Boost your health (Refuerza tu Salud): Design of a randomized controlled trial of a community health worker intervention to reduce inequities in COVID-19 and influenza vaccinations

Abstract

Introduction:

Low-income and underserved populations, especially racial and ethnic minorities, experience health disparities linked to social determinants. The COVID-19 pandemic amplified these disparities, necessitating effective strategies to address structural racism and related factors. Vaccination, crucial for mitigating infectious diseases, including COVID-19 and influenza, remains challenging among underserved populations. Community health worker (CHW) interventions show promise in addressing these disparities but have not undergone rigorous evaluation with a randomized controlled trial to increase vaccination uptake among underserved populations. This study develops and evaluates a CHW vaccination behavior (CHW-VB) intervention to increase COVID-19 and influenza vaccination among adult patients in primary care settings.

Methods:

Tailoring of the Boost Your Health (Refuerza tu Salud) intervention is grounded in behavior change theory and integrates input from a Community Advisory Board. The study employs a patient randomized controlled trial design to test the effectiveness the CHW-VB intervention compared with usual care across six Federally Qualified Health Centers (FQHCs) in New York. Patients are being screened for eligibility (vaccinated but not up to date with the COVID-19 vaccine and have at least one of seven common chronic illnesses) and 800 are assessed at baseline and three months. Outcomes include COVID-19 vaccine (primary) and influenza vaccine (secondary) uptake. The study also evaluates intervention implementation using the RE-AIM model.

Conclusion:

Boost Your Health aims to increase COVID-19 and influenza vaccination among racially/ethnically diverse, underserved populations with chronic illness through the CHW-VB intervention, targeting critical gaps in vaccination uptake to reduce health disparities and increase health equity.

Trial registration:

(ClinicalTrials.gov NCT06156254).

1. Introduction

Low-income and underserved populations, particularly racial and ethnic minorities, experience health inequities from detrimental social determinants of health. The COVID-19 pandemic exacerbated these inequities, while effective strategies to address structural racism and other contributing factors remain elusive.

Key factors contributing to substantial differences between non-Hispanic white and other racial and ethnic minority groups [1,2] include racism embedded in health care, housing, labor, and various policies that put racially/ethnically diverse populations at higher risk [3-5]. Other factors, including the disproportionate burden of chronic illnesses in these populations (e.g., diabetes, cardiovascular disease, asthma, obesity, and mental illness), are associated with worse COVID-19 and influenza outcomes. Further compounding inequities is disparate access to high-quality preventive care and chronic disease services that were limited even before the pandemic [6]. Also relevant are a general lack of cultural, language, and literacy-appropriate public education strategies along with misinformation and disinformation that contribute to low awareness of and motivation to partake in protective personal behaviors like vaccination [7-10].

Consequently, people of color have faced disproportionate infection, hospitalization, and mortality rates from infectious illnesses like influenza [11-15], and COVID-19 [1,2,16]. This underscores the critical importance of strategies to improve vaccination access and vaccine completion [17]. Equitable vaccination among racial/ethnic minority populations with chronic disease can help decrease disparities in severe illness due to COVID-19 and influenza [8,18]. Vaccination against COVID-19 and influenza is the safest and most effective public health strategy to directly protect those with chronic illness from infection, hospitalizations and deaths [19-23]. Since vaccination prevents infections, it should also reduce transmission of COVID-19 and influenza in the general population, thus reducing the likelihood of exposure for people of color with already compromised health and social circumstances.

Significant gaps in vaccination rates existed among populations of color early in the pandemic [8,18,24,25,26]. According to the National Immunization Survey, some of these early racial disparities have narrowed over time, but disparities in coverage for the updated COVID-19 vaccines persists, contributing to ongoing differences in COVID-19-related hospitalization and death rates [22]. Nationwide, White non-Hispanics had the highest vaccination rate for the 2023–2024 COVID-19 vaccine among adults (25 %), compared to Hispanics (17 %), non-Hispanic Blacks (21 %), non-Hispanic Asians (20 %), Native Hawaiian or Other Pacific Islanders (NHOPI) (15 %), and non-Hispanic American Indians or Alaska Natives (AI/AN) (20 %) [27]. This aligns with cumulative data showing that Black, Hispanic, AI/AN, and NHOPI people have experienced higher rates of COVID-19 cases and deaths compared to White people, even after adjusting for age differences by race and ethnicity [28]. In New York City (NYC) a similar pattern is observed with the NYC Department of Health’s most recent data reporting that Black and Hispanic residents, across all ages and boroughs received bivalent boosters at less than half the rates of all other groups [29], [30-34].

Interventions delivered by community health workers (CHWs) have been used successfully worldwide to address many health issues and have been proposed as an important strategy to tackle the pervasive effect of the pandemic on health consequences, particularly for underserved populations [35-39]. However, no randomized controlled trial has rigorously evaluated the effectiveness of a CHW intervention to increase vaccination among underserved populations. This study, entitled Boost Your Health (Refuerza tu Salud), will develop, and test a CHW vaccination behavior (CHW-VB) intervention to increase vaccination among medically underserved patients. The study aims are to:

1. Examine barriers and facilitators to adult COVID-19 and flu vaccination to inform strategies to tailor the intervention for a diverse and low-income population with chronic illness.

2. Test the effectiveness of a CHW-VB intervention compared with usual care using intent-to-treat analysis on COVID-19 vaccine acceptance and vaccination rate (primary outcome) and influenza vaccination rate (secondary outcome).

3. Explore whether measures of capability, motivation, and opportunity for vaccination mediate the effects of the CHW intervention on primary outcomes relative to usual care and/or whether effects are moderated by demographic characteristics and psychiatric/medical conditions.

4. Assess contextual factors affecting implementation and sustainability. Participants.

2. Methods

2.1. Design overview

This study is a randomized controlled trial (RCT) in which we randomly assign patients within six federally qualified health centers (FQHCs) in New York to a CHW intervention to enhance vaccination behavior (CHW-VB) or to usual care. Fig. 1 illustrates the study design and flow. We used 1:1 randomization for waves 1 and 3 with predetermined participant numbers, and a randomized block design with blocks of size 2 or 4 for wave 2, where FQHCs each had two sampling sites with unknown participant numbers, to maintain group balance.

Fig. 1.

Study Flow.

*CHW-VB=Community Health Worker-Vaccination Behaviour

2.2. Study settings and target population

RAND partnered with Clinical Directors Network (CDN), an Agency for Healthcare Research and Quality-designated Center of Excellence (P30) for Primary Care Practice-Based Research and Learning, to recruit and engage sites, patients, and stakeholders from FQHCs. These FQHCs are designated patient-centered medical homes (PCMHs) with comprehensive care management and integrated behavioral health care. They support over 339,500 low-income adult patients across 1 to 22 sites per network, predominantly serving non-white, underinsured, with 33–71 % Spanish-speaking preference. To ensure that we reach our goal sample size of 800 patients (n = 400 per arm), we targeted large FQHCs that have relatively low rates of COVID-19 vaccination uptake and high racial and ethnic diversity. We recruited six large FQHC practices from eight CDN FQHC networks.

2.3. Participant recruitment

Within each FQHC, patients are entered into the study through onsite or telephone screening by trained recruitment coordinators hired by CDN. Patients from each FQHC may be referred by primary care clinicians (PCCs) or self-refer (contact CDN staff from information on flyers deployed in waiting areas) into the study.

2.4. Inclusion and exclusion criteria

Inclusion:

Patients must be 18 years of age or older; speak English or Spanish; and have no obvious physical or cognitive impairment that would make them unable to complete the assessment (as indicated by confusion or inability to understand the questions); and consider the FQHC to be their usual source of care.

Exclusion:

Patients must not be experiencing active psychosis (as indicated by inability to concentrate, having delusions or hallucinations) or be at high risk of suicide (as indicated by having a current plan or means).

2.5. Participant eligibility

Patients are eligible for full participation if they have received a COVID-19 vaccine but have not received the updated COVID-19 vaccine AND self-report being told by a doctor that they have at least one of the following medical conditions: overweight or obese; high blood pressure/hypertension, diabetes, asthma, OR meet criteria for probable depression, generalized anxiety disorder, or posttraumatic stress disorder (PTSD).

2.6. Randomization

After identifying willing and eligible participants post-baseline, we will randomly assign them (1:1 within each FQHC) to either the CHW-VB intervention or usual care. To prevent contamination, we hired CHWs exclusively for the intervention arm who were external to the FHQC staff, thereby ensuring their exclusive interaction with the participants randomized to the experimental group.

2.7. Data collection

For the first formative aim, we worked closely with FQHC staff and Community Advisory Board (CAB) members to identify key community stakeholders to participate in 30-min interviews with the goal of identifying barriers and facilitators to vaccination and potential strategies for changing behavior toward vaccination. We conducted interviews with a diverse group of individuals across six centers, including unvaccinated patients, community leaders, and healthcare providers, totaling over 20 participants who each received compensation for their time. For Aims 2 (effectiveness) and 3 (mediation/moderation), we will collect data from patients via interviewer-administered surveys at baseline and 3-month follow-up (Table 1). For Aim 4, we will conduct exit interviews with varied community stakeholders (CHWs, patients and clinicians) at each of the six FQHCs to understand the implementation and sustainability of the intervention (e.g., which features worked, what challenges, and features that might be sustainable and scalable).

Table 1.

Study evaluation measures, data collection schedule, and data sources.

Type of Measure	Measure	Operational Definition	Data Source(s)	Data Collection
				Pre	3 Mo
Effectiveness: Primary Outcomes	COVID-19 uptake [84] COVID-19 vaccine acceptance	Whether participants report having received the updated COVID-19 in the last 3 months “If I was not up to date with the COVID-19 vaccines recommended for me, I would get the missing vaccinations.”	Patient survey, electronic health record (EHR), Healthix*		X
Effectiveness: Secondary Outcome	Influenza vaccine uptake CAPABILITY	Whether participants report having received a flu vaccine in the last 3 months.	Patient survey, EHR, Healthix		X
Other Measures and Mediators (if assessed pre- and post-intervention)	Vaccine knowledge [85] OPPORTUNITY	General knowledge about vaccinations, measured through a scale comprised of 14 items rated as “correct”, “incorrect”, or “don’t know”. Measures used in a previous study of influenza vaccination, rated in a 5-point visual-aided scale ranging from “few 0-%,” to “most or nearly all (81–100)”.	Patient survey	X	X
	Social norms [86]	Operationalization of: Perceived descriptive norms (e.g., “Thinking about people of [X group, race or ethnicity], how many of them do you think get the COVID-19 vaccine?”) Perceived subjective norms (e.g., “Thinking about people of your own race/ethnicity, how many of them do you think get a COVID-19 (flu) vaccine each year?”)	Patient survey	X
	Access [1]	3 items: How difficult is it to get a COVID-19 vaccine? List of barriers to getting the COVID-19 vaccine Rating of ability to get the COVID-19 vaccine “Would you say not at all difficult, a little difficult, somewhat difficult, or very difficult?” “Many things might make it difficult to get a COVID-19 vaccine. Please tell me if anything I list makes it difficult for you (e.g., getting an appointment online, not knowing where to get vaccinated, hard to get to vaccination sites, vaccination sites aren’t open at convenient times).” “How much do you agree with the following statement: ‘I can get a COVID-19 vaccine if I want to’” (rated on a 4-point agreement scale).	Patient survey	X
	MOTIVATION Vaccine intention [1]	How likely the individual is to get a COVID-19 vaccine. “Would you say you would definitely get a vaccine, probably get a vaccine, probably not get a vaccine, definitely not get a vaccine, or are not sure?” Single categorical item that describes each of the five stages of change:	Patient survey	X
	Readiness to change [87]	Pre-contemplation Contemplation Preparation Action Maintenance The assessment of COVID-19 risk involves:	Patient survey	X
	Perceived risk [84,88]	The likelihood that person will get COVID-19 Severity of the effects of COVID-19 when it does happen 6, Likert-rated items about COVID-19 risk. Reasons for not getting vaccinated includes concerns about vaccine safety, trust side effects, and efficacy.	Patient survey	X	X
	Trust [89,90]	10 general COVID-19 mistrust measure items rated on a 5-point Likert agreement scale, including questions on mistrust in public health information advanced by the government, and conspiracy theory-based Ratings of trust by different COVID-19 information sources.	Patient survey	X	X
Moderators	Sociodemographic characteristics	Preferred language (English or Spanish) English fluency Sex Age Education Race/ethnicity (including Hispanic subgroup) Household size Marital status Education level Employment status Insurance status and type	Eligibility screening	X
	Medical conditions	Whether participants report having been told by a doctor that they have one of these common medical conditions: High blood pressure or hypertension; heart disease such as angina, heart attack, heart failure, or stroke Diabetes Asthma Overweight or obese	Eligibility screening	X
Implementation Outcomes	Psychiatric conditions [91,92,93]	Whether participants report screening positive for probable depression, anxiety. or post-traumatic stress disorder (PTSD) using validated brief screeners: PHQ-8 (8-item Patient Health Questionnaire) GAD-7 (7-item Generalized Anxiety Disorder) PC-PTSD-5 (5-item Primary Care PTSD Screen)	Eligibility screening	X
	Reach	Exposure to key intervention components: Number of Contacts with the CHW Education about COVID-19 vaccines (1-pagers) Received patient action plan Motivational interviewing Behavioral activation approaches (discuss of pros/cons; share a story; other) Assistance with navigating barriers for an appointment	Patient recruitment log, CHW registry, survey		X
	Adoption	Barriers and facilitators to CHW intervention implementation	FQHC staff exit interview		X
	Implementation	Patient ratings of acceptability, helpfulness, relevance, and usefulness	Patient survey		X
	Maintenance	Whether intervention components are put into routine practice	Exit interviews with FQHC staff		X

^*New York City Department of Health and Mental Hygiene (NYCDOHMH).

3. Study conditions

3.1. Overview of intervention tailoring process

The research team tailored the Community Health Worker Intervention to Enhance Vaccination Behavior (CHW-VB) based on previous interventions [40-42] using a rigorous approach that integrated CBPR (community-based participatory research) methods, theoretical and practical models for behavioral change and was based on formative information about barriers to COVID-19 vaccination drawn from the literature and the key informant interviews previously mentioned.

The CBPR approach is anchored in partnership with the project’s CAB in all stages of the research, through regular hybrid in-person/online meetings for the formative assessment, co-designing the adaptation of the intervention, trouble-shooting implementation hurdles, and actively participating in the analysis, interpretation, and dissemination of the results.

3.2. Community health worker intervention to enhance vaccination behavior (CHW-VB)

The Boost Your Health intervention was developed following the Behavior Change Wheel framework (BCW) [43]. We conceptualized vaccination behavior using the Capability, Opportunity, and Motivation model for Behavior change (COM—B) [43,44]. COM-B is at the core of the BCW and has been applied to understand a variety of health behaviors, from infection-related behaviors such as hand washing and use of personal protective equipment, to sexually transmitted infections and non-recreational prescription medicine sharing [45,46]. The BCW has been adapted by the World Health Organization (WHO) and by many researchers to conceptualize contributing factors and behavioral and preventive health interventions related to COVID-19 [47-53] and influenza [54-58], including vaccination behaviors [47,59,60].

Fig. 2 shows the main components of the adapted COM-B model tailored to vaccination behavior. Capability refers to patients’ physical and psychological capability to engage in the target behavior which is influenced by both psychological and physiological mechanisms. Opportunity refers to the contextual factors that prompt the behavior to occur, specifically, physical versus social opportunities. Physical opportunity refers to structural factors that influence vaccination behavior, including access (e.g., physical, financial) to vaccines/vaccination appointments, convenience of vaccination access, and availability of vaccine information from trusted sources and with appropriate language and health literacy. Social opportunity refers to social processes such as perceived social norms and other supports that shape vaccination behavior. The model also describes the relationship among these components – capability and opportunity can influence motivation, and all three components directly influence vaccination behavior. Motivation refers to reflective, decision-making processes (e.g., vaccine confidence, trust, attitudes, risk assessment, and vaccine intention) as well as automatic processes (e.g., emotional responses) that motivate and lead to the target behavior.

Fig. 2.

Adapted Capacity, Opportunity, and Motivation (COM—B) Model.

Capability, Opportunity, and Motivation model for Behavior change.

This CHW-VB intervention is a multi-component intervention to provide education, encourage positive vaccination behaviors, and aid in navigating barriers to increase equal access to vaccination among adults of color with chronic illness. The CHW intervention consists of up to three educational sessions (“touchpoints”) in English or Spanish targeting the specific reason(s) why a patient is not up to date with their COVID-19 vaccine. According to the reason(s), the CHW uses strategies to educate, motivate, and help navigate any access barriers to getting vaccinated. CHWs use Motivational Interviewing (MI) techniques to encourage patients to get vaccinated. Patients also receive educational flyers designed by a local artist to address their COVID-19 vaccination knowledge gaps. At the end of each session, the CHW works with the patient to create an individualized Patient Action Plan with steps the patient can take to overcome their barriers to vaccination.

Fig. 3 presents the overall flow of the intervention. Patients are invited to meet with the CHW by phone. Within 1–2 weeks of recruitment and at the beginning of the first touchpoint, the CHW confirms whether the patient was vaccinated in the interim, and if so offers to “exit” the study without further implementation of the intervention. If not vaccinated, the CHW will proceed to implement up to 3 “touchpoints” depending on whether the patient received the updated vaccination or not.

Fig. 3.

Flow of the Boost Your Health Intervention.

For each touchpoint, the CHW session elicits or confirms reasons for not being vaccinated to better understand why the patient is not up to date on COVID-19 and influenza vaccines. The reasons for not being vaccinated may include one or more of the following categories of barriers:1) lack of knowledge of vaccine safety, 2) lack of knowledge of vaccine effectiveness, 3) lack of knowledge of health risks of COVID-19, 4) lack of access, 5) lack of affordability, 6) mistrust/distrust, and 7) inconvenience, among others. For each touchpoint the CHW decides to implement any of a number of intervention strategies (actions) based on the barriers cited by the patient or the CHW observes.

CHWs have main actions that are drawn from the COM-B model and the related behavior modification techniques at their disposal. The four actions are Educate (primarily targeting Capability using intervention function of Education), Encourage (primarily targeting Motivation through intervention function of Persuasion), Navigate (primarily targeting Opportunity through intervention function of Enablement) and Storytelling (primarily targeting Motivation through Modeling and Education). As we recognize the mistrust that is widespread in the study population, we emphasized the delivery of the behavior change techniques to achieve these broad intervention functions through the communication style of MI to facilitate rapport and trust building and to acknowledge and validate mistrust around vaccination.

3.3. Usual care

Patients randomized to this condition do not have access to the CHWs. They receive the care that they would usually receive independent of the study.

3.4. CHW training and supervision

We hired CHWs with diverse backgrounds, including fully bilingual English/ Spanish speaking CHWs with varied cultural/racial self-identification (Latino/a, Black, non-Hispanic White, and another or combined race/ethnicity). The CHWs are external to the FQHCs to minimize risks of contamination. We hired CHWs that have direct experience with the local communities surrounding the FQHCs, five or more years of working as a CHW or similar patient-facing capacity in a community or in health care setting, good interpersonal and communication skills, and basic word computer skills for using the patient registry to track and document patient encounters.

After a two-day interactive training and a standardized two-week supervised period, CHWs engage in weekly “learning collaboratives” with the research team. The training, which followed the CHW manual, covered study overviews, vaccination information, psychoeducational content, and behavioral modification technique role-play scenarios. Regular feedback from CHWs during these sessions and throughout the intervention refined the manual with the research team. CHWs manage intervention patients via a RAND-CDN-developed registry, tracking communication modes (in-person, video, audio) and documenting interventions per contact (e.g., MI). The registry supports scheduling and tracking of pending activities and producing case summaries for team reviews. A pilot test refined the registry’s usability before the intervention.

4. Assessment procedure and study measures

This study follows a hybrid type 1 effectiveness/implementation design, which includes patient participants to evaluate the effectiveness of the intervention. We also assess the implementation of the intervention through qualitative evaluation.

4.1. Screening measures

All chronic medical conditions are determined by self-report – if the patient has ever been told by a doctor that they have it. To screen for current (past month) mental health symptoms we administer the following validated self-report measures for depression (8-item version of the Patient Health Questionnaire, PHQ-8 [61]), anxiety (7-item Generalized Anxiety Disorder, GAD-7 [62]) and PTSD (5-item primary care screener, PC-PTSD-5 [63]). Cut points for probable depression, anxiety disorder, and PTSD are 10+, 10+, and 3+, respectively.

4.2. Outcomes assessment

We collect data from patients via web-based surveys (either interviewer-administered or self-administered, if needed) at baseline and at 3-month follow-up. Based on the previous trials in these settings conducted by the CDN staff, we expect no more than 10 % attrition. In most cases, we use previously tested measures or adapt measures from the literature that are available in both English and Spanish. We pretested survey items to assess the level of comprehension and, if necessary, eliminated items or substitute them with simpler measures. Table 1 summarizes the survey measures by time points.

4.2.1. Primary outcomes

Our primary outcomes are COVID-19 vaccine acceptance measured by self-report and updated COVID-19 vaccine update at 3 months. In addition to self-report, we confirm receipt of the vaccine by extracting vaccination information from the FQHC Electronic Health Record (her) data for patients who are vaccinated onsite at the FQHCs. Vaccination data are also extracted from Healthix, the largest Regional Health Information Organization (RHIO), which receives near real-time updates for COVID and influenza vaccination from the New York City Department of Health and Mental Hygiene (NYCDOHMH) Citywide Immunization Registry (CIR) [64], and from other facilities providing vaccinations, including hospitals, health centers, ambulatory care sites, community vaccination sites, and pharmacies.

4.2.2. Secondary outcome

Our secondary outcomes assess the effect of the intervention on influenza vaccine uptake at three months. As for COVID-19 vaccination, we rely on a combination of self-report, EHR, and Healthix/NYCDOHMH-CIR data.

4.2.3. Mediators

To explore mediation of the treatment-outcome relationship, we will examine whether measures of capability, motivation, and opportunity explain the effects of the intervention on COVID-19 and influenza vaccination outcomes. Candidate mediators are shown in Table 1.

4.2.4. Moderators

We will test baseline sociodemographic and health characteristics as potential moderators of the intervention effect. Candidate moderators are in Table 1.

4.3. Implementation evaluation

4.3.1. Approach and outcomes

We operationalize the five RE-AIM framework outcomes (reach, effectiveness, adoption, implementation, and maintenance) (Table 1) [66,67] using patient registries to evaluate the intervention’s implementation. Adoption will be evaluated through exit interviews with stakeholders from zAim 1 (formative), focusing on implementation barriers and facilitators.

5. Data analysis

5.1. Power calculation

For our power calculations, we assess differences in vaccination uptake proportions between the treatment and control arms. We will adhere to standard assumptions of 80 % power and a significance level of α = 0.05 for all calculations. The required sample size depends on an average follow-up vaccination rate of 50 %. Larger sample sizes are needed for proportions close to 50 %, due to higher uncertainty. Therefore, conservatively assuming a 50 % proportion in sample size calculations means the study is also powered to detect differences for any proportion that is smaller (or larger) than 50 %. Our study population is comprised of adults who have not received the updated COVID vaccination. Therefore, we expect to have follow-up vaccination rates below 50 %, for which we are powered for given our conservative sample size calculation assumptions.

We consider effect sizes ranging from 8 to 20 percentage points (Table 2), in line with related studies. For context, a meta-analysis on CHWs’ impact on childhood immunization showed a relative risk (RR) of 1.2, similar to a 10-percentage point difference around 50 % [68]. Other studies have found various effect sizes, from an RCT in a multicenter setting showing a 12 % increase in vaccine intention [66], to an RCT on infant vaccination where MI impacted vaccine coverage by 3.2 % - 7.3 % [69]. Larger effects (21–27 percentage points) were seen in a study by Krieger et al. [65] using a peer-based intervention for senior citizens’ pneumococcal and influenza vaccination rates.

Table 2.

Required sample sizes under various alternative conditions.^*

Percentage point diff	Percent treatment	Percent control	Percent attrition	Required n (per group)
20	60	40	0	97
20	60	40	5	102
20	60	40	10	107
12	56	44	0	272
12	56	44	5	286
12	56	44	10	302
10	55	45	0	392
10	55	45	5	412
10	55	45	10	435
8	54	46	0	612
8	54	46	5	645
8	54	46	10	680

^*For alpha (2-tailed) = 0.05 and 1-beta = 0.80.

Attrition is another crucial factor for sample size. Since we will have access to participants’ vaccination records, including those lost at follow-up, we anticipate minimal attrition impacting our primary outcome and thus considered attrition rates between 0 % and 10 %. Table 2 details the sample size requirements for various effect sizes and attrition rates. A baseline recruitment of 400 participants per group (800 total) will suffice to detect a minimum 10 percentage point effect size with no attrition and a minimum 10.4 percentage point effect with 10 % attrition, assuming a 50 % proportion for the treatment group.

5.2. Analysis overview

We will use a mixed methods analytic approach: qualitative analysis with data from stakeholder interviews, recruitment logs, and patient registries; and quantitative intent-to-treat analysis using patient interview data at baseline and 3-months.

5.2.1. Quantitative analysis

We will compile univariate descriptive statistics for all study variables and compare intervention outcomes with usual care using two-sample descriptive analyses. Primary analyses will focus on complete cases, with use of imputed data to evaluate sensitivity of our findings. We will use single imputation unless more than 5 % of respondents have item-level missingness, in which case we will use multiple imputation with the final number of imputed datasets driven by the rate if missingness. Imputations will be generated using sequential predictive mean matching with the R package MICE, and sensitivity to the missingness at random assumption will be explored [70] We will consolidate the imputed results to minimize uncertainty and provide thorough analysis for each study objective.

To evaluate intervention effectiveness, we will use logistic mixed-effects models for our two binary outcomes (COVID-19 vaccination and influenza vaccination rates) controlling for individual-level time-invariant variables (e.g., sex, age, baseline conditions) along with fixed effects for the FHQC and patient-level random effects (e.g. random intercepts and slopes). The random effects variance-covariance matrix will be unstructured and therefore capture correlations between random effects. Despite the potential for a simple intent-to-treat analysis due to randomization, we include covariates to refine our statistical efficiency. Model assumptions, like a linear relationship between predictors and the log odds of the outcomes, will be evaluated. Any necessary adjustments, such as transforming variables to account for non-linear relationships, will be made if violations are detected.

Since mediation alone cannot confirm causality, as a sensitivity test, we will consider performing propensity-score adjusted regressions by explicitly modeling mediator assignment and conduct weighted analysis of the models above. We will also test for moderating effects of time-invariant covariates. This will be done by fitting a regression model for each of the candidate moderating variables.

We will conduct mediation analysis [71,72] to examine how our primary outcomes relate to the treatment, considering mediating variables (see Table 1) such as risk perception. The mediation analysis will first involve confirming the association between the treatment and the outcome of interest, as well as the mediator of interest, using regression models. After confirming these associations, a regression model will be fitted that includes the treatment indicator variable and mediator variables, while controlling for the outcome at baseline and key covariates from the primary analysis. The significance of the coefficients on the treatment and mediator variables will indicate the extent to which the relationship between the treatment and outcome is mediated by the mediator.

To test the robustness of our findings, we will conduct two follow-up secondary analyses. First, we will estimate mediator assignment propensity scores and conduct weighted regression analysis of the models above, since mediation alone cannot confirm causality. Second, we will test for moderating effects of time-invariant covariates by fitting a regression model with an interaction term between the treatment variable and the time-invariant covariate for each of the candidate moderating time-invariant variables. Given multiple moderators and mediators under consideration, we will apply the Benjamini-Hochberg correction to control the false discovery rate when conducting multiple hypothesis tests.

5.2.2. Qualitative analysis

To assess contextual factors affecting implementation and sustainability, we will conduct rapid content analysis of stakeholder interviews [73-77]. Two team members will independently review interview notes to outline themes using a predefined template, creating a comprehensive codebook with theme descriptions, criteria, and examples. For exit interviews, we will use Dedoose version 8.3.35 (Los Angeles CA) [78] to code themes, initially practicing on 20 % of the data until achieving consistent theme identification. Similarly, we will employ template analysis [79] for patient logs and registries, categorizing data hierarchically, such as CHW contact type followed by duration of contacts.

6. Discussion

Racial and ethnic minorities with chronic illness suffer disproportionately from preventable infectious diseases. COVID-19 and influenza data suggest multiple determinants of vaccination inequities between non-Hispanic whites and other racial/ethnic groups including manifestations of structural racism which limit access to and motivation to seek preventive care.

The Boost Your Health intervention aims to increase COVID-19 and influenza vaccination rates among underserved, racially/ethnically diverse patients with chronic illnesses at FQHCs through a CHW-VB intervention. Leveraging CHWs’ community trust and experience, the intervention addresses structural racism and social determinants of health barriers faced by the target population. Increasing vaccination uptake presents challenges. Some patients may only need assistance or a reminder to get vaccinated, while those more hesitant may require targeted efforts.

The Boost Your Health approach is grounded in the behavior change literature while simultaneously incorporating input gathered from a community participatory approach to maximize effectiveness. We integrate the intervention with the community’s unique cultural, socio-economic, and health contexts, ensuring theoretical soundness and practical relevance. A key feature is our innovative blend of the BCW and MI principles in designing this CHW intervention to address vaccination behaviors. This aligns with recent advancements in the BCW approach, emphasizing that the delivery communication style of behavior change techniques is as crucial as the content delivered to patients [80]. Importantly, integrating MI style is critical for promoting health behavior change among marginalized racial and ethnic minority groups, especially amid widespread medical mistrust [81-83]. While MI has been discussed as a promising approach to address COVID-19 vaccine hesitancy, our integration of this style with a rigorously designed behavior change intervention is a novel and important contribution to the literature.

Vaccination is crucial, especially during a public health crisis, yet persistent misinformation and declining motivation have increased COVID-19 hesitancy. Our study aims to fill a key knowledge gap by evaluating CHWs’ effectiveness in boosting vaccine confidence, intention, and uptake among people of color with chronic illness, marking the first randomized controlled trial of its kind with a rigorous mixed method design. Our findings will guide future strategies to increase vaccination rates in underserved, racially/ethnically diverse populations, to reduce health disparities and increase health equity.

Data availability

No data was used for the research described in the article.