievement, and low income.³⁷ Because these characteristics align with our population, the label was modified to contain only values divisible by 10. In addition, the original images of labels used in items on unit price included brand names, which were replaced by labels without brand names to prevent bias toward specific brands.

The internal consistency reliability of the FLitT was acceptable with Cronbach α=0.76–0.92 (acceptable range: α= 0.7 or above)³⁰ for subscale (knowledge, self-efficacy, and behavior) totals and individual domains (plan and manage, select, prepare, and eat) within the behavior and self-efficacy subscales. However, the internal consistency reliability of the knowledge subscale (KR-20=0.51) is below the acceptable limit of 0.70.³⁰ Low internal consistency reliability indicates that some of the items or set of items are not correlated well with each other. This indicates there may be more than one dimension of nutrition knowledge within the FLitT. The knowledge subscale addresses multiple areas, including MyPlate, nutrient content, unit price, Nutrition Facts labels, and food safety. These different areas within the knowledge subscale may vary in difficulty. One expert reviewer commented that items based on unit price and Nutrition Facts labels may be too challenging for low-income adults. Additionally, only 14 participants reported “always” reading Nutrition Facts labels when selecting new foods. Some participants correctly answered question based on MyPlate and specific nutrients, but not items on unit price and Nutrition Facts label, which are relatively challenging. The varying difficulty across the domains may explain the low internal consistency.

The test-retest reliability of all domains and subscales of FLitT was moderate (ICC=0.50–0.75), good (ICC=0.75–0.90), or excellent (ICC≥0.90) for all domains and subscales.³¹ This may be in part due to testing fatigue. In the first admission of the survey, 22 of 98 participants chose the same response for all items on the last page of the survey. Also, compared to the self-efficacy subscale total, individual domains of the self-efficacy subscale show lower test-retest reliability, except the eat domain. The test-retest reliabilities of the select and prepare domains were the lower (ICC=0.52 and 0.50 respectively) than the plan/manage and eat domains (ICC=0.68 and 0.74 respectively).

When test-retest reliability of individual items was measured, the ICC for the item assessing confidence to “know when a fruit or vegetable is spoiled” within the select domain was not significant. This may be explained in part by the limited access to fruits and vegetables in low-income, primarily African American neighborhoods,³⁸ as well as the inconsistent offerings of fresh produce at food pantries, which may make answers less consistent. Similarly, the item assessing confidence related to preparing food using available tools had the lowest test-retest reliability within the prepare domain. This may be related to the limited access to basic cooking equipment reported by participants. Over half of participants (n=56) reported having access to a stove/hot plate, 59 reported access to a freezer, 31 had access to a blender/mixer. Again, a lack of access may cause participants to answer less consistently.

The floor and ceiling effects are defined as wen 15% or more of participants select the minimum or maximum response option. This is a concern, as the scale may be insufficient to distinguish small variations at the scale extremes.³² The floor effect, seen in items such as frequency of eating out at sit-down restaurants and skipping meals may be explained by population characteristics. Low-income households spend significantly less on eating out compared to higher income households.³⁹ Additionally, decreased education level has been found to be a predictor of frequency of skipping meals.⁴⁰ Other items showing floor and/or ceiling effects would likely improve with the addition of response options.⁴¹

The use of cognitive interviewing for face validity and response process validity was a strength. The cognitive interviews provided detailed information not necessarily available in focus groups or close-ended surveys. Open-ended questions, used during the cognitive interview, allowed the participant to guide the conversation, and the one-on-one interviews prevented social influences often seen in focus groups. Additionally, in an effort to limit sources of error, the surveys for testing reliability were scheduled during the same time of day for each participant and completed in a consistent setting.⁴² Similarly, the second survey was completed 1–3 weeks after the first survey. This decreased the opportunity for participants to learn new information or experience drastic changes in environment or circumstance. However, participants for face validity and response process validity and reliability testing were recruited via convenience sampling, limiting the generalizability of these findings. Likewise, response burden of completing the lengthy survey may have been associated with decreased test-retest reliability, as well as percent completion of the second survey.⁴³ Even though a large number of items were revised, a second round of cognitive interviews was not conducted and construct validity using factor analysis was not tested because of the small sample size.

CONCLUSIONS AND IMPLICATIONS

The FLitT was developed and tested for face validity and response process validity, content validity, and internal and test-retest reliability in low-income adults. Findings indicates that it is reliable tool to assess food literacy, but the internal consistency reliability of the knowledge subscale may need to be improved. Further research is needed to conduct additional round of face validity and response process validity and test construct validity using factor analysis with a larger size. Assessment of food literacy in low-income adults can serve as a foundation for developing tailored interventions and to evaluate the effect of interventions aimed toward increasing food literacy.