Factors associated with discontinued participation in a fruit and vegetable incentive program in Seattle, Washington (United States): The role of community-based organizations in engaging non-English speakers

Abstract

Objective

To identify factors associated with discontinued participation from Fresh Bucks – a fruit and vegetable (F&V) incentive program in Seattle, WA, USA – two years after initial enrollment, particularly among non-English speakers.

Methods

Administrative and survey data from first enrollment application (Jan-Mar 2020) and re-application (Jul-Oct 2021) were used to identify program discontinuation (n = 12,813). Linear probability models included age, race/ethnicity, preferred language, enrollment pathway (public vs. community-based organization [CBO]-assisted), F&V consumption (<three times/day, ≥three times/day), household income, presence of children, and food security. Pooled and language-stratified analyses were run.

Results

Rates of program discontinuation were 46% (total), 37.6% (English speakers), and 53.3% (non-English speakers). In pooled analyses, age ≥ 30 years (vs. 18–29); having children; income-to-poverty ratio (IPR) = 100–200% or IPR > 200% (vs. IPR < 100%); being non-Hispanic Asian (vs. non-Hispanic White); speaking Chinese or Spanish (vs. English); and being food secure predicted lower probability of program discontinuation. Being non-Hispanic Black (vs. non-Hispanic White) predicted higher probability of program discontinuation. CBO-assisted enrollment (vs. public) predicted lower probability of discontinuation for non-English speakers (B = -7.3 percentage points, 95%CI = −10.5, −4.1).

Conclusions

Age, race/ethnicity, preferred language, income, and food security predicted program discontinuation. Community engagement was important for retaining non-English speakers, offering insights for improving retention in F&V incentive programs.

Highlights

• •Nutrition incentive programs seek to reduce inequities in healthy food consumption.
• •Fresh Bucks provides $40/month to low-income Seattleites to buy produce.
• •In 2021 Fresh Bucks requested participants to re-apply to retain benefits.
• •Fresh Bucks discontinuation rates were higher for English than non-English speakers.
• •Community-based partnerships were important to retain non-English speakers.

1. Introduction

Fruits and vegetables (F&V) are critical components of a healthy dietary pattern, yet their consumption in the United States (U.S.) is low. In 2019, 12.3% and 10% of U.S. adults met fruit and vegetable intake recommendations, respectively, with significant variations by sociodemographic characteristics (Lee et al., 2022). Only 6.8% of U.S. adults in low-income households met vegetable recommendations, compared to 12.2% of adults in the highest income category (Lee et al., 2022). This disparity is partly explained by the higher cost of F&V, relative to more energy-dense foods (Yeh et al., 2008; Darmon and Drewnowski, 2015). Given that inadequate F&V consumption is associated with a higher burden of chronic disease (Xu et al., 2025) and that diet-related chronic diseases are more prevalent among lower-income groups (Abdalla et al., 2020), addressing income disparities in F&V consumption is a public health priority.

A strategy for addressing F&V consumption disparities is F&V incentive programs, which provide low-income households with dollar-to-dollar match incentives or direct funds for the purchase of F&V. Fresh Bucks is one such program, providing low-income households in Seattle, WA, U.S. with $40/month to purchase F&V in participating food outlets. Participant eligibility criteria include living within the City of Seattle limits and having a household income ≤80% of the Area Median Income. Since 2018 Fresh Bucks has been funded by revenue from the city's tax on sugar-sweetened beverages, which has allowed it to significantly expand, now serving ∼17,000 households (City of Seattle, Office of Sustainability and Environment, 2026).

Fresh Bucks allows eligible Seattle residents to apply for the program via their website. In addition, Fresh Bucks partners with community-based organizations (CBOs) to assist with enrolling and engaging traditionally underserved groups, including non-English speakers. Engagement of CBOs with Fresh Bucks participants beyond enrollment assistance ranges from answering their questions about the program or how to use its benefits, to assisting participants with their Fresh Bucks-related shopping at authorized retailers. Previous research found that Fresh Bucks participation reduced food insecurity and increased F&V consumption (Knox et al., 2025). Further, research showed that redemption of Fresh Bucks benefits was higher among Chinese-, Vietnamese-, and Spanish-speaking households, compared to English-speaking households (Wallace et al., 2026a).

Unlike other nutrition incentive programs which often have limited durations, Fresh Bucks allows participants to stay in the program long-term. In 2021, Fresh Bucks opened enrollment for new participants and requested existing participants to re-apply to maintain benefits starting January 1, 2022. Existing participants were notified of the re-application requirement via four mailers sent to their home addresses between July and September 2021; these mailers contained other program-related updates/information and were not exclusively focused on re-application information. The window for re-application via the website was October 11–29, 2021, while participants enrolled via CBOs were re-enrolled continuously between July 26–October 29, 2021.

The aims of this study were to identify factors associated with discontinuation of Fresh Bucks participation between 2020 and 2022, when participant re-application was required, and to assess if these factors varied by participants' preferred language. These research questions have important programmatic implications, informing the retention efforts of Fresh Bucks and other similar F&V incentive programs, particularly for historically underserved groups.

2. Methods

2.1. Study design and population

This is a cross-sectional study focused on Fresh Bucks participants who enrolled in the program in 2020 and were asked to re-apply in 2021 (n = 12,814). De-identified administrative data collected by Fresh Bucks at program enrollment (January–March 2020) and re-application (July–October 2021) were used to identify participants who discontinued the program (dependent variable). Baseline administrative data plus baseline survey data (linked together by Fresh Bucks staff and de-identified prior to sharing) were used to obtain independent variables. From a starting sample of n = 12,814, n = 1630 participants were removed because of missing independent variables, for a final analytical sample of n = 11,184 for multivariable models.

2.2. Measures

Independent variables obtained from administrative data included participants' age, race/ethnicity (non-Hispanic Asian, non-Hispanic Black, Hispanic/Latino, non-Hispanic White, non-Hispanic Other – hereafter Asian, Black, Hispanic/Latino, White, and Other), preferred language (Chinese, English, Spanish, Vietnamese, Another non-English language), having a child in the household (yes/no), household size, household income, and Fresh Bucks enrollment pathway, categorized as public (i.e., participants enrolled via the Fresh Bucks website) or CBO-assisted (i.e., participants enrolled via a Fresh Bucks-affiliated CBO). Household income-to-poverty ratio (IPR) was calculated using household income and size, plus the 2021 federal poverty guidelines.

Household food security status and participants' F&V consumption were obtained from the baseline survey. Household food security was assessed with the two-item Hunger Vital Sign™ (Hager et al., 2010); participants who responded positively to at least one question were considered food insecure. F&V consumption was assessed with a modified version of the Behavior Risk Factor Surveillance System Fruit and Vegetable Module Questionnaire (Lee and Moore, 2020), which excluded 100% fruit juice. The screener asks about frequency of consumption of fruits, green leafy or lettuce salads, fried potatoes, other potatoes, and other vegetables during the previous month, with response options ranging from “never” to “two or more times per day.” For analysis, all responses were converted into daily frequency of F&V consumption, categorized into <three times/day, ≥three times/day, or missing.

2.3. Statistical analysis

Characteristics of Fresh Bucks participants were summarized using descriptive statistics, for the full sample and stratified by program discontinuation. Probability of program discontinuation was estimated by fitting a linear probability model with robust standard errors in the full sample, with program discontinuation (yes/no; 1/0 transformed to 100/0 to express coefficients in percentage points) as dependent variable and the following independent variables: participants' age, race/ethnicity, preferred language, enrollment pathway, F&V consumption, presence of children in the household, household IPR, and food security. To assess if factors associated with program discontinuation varied by preferred language, analysis were stratified by language preference, running separate linear probability models with the same parametrization for those who preferred to speak English (hereafter called English speakers) and those who preferred to speak a non-English language (hereafter called non-English speakers). Linear probability models were chosen given their computational and interpretation advantages over logit or probit models (avoiding nonlinear transformations or calculations of marginal effects), with the results interpreted as risk differences as opposed to ratios (Cheung, 2007). Analyses were conducted in Stata 19, with p-value<0.05 denoting statistical significance.

Given the use of deidentified secondary data, this study was deemed not human subjects research by the University of Washington IRB (IRB ID: STUDY00015977).

3. Results

Table 1 presents Fresh Bucks participants' baseline sample characteristics by program discontinuation. Almost half (46%) of Fresh Bucks participants discontinued participation when prompted to re-apply. Age composition differed by program discontinuation, with a lower proportion of older (60+ years) participants in the discontinuation (32.5%) vs. the continuation (50.7%) group. The discontinuation group had a lower proportion of households with children (61.2%) than the continuation group (68.0%). There were also differences by race/ethnicity and preferred language, with a lower proportion of Asian (23.8% vs. 39.9%) and Chinese speakers (8.6% vs. 22.2%) and a higher proportion of Black (32.4% vs. 22.1%) and English speakers (64.4% vs. 48.8%) in the discontinuation vs. continuation groups. Food insecurity prevalence was 89% in the discontinuation group and 86.5% in the continuation group. There were no meaningful differences in program discontinuation by program enrollment pathway, household IPR, or F&V consumption.

Table 1.

Sample characteristics of participants of Fresh Bucks – a fruit and vegetable incentive program in Seattle, Washington (USA) – at baseline (Jan-Mar 2020), by Fresh Bucks discontinuation (n = 12,814).

	Continued participation N (%)	Discontinued participation N (%)	Total sample N (%)
N	6865	5949	12,814
Age (years)
18–29	513 (7.5)	990 (16.9)	1503 (11.9)
30–39	1061 (15.6)	1294 (22.0)	2355 (18.6)
40–49	942 (13.9)	940 (16.0)	1882 (14.8)
50–59	840 (12.4)	741 (12.6)	1581 (12.5)
60–69	1350 (19.9)	853 (14.5)	2203 (17.4)
70+	2095 (30.8)	1055 (18.0)	3150 (24.9)
Race/ethnicity
Non-Hispanic Asian	2647 (39.9)	1357 (23.8)	4004 (32.5)
Non-Hispanic Black	1465 (22.1)	1843 (32.4)	3308 (26.8)
Hispanic/Latino	750 (11.3)	667 (11.7)	1417 (11.5)
Non-Hispanic White	1232 (18.6)	1153 (20.2)	2385 (19.3)
Non-Hispanic Other	546 (8.2)	675 (11.9)	1221 (9.9)
Preferred language
Chinese	1526 (22.4)	514 (8.7)	2040 (16.0)
English	3351 (49.1)	3830 (64.8)	7181 (56.4)
Spanish	610 (8.9)	470 (8.0)	1080 (8.5)
Vietnamese	538 (7.9)	357 (6.0)	895 (7.0)
Another non-English	799 (11.7)	739 (12.5)	1538 (12.1)
Child in the household
Yes	4605 (68.0)	3598 (61.2)	8203 (64.9)
No	2165 (32.0)	2278 (38.8)	4443 (35.1)
Household income-to-poverty ratio
<100%	4072 (59.3)	3563 (59.9)	7635 (59.6)
100–200%	1634 (23.8)	1280 (21.5)	2914 (22.7)
>200%	1159 (16.9)	1106 (18.6)	2265 (17.7)
Enrollment pathway
Public	4068 (59.3)	3413 (57.4)	7481 (58.4)
Community based organization-assisted	2797 (40.7)	2536 (42.6)	5333 (41.6)
Food security status1
Food secure	839 (13.5)	594 (10.9)	1433 (12.3)
Food insecure	5357 (86.5)	4874 (89.1)	10,231 (87.7)
Fruit and vegetable consumption2
<3 times per day	5334 (77.7)	4773 (80.2)	10,107 (78.9)
≥3 times per day	1216 (17.7)	911 (15.3)	2127 (16.6)
Missing	315 (4.6)	265 (4.5)	580 (4.5)

¹Measured with the two-item Hunger Vital Sign™ (Hager et al., 2010); participants who responded positively to at least one of the two questions were considered to live in a food insecure household.

²Measured with a modified version of the Behavior Risk Factor Surveillance System Fruit and Vegetable Module Questionnaire (Lee and Moore, 2020), excluding 100% fruit juice.

Table 2 presents the linear probability models' results. Across specifications, the variable with the highest magnitude of association with discontinued Fresh Bucks participation was age, with probability of discontinuation significantly decreasing as age increased. Compared to those aged 18–29 years, being 70+ years old was associated with 25.8 percentage points lower probability of discontinuation in the main model (95% confidence interval [CI] = −29.4, −22.3), with similar magnitudes observed in the stratified analysis for both language preference groups. In the main model, race/ethnicity and preferred language were also significantly associated with program discontinuation: Self-identifying as Asian and reporting preference for speaking Chinese, Spanish, or Another non-English language were associated with 4.6 (95%CI_Asian = −8.4, −0.8), 14.9 (95%CI_Chinese = −18.9, −10.9), 10.4 (95%CI_Spanish = −16.2, −4.5), and 6.3 (95%CI_Another = −9.4, −3.1) percentage points lower probability of discontinuation, respectively, compared against White participants and English speakers. Conversely, self-identifying as Black, compared to White, was associated with 5.7 percentage points higher probability of program discontinuation (95%CI = 2.7, 8.6).

Table 2.

Results from linear probability models identifying factors associated with discontinuation of participation in Fresh Bucks – a fruit and vegetable incentive program in Seattle, Washington (USA) – between Jan-Mar 2020 and Jul-Dec 2021; main model and stratified by preferred language.

	Main model (n = 11,184)		Stratified models
			Preferred English (n = 6327)		Preferred non-English language (n = 4857)
	B	95% CI	B	95% CI	B	95% CI
Age (years)
18–29	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
30–39	−9.3**	−12.7, −5.9	−7.0**	−10.9, −3.1	−14.3**	−21.2, −7.4
40–49	−13.2**	−16.7, −9.7	−11.7**	−15.9, −7.6	−15.9**	−22.8, −9.0
50–59	−17.7**	−21.4, −14.1	−18.5**	−22.8, −14.2	−16.8**	−24.2, −9.5
60–69	−23.7**	−27.2, −20.2	−24.9**	−29.2, −20.7	−22.4**	−29.7, −15.2
70+	−25.8**	−29.4, −22.3	−26.8**	−31.3, −22.4	−25.6**	−32.8, −18.3
Race/ethnicity
Non-Hispanic Asian	−4.6*	−8.4, −0.8	−7.0**	−11.1, −2.8	−7.0	−16.0, 2.1
Non-Hispanic Black	5.7**	2.7, 8.6	4.0*	0.8, 7.2	11.0*	1.2, 20.9
Hispanic/Latino	2.4	−3.0, 7.9	2.6	−3.0, 8.3	4.8	−5.1, 14.8
Non-Hispanic White	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
Non-Hispanic Other	5.2**	1.6, 8.9	3.1	−0.8, 7.0	12.8*	0.9, 24.8
Preferred language
Chinese	−14.9**	−18.9, −10.9
English	Ref.	Ref.
Spanish	−10.4**	−16.2, −4.5
Vietnamese	−4.2	−8.9, 0.6
Another non-English	−6.3**	−9.4, −3.1
Child in the household
Yes	−3.8**	−6.2, −1.4	−3.8*	−6.7, −0.8	−3.2	−7.3, 0.9
No	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
Household income-to-poverty ratio
<100%	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
100–200%	−6.1**	−8.4, −3.8	−7.6**	−10.5, −4.6	−1.7	−5.5, 2.1
>200%	−4.4**	−7.0, −1.8	−5.0**	−8.1, −1.9	−3.8	−8.5, 0.9
Enrollment pathway
Public	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
Community-based organization-assisted	−1.4	−3.4, 0.6	2.5	−0.1, 5.1	−7.6**	−10.7, −4.5
Food security status1
Food insecure	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
Food secure	−3.6*	−6.4, −0.8	−3.2	−7.1, 0.8	−4.8*	−8.7, −0.9
Fruit and vegetable consumption2
<3 times per day	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
≥3 times per day	0.6	−1.8, 3.0	−0.1	−3.5, 3.2	1.2	−2.2, 4.5
Missing	10.0	−3.2, 23.3	10.8	−5.2, 26.8	8.7	−15.9, 33.4

* p < 0.05, ** p < 0.01.

¹Measured with the two-item Hunger Vital Sign™ (Hager et al., 2010); participants who responded positively to at least one of the two questions were considered to live in a food insecure household.

²Measured with a modified version of the Behavior Risk Factor Surveillance System Fruit and Vegetable Module Questionnaire (Lee and Moore, 2020), excluding 100% fruit juice.

Having a child in the household (B_Child = −3.8, 95%CI = −6.2, −1.4), a higher IPR (B_{(100–200% IPR)} = −6.1, 95%CI = −8.4, −3.8; B_{(>200% IPR)} = −4.4, 95%CI = −7.0, −1.8), and food security (B_{food secure} = −3.6, 95%CI = −6.4, −0.8) were associated with lower probability of program discontinuation, compared with not having a child, IPR < 100%, and food insecurity, respectively. Enrollment pathway was not associated with program discontinuation in the main model.

In models stratified by language preference, age, race/ethnicity, having a child in the household, and IPR were significantly associated with program discontinuation for English speakers, following the same direction and similar magnitude as the full sample (Table 2). For non-English speakers, age and food security were significantly associated with program discontinuation in the same direction and similar magnitude as the full sample. Program enrollment pathway emerged as a significant factor associated with program discontinuation for non-English speakers, for whom CBO-assisted enrollment (vs. public) was associated with 7.6 percentage points lower probability of program discontinuation (95%CI = −10.7, −4.5) (Table 2).

4. Discussion

This study aimed to identify factors associated with discontinuation of Fresh Bucks participation when participants were requested to re-apply to retain their benefits. Further, based on previous research indicating higher program engagement among Chinese, Vietnamese, and Spanish speakers (vs. English speakers) (Wallace et al., 2026a), this study assessed if factors associated with program discontinuation varied by participants' preferred language. Results show that 46% of program participants discontinued participation during the evaluated period. Older age, being Asian, non-English speakers, having a child in the household, living above poverty, and being food secure at baseline were associated with lower probability of program discontinuation. In addition, CBO-assisted enrollment, compared to public enrollment, was associated with lower probability of program discontinuation among non-English speakers.

The finding that almost half of Fresh Bucks participants discontinued participation when prompted to re-apply is an important one. Given the study data it is not possible to ascertain if program discontinuation was due to death, loss of eligibility (i.e., income changes, relocation outside Seattle), communication barriers, or administrative burden. Administrative burden, including re-certification requests, is a known determinant of discontinued participation for food and nutrition assistance programs, even when financial need persists (Gray, 2019; Fox et al., 2023; Kenney et al., 2022). The timeline of this study corresponds with the COVID-19 pandemic, which may have influenced the rate of Fresh Bucks program discontinuation in two ways. First, pandemic-related social distancing rules made it difficult to communicate and engage with Fresh Bucks participants, particularly via CBOs, as many in-person events and engagements were cancelled. This may have hindered the flow of information regarding the need to re-apply and prevented people from re-applying via CBOs. Second, there was an unprecedented increase in financial and food assistance from the federal government during COVID-19 (Patton et al., n.d), which remained in place in Washington state until 2022–2023 (Washington State Deparment of Social and Health Services, Economic Services Administration, 2026), past the October 2021 deadline for Fresh Bucks re-application. Local governments and non-profit organizations also launched and/or expanded programs to address the pandemic-induced increase in food insecurity, for example by delivering food boxes to low-income households (United Way King County, n.d). It is possible that having access to food via other programs and/or receiving larger-than-usual benefits from usual programs influenced the decision of people to re-apply for Fresh Bucks. Previous research investigating reasons why families discontinue participation in the Special Supplemental Nutrition Program for Women, Infants and Children (WIC) while still eligible have identified receipt of food from other sources (e.g., Supplemental Nutrition Assistance Program [SNAP], food banks) as an important factor (Terry et al., 2024), particularly when the other program benefits are perceived as more generous or flexible (Weber et al., 2019). On the other hand, food inflation was still very high in the late quarter of 2021 (U.S. Bureau of Labor Statistics, 2026), and multiple program participation was common for certain low-income households during the pandemic (Chaparro et al., 2025a).

The association between age and Fresh Bucks program discontinuation was the one with the highest magnitude, with probability of program discontinuation decreasing in a somewhat linear fashion as age increased. Participants aged 70+ years had 26 percentage points lower probability of program discontinuation than those aged 18–29 years. It is important to note that some Fresh Bucks-affiliated CBOs focus on older adult services, which may partly explain their high engagement with Fresh Bucks. A previous study of Fresh Bucks participants found that adults 60+ years had 3.5 times the odds of redeeming >75% of their Fresh Bucks benefits (vs. <25% of the benefit), compared to 18–39-year-olds (Wallace et al., 2026b). Combined, these results suggest that older adults are highly engaged in the program. This is unlike other food assistance programs such as SNAP, for which participation rates of adults 60+ years are significantly lower than for those aged <60 years (55% vs. 88% in FY2022) (Vigil and Rahimi, 2024).

Those who were food secure, compared to food insecure, and those who were above the poverty threshold, compared to below the poverty threshold, were found to have lower probability of Fresh Bucks program discontinuation. These results are counterintuitive, since one would anticipate that those most in need would decide to continue participating in Fresh Bucks when given the opportunity (though it is important to underscore that all Fresh Bucks participants are low income). Contrary to this study's results, Kenney et al. (2022) found that lower-income households with children were more likely to leave SNAP due to administrative barriers. On the other hand, Gray (2019) found different responses by low-income households to different types of administrative burden, with higher-income households with children (within the definition of ‘low-income’) more likely to discontinue SNAP if faced with increased paperwork burden. Research on WIC participation is mixed; living below the poverty level was associated with lower redemption of WIC foods (Anderson and Whaley, 2025) – and lower redemption of WIC foods with higher likelihood of WIC program discontinuation (Anderson et al., 2023) – among WIC participants in Southern California, whereas a lower income was associated with longer duration of WIC participation in a longitudinal analysis of national data (Anderson et al., 2022). Another explanation for this study's results may be that these most-in-need groups were receiving other income and/or food assistance in addition to Fresh Bucks which, as stated earlier, may have been more generous than usual given pandemic-related augmentations and may have led to a reduced need to re-apply for Fresh Bucks. Lastly, lowest-income households may have been more likely to be priced out of Seattle, leading them to move out of the Fresh Bucks geographic eligibility area. However, eviction moratoriums and other rental assistance implemented during COVID-19 (Rental Housing Association of WA, 2026) may have prevented low-income households from needing to move out of Seattle during this time.

Race/ethnicity was associated with program discontinuation. Compared to White participants, Asian participants had lower, and Black participants had higher, probability of program discontinuation. Previous research on Fresh Bucks benefit redemption reports similar findings, with Asian households having higher, and Black households having lower, benefit redemption rates, than White households (Wallace et al., 2026a). Research on WIC food benefit redemption also found lower redemption rates among Black families (Chaparro et al., 2025b). On the other hand, a rigorous evaluation on the impact of Fresh Bucks found that Black participants increased their F&V consumption because of Fresh Bucks participation (Knox et al., 2025). While structural disadvantages faced by Black households may partly explain this study's results (Phelan and Link, 2015), to inform retention efforts, future research should investigate why Black participants were more likely to discontinue Fresh Bucks participation when prompted to re-apply.

Preferring to speak Chinese, Spanish, or another non-English language (vs. English) was associated with lower probability of program discontinuation. Similar findings were reported previously in terms of Fresh Bucks benefit redemption (Wallace et al., 2026a). It is important to note that non-English speakers are a target population of Fresh Bucks-affiliated CBOs, with some CBOs focusing on specific cultural groups and, thus, providing culturally- and linguistically-tailored support to those groups. In fact, when the sample was stratified by language preference, CBO-assisted enrollment (vs. public) emerged as a significant factor associated with lower probability of program discontinuation among non-English speakers. Previous qualitative studies investigating barriers and facilitators of implementation of Food Is Medicine programs in clinical settings (e.g., produce prescription programs, medically tailored meals) identified community engagement/partnership as an important facilitator (Houghtaling et al., 2025; Houghtaling et al., 2024; McWhorter et al., 2023). Furthermore, a process evaluation of Health Bucks, a farmers' market coupon incentive program in New York City, reported that using CBOs for coupon distribution may have increased awareness of the program as well as facilitated nutrition education activities carried out by CBOs (Payne et al., 2013). Other public health research has also highlighted the importance of partnering with CBOs to recruit and retain underserved populations in health promoting programs (Steinman et al., 2022; O'Mara-Eves et al., 2015).

This study has strengths and limitations. Strengths include the availability of longitudinal administrative program data, which allowed for the identification of those who discontinued participation when requested to re-apply. The large sample size and diversity of Fresh Bucks participants in terms of sociodemographic characteristics are also strengths, providing sufficient power to assess how probability of program discontinuation varied by these characteristics. Another strength is the use of food security (Hager et al., 2010) and F&V consumption (Lee and Moore, 2020) measurement tools that have been previously validated. In terms of limitations, the data used did not provide reason(s) for program discontinuation (e.g., administrative burden vs. death, loss of eligibility, communication barriers, other). The study is based on one city-wide F&V incentive program, so results may not be generalizable to other incentive programs with different target populations, eligibility criteria, benefit amounts, and/or incentive scheme. It is important to note, though, that most of the study results are similar to those previously reported across a variety of food and nutrition assistance programs. Lastly, the overlap of the study dates with the COVID-19 pandemic is an important limitation; it is possible that the factors identified as associated with program discontinuation would be different if assessed at a different time. However, results from this study are highly relevant in today's context in which funding for, and access to, food and nutrition assistance at the local, state, and federal levels are being constrained. For Fresh Bucks and other similar programs, knowing which sociodemographic groups would be most likely to discontinue participation if requested to re-apply (due to funding or other constraints) is crucial information for resource prioritization and retention efforts.

5. Conclusions

While age, race/ethnicity, language, and income were key factors associated with Fresh Bucks program discontinuation, Fresh Bucks was able to retain a substantial share of the city's most socially and economically vulnerable households, especially older adults and non-English speakers, priority groups who are targeted for enrollment and engagement via Fresh Bucks-affiliated CBOs. These findings highlight the importance of using community engagement strategies to maximize the long-term impact of healthy food voucher programs for historically underserved populations.

CRediT authorship contribution statement

M. Pia Chaparro: Writing – original draft, Methodology, Investigation, Conceptualization. Jong Min Jung: Writing – review & editing, Software, Formal analysis. Jamie Wallace: Writing – review & editing, Software, Investigation, Formal analysis, Data curation. Jessica Jones-Smith: Writing – review & editing, Methodology, Investigation, Funding acquisition, Conceptualization. Barbara Baquero: Writing – review & editing, Investigation, Funding acquisition, Conceptualization. Melissa Knox: Writing – review & editing, Supervision, Software, Project administration, Methodology, Investigation, Funding acquisition, Conceptualization.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Jessica Jones-Smith, Barbara Baquero, Melissa Knox (mPIs) reports financial support was provided by National Institutes of Health. M. Pia Chaparro reports a relationship with US Department of Agriculture that includes: funding grants. Jessica Jones-Smith, Jamie Wallace, Melissa Knox reports a relationship with Bloomberg Family Foundation Inc. that includes: funding grants. Jessica Jones-Smith, Melissa Knox reports a relationship with City of Seattle that includes: funding grants. M. Pia Chaparro reports a relationship with American Heart Association Inc. that includes: funding grants. Jessica Jones-Smith, Barbara Baquero, Melissa Knox reports a relationship with National Institutes of Health that includes: funding grants. Barbara Baquero reports a relationship with Latino Community Health Board that includes: board membership. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

The authors do not have permission to share data.