ARTICLE IN PRESS

Research Brief
A Pediatric Fruit and Vegetable Prescription Program
Increases Food Security in Low-Income Households
D1X XRonit A. Ridberg, D2X XPhD, MS1; D3X XJanice F. Bell, D4X XPhD, MN, MPH1; D5X XKathryn E. Merritt, D6X XMPH2;
D7X XDiane M. Harris, D8X XPhD, MPH3; D9X XHeather M. Young, D10X XPhD, RN, FAAN1;
D1X XDaniel J. Tancredi, D12X XPhD4,5

ABSTRACT
Objective: To assess change in household food security associated with participation in a pediatric fruit/
vegetable prescription program.
Methods: The researchers analyzed clinic-based, fruit/vegetable prescription program data for 578 low-
income families, collected in 2013 2015, and calculated changes in food security (summative score; high/
low/very low; and individual US Department of Agriculture measures).
Results: Of participating households, 72% increased their summative score over the course of the program.
In adjusted regression models, participants had higher change scores with 5 or 6 clinical visits, compared
with 1 or 2 visits (b = .07; 95% confidence interval, 0.01 0.14), and college education of the primary care-
taker, compared with less than college (b = .05; 95% confidence interval, 0.01 0.09). Select clinic sites
(but neither visit nor redemption proportions) significantly contributed to change score variance. All US
Department of Agriculture measures saw significant increases.
Conclusions and Implications: Fruit/vegetable prescription programs may help providers address
patients’ food insecurity. Further research using experimental designs and implementation science could
build the case to incorporate programs into practice.
Key Words: food security, fruit and vegetable prescription program, nutrition incentive programs (J Nutr
Educ Behav. 2018; 000:1 8.)
Accepted August 3, 2018.

INTRODUCTION performance,3 social development,4 incentives at the point of purchase.

and health care use.5 Food-insecure Incentive programs have increased
In 2016, 15.6 million American adults have higher probabilities of FV purchasing,8 increased and diver-
households experienced food insecu- chronic disease6 and food-insecure sified FV consumption,9,10 and
rity (defined by the US Department of households have higher annual improved food security (FS)11 for par-
Agriculture [USDA] as lacking consis- health care costs.7 ticipants. Ongoing cost-effectiveness
tent, dependable access to enough Nutrition incentive programs that studies suggest that nationwide FV
food for active, healthy living1), with match federal benefits with coupons subsidies could reduce the incidence
higher rates among households led or discounts at farmers’ markets or of type 2 diabetes, myocardial infarc-
by Hispanic or black individuals and grocery stores may mitigate the sever- tion, and stroke and provide disease-
among lower-income households.1 ity of food insecurity by improving related, long-term cost savings.12
Food insecurity has a negative impact access and affordability to fruits and Only a few studies in the literature
on children’s health,2 academic vegetables (FV) with financial have evaluated FV prescription pro-
grams, a collaborative nutrition
incentive model in which health care
1
Betty Irene Moore School of Nursing, University of California Davis, Sacramento, CA providers dispense FV prescriptions
2
Wholesome Wave, Berkeley, CA to select patient populations in the
3
Division of Nutrition, Physical Activity, and Obesity, Centers for Disease Control and Pre- form of coupons or vouchers for use
vention, Atlanta, GA at participating outlets, supple-
4
Center for Healthcare Policy and Research, University of California, Davis, Sacramento, CA mented by nutrition education in
5
Department of Pediatrics, University of California, Davis, Sacramento, CA the clinic. Outcome measures in
Conflict of Interest Disclosure: The authors’ conflict of interest disclosures can be found online these studies, all of which focus on
with this article on www.jneb.org. adults, included decreased hemoglo-
Address for correspondence: Ronit A. Ridberg, PhD, MS, Betty Irene Moore School of bin A1c levels,13 increased FV con-
Nursing, University of California, Davis, 2450 48th St, Sacramento, CA 95817; Phone: (916) sumption,14 and decreased body
832-7785; Fax: (916) 734-3257; E-mail: raridberg@ucdavis.edu mass index.15 Importantly, although
Ó 2018 Society for Nutrition Education and Behavior. Published by Elsevier Inc. All rights health care experts place growing
reserved. emphasis on social determinants of
https://doi.org/10.1016/j.jneb.2018.08.003 health and the industry focuses on

Journal of Nutrition Education and Behavior  Volume 000, Number 000, 2018 1

2 Ridberg et al
innovations in payment and service
delivery to the authors’ knowledge,
no prescription studies to date have
assessed changes in household-level
FS. The current study addresses this
gap using a multiyear data set from
the national nonprofit Wholesome
Wave’s pediatric Fruit and Vegetable
Prescription Program (FVRx) specifi-
cally to examine FS behaviors and
status for participating households.
Findings may inform policy debate
and funding decisions regarding pre-
scription programs’ potential to
increase FS and improve public health.
METHODS
The Wholesome Wave FVRx program
was a 4- to 6-month intervention
offered since 2011 in select cities and
states across the country. In pediatric
programs during 2013 2015, health
care providers at federally qualified
health centers enrolled children and
youth (aged 2 18 years; 1/house-
hold) who were clinically obese or
overweight; inclusion criteria con-
sisted of this diagnosis based on body
mass index weight-for-age; parent
consent and patient willingness to
participate; and the ability to make at
least 3 program visits.16 Program
materials and training were provided
by Wholesome Wave and enrollment
was managed by the clinical sites.
Participants received nutrition edu-
cation by a clinician, nutritionist, or
trained health educator at each clini-
cal visit or in a class setting (approxi-
mately monthly), including
guidance on FV consumption and
replacement of unhealthy foods with
fresh FV.16 Providers distributed pre-
scriptions allocated by household
size ($0.50 to $1.00/person per day:
for example, $28/wk for a family of
4) and shared details of partnering
farmers’ markets, where prescriptions
were redeemed for produce. (More
detailed background about the FVRx
program and its institutional part-
ners is provided in reports available
on the Wholesome Wave website).16
The program collected household
demographic data from parents and
caregivers at the first program visit,
alongside a survey including FS meas-
ures, which was repeated at the last
visit or in a follow-up phone call. The
current study used 2013 2015 FVRx
ARTICLE IN PRESS
Journal of Nutrition Education and Behavior  Volume 000, Number 000, 2018
data collected at 9 clinical sites. The
study was deemed exempt by the
University of California Davis Insti-
tutional Review Board.
Sample
Of 900 households that participated
during 1 of 3 program years in 9 clini-
cal sites (1 each in Maine, Massachu-
setts, New Mexico, Rhode Island, and
the District of Columbia and 4 in New
York), households were excluded if
enrolled children were aged <2 or
>
-- 18 years at the first clinical visit
(n = 6), explicitly dropped out of the
program (n = 32), or were repeating the
program (n = 87), or the households
did not complete the FS screening at
the program’s end (n = 197). Of those
excluded from the analysis, propor-
tions were higher for households
reporting black, African, or Caribbean
American children. Three sites
accounting for nearly half of the sam-
ple had particularly high attrition rates
ranging from 20% to 26% (other sites
ranged from 1% to 10%).
Measures
Change was measured in 5 house-
hold-level behaviors, adapted from
the USDA’s 18-item Household Food
Security Survey.1 Four items asked
the respondent about the frequency
over the previous 3 months (often,
sometimes, or never) with which
they worried that household food
would run out before they had
money to buy more; could not afford
balanced meals; cut or skipped meals
owing to lack of money for food; or
reported that children did not eat for
a whole day owing to lack of money
for food. An additional question
asked the respondent to evaluate the
types and quantities of foods avail-
able in the household over the past 3
months: enough of the kinds of food we
like to eat, enough but not the kinds of
food we like to eat, sometimes not enough
to eat, or often not enough to eat.
Because these 5 items did not
comprise a validated tool on their
own, a summative scale was created
to calculate a change score. Items in
this scale had different response sets;
thus, to make the items commensu-
rate, each was normalized to a 0 1
scale using the equation Y_n
= (Y_o Y_min)/(Y_max Y_min),
in which Y_min and Y_max were
the lowest and highest possible
responses and Y_o and Y_n were the
original and normalized scores,
respectively. As long as the house-
hold had answers to at least 3 of 5 FS
measures, a summative FS score was
calculated by averaging the 5 nor-
malized measures. Internal consis-
tency of the 5 items used to form
the summative score was good
(Cronbach a = .76) and unidimen-
sionality was verified with Horn’s
parallel analysis.17 An FS change
score was then calculated as the pri-
mary outcome by subtracting the
summative score at first visit from
the summative score at last visit
(continuous variable, 0 1).
Households with answers to all 5
questions (n = 548) were also assigned
the FS categories favored by the USDA
(high/marginal, low, or very low) by
adapting the raw scoring methodology
from their 6-item short form, in which
responses of sometimes or often are
counted as affirmative answers and
then summed (0 1 = high/marginal;
2 4 = low; and 5 6 = very low).18 The
5-item scale was scored conservatively:
0/1 = high/marginal; 2/3 = low; and 4/
5 = very low.
Sociodemographic characteristics
collected at the first visit (Table 1)
included participant gender, age, and
race/ethnicity; household size;
health insurance type; household
enrollment in federal food assistance
programs, and highest education of
the participant’s mother or primary
caretaker. Program covariates in-
cluded clinical site, program year,
number of clinical visits, and value
and proportion of FVRx prescription
redeemed. Redemption data were
tracked in spreadsheets by farmers’
market managers and later compared
with total prescribed to calculate pro-
portion redeemed.
Statistical Analysis
Stata software (version 13, StataCorp
LP, College Station, TX; 2013) was
used for all statistical analyses. The
researchers used descriptive statistics
to summarize the distribution of study
variables. Paired t tests and McNemar
paired tests19 with Bonferroni correc-
tions compared all FS outcomes
 ARTICLE IN PRESS
Journal of Nutrition Education and Behavior  Volume 000, Number 000, 2018 Ridberg et al 3

between the first and last visits. In

fully adjusted linear regression mod- Table 1. Descriptive Characteristics of Wholesome Wave’s Fruit and Veg-
els, the FS change score was modeled etable Prescription Program Households (1 Participant Each),
as a function of the independent pro- 2013 2015
gram variables, including clinical vis-
its and FVRx redemption over the Household/Participant Characteristics n (%)
course of the program. Independent Total sample (households) 578 (100)
variables measuring exposure to the Year
program were quantified using both
absolute values (ie, number of visits, 2013 201 (34.8)
prescription value totals) as well as rel- 2014 192 (33.2)
ative amounts expressed as a percent- 2015 185 (32.0)
age of maximum theoretical dose (ie,
Participant age at enrollment, y
number of visits as a percentage of
those possible, proportion of FVRx 2 8 206 (35.7)
prescription redeemed). Models were 9 13 270 (46.7)
fit using alternative parameterizations 14 18 102 (17.7)
of the independent variables measur-
Female participant 303 (52.4)
ing exposure. All models used a robust
estimator to account for within-clini- Participant race/ethnicity
cal site correlations while controlling Hispanic or Latino 374 (64.7)
for clinical site fixed effects.20 Models White (not Hispanic) 95 (16.4)
were fit with and without interaction
Black, African, or Caribbean American (not Hispanic) 85 (14.7)
terms (each dosage variable £ clinical
site) using Akaike information crite- Mixed race or other race 24 (4.2)
rion. Statistical significance (P < .05) Health insurance
and parameter estimates (95% confi- Medicaid/public 539 (93.6)
dence intervals [CI]) are reported.
Private insurance 28 (4.9)
RESULTS Uninsured or other insurance 9 (1.2)
Total health clinic visits, n
Most of the enrolled children in the 1 2 57 (9.9)
578 households were aged <13 years
3 4 457 (79.1)
(77%); half were female (52%). More
than two thirds were reported as 5 6 64 (11.1)
Latino (65%), followed by white, not Total FVRx coupons redeemed/household (mean [SD]) $375 ($226)
Hispanic (16%) and black, African or Proportion of FVRx redeemed (mean [SD]) 54% (38%)
Caribbean-American, and non-His-
panic (15%) (Table 1). About half of Days between first and last visits (mean [SD]) 89 (23)
participating households included Highest education of mother/primary caretaker
<--4 members (52%) and 94% were High school classes, degree, or General Equivalency 316 (54.6)
covered by Medicaid or public insur-
Diploma
ance. A total of 72% of households
were enrolled in federal nutrition Some college or more 143 (24.7)
programs (ie, Supplemental Nutrition Household size, n
Assistance Program, Special Supplemen- <
tal Nutrition Program for Women, --4 300 (52.0)
Infants, and Children). In most house- 5 278 (48.1)
holds, the highest education Household enrolled in Supplemental Nutrition Assistance 417 (72%)
obtained by the participant’s mother Program or Special Supplemental Nutrition Program for
or primary caretaker was some high Women, Infants, and Children
school, a high school degree, or a
Largest sites by household (%)
General Equivalency Degree (55%).
Average household redemption of Site 3 131 (22.7)
FVRx prescriptions was $375 (mean Site 5 75 (13.0)
redemption proportion = 54% [SD, Site 8 102 (17.7)
38%]) and most (79%) households
made 3 4 clinical visits over an aver- FVRx indicates Fruit and Vegetable Prescription Program.
age of 3 months. The 3 largest sites
enrolled 23%, 18%, and 13% of total
participants.
 ARTICLE IN PRESS
4 Ridberg et al Journal of Nutrition Education and Behavior  Volume 000, Number 000, 2018

Table 2. Food Security Outcomes, According to Answers Regarding the Past 3 Mo

First Visit Last Visit

Household Food Security n (%) n (%) P
Mean summative food security score 0.72 0.81 < .001
Household food security status < .001
High or marginal 317 (58) 419 (76)
Low 183 (33) 121 (22)
Very low 48 (9) 8 (1)
Worried food would run out < .001
Sometimes or often 359 (66) 292 (53)
Never 189 (34) 256 (47)
Could not afford balanced meals < .001
Sometimes or often 355 (65) 281 (51)
Never 193 (35) 267 (49)
Cut or skipped meals owing to lack of money for food < .001
Sometimes or often 225 (41) 136 (25)
Never 323 (59) 412 (75)
Children did not eat a whole day owing to lack of money for food < .001
Sometimes or often 84 (15) 42 (8)
Never 464 (85) 506 (92)
Types of food in the households < .001
Enough of the kinds of foods we like to eat 174 (32) 338 (62)
Enough but not the kinds of food we like to eat 243 (44) 159 (29)
Sometimes not enough to eat 122 (22) 46 (8)
Often not enough to eat 9 (2) 5 (1)
Notes: Paired t tests were used to compare summative food security score at first and last visits; McNemar tests were used to
compare each individual measure as well as food security status. McNemar test P was computed after dichotomizing the items
as sometimes/often vs never. Results include application of Bonferroni correction. Calculation of summative score required
answers to at least 3 of 5 measures (n = 578), food security status categories required answers to all 5 measures (n = 548), and
individual measures’ response numbers vary as shown.

The normalized mean summative
FS score at the beginning of the pro-
gram was 0.72 (Table 2). More than
two thirds of households (72%)
increased the summative FS score
from beginning to end and the total
sample’s mean change was 0.09 (SD,
0.20). The percentage of participating
families with high/marginal FS
increased from 58% to 76%, those
experiencing low FS decreased from
33% to 22%, and those with very low
FS decreased from 9% to 1%. Individ-
ual measures improved significantly
at the last program visit compared
with the first one (P < .001 for each
measure using McNemar paired tests
after application of Bonferroni cor-
rection): fewer households reported
that they sometimes or often worried
food would run out (53% vs 66%;
chi-square with 1 degree of freedom,
x2(1) = 26.9), could not afford bal-
anced meals (51% vs 65%,
x2(1) = 31.8), cut or skipped meals
(25% vs 41%, x2(1) = 44.3), children
did not eat for a whole day (8% vs
15%, x2(1) = 21.5). Finally, house-
holds reporting that they had
enough of foods they liked to eat, as
well as the kinds of foods, nearly dou-
bled at the last visit compared to the
first one (62% vs 32%, x2(1) = 45.7).
Multiple regression models used
visit numbers and FVRx redemption
proportion as separate independent
variables (models 1 and 2) (Table 3)
whereas both exposure variables
were included in model 3 to yield a
more interpretable effect size for the
relative program exposure parameter.
Models without interaction terms
noted in Methods were a better fit
and therefore are presented in Results
when describing adjusted mean dif-
ferences associated with changes in
visit number, holding other terms
constant. Compared with families
with 1 2 visits, the mean FS change
score was 0.07 higher for families
with 5 6 visits (95% CI, 0.01 0.14).
Households whose primary caretaker
had attained higher education had a
greater change score on average
(model 1: b = .05; 95% CI, 0.01 0.09;
Models 2 and 3: b = .05; 95% CI,
0.01 0.10). Participant households
at site 6, compared with those at site
3 (the site with the most participants
and the reference site21) had a sum-
mative change score 0.10 higher on
average (95% CI, 0.01 0.20; model
1). Neither visit proportion (visits
made / total possible; results not
included) nor redemption propor-
tion (FVRx redeemed / total pre-
scribed) was associated with FS
change score. Each model accounted
for 10% of variance in the change
score, which indicated that there
may have been important omitted
covariates.
 ARTICLE IN PRESS
Journal of Nutrition Education and Behavior  Volume 000, Number 000, 2018 Ridberg et al 5

Table 3. Program Visits, FVRx Redemption Proportion and Sociodemographic Multiple Regression Coefficients for
Food Security Change Score Outcome
Model 1 Model 2 Model 3
(R2 = 0.10) (R2 = 0.10) (R2 = 0.10)

Household/Participant Characteristics b 95% CI b 95% CI b 95% CI

Clinical visits, n (reference 2 visits)
3 4 .03 0.01 to 0.08 .02 0.04 to 0.07
5 6 .07* 0.01 to 0.14 .06 0.02 to 0.13
Proportion of FVRx redemption/household .01 0.04 to 0.17 .02 0.03 to 0.07
Highest education of mother/primary .05* 0.01 to 0.09 .05* 0.01 to 0.10 .05* 0.01 to 0.10
caretaker: some college or more
Race (reference Latino)
Black, African, or Caribbean American .02 0.09 to 0.06 .01 0.10 to 0.07 .02 0.10 to 0.06
(not Hispanic)
White (not Hispanic) .01 0.08 to 0.06 .01 0.08 to 0.07 .01 0.07 to 0.07
Mixed race or other race .07 0.01 to 0.15 .08 0.01 to 0.17 .08 0.01 to 0.17
Age of enrolled child participant, y
(reference 2 3)
4 8 .00 0.09 to 0.08 .00 0.09 to 0.08 .01 0.09 to 0.08
9 13 .06 0.14 to 0.03 .05 0.14 to 0.04 .05 0.14 to 0.03
14 18 .00 0.09 to 0.09 .01 0.08 to 0.10 .01 0.09 to 0.10
Male participant .00 0.03 to 0.03 .00 0.03 to 0.04 .00 0.03 to 0.04
Household size .00 0.01 to 0.01 .00 0.01 to 0.01 .00 0.02 to 0.01
Federal food program participation .02 0.02 to 0.06 .03 0.01 to 0.07 .03 0.01 to 0.07
Program year (reference 2013)
2014 .00 0.04 to 0.05 .00 0.05 to 0.04 .01 0.04 to 0.05
2015 .01 0.06 to 0.04 .03 0.09 to 0.02 .03 0.08 to 0.03
Service location (reference site 3)
Site 1 .04 0.2 to 0.10 .05 0.14 to 0.04 .04 0.14 to 0.06
Site 2 .03 0.06 to 0.12 .02 0.06 to 0.11 .03 0.06 to 0.12
Site 4 .05 0.14 to 0.03 .06 0.14 to 0.03 .05 0.14 to 0.04
Site 5 .04 0.10 to 0.02 .04 0.10 to 0.01 0.04 0.11 to 0.03
Site 6 .10* 0.01 to 0.20 .09 0.01 to 0.18 .10 0.01 to 0.20
Site 7 .09 0.21 to 0.01 .07 0.18 to 0.04 0.06 0.17 to 0.05
Site 8 .00 0.07 to 0.07 .01 0.07 to 0.07 0.01 0.08 to 0.07
Site 9 .05 0.03 to 0.13 .07 0.01 to 0.14 0.05 0.03 to 0.13
*P < .05.
CI indicates confidence interval; FVRx, Fruit and Vegetable Prescription Program; R2, multivariate coefficient.
Notes: Models 1 and 2 capture alternative operationalization of program exposure (number of visits; proportion of prescription
redemption) as variables were correlated at a significant but modest level (r = .159; P < .05). Both measures of exposure are
therefore included in model 3 to yield a more interpretable effect size for the relative program exposure parameter. Robust SEs
were used, to account for heteroscedasticity, residual within-site effects, or other violations of model assumptions.

DISCUSSION insecure behavior measures among program participation. Only 1 of the

To the authors’ knowledge, this was
the first study to assess changes in FS
as a result of participation in a pediat-
ric fruit and vegetable prescription
program. The study found significant
improvements in a summative 5-item
FS scale as well as 5 individual food-
578 low-income households. More-
over, highest program exposure (spe-
cifically, 5 or 6 clinical visits) was
associated with the FS change score,
which suggests that increased levels
of participation could indicate a larger
degree of change in household FS
from the beginning to the end of
other 3 published prescription studies
suggested a dose to outcomes effect:
Diabetic adult patients with more fre-
quent use of the farmers’ market had
higher odds of increasing FV con-
sumption.14 Numerous aspects of the
clinical visit (eg, additional provider
touch-point, nutrition education, the

6 Ridberg et al
prescription refill) or some combina-
tion of factors could have been the
impetus for increased change. For
example, repeated discussions with
health care providers about nutrition
and healthy eating goals reinforced in
each program visit may have resulted
in increased knowledge about where
to obtain and how to prepare fresh
FV, or a strengthened skill set to
stretch a food budget.
The percentage of households
experiencing very low FS was greatly
reduced and those with high/mar-
ginal FS increased > -- 30%, demon-
strating the desired program impact,
which was consistent with Supple-
mental Nutrition Assistance Pro-
gram based incentive programs.11
Particularly encouraging is that
nearly 50% fewer households
reported a child not eating for a
whole day at the last visit, which sug-
gested that prescription program
benefits reached individual family
members as well as the household as
a whole. Program results directly met
select Healthy People 2020 nutrition
goals to increase household FS, child
FS, and nutrition/diet counseling in
pediatric visits.22
The sample started the program, on
average, at a lower level of FS (58%
high/marginal compared with 88%
nationally1) which may be attributed
to household characteristics indepen-
dently associated with lower FS in
national polling, including income
and race/ethnicity of the head of the
household, or that children lived in
the household.1 The significant differ-
ences in FS change based on the educa-
tion of the primary caregiver may
warrant an evaluation of literacy levels
of program educational components,
an appraisal of capturing household
income level, and comparative effec-
tiveness studies to disentangle the
effects of these vs program dosage.
Despite similar training, enroll-
ment methods, educational materi-
als, and survey instruments at all
sites, site-specific differences may
have arisen based on (1) food access
(eg, the number and location of par-
ticipating markets and desirability of
produce selection); (2) capacity (eg,
budgets, staff time, and technical
assistance); (3) patient population
(eg, within-site homogeneity of
children’s race or ethnicity precludes
ARTICLE IN PRESS
Journal of Nutrition Education and Behavior  Volume 000, Number 000, 2018
comparisons of the findings by race/
ethnicity); and (4) delivery of nutri-
tion education (eg, group classes vs
one-on-one or delivered by a nutri-
tionist vs a pediatrician). According
to the former chief programs officer
at Wholesome Wave (S. Cornell, writ-
ten communication, January, 2018),
the site with significantly higher FS
change score (site 6) had a larger bud-
get, received more technical assis-
tance, and had multiple market
access points for prescription
redemption, each of which could
have contributed to the resulting FS
changes. Mean FVRx redemption at
site 6 was about $446, compared
with the average at all other sites of
$367, and 98% of households at the
site made 3 4 visits; other variables
such as levels of mothers’ education
were not significantly different from
those of other sites. Future research
on FV prescription interventions
might explore program fidelity by
site and measure implementation cli-
mate (a construct related to expecta-
tions, support, and rewards, and
which is positively associated with
implementation effectiveness)23 with
each participating clinical location.
These findings from a novel pro-
gram capitalize on existing data;
nevertheless, major inherent
challenges include potential self-
selection bias and the lack of a con-
current control group. Families that
completed the FS survey at the last
visit may have been more engaged
and motivated overall and therefore
had higher improvement scores.
Without a separate comparison
group extrapolating no dose, it is
impossible to determine whether FS
status changes resulted from the pro-
gram or other factors. In addition,
this relatively low-income and
majority Latino or Hispanic sample
may have disproportionately
included individuals with low health
literacy (because both demographic
characteristics have been associated
with lower than average health liter-
acy rates),24 which could have impli-
cations for the quality of self-
reported data as well as the level of
participant engagement in the pro-
gram.25 Social desirability bias may
contribute to positive response ten-
dency, particularly regarding some-
thing as sensitive as whether one’s
children have food to eat, possibly
understating baseline FS status and/
or overstating program impact.26
These potential unmeasured con-
founders may limit generalization of
the findings to other groups. Never-
theless, the study is unique in provid-
ing aggregated results from a
program implemented in the field
instead of a research setting and sets
an important foundation for future
inquiry by demonstrating improved
FS in individuals and households.
IMPLICATIONS FOR
RESEARCH AND PRACTICE
Fruit and vegetable prescription pro-
grams in clinical settings have the
potential to increase FS in low-
income households, with effects
reaching children. Whereas tying
incentive programs to federal supple-
mental nutrition programs can use
established infrastructure, an addi-
tional incentive delivery system via
health care settings may be appropri-
ate. Health care providers are in a
favorable yet underused position to
recognize, screen for, and offer
resources to mitigate food
insecurity.27 29 More robustly con-
trolled trials are warranted to pursue
the question of causality; future com-
parative effectiveness studies are rec-
ommended to disentangle the effects
of the program’s cash incentive and
in-clinic nutrition education compo-
nents, ideally in diverse settings,
using a complete USDA FS instru-
ment and with true experimental
designs.
ACKNOWLEDGMENTS
Support for the first author was pro-
vided by the Betty Irene Moore
School of Nursing at University of
California, Davis. This research
received no specific grant from any
funding agency, commercial, or not-
for-profit sectors. Wholesome Wave
received FVRx program funding from
the Laurie M. Tisch Illumination
Fund, New World Foundation, New-
man’s Own Foundation, Vincent J.
Coates Foundation, and Sampson
Foundation, among others. These
funders had no role in the design,
analysis or writing of the manuscript.
The authors greatly appreciate the

Journal of Nutrition Education and Behavior  Volume 000, Number 000, 2018
helpful guidance and comments of
Sheri Zidenberg-Cherr, PhD, in pre-
paring the manuscript. They grate-
fully acknowledge Michel Nischan
and Shikha Anand for founding con-
tributions to the FVRx program;
Wholesome Wave staff Skye Cornell
and Ashley Fitch; and organizational
partners and clinic sites for program
implementation and data collection.
The findings and conclusions pre-
sented in this article are those of the
authors and do not necessarily repre-
sent official reporting of Wholesome
Wave or the official position of the
Centers for Disease Control and Pre-
vention.
REFERENCES
1. Coleman-Jensen A, Rabbitt Mp,
Gregory CA, Singh A. Household
Food Security in the United States in
2016, ERR-237, US Department of
Agriculture, Economic Research Ser-
vice . 2017. https://www.ers.usda.
gov/webdocs/publications/84973/err
-237.pdf?v=0. Accessed August 30,
2018.
2. Alaimo K, Olson CM, Frongillo Jr EA,
Briefel RR. Food insufficiency, family
income, and health in US preschool
and school-aged children. Am J Public
Health. 2001;91:781-786.
3. Jyoti DF, Frongillo EA, Jones SJ. Food
insecurity affects school children’s aca-
demic performance, weight gain, and
social skills. J Nutrition. 2005;135:2831-
2839.
4. Alaimo K, Olson CM, Frongillo EA.
Food insufficiency and American
school-aged children’s cognitive, aca-
demic, and psychosocial development.
Pediatrics. 2001;108:44-53.
5. Ma CT, Gee L, Kushel MB. Associa-
tions between housing instability and
food insecurity with health care access
in low-income children. Ambul Pediatr.
2008;8:50-57.
6. Gregory CA, Coleman-Jensen A.
Food insecurity, chronic disease, and
health among working-age adults,
ERR-235, US Department of Agri-
culture, Economic Research Service.
2017. https://www.ers.usda.gov/web-
docs/publications/84467/err-235.pdf?
v=0. Accessed August 30, 2018.
7. Tarasuk V, Cheng J, de Oliveira C,
Dachner N, Gundersen C, Kurdyak P.
ARTICLE IN PRESS
Association between household food
insecurity and annual health care costs.
Can Med Assoc J. 2015;187:E429-E436.
8. Olsho LE, Payne GH, Walker DK, Bar-
onberg S, Jernigan J, Abrami A. Impacts
of a farmers’ market incentive pro-
gramme on fruit and vegetable access,
purchase and consumption. Public
Health Nutr. 2015;18:1-10.
9. Olsho LE, Klerman JA, Wilde PE, Bart-
lett S. Financial incentives increase fruit
and vegetable intake among Supplemental
Nutrition Assistance Program participants: a
randomized controlled trial of the
USDA Healthy Incentives Pilot. Am J
Clin Nutr. 2016;104:423-435.
10. Young CR, Aquilante JL, Solomon S,
et al. Improving fruit and vegetable
consumption among low-income cus-
tomers at farmers markets: Philly Food
Bucks, Philadelphia, Pennsylvania. Prev
Chronic Dis. 2013;10:E166.
11. Savoie-Roskos M, Durward C, Jeweks
M, LeBlanc H. Reducing food insecu-
rity and improving fruit and vegetable
intake among farmers’ market incentive
program participants. J Nutr Educ Behav.
2016;48:70-76.
12. Choi SE, Seligman H, Basu S. Cost
effectiveness of subsidizing fruit and
vegetable purchases through the Supple-
mental Nutrition Assistance Program. Am J
Prev Med. 2017;52:e147-e155.
13. Bryce R, Guajardo C, Ilarraza D, et al.
Participation in a farmers’ market fruit
and vegetable prescription program at a
federally qualified health center
improves hemoglobin A1C in low
income uncontrolled diabetics. Prev
Med Rep. 2017;7:176-179.
14. Freedman DA, Choi SK, Hurley T,
Anadu E, Hebert JR. A farmers’ mar-
ket at a federally qualified health center
improves fruit and vegetable intake
among low-income diabetics. Prev Med.
2013;56:288-292.
15. Cavanagh M, Jurkowski J, Bozlak C,
Hastings J, Klein A. Veggie Rx: an out-
come evaluation of a healthy food
incentive programme. Public Health
Nutr. 2016;20:1-6.
16. Wholesome Wave. Wholesome Wave’s
Fruit and Vegetable Prescription Program
2013 report. https://www.wholesome-
wave.org/sites/default/files/wp-content/
uploads/2014/10/2013-FVRxReport_-
Final-1.pdf. Accessed August 30, 2018.
17. Horn JL. A rationale and test for the
number of factors in factor analysis.
Psychometrika. 1965;30:179-185.
Ridberg et al 7
18. Bickel G, Nord M, Price C, Hamilton
W, Cook J. Guide to Measuring House-
hold Food Security. Alexandra, VA: US
Dept of Agriculture, Food and Nutri-
tion Service; 2000.
19. Lachin JM. Logistic regression models.
In: Balding DJ, Cressie NAC, Fitz-
maurice GM, et al, editors.Biostatistical
Methods: The Assessment of Relative
Risks. 2nd ed. 2011:283 380.
20. McCulloch CE, Searle SR, Neuhaus
JM. Generalized Linear Mixed Models.
2nd edition. Hoboken, NJ: John Wiley
& Sons, Inc.; 2008.
21. Polissar L, Diehr P. Regression analysis
in health services research: the use of
dummy variables. Med Care. 1982;20:
959-966.
22. Healthy People 2020. Washington,
DC: US Department of Health and
Human Services, Office of Disease Pre-
vention and Health Promotion; 2018.
https://www.healthypeople.gov/2020/
topics-objectives/topic/nutrition-and-
weight-status/objectives. Accessed April
22, 2018.
23. Jacobs SR, Weiner BJ, Bunger AC. Con-
text matters: measuring implementation
climate among individuals and groups.
Implementation Science. 2014;9:46.
24. Kutner M, Greenburg E, Jin Y, Paulsen
C. The Health Literacy of America’s
Adults: Results from the 2003 National
Assessment of Adult Literacy. NCES
2006-483. National Center for Educa-
tion Statistics; 2006. https://nces.ed.
gov/pubs2006/2006483.pdf. Accessed
August 30, 2018.
25. Berkman ND, Sheridan SL, Donahue
KE, Halpern DJ, Crotty K. Low health
literacy and health outcomes: an
updated systematic review. Ann Intern
Med. 2011;155:97-107.
26. Nederhof AJ. Methods of coping with
social desirability bias: a review. Eur J
Soc Psychol. 1985;15:263-280.
27. Gitterman BA, Chilton LA, Cotton
WH, et al. Promoting food security for
all children. Pediatrics. 2015;136:e1431-
e1438.
28. Hoisington A, Braverman M, Har-
gunani D, Adams E, Alto C. Health
care providers’ attention to food
insecurity in households with chil-
dren. Prev Med. 2012;55:219-222.
29. Lundeen EA, Siegel KR, Calhoun H,
et al. Peer reviewed: clinical-com-
munity partnerships to identify patients
with food insecurity and address food
needs. Prev Chronic Dis. 2017: 14.
 ARTICLE IN PRESS
7.e1 Ridberg et al Journal of Nutrition Education and Behavior  Volume 000, Number 000, 2018

CONFLICT OF INTEREST
K. E. Merritt was the research man-
ager at Wholesome Wave during the
creation of the manuscript. The other
authors have not stated any conflicts
of interest.