Does a food insecurity intervention improve perinatal outcomes for mother and child? A randomized control study protocol of the Fresh Rx: Nourishing Healthy Starts program

Abstract

Background

Pregnancy and postpartum periods represent critical times to support nutrition and household food security, especially for families with limited or strained economic resources. The Fresh Rx: Nourishing Healthy Starts study uses a randomized design to examine a comprehensive, holistic “food is medicine” program targeting food insecure expectant mothers in an area with high rates of inequitable adverse maternal and birth outcomes. Participants receive weekly fresh food deliveries from a Combined Community Supported Agriculture partner, cooking and nutrition supports, access to culinary skills education from a trained chef, and counseling and care management. In addition to providing basic cooking tools and pantry staples, the program links participants with locally available community resources and federal food assistance programs such as the Supplemental Nutrition Assistance Program and Special Supplemental Nutrition Program for Women, Infants, and Children.

Design and methods

This study protocol establishes a randomized control design with a control and two treatment groups to isolate distinct program components and examine variation across outcomes including household food security, maternal depressive symptoms, gestational age, and birthweight. The study design and protocol follow Consolidated Standards of Reporting Trials (CONSORT) principles to emphasize transparency and replicability.

Expected impact of the study for Public Health

Guided by the Cycle of Food Insecurity and Chronic Disease, study implications for public health are significant given the program’s potential for improving birth outcomes, reducing per capita healthcare costs, and filling a notable policy and program gap during the critical perinatal period for mother and child alike.

Introduction

While hunger represents a physiological condition or physical state resulting from inadequate intake of nutrients, food security utilizes a broader lens of the economic and social factors that impact households with wide ranging consequences. The U.S. Department of Agriculture (USDA), which administers and oversees Federal Nutrition Assistance Programs, defines food insecurity as “a household-level economic and social condition of limited or uncertain access to adequate food.” ¹ During 2020, 10.5% of U.S. households were food insecure, including 13.6% of households with children. ²

Food security also exists at the intersection of U.S. public health and healthcare, which lags far behind peer countries across a wide range of measures and outcomes. ³ As a social determinant of health under the broader umbrella of economic security, ⁴ food insecurity increases incidence of adverse health conditions and outcomes. ⁵ Underinvestment in public health and entrenched health inequities extending from race and racism alongside geographic segregation also result in food insecurity and significant disparities in life expectancy. ⁶ A public health approach requires that researchers move beyond understanding the circumstance and barriers to food security and test new community level approaches that consider circumstances, barriers, and opportunities beyond the amount and quality of food that households consume.

Core to a public health approach to food insecurity is its cyclical nature with chronic disease. Chronic disease incidence contributes to increased food insecurity and vice-versa. ⁷ Food insecurity leads to increased stress for families and chronic disease incidence, which contributes to increased health care expenditures, potential for reduced employment, decreased income, and increased spending tradeoffs on basic needs. These in turn amplify and extend household food insecurity, as described by the Cycle of Food Insecurity and Chronic Disease. ⁷ Similarly, under the Family Stress Model, families may experience emotional distress when making trade-offs in meeting their basic needs, including food. ⁸ Interventions to address food insecurity during pregnancy are thus a critical chance to improve primary health and interconnected outcomes for mother and child.

Pregnancy and postpartum periods present critical times to support nutrition and household food security, especially for families with limited or strained economic resources. During pregnancy, food insecurity can cause increased nutritional deficiencies, ⁹ preterm labor and low birthweights, ¹⁰ and birth defects. ¹¹ A strong relationship has also been found between low birthweight and adverse outcomes in later childhood such as repeating grades, assignment to special education services, and greater likelihood of involvement with the criminal justice system.^12,13 Additionally, stress, depression, and anxiety are more common among food insecure mothers,^14,15 and research has suggested a bidirectional relationship between maternal depression and food insecurity. ¹⁶

Notably, while food insecurity has a negative relationship with healthy pregnancy outcomes, increased nutrition reduces adverse birth outcomes and can moderate risks of chronic disease later in life. ¹⁷

In addition to high rates of infant mortality, the United States ranks last among 20 peer countries in low-birthweight and pre-term deliveries. ¹⁸ Furthermore, Black and Hispanic families in the U.S. have increased risk of food insecurity^10,19 as well as less access to prenatal care, lower birthweight infants, and higher infant and maternal mortality. ⁶ For example, Black women in Missouri have a pregnancy-related mortality ratio more than four times higher than White women. ²⁰ These outcomes have profound short and long-term consequences for families, communities, and public health.

Healthcare institutions and policymakers increasingly have considered “food is medicine” strategies to improve outcomes while reducing per capita costs. ²¹ These have included prevention and treatment approaches such as medically tailored meal programs and supplemental food programs ²² for chronic health conditions such as diabetes. ²³ To support household and community-level food security, broader foundational policies and programs (e.g. the Supplemental Nutrition Assistance Program) must be in place for food is medicine approaches to build upon and achieve their intended effects.

In this paper, we present the study protocol for a randomized design of a comprehensive food is medicine intervention to improve nutrition and health outcomes, address health inequities, and reduce healthcare expenditures. The unique program and study design protocol expand upon an increasing number of food prescription or delivery programs^24–26 because of its target population (food insecure pregnant women), extensive food is medicine program model, integration with the health care system, and outcomes of interest.

Design and methods

Program description

The Fresh Rx: Nourishing Healthy Starts program (hereafter “Fresh Rx NHS”) offers a comprehensive holistic intervention to address household food insecurity for pregnant women in an area with high rates of infant and maternal mortality, low birthweight, and preterm births. The program is based in St. Louis, Missouri, a U.S. Midwestern city with marked racial segregation. From enrollment through 60 days post-partum, participants receive weekly food shares in the form of meal kits of locally-sourced protein, dairy, fruits, and vegetables designed for supporting nutrition during pregnancy from a Combined Community Supported Agriculture provider. Weekly food shares include recipes and ingredients to make three meals of four to six servings each increasing with household size in acknowledgment of food sharing as well as older household members limiting their food intake to shield or buffer younger members from food insecurity. In addition to weekly fresh food packages, participants receive access to nutrition and health education, counseling, coaching from a registered dietitian and a trained chef, and care coordination and supportive services from a social worker. Participants are linked with Federal Nutrition Assistance Programs such as the Supplemental Nutrition Assistance Program (SNAP) and Special Supplemental Nutrition Program for Women, Infants, and Children (WIC), resources from local community organizations (e.g., job training, financial coaching), and home meal preparation essentials based on a home kitchen assessment. All services provided by program staff incorporate a participant-centered and trauma-informed approach that recognizes food insecurity as a form of trauma. ²⁷

Community collaboration and participation is instrumental to program and research design and aims to enhance appropriateness, relevance, cultural competence, and implementation. ²⁸ Drawing from best practices and evidence on community-collaborative research^29,30 and absent any formative guidelines for designing locally tailored, culturally appropriate integrated food and nutrition interventions, ³¹ we formed a Participant Advisory Council comprised of Fresh Rx NHS participants during a program pilot. This group guided program and study design considerations including local barriers, issues of structural racism at the root of birth disparities, food delivery logistics, and ideas for including more familiar and resonant meal choices.

Study recruitment and inclusion criteria

Potential participants are identified by a health plan care manager during an initial outreach by telephone. Staff explain that all health plan members are asked about food access during pregnancy because of its importance for healthy births and administer the validated USDA Hunger Vital Sign™ two-item screening assessment. ³² To be eligible for the study, participants must respond “often” or “sometimes” in response to one or both assessment questions: “Within the past 12 months, we worried whether our food would run out before we got money to buy more” and “Within the past 12 months, the food we bought just didn’t last and we didn’t have money to get more.” Individuals screening positive for household food insecurity receive a referral from their health plan via phone call, text message, email connection, or online form to the local hunger relief organization providing the program intervention.

Eligible participants must be pregnant; less than 24 weeks gestation; currently receiving and planning to continue care throughout pregnancy; English speaking; older than 14 years of age, and a Missouri resident enrolled in one of three state Medicaid Plans. After eligibility is established, interested participants complete an IRB-approved informed consent form, receive a program orientation, and complete a comprehensive intake assessment read aloud to capture baseline measures.

Study design

To measure Fresh Rx NHS program impacts, we designed and implemented a randomized control study design. Following Consolidated Standards of Reporting Trials (CONSORT), ³³ the study emphasizes transparency and replicability in its methodology and planned analysis measuring associations between program participation, sociodemographic characteristics, and maternal health and birth outcomes. In the randomized design for Fresh Rx NHS, participants who screen positive for food insecurity and consent to participating in the study have an equal chance of assignment into each of the control and treatment groups. The study will enroll 250 women in the control and each of two treatment arms, which will allow for observation of program effects and differences in key identified outcomes. Random assignment into one of three groups occurs after participants consent to participate in the study (Figure 1).

Figure 1.

Randomization and study treatment arms.

Each study arm receives distinct services to allow analysis of different study elements. The control group receives an “enhanced standard of care,” which involves the standard case management services participants would receive from their Medicaid case manager in addition to support enrolling in nutrition assistance programs such as SNAP and WIC and information about local food pantry availability. In addition to the enhanced standard of care, Treatment Group One receives meal kit deliveries, nutrition and cooking education, and essential cooking tools. Treatment Group Two receives all services and program components available to Treatment Group One, plus ongoing holistic wraparound social worker support. The guiding research questions of the experiment are as follows:

1. What is the impact of providing direct food delivery and nutrition support on the health outcomes of food insecure pregnant women in the study treatment groups?

2. To what extent does the provision of additional counseling and social support to Treatment Group Two enhance program efficacy and outcomes?

3. To what extent does program provision reduce participant per capita healthcare costs and adverse pregnancy outcomes?

Data collection

To examine the primary research questions, the Fresh Rx NHS study collects a variety of data from participant surveys, interviews, and healthcare claims data. Data are collected at baseline (pre-test), during the third trimester of the pregnancy, 60 days postpartum, and 1 year post-partum. Participants complete a program satisfaction survey, offering insights on their program experience and perceptions. In addition to the anticipated benefits of program involvement, participants receive a $25 gift card for completing each of four program assessments. Health care claims data will be provided by participating Missouri Medicaid managed care organizations at quarterly intervals throughout the study. We utilize single masking (outcomes assessor) for the study as program staff need treatment assignment information in order to provide the associated services, and participants need to be aware of services they are eligible to receive. The study and data collection began in August 2021, and we expect it to continue for 24–30 months.

Assessment tools and measures

Primary study outcome measures include changes in household food insecurity, perinatal depressive symptoms, gestational age at birth, and birthweight. Secondary outcome measures include changes in food spending, self-reported physical and mental health, healthcare utilization indicators (e.g. number of prenatal appointments), financial health indicators (e.g. difficulty paying bills), public program participation, and an array of maternal and child health indicators collected through healthcare claims data (e.g., days in Neonatal Intensive Care Unit, maternal anemia diagnosis). Process outcomes of interest include receipt of weekly food deliveries, program attrition, cooking class participation, and supplemental nutritional coaching supports.

Conducted with the Participant Advisory Council, researchers used cognitive interviewing to test program assessment tools for plain, accessible language. Interviews gaged appropriateness and comprehension of the informed consent and myriad of assessment tools in place for program and evaluation purposes. Utilizing a concurrent interview design approach, ³⁴ interviews tested assessment and survey administration with 10 program pilot participants. This offered feedback regarding participant experience and comprehension as well as an opportunity for program staff and researchers to determine question effectiveness and minimize unnecessary or burdensome participant assessment components. ³⁵

Analysis plan

Study analyses will draw from 750 total Fresh Rx NHS participants across three arms (n = 250 each). The analysis plan includes benchmark sample comparisons to known population characteristics and descriptive statistics on the baseline population characteristics. Additionally, balance tests among the treatment groups and population outcomes at follow up will include two-tailed t-tests for comparing means of continuous variables and bivariate and multivariate analyses on treatment assignment effects. Linear mixed models will test our hypotheses between our control and treatment groups.

Program outcome analyses will examine treatment effects through multiple approaches. First, we will follow the Intention-to-Treat (ITT) approach ³⁶ to account for health impacts and variation in participation and other characteristics of program enrollees. Here, participants who discontinue will be included in analysis and invited to complete follow-up surveys. Additionally, a treatment-on-treated approach will remove participants who attrite out of the program to understand the impact of treatment receipt. Finally, a dosage-response approach takes into account various levels of program engagement (e.g. meal kit delivery) and will be used to understand how treatment receipt level relates to outcomes. In all of three approaches, we will include moderation models to understand how program impacts differ by key household characteristics (household size, race, etc.).

Previous research has found a possible spillover effect where SNAP benefits can offer greater ability to spend on other household spending needs. For example, one recent study utilized a difference-in-differences estimation using data from the Consumer Expenditure Survey to track spending changes after a SNAP benefit increase in 2009. ³⁷ We look to build on previous findings and identify associations through survey questions administered throughout the program regarding basic needs and hardships between treatment groups and the degree of anticipated spillover effects. Descriptive approaches will identify usage patterns, program barriers, and program opportunities within the Fresh Rx NHS program.

Study analyses will also compare Fresh Rx NHS participants and results to a comparable population through health plan partnerships as well as publicly available rates of low birthweight and pre-term births in St. Louis. Weights may be used so that participant outcomes can be extrapolated to represent people in the larger metropolitan area who are Medicaid-eligible and pregnant. Sensitivity analysis will also test how various sources of uncertainty in our models contribute to overall uncertainty around the relationship between independent and dependent variables.

Notably, eligibility for the Fresh Rx NHS program is binary—a potential participant is food insecure or not based on responses to the two-item screener. However, the experimental and analysis design consider additional categories and depth of food insecurity using the 18-item assessment in each data collection period, which will be instrumental in determining program outcomes. Measuring change in household food insecurity over time will allow comparisons between levels of food insecurity with improved maternal health and birth outcomes. To look beyond binary food insecurity, our analysis will either use four discrete categories (high, marginal, low, and very low food security) and/or continuous variables that indicate the depth of food insecurity. ³⁸

Discussion

From its randomized control study design and, unique intervention point in relation to the cycle of food insecurity and chronic disease during pregnancy, the Fresh Rx NHS study seeks to add meaningful knowledge for addressing inequities in health outcomes. We hypothesize that Fresh Rx NHS will result in less incidence and lower severity of household food insecurity, improved maternal health and birth outcomes (e.g. fewer low birthweight or pre-term births), and per capita health care cost savings, with the greatest effects observed for Treatment Group Two. Here, we hypothesize that case management and access to counseling supports will pair with food access to increase participant engagement and utilization. We expect all groups to have higher rates of SNAP and WIC participation, both of which have been found to improve birth outcomes.^39,40

In a review of existing literature, we found that food policy and program study designs are typically evaluative, descriptive, quasi-experimental, or correlational, which limit the ability to understand unbiased impacts or draw causal inferences. Even though researchers can account for selection bias, Heckman and Smith ⁴¹ argue against experimental and randomized control design as the gold standard of scientific policy research due to a chasm between theoretical capabilities and practical results. Consequently, experiments with randomized designs might identify programs that “work” or “don’t work” but fall short in generating new knowledge of why programs succeed or fail (Heckman and Smith ⁴¹ , p. 108). The Fresh Rx NHS study addresses this by tracking participation throughout the program and systematically obtaining participant feedback to test assumptions of what elements do and do not work. Through the Participant Advisory Council and satisfaction surveys on participant experience, a program pilot identified opportunities to adapt and incorporate input into the study. For example, previously in-person cooking classes moved to recorded videos, food pickup expanded to include a delivery option by a Fresh Rx NHS program alum, and the package of fresh, locally grown food transitioned from an assorted box to a meal kit model.

Interpretation and application of study findings should be cognizant of the potential for selection bias in the sample, as participants who opted into the program may not be reflective of those who opted out. While the sample size will allow for identifying results for common indicators such as birthweight and gestational age, rarer occurrences may not have the appropriate statistical power to measure meaningful effect sizes. Findings should also be understood and discussed in the context of a program and study area of a largely urban and metropolitan area.

With the public health field embracing and promoting a social determinants approach to improving health outcomes, health plan stakeholders have begun to advance a range of initiatives and long-term commitments to do more than care management and reimbursement for services, expanding the traditional definition of medically necessary to encompass interventions and investments in the social determinants increasingly recognized for driving health. Health plans have significant incentive to increase their member participation in public benefits and services not traditionally billable to Medicaid, however prevention only accounts for 2.9% of total healthcare spending in the U.S. ⁴² Findings from the study can have significant implications for maternal and birth outcomes alongside healthcare system savings both in the immediate and long term. Following participant program participation and outcomes through 1 year post-partum may also inform state-level policy discussions around the value of extending eligibility beyond the typical post-partum window of 60 days.

Conclusion

With the severe consequences of food insecurity, there is an urgent need to understand how health inequities manifest due to environmental, systemic, and structural factors and to establish new evidence on how best to support nutrition during the critical period of pregnancy. While food is medicine approaches to health gain increased attention, few studies have empirically explored their impacts. We are engaging in one of the first randomized control studies with food-insecure pregnant women that utilizes a comprehensive array of nutritional services, including at-home meal kit deliveries. As a key social determinant of health and with a cyclical relationship to chronic disease, food insecurity should be understood and addressed through cross-sector partnership ⁴³ and with a broader lens than food access with significant implications for improving outcomes and addressing health inequities.