This study was approved by the UT Southwestern Medical Center Institutional Review Board (Identifier: STU-2022-0809). A step-by-step description of the study protocol is available in Supplementary Material 1.

Recruitment

Clients were recruited from February 12th -April 4th, 2023, by study staff who sat at a table with study signage in the middle of Crossroads waiting room during open food distribution hours (Monday-Thursday 8:30 AM–1:30 PM, first Saturday 8:30–11:30 AM). If clients were not approaching the table, staff would make a brief announcement to let clients know about the study. Fliers were also present throughout the pantry and clients could scan a QR code on the flier to complete the eligibility screener if study staff were with other participants.

Eligibility

Study inclusion criteria included adults 18 or older; able to read, write, and/or speak English or Spanish fluently; able to provide informed consent; clients who had used Crossroads pantry at least once; and able and willing to participate in the study, which included willingness to return in two-weeks to complete the follow-up appointment. Study exclusion criteria were dietary restrictions, allergies, or sensitivities that could put the participant at-risk of harm from consuming study foods. Some restrictions could be accommodated (e.g., vegetarian), while strict dietary needs (e.g., ketogenic diet, vegan diet), dietary disorders (e.g., celiac disease), and food allergies (e.g., dairy) were excluded as we could not guarantee no cross-contamination between study foods and products that could trigger an allergy or illness. Five people out of the 102 screened reported one of these dietary restrictions (5%). Clients were screened for eligibility by completing a 10-item survey on a study tablet or from their electronic device. Staff assisted clients uncomfortable with the tablet by reading questions and entering responses. The eligibility survey took an average of five minutes.

Power

This was the first of its kind pilot study, therefore there were no a priori effect sizes to calculate a power analysis. Using G*Power 3.1, we determined a small to medium size between-within interaction effect (f = 0.17) can be detected with a sample of 70 participants, thus we planned to stop enrollment at 70 clients.

Study procedures

Consent process

After eligibility was confirmed, clients could choose to participate in the first appointment immediately or to schedule the appointment for a future date. At the first appointment clients were read the consent form in English or Spanish and were asked if they had questions about the study. Once questions were answered or if there were no questions, clients were consented to participate. The time and date of verbal consent were recorded by study staff. Clients were consented and enrolled in the study by authors CH and JT.

Baseline measures

Participants completed a baseline questionnaire which included questions on demographics (age, race/ethnicity, adults and children in the home, annual income, and years of education). Two-items on medical insurance coverage were asked “Was there any time during the past two years when you did not seek medical care because it was too expensive, or health insurance did not cover it?” and “In the past two years have you always had health insurance or other medical coverage for health care?” which had Yes/No responses [21]. To measure food security we used the United States Department of Agriculture (USDA) six-item household food security scale which was coded as a continuous variable (0–6, high to very low food security) for analysis of variance and a categorical variable for participant characteristics (0–1 = High or marginal food security, 2–4 = Low food security, 5–6 = Very low food security) [22]. Perceived diet quality was measured using a single validated item that asked participants to rate their diet as “excellent” “very good” “good” “fair” or “poor” (5 = Excellent, 1 = Poor) [23].

Randomization

Participants were randomized by first author (KH) using simple randomization by generating random study identification numbers and randomly assigning those identification numbers to Group 1 (Meal kits) or Group 2 (No-prep meals) using a 1:1 allocation ratio. Assignments were concealed until the participant had to select study foods, at which point there was no way to conceal assignment.

Hedonic liking measurement

The questionnaire was programmed to bring up questions on hedonic liking that aligned with the participants group assignment; participants in Group 1 saw pictures of meal kits, while participants in Group 2 saw pictures of no-prep meals. However, participants were not told they were answering questions pertaining to their group assignment and thus were unlikely to be aware of their assignment at this point. This section of the questionnaire was also programmed to randomize question order (e.g., the order in which meals were shown) to reduce the likelihood of an order effect. Hedonic liking of study meals was measured by showing the name and picture of the meal and asking participants to respond to a 9-point bipolar scale with four measures of liking, four measures of dislike, and a neutral item, higher scores indicated higher hedonic liking [24].

Selection of study meals and nutritional content of meals

After participants finished the hedonic liking questionnaire, they were presented with laminated cards that displayed pictures of meals available depending on the participants group assignment. In both groups clients were able to select up to 84 servings of study meals as it was enough for a household of three (the average household size in Dallas County) [25] to have two meals per day each day of the two week study period. Meal kits were matched in quantity and content as closely as possible to the no-prep meals, such that a meal kit would include recipes and ingredients equivalent to three servings of a no-prep meal. We also aimed to replicate the no-prep meals standard weight of meat and grains as closely as possible when creating recipes and selecting ingredients for the meal kits. Participants were able to select from a menu of 14 breakfasts and 14 dinners and could select up to six servings or two meal kits of each available meal as an inventory control measure. Table 1 provides a brief description of each meal and the average nutritional content of each meal across groups. Nutritional content for the no-prep meals came directly from the no-prep meal distributor. Axxya Nutritionist Pro™ v7.9 software was used to conduct nutritional analysis of each meal kit. Ingredients in the meal kits had a mix of perishable (e.g., carrots, cheese, whole wheat bread), semi-perishable (e.g., potatoes, precooked chicken), and non-perishable (e.g., brown rice, low-sodium black beans, frozen produce) items. A full description of the ingredients in each meal kit is in Supplementary Material 2. Clients indicated food selections to a study staff member who entered the selections into an excel spreadsheet that calculated the total number of items and servings selected.

While clients went through Crossroads typical ordering procedures with pantry staff, study staff collated the participants study selections. Study staff were trained on safe food storage and handling practices by Crossroads staff prior to the onset of the study and followed all food safety guidelines set forth by Feeding America, which are inclusive of and more stringent than rules governing grocery retailers, food manufacturers and restaurants in the US [26]. No-prep meals were retrieved from a walk-in freezer and brought to the participant. Ingredients for meal kits were bundled in a recyclable paper bag and a recipe was stapled or taped onto the outside of the bag to indicate how to prepare the meal. An electronic copy of the recipes was provided by email to participants in the meal kit group upon request. Clients were instructed that no-prep meals could be stored in a refrigerator or freezer and eaten after thawing or reheating depending on consumer preference.

Crossroads typical pantry procedures

Clients can visit Crossroads pantry once each month and select food for up to 21 meals for each person in their household. Clients select food with a pantry staff member on a computer with a live Salesforce inventory system (Salesforce Inc., San Francisco, CA). The Salesforce interface uses a point system created by registered dietician nutritionists, that considers the age, gender, and activity level of each member of the household and indicates how many of each type of food group clients can select based on their nutritional needs. Clients are told how many points they have, and each available food item has a point value. After clients make food selections, their list is printed, and they take a grocery cart through the physical pantry aisles, like a grocery store. At checkout, pantry provisions are bagged, and a volunteer assists the client to their vehicle. Participants followed these same procedures during the study, except that study staff and/or student research assistants met the clients at pantry checkout and helped the participant load their vehicle, at which point clients were able to ask any questions and were reminded of the date and time of their follow-up appointment.

Follow-up questionnaire

Participants completed the same hedonic liking, perceived diet quality, and food security measures at follow-up in-person, except the wording was slightly modified to the temporal window (e.g., “How would you rate your dietary quality over the past 2-weeks?”). Participants were also asked questions on intervention satisfaction adapted from prior studies with pantry clients: “The food has been helpful” “The food provided was food my household likes to eat” “The food provided was good quality” “Enough food was provided” “I know how to prepare the foods” with “Strongly agree” “Agree” “Disagree” and “Strongly disagree” response options [27]. The items were summed to create an Intervention Satisfaction variable with the lowest possible score of 4 if all responses were “Strongly disagree” and the highest possible score of 20 if all responses were “Strongly agree.”

To measure intervention fidelity, we asked clients in both groups “Thinking about the times when you ate the study meals. Did you add additional ingredients or eat additional food when eating the study meals?” and clients were instructed to select all of the following options that applied: “Yes, I added extra seasoning or condiments (e.g., hot sauce, ketchup, mustard, salt, pepper, etc.)” “Yes, I added extra food items into the meal (e.g., I mixed extra meat, egg, protein, vegetables, fruit, etc. into the meal)” “Yes, I ate separate food or snacks in addition to the meal” or they could indicate “No, I ate the meal as is” as a single response. We asked, “When thinking about the size or quantity of the food in each study meal, was the size…” with response options “Too little (not enough food)” “Just right (the right amount of food)” and “Too big (too much food),” we asked for breakfast and lunch/dinner meals separately. Participants that received meal kits were asked “Did you use the recipe provided on the recipe card attached to each meal kit?” with Yes/No options. Participants were also asked “Over the past year, how many months have you gotten food from this food pantry” and could select up to 12 from a dropdown menu, indicating they had come every month within the past year.

Missed appointment protocol

If a participant missed the follow-up appointment, we had a protocol to contact the participant using their preferred method of contact twice on the day of the missed appointment and once per day for the two days following the appointment. Thereafter, we would contact once per week for the next four weeks.

Intervention costs

We added the total spent to purchase all the ingredients and supplies needed to create the meal kits and the total amount spent to purchase no-prep meals and divided those numbers by the number of meals provided in each group. Half of the no-prep meals were generously donated by the distributor; therefore, we calculated costs with and without the donation.

Analytic plan

IBM SPSS Statistics for Windows, Version 28.0. Armonk, NY: IBM Corp was used to conduct data analyses. Unpaired t-tests were used to determine differences between nutrient composition by group. Descriptive statistics (e.g., means, percentages, chi-squared tests of independence) were used to describe participant characteristics, categorical baseline differences, and costs associated with purchasing each study meal. Fisher’s exact tests were used if categorical items had small cell counts (e.g., gender, items on size/quantity of meals). One way analysis of variance (ANOVA) was used to test baseline differences in continuous variables (e.g., age, education) and number of servings selected by group. Distributions and QQ plots were checked. Two-way repeated-measures ANOVA was used to test for group and time effects on study outcomes of interest (hedonic liking of study meals, perceived diet quality, food security) and group-by-time interactions. Bonferroni correction was used for analysis on the hedonic liking of study meals as assessing differences by group over time for each meal leads to 28 comparison tests and high risk for Type 1 error. Therefore, Bonferroni adjusted α is set at 0.00179 for each individual meal, meaning the null hypothesis should only be rejected if the p-value is < 0.00179. One way ANOVA was used to test for follow-up differences in continuous variables (e.g., Intervention Satisfaction).