Understanding multi-level policy implementation in the national school lunch and breakfast programs: a mixed-methods and agent-based modeling protocol

Abstract

Background

The Healthy, Hunger-Free Kids Act (2010) improved the nutritional quality of school meals in the U.S. by aligning the National School Lunch and Breakfast Programs with updated dietary guidelines. However, 2018 federal flexibilities in sodium, whole grains, and milk standards shifted key implementation decisions to local school districts. This created variability in uptake and potential inequities in diet quality and health outcomes, particularly among students from low-resource settings. Key drivers of differential uptake, and decision-making by local districts, is not well understood.

Methods

In this three-phase study, we will use a mixed-method approach to examine the implementation of sodium, whole grains, and milk flexibilities across the many levels involved in implementing the National School Lunch and School Breakfast Programs and describe the impact of flexibility implementation on child health-related outcomes. In phase 1, we will conduct a nationally representative survey of school district food service directors to assess current flexibility implementation practices and determinants of decision-making. We will combine these primary data with publicly available state level and school district data to understand how decisions relate to school meal participation and health-related outcomes. Phase 2 involves interviewing school food industry actors to understand how National School Lunch and School Breakfast Programs policy changes influence decisions that change food supply and distribution. In phase 3, we leverage insights from the actors surveyed and interviewed in phases 1 and 2 to develop an agent-based model. We will use this model to analyze potential effects of different policy changes on child health, examination of ways to improve effective and equitable district-level implementation of the National School Lunch and School Breakfast Program, and to develop tools to inform real-world policymaking.

Discussion

This study offers critical insights into the complexity of school food systems, emphasizing the role of decision making by actors involved in National School Lunch and School Breakfast Program implementation. Results will inform evidence-based strategies to support equitable implementation of school nutrition policies, expand policy implementation science by providing insight into the mechanisms that shape variation in implementation and health-related outcomes, and contribute methodological innovation using an agent-based model as a virtual policy laboratory.

Contributions to literature.

• This study uses agent-based modeling (ABM) to understand the dynamic interactions and decision-making processes in implementing school nutrition policies.

• This study integrates the Consolidated Framework for Implementation Research (CFIR) and Racial Equity and Policy Framework (REAP) to systematically address determinants and equity implications in policy implementation.

• This study offers critical insights into the complexity of school food systems, emphasizing the role of decision making by diverse stakeholders and providing actionable recommendations for more effective and equitable policy implementation.

Background

Cardiovascular diseases (CVD) are the leading cause of death in the US [1, 2]. Although the clinical manifestations of CVD most commonly begin to appear in middle age, the underlying pathophysiological processes are evident in childhood and adolescence [3, 4]. Several routinely measured CVD risk factors, including obesity and overweight, are known to cause CVD and prevention of these factors in adolescence can substantially minimize adult CVD [5, 6].

Unfortunately, the prevalence of youth obesity and overweight in the US has nearly doubled in the past 20 years [7, 8]. The interaction of environmental, behavioral, and structural factors has led to the sharp rise in obesity among US youth. Social determinants such as poverty, food insecurity, and limited access to healthy foods have affected youth obesity rates, especially among youth who experience persistent health inequities. For instance, the prevalence of obesity among Hispanic (25.6%) and Non-Hispanic Black (24.2%) youth is almost double that among non-Hispanic White youth [9]. Income status also impacts youth obesity, with obesity prevalence highest among children aged 2–17 in low to middle-income groups [10]. In addition, increased consumption of energy-dense, nutrient-poor foods and sugar-sweetened beverages, coupled with declines in physical activity, contribute to excess weight gain [11, 12].

Schools represent an important setting for policy approaches to prevent obesity and other CVD risk factors among all youth populations. More than 95% of youth (aged 5–17) in the US are enrolled in schools [13]. No other institution has as much continuous and intensive contact with youth during this critical time of development and growth [13–15]. While attending school, youth can consume up to two meals daily, plus snacks through the National School Lunch and School Breakfast Programs (NSLP and SBP). Established in 1946 and 1966, respectively, the NSLP and SBP provide low-cost or free meals to millions of children each school day [16, 17]. These programs and the policies that guide implementation affect children’s diets [18–21] and weight status [22–25].

The passage of the Healthy, Hunger-Free Kids Act (HHFKA) of 2010 modernized the NSLP and SBP by aligning school meal requirements with evidence-based dietary guidelines, resulting in the United States Department of Agriculture’s (USDA) final rule, Nutrition Standards in the National School Lunch and Breakfast Programs. Preliminary research has established the effectiveness of the updated nutrition standards (as originally intended) with positive and significant influence on school meal nutrition quality and other HHFKA provisions more broadly on student health and weight outcomes [14, 26–31]. Despite this success, since its inception, the evidence-based components have been challenged. In 2018, a significant policy change allowed for flexibilities, [32] relaxing requirements for schools to meet nutrition standards for milk, whole grains, and sodium [33–35]. The decision to implement these flexibilities was left to schools and districts (i.e., school food authority). Studies have shown the potential for negative health impacts from the implementation of sodium, wheat, and milk flexibilities including increasing the risk of CVD, diabetes, and stroke [36–44], especially among low socioeconomic status (SES) students.

There are several important actors involved in the implementation of the NSLP/SBP and the corresponding policies that guide the programs. The NSLP/SBP are part of a complex system involving food suppliers, school food service directors and other school leadership, and student consumers. Any policy change (or allowed flexibility) to the NSLP/SBP impacts these diverse actors involved within each part of this system and vice-versa, the actors work in both unique and shared ways to influence school food policy [45]. The decisions they make in response to a policy change shape implementation and ultimately patterns of youth food consumption and nutrition that may diverge from the intended goals of the policy. These decisions can therefore drive important public health outcomes and the potential for inequities [46].

To inform more effective and equitable policy development and implementation, it is imperative to understand decision-making by key actors across the school food system and the potential for dynamic heterogeneous responses to policies, with the goal of ultimately eliminating school nutrition-related health inequities including those associated with CVD risk factors among youth.

To our knowledge, this will be the first study to assess and document the implementation of the wheat, sodium, and milk flexibilities in school districts nationwide. While our pilot study [47] assessed implementation of wheat, sodium, and milk flexibilities, it examined school districts in Missouri and did not assess the impact of implementation on health outcomes. Few studies have examined equitable implementation outcomes, and none have investigated flexibilities and their implications across districts on a national level with attention to structural and social determinants of health. In addition, limited research exists that examines the response of the school food system to changes in federal policy. While several studies [26, 48–52] have documented factors that guide food service director decisions related to mandatory federal guidelines, little research exists about decision making when flexibilities are allowed and the decision to implement is left solely on the food service authority. Decision making by the food supply actors in response to federal level policy changes has also been understudied. There is a significant gap in literature related to other food supply influences (food industry, vendor availability, commodity offerings, contract structures) as determinants of district decisions.

In this three-phase study, we take a comprehensive approach to describe the impact of NSLP/SBP policy flexibilities on the many levels involved in implementation. In phase 1, we will conduct a survey among a nationally representative sample of school district food service directors to assess current flexibility implementation practices and determinants of decision-making. We will also conduct interviews with food service directors. We will combine these primary data with publicly available state level and school district data to understand how decisions relate to school meal participation and health-related outcomes. Phase 2 involves interviewing food industry actors to understand how NSLP/SBP policy changes result in decisions to change food supply and distribution. In phase 3, we leverage insights from the actors surveyed and interviewed in phases 1 and 2 to develop an agent-based model (ABM). We will use this model to conduct analyses of potential effects of different policy changes on child health, examination of ways to improve effective and equitable district-level implementation of the NSLP/SBP, and to develop tools to inform real-world policymaking.

Methods/design

Overview of study design

This multiphase study, funded by the National Institutes of Health, National Heart, Lung, and Blood Institute (5R01HL178372-02), seeks to understand how decision-making in response to policy changes among actors in the school nutrition system could lead to school nutrition-related health inequities. We have designed the study using elements of both political, public health, and implementation sciences to evaluate the impact of federal school food policy changes on the many levels involved in implementation. We will use a mixed method approach [53] to complete three overlapping and complementary phases. Figure 1 depicts a visual of the study schema and the components of each study phase.

Fig. 1.

Study activities

Conceptual framework

The theoretical basis for our study is the Consolidated Framework for Implementation Research (CFIR) and the Racial Equity and Policy Framework (REAP) [54]. The CFIR framework includes five study relevant domains: Outer Setting, Inner Setting, Intervention Characteristics, [55]. Characteristics of Implementers, and Implementation Process. The measures associated with each domain will be collected through a semi-structured interview and survey administered to internal agents (food service directors) and external agents (food industry). We will also utilize components of REAP to highlight important equity implications of health policies. We will specifically examine inner and outer contexts (school culture, student demographics, and decision-making) to assess potential determinants of health inequities. In all, combining these models allows for a comprehensive understanding of how policy, defined at the federal level, is implemented throughout a complex system and further defines how decision making in some school districts may lead to less healthy food served to some populations (i.e., low SES). Using ABM, we will also identify the interaction of factors influencing policy implementation and potential leverage points for improving health outcomes for all students (Fig. 2).

Fig. 2.

Conceptual model

Phase 1: Assessment of implementation processes, determinants of decision-making among school food service directors, and related health outcomes

Phase 1 will involve multiple methods to understand current implementation processes and determinants of decision-making among food service directors, as well as to understand the impact of these decisions on school meal uptake and child health. Methods include surveying a nationally representative sample of US school district food service directors, collecting school district and state-level data, and interviewing food service directors.

Survey and semi-structured interview tool development

We will develop a survey and an interview guide using the Consolidated Framework of Implementation Research to understand five domains potentially involved in deciding whether to implement the flexibilities. These include: (1) the Inner Setting (i.e., structure, networks and communication, implementation climate, culture; (2) Intervention Characteristics (i.e., evidence strength and quality, complexity, cost); (3) Characteristics of the Individuals involved (i.e., knowledge and beliefs about the intervention and self-efficacy); (4) the Implementation Process (i.e. planning, engaging, executing, reflecting/evaluating); and (5) the Outer Setting (i.e., local environment, partnerships, policy). Survey questions will focus on each CFIR domain with the goal of identifying influential factors of policy adoption. Prior to data collection, we will conduct cognitive response testing with 10 food service directors through pilot interviews to ensure any survey items are clearly worded and assess the intended constructs [56]. Survey items will be scored on a Likert scale to identify if they are influential to flexibility adoption and decision-making. Items will be analyzed separately before analyzing internal consistency to examine whether items are consistent within the CFIR domains.

In addition, a random subsample of food service directors will be asked to participate in follow-up interviews to further contextualize the survey data by elucidating how influential key determinants are to decision-making and adoption of flexibilities and to identify other determinants not originally considered. We will also consider equity dimensions outlined in the REAP Framework [54] and within the specified CFIR domains to specifically address 1) disproportionality and culturally relevant factors of the district population served, 2) decentralization (i.e., the power of actors to affect a wide variety of policy outcomes), 3) voice (ability of communities to shape the policy environment), and 4) societal context including social determinants of health. Further, questions regarding sociopolitical forces will be developed addressing specific policies (e.g., school board decisions or state specific policies) which may impact districts’ abilities to meet the needs of student stakeholders and may drive some decisions regarding adoption of flexibilities. The follow-up interview guide will also be tested prior to administration.

Quantitative data collection

Survey sample selection, recruitment, and data collection

The quantitative component of Phase 1 will rely on a survey administered to a nationally representative sample of school district food service directors through collaboration with a partner organization. This organization has 6,212 food service members across 4,024 districts. The membership of the partner organization has geographical diversity and also serves school districts with diverse student populations— 22% of membership serves > 65% free and reduced priced lunch students. We anticipate collecting information from 25% of food service directors from member districts (roughly 1,000 districts. Data will be collected using an online Qualtrics survey [57] that will be delivered via email, sent by our partner organization.

Other quantitative data

In addition, we will use state school data from the National Center for Education Statistics (NCES) to weight responses to approximate a nationally representative sample. Specifically, we will use school district ID codes to merge our survey data with three sources of data: (1) demographic and attendance data from the NCES’s Common Core of Data (CCD). The CCD will contain information on race and ethnicity, SES (Free and Reduced-Price Lunch Eligibility), geography (urbanicity), and district type (Public, Charter, etc.), attendance, and school lunch revenue; (2) discipline data (e.g. suspensions, etc.) from the Office of Civil Rights (OCR); and achievement Data from the Stanford Education Data Archive (SEDA).

Quantitative measures and data analyses

Our survey will allow us to capture policy flexibilities adoption in a given year through a binary response. Stemming from our conceptual framework, we will explore how the implementation climate (inner setting) and the characteristics associated with the intervention and individuals relate to the implementation of flexibilities. Next, we will examine how these direct relationships are indirectly explained (i.e. mediated) by the implementation process. Finally, we will examine how direct and indirect relationships vary (i.e., are moderated) across various social and demographic characteristics in the outer setting. Similar to previous survey research using a CFIR framework, the implementation climate (inner setting), the characteristics associated with the intervention and individuals, and the implementation process can be conceptualized as latent constructs, while elements in the outer setting can be conceptualized as observed variables. Thus, following previous CFIR research, we will leverage a Structural Equation Modeling (SEM) approach, which simultaneously tests the significance and strength of multiple hypothesized relationships among both latent and observed variables [58, 59].

Our analytic strategy will proceed in three stages. Within each stage a series of model fitting techniques will be used to confirm or adjust our theoretical model. First, we will use confirmatory factor analysis to create valid latent constructs of the inner setting (e.g., implementation climate, communication, organizational culture, etc.), characteristics of the intervention (e.g., perceptions of complexity, cost, relative advantage, etc.), characteristics of the individuals within the organization (e.g., self-efficacy, knowledge and beliefs about the intervention, etc.), and the implementation process. Second, SEM will describe how inner and outer settings, characteristics of intervention and individuals, and the implementation process are directly related to the implementation of flexibilities. In addition, SEM will describe how inner and outer settings, as well as characteristics of intervention and individuals are indirectly related to the implementation of flexibilities through the implementation process (mediating relationship). Third, using moderated mediation analyses, we will examine how the relationships among inner settings, characteristics of intervention and individuals, the implementation process, and the implementation of flexibilities, are moderated by outer setting characteristics [60].

To demonstrate the impacts of the implementation of flexibilities on student health, we will use secondary data obtained from the Standard Education Data Archive (SEDA), the Office of Civil Rights (OCR), and the Common Core Data (CCD) to explore the relationship between the implementation of flexibilities and meals served. We will also examine the relationship between the implementation of flexibilities and attendance and absences, which can be considered a proxy for health [61–66]. Indeed, attendance, which is often associated with chronic absenteeism, has been related to both poor mental and physical health [67–70]. In addition, childhood overweight or obesity is a risk factor for chronic absenteeism. [69]. In doing so, we will leverage a quasi-experimental design by combining propensity score weighting (PSW) and Difference-in-Difference (DiD) modeling, as we have done in our prior work on school meal policies [71]. PSW, which uses inverse probability treatment weights to limit selection bias in the estimation of treatment effects [72], will be used to balance implementation adopters and non-adopters on observed district (i.e., inner setting) and demo-graphic characteristics (i.e., outer setting). With the treatment (adopters) and control (non-adopters) groups effectively balanced on pre-treatment measures, we will use a DiD approach to estimate treatment effects of the policy (flexibilities) adoption. By establishing similar pre-treatment trends through PSW, DiD leverages both within-district changes over time and between-district changes over time in order to identify treatment effects in the years following policy adoption, [73].

$${\mathrm{\delta }}_{\mathrm{DD}}={{\mathrm{Y}}_1}^{\mathrm{T}}-{{\mathrm{Y}}_0}^{\mathrm{T}}-({{\mathrm{Y}}_1}^{\mathrm{C}}-{{\mathrm{Y}}_0}^{\mathrm{C}})$$

where Y₁ ^T represents the number of meals served in the treatment group at time two; Y₀ ^T represents number of meals served in the treatment group at time one; Y₁ ^C represents the number of meals served in the control group at time two; and Y₀ ^C represents the number of meals served in the control group at time one.

In order to conduct this analysis, we will merge nationwide data from the SEDA, OCR, and CCD. Specifically, these data will include academic achievement on state tests, suspensions, the number of students (regular, free-meal-eligible, reduced-price-eligible), school meal expenditures (which can be used to estimate the number of meals served), and average daily attendance from 2 years prior to the introduction of the flexibilities and 2 years after the introduction of flexibilities. While survey responses may not be nationally representative of the entire population, we will be able to leverage analytic weights from the CCD to expand the generalizability of our findings.

For the factor analysis we rely on Westland’s (2010) methodology to arrive at a lower bound a priori sample size estimate. Given a minimum of four latent constructs and one observed variable (as seen in our conceptual model), as well as a p-value of 0.05 and a power of 80%, our analyses will require a sample of 991 to detect an effect size of 0.2. For our DiD estimates, we expect a moderate treatment effect, or an effect size of 0.2 standard deviation on core outcomes (e.g., meals served and students’ attendance). For DiD analysis we rely on Cohen’s (1988) and Champely’s (2020) methodology to arrive at a lower bound a priori sample size estimate [74–76]. Given 0.05 level of significance and 80% power, as well as (a) 5 time points for the trend line (i.e., one and two years before implementation, implementation year, and one and two years after implementation), (b) a two-group design (i.e., flexibility adopters and non-adopters), (c) a random effects structure with a random slope, residual term, and autocorrelated residuals, (d) a 3% inflation factor, and (e) a 5% attrition rate between successive time points, we calculate our optimal sample size as follows:

$$N=\frac{n·\text{inf}lationrate·\#samplegroup}{{\left(1,-,a,t,t,r,i,t,i,o,n,r,a,t,e\right)}^{\#surveypoints-1}}\approx 996$$

Given that our range of expected districts starts at 1,000, our analyses will be adequately powered.

Qualitative data collection and analyses

Key informant sample and recruitment

A subsample of 5–8 food service directors from each of seven DHHS regions (total sample = ~ 35–42) will be recruited to participate in 45-min follow-up interviews. All interviews will be audio recorded, transcribed, and reviewed for completeness and accuracy.

Data coding and analysis

All transcripts will be analyzed using NVivo 15. Codebooks will be developed based on the interview guides and will be reviewed by the study team. Coding will be primarily deductive through the CFIR, but we will remain somewhat inductive to allow for new themes to emerge from data. Each transcript will be coded independently by two team members to organize the text into categories that summarize the concept or meaning articulated according to the codebooks [77–79]. For any codes for which coders do not reach agreement and consensus is not reached, an additional study team member will determine the appropriate code. Themes from the coded transcripts will be summarized. Data analysis may also include quantification or other forms of data aggregation, including the use of data matrices [80]. All data will be summarized to inform Phase 3 ABM development.

Phase 2: Assessment of decision making related to school food supply system

Phase 2 involves semi-structured qualitative interviews with school food supply actors (manufacturers, vendors, and USDA) to further assess the Outer Setting that school meal programs must operate within. We will modify a previously designed CFIR-informed semi-structured interview tool [47].

Semi-structured interview tool development

We will develop an interview guide using the CFIR to understand the outer setting [55, 81–83]. CFIR describes the outer setting as the economic, political, and social context within which an organization (school districts) resides. When the mandatory HHFKA guidelines were initiated, schools had no option to decide whether to implement healthier nutrition standards according to federal policy. With the flexibilities, the outer setting is malleable, and schools are given decision-making authority. However, even with the ability to decide whether to implement flexibilities, schools must rely on school food manufacturers and overall school food supply chain to develop new products and respond to new demand. Re-formulation and product development take time and research investment. Some manufacturers may be better equipped to make these investments and respond to changes in federal policy and resulting demand from schools. Others may not. The outer context has been an area neglected in school food policy research yet is extremely important as federal guidelines change and flexibilities are granted.

Qualitative data collection and analysis

Key informant sample and recruitment

We will focus on three main entities because of their influence on the school food supply chain: (1) food vendors and suppliers, (2) commercial food manufacturers, and (3) the USDA Foods in Schools program. We will interview 35–42 informants: 3–5 from the USDA Food in Schools Program and 35–40 (5 informants per each of 7 regions) from vendors and major commercial manufacturers. Previous research has shown saturation may be achieved with as few as 10–12 interviews when the subject being investigated has a narrow focus and the interviewees are somewhat homogenous [84]. We will determine the vendor and manufacturer informants based on regional product development and delivery since not all vendors and manufacturers serve all schools across the US. We will also choose vendors and manufacturers of different types of food (meat/processed/milk, etc.). This will ensure examination of differences regionally and across food type. The USDA Foods in Schools program purchases commodities for use by schools participating in the NSLP/SBP. USDA commodities make up 15–25% of foods served at school [85]. While USDA contracts with specific manufacturers to supply certain commodities, it is important to understand the selection process since commodities are provided to all schools.

Data coding and analysis

Coding and analysis will follow the same protocol as outlined in Phase 2 qualitative coding and analysis.

Phase 3: Develop an ABM to demonstrate how changes in federal school food policy affect the implementation of the NSLP/SBP at the school district level

For Phase 3, we will design, develop, parameterize, test, and iteratively improve the explanatory power of an ABM of NSLP/ SBP implementation. This model will simulate school district decisions on the use of nutrition standard flexibilities over time. It will incorporate data collected during Phase 1 and 2 activities to characterize school district decision-makers, the inner and outer settings that they experience, and the dynamic interactions between decisions, settings, and decision-makers. We will then use this model to explore counterfactual scenarios in which effectiveness and equity of implementation might be meaningfully and sustainably improved, and present recommendations based upon model findings in user-friendly, context-specific formats to relevant policymakers and stakeholders. The creation and use of ABMs will follow established best practices [86], and a proven process that has yielded successful applications of ABM in other topic areas by members of our team [87–91]. Our activities during this phase will consist of the following specific steps: model design, development, and testing.

Model design and initial face validity

A key focus of the model will be on representing the decision-making process (of schools/food service directors, manufacturers, districts) and thus the extensive data collected during Phases 1 and 2 on these processes will be rich and invaluable input. We will build the model with these inputs and evaluate the initial model design using feedback from the full research team as well as external practitioner partners to ensure it has sufficient face validity before proceeding. During this phase, we will explicate assumptions that underlie the model’s conceptual design and operation; these assumptions will be rigorously explored during later modeling steps.

Development and initial testing

Once designed, the ABM will then be instantiated in computational architecture. Computational code will be developed and maintained by our modeling team, and we will follow best practices from computer science for software design. This includes initial model testing for internal consistency to identify and address flaws in design or development. At the end of this step, we expect to have a well-specified, fully operationalized, dynamic computational representation of district decision-making processes that results in the use of flexibilities as well as a data output framework that will facilitate rigorous empirical testing during subsequent steps.

Model parameterization, testing, and iteration

The model will be parameterized such that it can meaningfully represent specific real-world contexts (e.g., decision-makers and the settings that they experience). These model inputs will be obtained from literature, preliminary studies, and data collected during Phase 1 and 2 activities. Parameter values will be derived from direct analysis of data as well as model calibration. After initial parameterization, we will test the model’s explanatory power and refine iteratively as needed. Assessment results will be shared and discussed with the full research team. If the model does not generate a sufficient match with observed real-world outcomes, the research team will revisit the assumptions made during design and the decisions made during development and initial parameterization in order to refine and iteratively improve the model. At the end of this process, we expect to have a well-tested, explanatory model which can explicate and quantify the overlapping mechanisms that underlie existing patterns in school district NSLP and SBP decision-making and implementation. This model and parameter input set will form the “baseline” or “control” against which counterfactuals are compared in the final step.

Experimentation and sensitivity analyses

In this step, we will use our ABM to explore ways that effectiveness and equity of implementation might be meaningfully and sustainably improved. Our ABM will provide a valuable opportunity to gain insights into potential policies and practices that have not been attempted in the real world (or have not been attempted in particular contexts. We will work with our full research team and solicit input from research partners to identify a large but finite set of experimental conditions to explore. These will be comprised of variations in types of interventions/policy flexibilities (and combinations of interventions/flexibilities), intervention targets (i.e., school districts affected), and intervention effect magnitudes (thus allowing us to identify “tipping points”). This step will culminate with rigorous sensitivity analyses to assess robustness of the explanatory power of the baseline model and findings from experimentation will be assessed. Findings from our experimentation and sensitivity analyses will be presented to the full research team and external research partners for feedback about how they can best be interpreted and translated into actionable insights. At the end of this step, we expect to have material that we can present in user-friendly, context-specific formats to relevant policymakers and stakeholders during dissemination.

Dissemination plan

We will take a systematic approach when disseminating project findings, following well-established principles of audience segmentation and recognizing different audiences require specific communication messages and channels [92, 93]. We will focus on three primary audiences: researchers, practitioners, advocates, industry partners, and policy makers. Scientific researchers will be reached via publications in high-impact, peer-reviewed journals. We will also present at top-flight research meetings such as the Academy of Nutrition and Dietetics Food & Nutrition Conference and the Annual Conference on Science of Dissemination and Implementation in Health. Practitioners, advocates, and food industry partners will be reached via materials distributed through a variety of forums. To reach food service personnel and advocates, we will utilize current dissemination and learning forums led by the School Nutrition Association,, Healthy Eating Research Collaborative, and American School Health Association. We will also develop publications in practice-oriented journals and present at meetings with high practitioner attendance. To reach those in the food industry, we will attend the American Commodity Distribution Association Annual Conference and work with the association’s board members to disseminate findings (Phase 2). To reach policy makers, we will develop separate materials tailored to their needs such as short 1-page issue briefs to summarize key findings from our study and disseminate through venues such as district and state Boards of Education, National Conference of State Legislatures, National Association of County and City Health Officials, and the National League of Cities. We will also engage officials at USDA to present our data and inform final rule considerations.

Study status

The study team has completed early phase 1 activities, including an extensive literature review, determination of key metrics and variables, and preliminary development of the survey and interview scripts. In pursuit of Aim 3, we have developed an initial ABM design which we will continue to update and revise as we gather data from Phases1 and 2.

Discussion

Innovation

This study provides an innovative approach to understanding system level policy implementation. The application of computational modeling in implementation science and federal school lunch policy is I novel, especially for theory building and for the study of emerging issues. ABMs have not yet been used to study the implementation of federal-level public health policy as we describe here. In addition, the National Academies report, Redesigning the Process for Establishing the Dietary Guidelines for Americans, recommends the use of systems science in understanding the complexities of the food system and in linking diet and health [94].

In addition, this study will be among the first to apply constructs from the CFIR to the implementation of a US federal food policy at the local level. Use of the CFIR will extend the application of results across different levels (legislature, industry, school) involved in the decision to implement flexibilities and other policy changes. This study also utilizes a health equity framework that has been identified as useful specifically in highlighting important equity components in policy [54]. We extend the use of this framework into the school setting and identify inequities among food served to students in some school districts compared to others. This will inform more equitable school food policy development and implementation.

Finally, this study engages in system-wide research to understand the implementation of NSLP/SBP flexibilities. School food research typically focuses on the school or student only. This research adds value and much-needed insights through involving the entire school food system from government policymaking to school food supply chain to implementation in school districts and ultimately impacts on youth health.

Limitations

The study has a few limitations. One of the main limitations of this study is the potential difficulty in recruiting survey participants. We plan to mitigate this by leveraging incentives, personalized follow-ups, and our close partnership with organizations that regularly work with school food service staff. Although we are working with a national organization, it is possible that survey participants are not nationally representative of the population. To address this, we will utilize district weights from the CCD to increase the generalizability of our study. Additionally, recruiting a sufficient number of participants to collect adequate data for generalizability remains a concern. However, based on our past success in recruiting participants and employing personalized follow-ups, we anticipate reaching saturation of information and ensuring generalizability.

Another limitation is the inherent constraint of the initial conceptual model, which will be designed to reflect key influences theorized and observed in existing literature, along with data from Phases 1 and 2. This model may need to be expanded to accommodate additional elements identified during model development and testing. However, as noted above, our team has extensive experience successfully iteratively testing and improving ABMs during multi-year research projects.

Conclusion and impact

This study aims to improve child health outcomes by understanding decision-making within the school district nutrition system in response to policy changes. Our results will provide critical insights into the complex system of school nutrition policy implementation and its subsequent impact on youth health, particularly among communities experiencing health disparities. By elucidating the decision-making processes of key actors and identifying leverage points for equitable policy intervention, this research will serve as a valuable tool for shaping more effective and inclusive national school food policies. The outcomes of this study have the potential to drive meaningful changes in reducing CVD risk factors and health disparities among youth in the US. These findings will also contribute significantly to the existing body of research on effective and equitable policy development and implementation.

Abbreviations

CVD

Cardiovascular Diseases

HHFKA

Healthy Hunger-Free Kids Act

NSLP/SBP

National School Lunch Program / School Breakfast Program

SES

Socioeconomic Status

ABM

Agent-Based Model

CFIR

Consolidated Framework for Implementation Research

REAP

Racial Equity and Policy Framework

NCES

National Center for Education Statistics

CCD

Common Core of Data

OCR

Office of Civil Rights

SEDA

Stanford Education Data Archive

SEM

Structural Equation Modeling

PSW

Propensity Score Weighting

DiD

Difference-in-Differences

USDA

United States Department of Agriculture

Authors’ contributions

SMR is the principal investigator of this study. SMR, DF,GM, JG, JJ, MK, PA, RAH, and TC contributed to the initial conceptualization and design of the study and continuously offer guidance and support. SMR led the initial grant writing. JG coordinated the study and input from the study team on the protocol. All authors provided substantial edits and revisions to this paper. All authors read and approved the final manuscript.

Funding

This study is funded by the National Heart, Lung, and Blood Institute (5R01HL178372-02). The findings and conclusions in this paper are those of the authors and do not necessarily represent the official positions of the National Institutes of Health. In addition, this work was made possible with support from Washington University in St. Louis WU-CDTR (Grant Number P30DK092950 from the NIDDK) and the Foundation for Barnes-Jewish Hospital.

Data availability

Data resulting from this research will be shared via the Washington University School of Medicine (WUSM) institutional data repository, Digital Commons Data@Becker.

Declarations

Ethics approval and consent to participate

The study was approved by the Washington University in St. Louis Institutional Review Board in February 2025.

Consent for publication

Not applicable.

Competing interests

First author, Sarah Moreland-Russell is an Associate Editor for Implementation Science Communications. All other authors declare that they have no competing interests.