How does eligibility for GusNIP produce prescriptions relate to fruit and vegetable purchases and what factors shape the relationship? A protocol for a secondary analysis of nationally representative data in the USA

Abstract

Introduction

US Department of Agriculture (USDA) Gus Schumacher Nutrition Incentive Programme (GusNIP) produce prescription programme (PPR) ‘prescriptions’ provide eligible participants with low income, risk for diet-related chronic disease and food insecurity a healthcare issued incentive to purchase lower to no cost fruits and vegetables (FVs). However, GusNIP requirements specify that PPR prescriptions can only be redeemed for fresh (not frozen, canned or dried) FVs. This requirement may prevent participants from fully engaging in or benefiting from GusNIP PPR, given communities with lower healthy food access may have reduced fresh FV accessibility.

Methods and analysis

We will use the nationally representative 2012–2013 National Household Food Acquisition and Purchase Survey (FoodAPS) and complementary FoodAPS Geography Component data in a secondary data analysis to examine how household GusNIP PPR eligibility relates to the quantity and variety of fresh, frozen, canned and dried FV purchases and to what extent individual, household and food environment factors shape the relationship. FoodAPS data include household food purchasing and acquisition information across a 7 day period from 14 317 individuals among 4826 households and was collected between April 2012 and January 2013. The FoodAPS Geography Component provides information about the local community/environment relative to FoodAPS households. This study will examine the correlation or association of selected variables between different quantities and varieties of fresh, frozen, canned and dried FVs, as well as correlations among multilevel predictors.

Ethics and dissemination

We are following data integrity standards as outlined by agreements with the USDA Economic Research Service. All results of analyses will undergo a thorough disclosure review to ensure no identifiable data are shared. Results will be disseminated to research, practice and policy communities using an Open Access peer-reviewed manuscript(s), scientific and practice presentations, and a public facing report and infographic.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• We will use the 2012–2013 National Household Food Acquisition and Purchase Survey (FoodAPS), a nationally representative data set that is well suited for policy investigations on dietary purchases and acquisitions among households with lower income in the USA.

• The FoodAPS data collection methodology included a period of 7 days, which allowed for a rigorous view of household food purchase and acquisition data, including fruits and vegetables (FVs), compared with other available nationally representative US data sets.

• The FoodAPS data we plan to use as dependent variables include household purchases of fresh, frozen, canned and dried FVs, which does not necessarily correspond to dietary practices.

• The FoodAPS purchasing data pertaining to fresh, frozen, canned or dried FVs do not allow for discerning purchases with or without added sugars, which would likely be a requirement should the Gus Schumacher Nutrition Incentive Programme (GusNIP) produce prescription programme (PPR) be expanded to include these products.

• The FoodAPS data were collected in 2012–2013 and therefore may not reflect current US food purchasing practices among those eligible for GusNIP PPR.

Introduction

Achieving ‘nutrition security’ or the ability for all Americans to obtain an adequate quantity and variety of foods and beverages recommended by the Dietary Guidelines for Americans (DGA), including fruits and vegetables (FVs), is a US priority.^1–4 The 2020–2025 DGA recommends US adults consume a variety of FVs in adequate amounts, meaning at least one-and-a-half to two cups of fruit and two to three cups of vegetables each day.⁴ However, most Americans do not consume enough FVs as recommended by the DGA.⁴ There are also persisting disparities in FV accessibility, purchasing and intake among racial and ethnic minority groups and populations with low income that require unique policy, systems and environmental strategies to achieve nutrition security.^5–10 One example gaining momentum is the Food is Medicine movement.^11–13

As part of Food is Medicine, the US Department of Agriculture (USDA) has funded the Gus Schumacher Nutrition Incentive Programme (GusNIP) since 2019. This programme includes two types of financial incentive projects, nutrition incentive (NI) and produce prescription programmes (PPR), to incentivise households with low income to purchase FVs.¹⁴ GusNIP complements larger federal nutrition assistance programmes, such as the Supplemental Nutrition Assistance Programme (SNAP), to improve food security and dietary quality among eligible individuals and households with low income and with greater risk for diet-related chronic disease.^14–16 GusNIP NI programmes allow households with low income to redeem matched incentives (eg, spend $5.00 to get $5.00) for any variety of whole or cut fresh, canned, dried or frozen FVs without added sugars, fats or oils, or sodium,¹⁷ while PPR participants can use healthcare ‘prescriptions’ to source fresh whole or cut FVs only.¹³ Incentive redemption occurs at participating sites that include farm direct and/or brick and mortar retailers, such as farmers markets and grocers.^{13 17}

The GusNIP PPR policy that limits redemptions to only fresh FVs may prevent PPR participants from engaging in or fully benefiting from the programme for several reasons. Evidence supports consuming a variety of fresh, frozen, canned and dried FVs as part of a nutritious diet. For example, frozen and canned options may be equal to or have higher phytochemical integrity, or the compounds thought to protect against diet-related conditions, than fresh FVs.^{18 19} Frozen, canned and dried FVs can also offer culturally appropriate options at a reduced time cost regarding preparation, which is a key component of ‘affordability’ among households with low income.²⁰ Moreover, supermarkets or grocery stores that stock a variety of high quality and fresh FVs tend to be less prominent in communities with documented health disparities compared with communities comprised of predominantly white, higher income and/or urban residents.^{6 9 21} Transportation barriers may further compromise the ability to access fresh FVs.²² Therefore, it is critical to understand if the GusNIP PPR policy is equitable or appropriate, especially among households and communities with lower healthy food access.

Methods and analysis

Study design

We aim to answer the following primary research questions using a secondary data analysis of the 2012–2013 National Household Food Acquisition and Purchase Survey (FoodAPS)²³ and complementary FoodAPS Geography Component data:^24–26 (1) ‘How does GusNIP PPR eligibility relate to the quantity and variety of FV purchases?’ and (2) ‘To what extent do individual, household and food environment factors shape the relationship?’ We hypothesise that households comprised of individuals eligible to participate in GusNIP PPR will be associated with fewer fresh FV purchases, more frozen, canned and dried FV purchases, and lower quantity and variety of FV purchases overall in comparison to ineligible households. We further hypothesise that this relationship will be moderated by individual, household and environmental factors that characterise health disparities and/or communities with low support for FV access.

This protocol was posted publicly on Open Science Framework (DOI 10.17605/OSF.IO/UMA86) in advance of inferential analyses, which are planned to begin in May 2024 and continue through July 2024. The University of Nebraska Medical Centre Institutional Review Board declared this research exempt.

Patient and public involvement

Given the nature of secondary data analysis, FoodAPS respondents were not involved in the approach to the present research. However, the overarching concept to the research was informed, in part, by feedback from GusNIP audiences based on our experiences at the GusNIP Nutrition Incentive Programme Training, Technical Assistance, Evaluation and Information Centre (NTAE). The NTAE provides GusNIP NI and PPR grantees with implementation and research and evaluation support.

Setting

The FoodAPS was designed by the USDA Economic Research Service for research aiming to examine influential factors on household food purchasing and acquisition practices and includes data collected between April 2012 and January 2013 from 14 317 individuals among 4826 households.²³ The FoodAPS data were sourced from a nationally representative and stratified sample of US households, including: (1) households participating in SNAP; (2) SNAP eligible households with low income not participating in SNAP; (3) non-SNAP households with income ≥100% and ≤185% of the federal poverty threshold and (4) non-SNAP households with income ≥185% of the federal poverty threshold.²³

FoodAPS data

Individual and household-level data, including food purchased or otherwise acquired over a 7 day period, were collected from primary FoodAPS respondents, who were the main food shoppers or meal planners for the household.^{23 27} FoodAPS respondents also provided information on all household members.^{23 27}

Food and beverage purchases and acquisitions (eg, from non-retail sources) during the 7 day FoodAPS data collection period were recorded using purchase scanning/entry, receipts and food book records.²³ The relatively short timeframe for FoodAPS data collection, combined with the comprehensive nature of record keeping (ie, triangulation of multiple purchase records and interviewer reminders), led to a good representation of household purchasing practices among households with both high and lower income relative to other large data sets.²³ This is especially true regarding FV purchases which tend to be under-reported in scanner data;²⁸ therefore, FoodAPS is the ideal data set available for this research despite the data collection period being dated. Additional details regarding FoodAPS data collection procedures and questionnaires are available elsewhere.^{23 27}

The FoodAPS Geography Component was specifically designed to allow for an understanding of the influence of the local community/food environment on FoodAPS household dietary purchases using data from sources such as the USDA Food Environment Atlas, the USDA Food Access Research Atlas and US Census Data at the census block, census tract or county level.^24–26 Additionally, researchers with the USDA Economic Research Service recently (March 2024) published a constructed household-level variable for use with FoodAPS.²⁹ The Food Retail Environment Healthfulness Quality (FREHQ) measure uses an inverse distance weighting technique to provide a variable pertaining to the healthfulness of the food retail environment within a 20-mile radius of FoodAPS households, considering distance, the healthfulness of different types of food retail sites (eg, grocery compared with dollar stores) using a store-level Healthy Eating Index (HEI) 2020 score and household access to a car.²⁹ We have an opportunity to be among the first to apply the new household-level FREHQ measure to understand how the food retail environment influences FV purchasing among the FoodAPS households considered eligible to participate in GusNIP PPR.

A description of the specific FoodAPS variables we will use to understand how eligibility for GusNIP PPR relates to FV purchases and how individual, household and community factors shape the relationship are detailed below.

Independent variable

To answer how household eligibility for GusNIP PPR relates to the quantity and variety of FV purchases, we will create a bivariate household eligibility variable (Yes or No) using available FoodAPS datapoints. GusNIP PPR eligibility criteria includes income and dietary components: (1) persons must be eligible for SNAP or medical assistance such as Medicaid (income eligibility); and, (2) persons must currently have or be at risk for a ‘diet-related health condition’, including food insecurity (dietary eligibility).¹³ Therefore, we will consider FoodAPS households eligible for GusNIP PPR if at least one income eligibility variable and at least one dietary eligibility variable are present, as described below.

To estimate income eligibility for GusNIP PPR among FoodAPS households, we selected variables related to: current or prior (within 12 months) household participation in SNAP; Social Security Income; and household income at or below 130% of the federal poverty level. Because the FoodAPS data set does not include a variable on household participation in Medicaid, we selected available variables to estimate household income eligibility for Medicaid among pregnant women and parents in years 2012/2013 due to higher poverty level thresholds compared with SNAP that varied by US state.^{30 31} Non-parent and non-pregnant adults were not eligible for Medicaid in most states until Medicaid expansion with the Affordable Care Act in 2014.^{32 33}

To estimate dietary eligibility for GusNIP PPR among FoodAPS households, we selected variables related to: the presence of household food insecurity; individuals in the household with reported overweight or obesity; perceived suboptimal dietary quality among FoodAPS respondents or the household overall; perceived suboptimal household intake of FVs; and perceived poor health status among any household member.

Dependent variables

FV quantity

The purchasing data collected among FoodAPS households for fresh, frozen, canned and dried FVs were used by the USDA Economic Research Service to indicate or estimate grams purchased by specified food group.²³ We will use these variables as the dependent variables in our analysis, regarding household quantity purchased of fresh, frozen, canned and dried FVs for home consumption, including grams of the following:

• Fresh whole fruit.

• Fresh vegetables (ie, summative value for groups: fresh starchy vegetables; fresh tomatoes; fresh dark green vegetables; fresh red and orange vegetables; fresh other/mixed vegetables; and fresh beans, lentils, legumes).

• Frozen whole fruit.

• Frozen vegetables (ie, summative value for groups: frozen starchy vegetables; frozen dark green vegetables; frozen red and orange vegetables; frozen other/mixed vegetables; and frozen beans, lentils, legumes).

• Canned whole fruit.

• Canned vegetables (ie, summative value for groups: canned starchy vegetables; canned tomatoes; canned dark green vegetables; canned red and orange vegetables; canned other/mixed vegetables; and canned beans, lentils, legumes).

• Dried whole fruit (there is no group available for dried vegetables).

• Total quantity of FVs purchased (ie, summative value of above variables).

FV variety

We will also use the variables noted above by FV group to construct counts or the variety of fresh, frozen, canned and dried FVs purchased for home consumption among FoodAPS households, with some changes. The variable for grams of fresh whole fruit as noted above is not as nuanced as a separate, available variable in cup equivalents, that splits fresh whole fruit into two categories, which we selected for the variety dependent variables (shown below). Furthermore, given the canned and frozen fruit group variables have limited varieties indicated, compared with the vegetable groups, we combine FVs for the variety dependent variables. Last, as the only dried group is for fruit, we merge this with canned FV variety.

Documented purchases among FoodAPS households within each of the selected groups will be scored as one ‘variety’. The dependent variety variables are shown below with the possible range in scores noted:

• Fresh FV variety, including groups: fresh whole citrus, melons and berries; whole fruit, excluding citrus, melons and berries; fresh starchy vegetables; fresh tomatoes; fresh dark green vegetables; fresh red and orange vegetables; fresh other/mixed vegetables; and fresh beans, lentils, legumes. The possible range in fresh FV varieties purchased among FoodAPS households is 0–8.

• Frozen FV variety, including groups: frozen whole fruit; frozen starchy vegetables; frozen dark green vegetables; frozen red and orange vegetables; frozen other/mixed vegetables; and frozen beans, lentils, legumes. The possible range in frozen FV varieties purchased among FoodAPS households is 0–6.

• Canned (and dried) FV variety, including groups: canned whole fruit; canned starchy vegetables; canned tomatoes; canned dark green vegetables; canned red and orange vegetables; canned other/mixed vegetables; canned beans, lentils, legumes; and dried whole fruit. The possible range in canned/dried FV varieties purchased among FoodAPS households is 0–8.

• Total FV variety, including the summative value of all noted above, with a possible range of 0–22.

Sensitivity analysis variables

The FoodAPS also collected data on FVs purchased for away-from-home use, such as at restaurants.²³ These variables are documented similar to those noted above. We will use this information in a sensitivity analysis (described below) regarding the quantity and variety of fresh, canned and dried FVs purchased among FoodAPS households. Frozen FV groups are not available in the FoodAPS data for food away-from-home purchases.

Moderating and confounding variables

We also selected individual, household and community-level variables for inclusion in multivariable adjusted models (described below) based on knowledge of factors that influence (eg, moderate and confound) food purchasing (table 1).6 24–26 29 34 35 We plan to collapse or recode certain FoodAPS variables, such as race and ethnicity, to align with current scientific practices,³⁶ and for use in our statistical models. See table 1.

Table 1.

Moderating and confounding variables used in a secondary analysis of nationally representative FoodAPS data to examine how eligibility for GusNIP produce prescription programmes relate to fruit and vegetable purchases in the USA

Variable	Categories/description
Individual-level variables
Sex assigned at birth of primary food shopper	Male Female
Age of primary food shopper	Age in years
Race/ethnicity of primary food shopper	American Indian or Alaska Native, non-Hispanic Asian, non-Hispanic Black/African American, non-Hispanic Native Hawaiian or Other Pacific Islander, non-Hispanic White, non-Hispanic Another race, non-Hispanic Multiple races, non-Hispanic Hispanic
Education of primary food shopper	High school education or less Some college or associate degree Bachelor or higher degree
Marital status of primary food shopper	Married Previously married Never married
Household-level variables
Household size	Number of all guests and family members staying in the household
Number of children	Number of all children staying in the household
Supplemental Nutrition Assistance Programme (SNAP) benefit issuance	Days since SNAP benefits were last received
Special Supplemental Nutrition Assistance Programme for Women, Infants and Children (WIC) participation	No Yes
Household member following a vegetarian diet	No Yes
Household member following a diet	No Yes
Household vegetable garden	No Yes
Household receiving fruits and vegetables from someone else’s garden	No Yes
Household gets food from farm stand or market	No Yes
Primary store used for food shopping	Traditional brick-and-mortar (eg, supermarket, grocery store) Farm direct (eg, farmers market) Non-traditional brick-and-mortar (eg, convenience store, dollar store)
Driving distance	Miles between home and primary store
Car access	No Yes
Alternate store used for food shopping	Traditional brick-and-mortar (eg, supermarket, grocery store) Farm direct (eg, farmers market) Non-traditional brick-and-mortar (eg, convenience store, dollar store)
Household in rural area	No Yes
Season of FoodAPS data collection	October–January April–June July–September
Meals prepared at home	Number of dinner meals within the prior 7 days that were prepared at home
Community/food environment variables
Food Retail Environment Healthfulness Quality (FREHQ)	Weighted value of how ‘healthy’ the traditional (eg, supermarket) and non-traditional (eg, convenience store, dollar store) retail environment within a 20 mile radius is relative to National Household Food Acquisition and Purchase Survey (FoodAPS) households, considering distance, healthfulness score attributed to store type and household access to a car.
Farmers market availability	Number of farmers markets within 0.5, 1, 10 or 20 miles of block group
Limited-service restaurant availability	Number of limited-service restaurants within 1 mile of block group centroid
Full-service restaurant availability	Number of full-service restaurants within 1 mile of block group centroid
Area poverty rate	Household poverty rate in block group
Occupied housing	Occupied housing units in block group
Block group has at least 100 households without access to a vehicle	No Yes
Area educational attainment	Share of population over the age of 25 with high school education in the county
US Department of Agriculture Food Access Research Atlas low income and low access designation	No Yes
Food cost	Total Thrifty Food Plan basket cost for a family of four
Region of the USA	West Midwest South Northeast

Statistical methods

We will classify FoodAPS households into one of two categories: (1) eligible for GusNIP PPR and (2) not eligible for GusNIP PPR. We will then generate descriptive statistics for the individual, household, community/food environment variables as well as the dependent quantity (fresh whole fruit; fresh vegetables; frozen whole fruit; frozen vegetables; canned whole fruit; canned vegetables; dried whole fruit; and total grams FVs) and variety (fresh FVs; frozen FVs; canned/dried FVs; and total FV variety) variables by household eligibility for GusNIP PPR. We will describe the numeric variables using weighted means and SD by eligibility status, and will use survey regression models to compare these outcomes between those who were eligible and those who were ineligible for GusNIP PPR, while accounting for the household weights and the complex survey design of the FoodAPS.²³ For the discrete variables, we will use the survey frequency procedure to present the weighted proportions by GusNIP PPR eligibility and we will conduct the group comparisons using the χ² test in the survey frequency procedure. We will examine the correlation or association of the outcomes between different FV quantities and varieties (eg, fresh quantity with canned quantity and frozen quantity; fresh variety with frozen variety), as well as correlations among the predictors. Our main objective is to identify the multilevel factors that are associated with FV purchases (quantity and variety), and whether household GusNIP PPR eligibility had differential impact on participating subgroups as defined by the individual, household and community/food environment characteristics.

We will perform a bivariate analysis between the predictor variables and each FV quantity purchased variable. We will consider the predictors that have a p-value<0.2 in the regression analysis. For each FV quantity outcome, we will apply a regression model to model the outcome on the prescreened predictors and will use backward elimination to choose the final predictors in the model. We expect the fresh, frozen, canned and dried quantity variables to be negatively correlated with each other, and thus we will control for the quantity of the other FV types when performing analyses. We will assess the model assumptions by examining the residual plots and transform the variables before the modelling as necessary. To examine whether the impact of GusNIP PPR eligibility on the quantity and variety outcomes was moderated by the individual, household and community/food environment variables, we will analyse in separate models by examining the interaction between PPR eligibility and these variables. In case there are significant interactions, we will perform post-hoc analysis to evaluate the different impacts within subgroups defined by these characteristics. For the variety outcomes, which are count data, we will use generalised linear models with a log link function (ie, Poisson distribution) and follow a similar model building scheme as detailed above. We will examine the food-at-home quantity and variety outcomes as the primary analyses and perform sensitivity analyses by combining the food-at-home and food-away-from-home data. We will assess the impact of the missingness by examining the patterns of the missing data and if the missingness is substantial, we will use multiple imputation methods to address the missing data issue and will perform sensitivity analyses as necessary. For all models, we will consider the FoodAPS design features including household weights and the clustering in primary sampling units.³⁷ All statistical analyses will be conducted with the SAS software (V.9.4).

Discussion

Strengths and limitations

We plan to use the robust FoodAPS data set for a secondary data analysis to examine the potential limitations of the USDA GusNIP PPR policy that requires PPR participants with low income and risk for diet-related conditions to redeem prescriptions for fresh FVs only. We are uniquely positioned to conduct this research and interpret results to inform GusNIP-relevant policy decisions as representatives of the GusNIP NTAE. Despite limitations, the FoodAPS is the best available data set for this research. The nationally representative FoodAPS sampling methodology used a stratified sampling technique by income and also used several dietary assessment methods (eg, receipts, scanning) to gain a robust picture of household purchasing patterns over the period of a week.²³ Other available scanner data sets for household food purchasing typically under-report FV purchases, compared with the FoodAPS methodology.²⁸

However, we acknowledge important limitations of this work. The timeframe of FoodAPS data collection period (2012–2013) is dated and the purchasing data may not reflect current household purchasing practices among households eligible for GusNIP PPR. As a recent example, the COVID-19 pandemic influenced household members’ interaction with the food environment and resulted in increased food prices.^{38 39} Furthermore, household food purchases do not necessarily equate to individual dietary practices³² and FoodAPS food records tended to drop off towards the end of the data collection period, as is typical of research engagement.²⁶ Detailed documentation is available that addresses data inconsistencies and limitations pertaining to FoodAPS.²³ Additionally, purchased FVs cannot be identified in FoodAPS by the products with or without added sugars, which is a limitation that does not allow for us to examine FV purchasing by products with and without added sugars, fats and sodium (ie, the GusNIP NI requirements).¹⁷ If the GusNIP PPR policy were extended to allow frozen, canned and dried FVs, it is likely that these would need to include no added ingredients. However, FoodAPS data predates GusNIP, and households would likely not have been incentivised to purchase frozen, canned or dried FVs without added sugars, fats and sodium. Last, we are limited by the variables available to us through FoodAPS and certain factors that influence a household’s likelihood to purchase different types of FVs (eg, household member disability status) may be excluded from our analysis.

Ethics statements

Patient consent for publication

Not applicable.