We will take a strengths and assets–based approach, underpinned by the social model of disability and by intersectional considerations to challenge discrimination [31,32]. Our exploration of the social determinants of health and well-being is framed by the social ecological model [33,34] (Figure 1).

Medical or biopsychosocial models of disability have led to the continued disenfranchisement and marginalization of people with physiological impairments, through the conflation of pathoanatomical diagnostic criteria with disability itself [35,36]. In other words, the two are inseparable and the person with the diagnosis is only seen as dysfunctional. This leads to ableism (discrimination in favor of nondisabled people) [37] and disablism, defined as “discriminatory, oppressive, or abusive behaviors arising from the belief that disabled people are inferior to others” [38] (page 9).

The CICADA study resists the use of these deficit-focused disability models, instead taking as its starting point the social model of disability because of its currency, usefulness in driving transformative outputs, and relevance to much-needed revisions in continuing discriminatory statute and law [39]. This continued discrimination persists despite the United Nations Convention on the Rights of Persons with Disabilities [40]. In the social model, impairments, as physiological problems of the body, are decoupled from disability, which results from exclusionary social oppression and prejudices [37]. Society, by accommodating impairments through the removal of iniquitous barriers to inclusion, can enable the full participation of all, across every area of life.

While important, the social model has limitations that are the topic of much discussion [41,42]. For example, Siebers [41] argues for: (1) more complex understandings of embodied variation and (2) more dynamic problematization of the liminal spaces occupied by lived reality than afforded by the social model. Arendt’s [42] criticism of the social model is that more account needs to be taken of the ways people with impairments internalize and make meaning of their lived experiences. We will therefore also draw on alternative nondeficit models in our analyses. Our overall interest is in how individuals with chronic conditions or impairments experience and make meaning of the world through their embodiment within it at the intersection with various other simultaneously and variably interacting social factors.

Rather than separately considering the multiple social categories of “identity, difference, and disadvantage” [43] (p.171) such as gender, racial/ethnic minoritization, disability, and occupation, we consider them as coexisting interacting systems of oppression. In other words, they work together (are mutually constitutive) under discriminatory institutional and structural conditions to create [44-47] lower levels of physical and mental health, poor access to quality health care, and poorer health outcomes [48-51]. Citizenship status adds an infrequently considered further important layer of complexity that we explore in the CICADA study [50]. There is a particular lack of research on the ways that health outcomes are shaped for undocumented migrants through their structural construction as “illegal” [51] within a hostile environment [52] with “no recourse” to welfare and housing support.

We consider both individual experiences of day-to-day discrimination and the wider context. We use the term “minoritized ethnic groups” to emphasize the stigmatization and oppression that a racialized society bestows on particular ethnic groups as racialized “others” [53], rather than to necessarily ascribe to them a “minority” status. An alternative term, “racialized communities,” is also used in some study documents to indicate the nature of this oppression.

Shifting identities among people who have recently migrated, in the face of racism, can include the racialization process of “becoming White.” This process tends to be neglected in the health literature [54] and is one that we also consider. We hypothesize negative consequences for recent immigrants of Arab or Central and East European ancestry, who may experience the tensions of being symbolically included in a White ethnic category but are excluded from many of its benefits [54], in a manner that is often invisible because of the lack of its exploration.

Through considerations such as these, built into the study design, intersectionality theory will allow us to develop complex nuanced insights into differences, while minimizing the risks of essentializing some combinations as inherently problematic or considering the minoritized experience as homogenous.

Intersectionality conceptualizes the ways an individual’s social interactions are shaped by their multiple subject positions (eg, as a female, recent migrant, disabled person). The social ecological model [33,34] highlights the ways this individual is positioned at the center of a system of mutually influencing sets of social determinants, incorporating their personal, community, regional, and national (policy and society) ecosystems of norms and practices. Embodied experiences of migration, citizenship, chronic conditions, and impairments are necessarily intersectional with areas of potential discrimination and oppression across the different levels of the social ecological model. Hence, there is a need for a range of comparisons and involvement of multiple stakeholders in our study to ensure that any potential strategies and recommendations we develop will apply both within and across the different levels [55]. This also fits with the new UK National Health Service (NHS) tiered Integrated Care Plan [56], which is relevant as the CICADA study is set in the United Kingdom.

Intersectional interactions across the levels of the social ecological model are in constant flux, which Bronfenbrenner [57] represented by the chronosystem in development of his original model. Recognizing the importance of these changes over time, our study is longitudinal. Our work is also underpinned by the Consolidated Framework for Implementation Research (CFIR) [58]. This is an amalgamation of various implementation theories that target different levels of the social ecological model, and we use it to comprehensively explore the feasibility of implementation of our recommendations and other outputs. The CFIR is easy to operationalize, flexible (the user selects relevant themes from a pool of 39), and facilitates actionable findings across multilevel implementation contexts.

The lefthand column of Table 1 shows the eight topics that run throughout the study. These topics have informed the reviews, topic guides, research workshops, and surveys. Instruments included within our surveys are shown in the righthand column of Table 1 as an example; they are mostly validated by their developers in different newly migrated populations, and thus are particularly appropriate for use in this study.

The primary online survey outcomes are: (1) formal/informal care measured using de novo questions and the World Health Organization (WHO) Quality of Care and Support-Disability questionnaire (QOCS-ID) at 4, 10, and 16 months; (2) quality of life measured using the World Health Organization Quality of Life tool (WHOQOL-BREF-ID) at 4, 10, and 16 months; (3) control of life measured using the “control of life” validated questionnaire at 4, 10, and 16 months; (4) physical and mental health measured using the WHO Anxiety and Depression Scale (ADS) at 4, 10, and 16 months; the Vulnerability Assessment Framework at 10 and 16 months; and the Global Mental Health Assessment Tool at 10 and 16 months; and (5) social networks measured using an adapted Close Persons questionnaire (for online work) at 4, 10, and 16 months in the online survey and as part of the semistructured interview (using closed questions, open questions, photographs taken by participants to represent their networks without any personal identifying information, and maps drawn of networks).

Secondary outcomes include fear of death measured using the Templer Death Anxiety Scale [76] at 10 and 16 months by online survey. Thematic qualitative data from survey free text, interviews, and workshops will be analyzed, including (1) patient experiences of health and social care, and other forms of formal and informal support during the pandemic, and their perspectives on the impacts on their health; (2) consideration of the impacts of their identity (eg, disabled, from a specific ethnicity, from a low-income background) on these experiences; (3) consideration of their beliefs (health beliefs, COVID-19–related beliefs, vaccination beliefs) and how these affect other themes; and (4) consideration of coping mechanisms, and strategies and assets used in relation to their access to and use of resources, services, and support as this affects their health and well-being.

Surveys will vary by sweep. Key outcome and exposure variables that we expect to change over time will be measured in all three sweeps to study trajectories. Theoretically stable concepts (eg, tolerance to uncertainty, demographic characteristics) will be measured only in one sweep. Key topics may be added to sweeps 2 and 3 that have been identified through our other work.

Our two-stage review work will: (1) create a systematic (living) map to summarize the pandemic-relevant experiences of living with impairments/chronic conditions and/or being from a minoritized ethnic group across the topics listed in Table 1, and (2) undertake an in-depth analysis and synthesis on specific aspects determined according to the map.

This review will ground the research in current evidence and generate themes that can be incorporated in the primary data collection design.

In both stages, two reviewers will independently screen titles, abstracts, and full texts against inclusion criteria, and extract data. They will compare a subset of this work to check for consistency as quality control, with any disagreements to be resolved by a third reviewer. We will assess risk of bias using Cochrane-recommended checklists, also noting the provenance and publication status of sources. Data extraction will be managed in EPPI-Reviewer software and will reflect the inclusion criteria and the designated aims of the review.

The review inclusion criteria, using a modified SPIDER (Sample, Phenomenon of Interest, Design, Evaluation topics, Research source, Setting) [77] framework are summarized in Textbox 2.

The outcomes and the focus of the granular review will depend on the evidence available, and gaps in the evidence will be highlighted for future study. Reporting will follow PRISMA (Preferred Reporting Items of Systematic Reviews and Meta-analyses) guidelines. The review is registered with Prospero (CRD42021262590).

Reporting of the data will depend on the types of included studies (eg, descriptive statistics, narrative synthesis, and diagrams). We will perform subgroup analyses where appropriate.

The primary survey is online and will collect quantitative and qualitative (free text) data. Survey sampling across the four nations of the United Kingdom will be open to any adult living in the United Kingdom, but purposively targeted via selected sites and networks to encompass all conditions/impairments and organizations supporting minoritized ethnic groups, including recently arrived and undocumented migrants. Sampling will not depend on individual patient data such as those that could be obtained via electronic health records to ensure we include people who are self-diagnosed or who perceive themselves to have a different diagnosis to the one held in the electronic record, as well as participants not registered with primary care. We will be mindful of the different biases this may cause, and will collect data on whether or not a diagnosis exists and whether the person agrees with this diagnosis for in-depth understanding.

The three survey sweeps will be evenly spaced over 15 months, with each sweep open for 1 month. Recruitment will predominantly involve distribution of a survey link via social media, and specialist and national networks (such as academic, health service, third sector), as well as mailing lists and large databases of adults interested in health research across the United Kingdom. We recognize that this strategy, being nonrandomized, will be biased such as toward those already interested in research participation or who are active users of third-sector sites and have online access. We will make available print copies for community groups involving participants lacking internet access. We will compare respondent demographics to whole population estimates where possible to explore representativeness (although formal data are limited).

Our new longitudinal survey will enable a description of the trajectories of key variables and outcomes and the links between them. Free-text data will also be analyzed for patient experience. The survey will not be used to test a particular treatment or focus on a single effect.

In our basic structural equation modeling (SEM), we have six core latent variables (factors) per sweep: (1) quality of life, (2) control of life, (3)access to care, (4) coping mechanisms, (5)mental health, and (6)social networks.

Considering statistical power in SEM [79], the required sample size increases with the number of latent variables, but at a decreasing rate (ie, the required sample size difference between a model with one vs two latent variables is larger than that between a model with three vs two latent variables). The required sample size also decreases strongly as the loadings on latent variables increase (ie, the magnitude of the association between latent and observed variables, where values below 0.9 are generally taken to show a confounding effect). Moreover, the power increases as the number of items (questions) used to measure each latent variable increases.

In terms of our study, each latent variable will be measured by several items (the average number being more than 8). In a worst-case scenario with average loadings of around 0.5 and an item missingness of 20% (as suggested from ActEarly work), a sample size of 800 per subgroup per sweep will yield useful analyses. We have four main subgroups (ie, self-identifying as minoritized ethnicities with a chronic condition/impairment, minoritized ethnicities without a chronic condition/impairment, of White British ancestry with a chronic condition/impairment, and of White British ancestry without a chronic condition/impairment). Using these four comparator groups enables us to fully understand the nature of relationships between different variables and the influence of chronic condition/impairment and minoritized ethnic group categories on each, both separately and combined. Thus, the required sample size is calculated at 800×4=3200, although we aim for 5000 for a more robust sample [79].

A descriptive statistical summary will be updated with each sweep. More in-depth analysis, using SPSS, R, Stata, or Python, will exploit all three sweeps of the data, with the following research questions:

For research questions 1 and 2, we will exploit the longitudinal nature of the data using latent growth modeling (LGM), with multiple group analysis, varying different combinations to consider the effect of intersectionalities on outcomes. Depending on geographical coverage and the numbers recruited, the LGM estimation for research question 2 could be carried out at the within-area level to examine the causal impact of the change in COVID-19 prevalence and pandemic social distancing across the sweeps on each key outcome, under the assumption that this is exogenous. We will examine the plausibility of this assumption and detail possible sources of endogeneity. This will be important for policy given our unique subgroups focus. For example, our data could help clarify why specific groups may find it unfeasible to adhere to recommended behavioral responses. For research question 3, we will estimate a developmental cascade model, including all three data sweeps and key variables, to explore how the key variables are associated with one another, both within survey sweeps and over time. We will fit the LGM models using SEM; this offers useful tools for dealing with missing data due to nonresponse and attrition.

The social network support module of our model will consider how respondents are connected to others who provide support to the respondent (eg, through relations such as friendship, kinship, exchanges, activities) [80,81]. Characteristics such as the participant network’s size, composition, and resources available will result in a latent “network capital” variable created through measurement analysis within the SEM as a novel contribution. These network metrics will be used to provide a descriptive presentation of the network(s) and any changes over time.

We will require completion of almost every question on every page for participants to proceed so that we can undertake the association analyses required. This means that there should generally be no missing items in any measures. However, this requirement may lead to completion attrition, with respondents giving up and logging off. We will try to mitigate this possibility in the design, which will allow participants to save responses and return to the survey later, and the survey will be developed and piloted with 30 people from our advisory group and patient advisory group (PAG). There is also the risk of attrition between sweeps. Participants will be asked to provide an email address upon enrolling online. The RedCap online secure system that we will use will automatically recontact them for sweep 2/3 follow-up questionnaires (with reminders). This automatic process makes for efficient and secure second and third sweep recruitment to reduce the risk of missing respondents. Lotteries appear effective in some online surveys [82] and we are including a £50 (~US $63) Amazon voucher as an incentive given at random.

To handle missing data, and address panel attrition and item nonresponse, we will use modern methods, including full information maximum likelihood; multiple imputation with chained equations that produce unbiased estimates under assumptions of missing at random (ie, missingness dependent on observable data only) and multivariate normality; and pattern mixture models that address missing not at random (ie, missingness dependent on unobserved data), assuming correct model specification [83]. These techniques, under certain assumptions, ameliorate loss of statistical power due to missing data and possible biases due to systematic missingness.

Our interview sampling frame follows an intersectional approach that allows us to consider and compare assumed homogeneity: (1) across chronic conditions/impairments irrespective of heritage, and (2) across ancestries irrespective of chronic condition/impairment. The aim of this is to tease out intersectional factors and heterogeneity. To achieve this, we will use a purposive quota sampling approach. At analysis, the focus may switch to other commonalities such as shared barriers or enablers in accessing health and social care resources.

Possible attrition between sweeps (up to 20% based on ActEarly experience) may require further recruitment if theme/pattern saturation is not reached, or early saturation may lead us to more theoretical sampling. The main interview inclusion criteria are summarized in Textbox 3.

We recognize the heterogeneity within these groups, and the way these categories are laden with assumptions such as with regard to multimorbidities, the concept of being British, the ancestry of people who are born in particular countries, and the apparent essentialization of specific groups. However, to ensure in-depth data while keeping sample numbers to a feasible level, we decided to use these problematic categorizations as tools to better organize our research so that we can then unpack and critique them [84]. The groups have been chosen to reflect recent migration waves to the United Kingdom (albeit that some people from these groups may have lived in the United Kingdom for decades) and to capture those groups most at risk of hospitalization or death from COVID-19.

The main interview exclusion criteria are: (1) student migrants, as they are likely to have structured educational institution support; and (2) residents of detention centers/closed facilities linked to national migration policies (eg, new asylum-seekers/refugees, displaced or trafficked persons), as these are deemed complex cases with specific considerations.

We sample from five interview sites within England for maximal sampling diversity in migrant population density, proportion of EU to non-EU migrants, and reasons for migration to enhance transferability, which we will explore against our four-nations survey findings. This is important as we only sample in England for qualitative work due to differences in the devolved nations’ responses to the pandemic and their health and social care systems. Sites for our qualitative research are London, Southeast England, Northeast England, West Midlands, and Yorkshire. While this means that some of our findings may be more relevant to England, we expect the principles to be similar across the four nations, and we will consider this in our reporting and outputs.

We will recruit interview participants from advertisements/links distributed through a range of platforms and networks, as well as through local lay coresearchers. We will rely on participant self-identification of citizenship status and condition/impairment. Posters, advertisements, and snowball sampling will target those who lack resources or technology to be recruited via online messages; they can contact us by telephone or email. Collaborators will provide recruitment and data collection support through organizations such as Born in Bradford (BiB) in Yorkshire, a long COVID center [85] in Gateshead, the Bromley-by-Bow community center in East London, and migrant charities in London and Canterbury.

Respondents can choose to have their interviews by phone, remote video methods, or face to face, depending on extant pandemic restrictions. Our PAG leads will help train lay community members to undertake some interviews locally at our five sites, supported by the core team. Interviews will be recorded. Attention will be given to making the interviews fully accessible and inclusive, and all researchers will be vigilant to the participant needs such as requiring frequent breaks.

All potential participants will be informed about the study in plain English (read to them if needed) and told that interviews will be in English by default. Where a participant feels more comfortable being interviewed in another language, if a researcher fluent in that language is available, this will be arranged. Translated and accessible study documents will be provided if required to ensure that participants are able to give fully informed consent.

We will probe in interviews for the same topics as covered in the survey (Table 1). For social networks, we will discuss the data from the following participant preinterview tasks: (1) a brief questionnaire, data from which we will translate into network “maps” using Network Canvas software; and (2) photographs and sketch-maps of the local area where people live, and the places significant to their health care and social interactions, using their smartphones or disposable cameras that we will provide. These data will also be thematically analyzed. This ethnographic approach facilitates a safe social space to communicate difficult issues and has been used to explore migrant resettlement [86].

The two research workshops and four cocreation meetings will all be participatory and designed for participants to work in equitable partnership. They will be led by a core team member and a PAG member. They will aim for outputs relevant to the “real world” that will maintain participant voices and will ensure the research outputs can be implemented. Each session will last approximately 4 hours (2 hours if held remotely).

The make-up/number of research workshop groups per sweep (2 and 3) will be determined by considering any typologies (patterns in intersectionalities and outcomes) from sweep 1 data. Participants will be recruited from sweep 1 interviews.

Sweep 2 workshops will discuss scenarios, or structured vignettes, shown as short videos recorded by community members reading scripts. Content will be developed from sweep 1 data into a pandemic-relevant story, with accessible transcripts provided in advance. Discussion will consider changes from previous findings. Sweep 3 workshops will follow a similar pattern with updated vignettes. We will also use participatory scenario planning [87], a policy tool whereby participants are encouraged to explore alternative futures, their impacts, and relevant action plans. To ensure inclusivity, we will work with our PAG group on workshop accessibility and will offer repeat interviews as an alternative.

Cocreation meeting participants will include patients and carers (our aim is that they will be representative of our interview participants), as well as key stakeholders in their support and care. They will be recruited via professional or dedicated community networks such as government and policy, welfare, social and health care staff migrant, settlement, and racialization-specific services; third-sector organizations; and community leaders. We will aim for two representatives from each of these groups per workshop, thus with approximately 20 people in each meeting. We will be assisted in this process by our PAG.

The cocreation meetings will involve discussion of findings from each data sweep and their cocreated translation into outputs to feed into the next stages or final study outputs, depending on what is appropriate at the time each workshop is held. To enable inclusion and stimulate discussion and outputs, we will use arts-based and participatory approaches such as miro.com, Collaborative Poetics materials [88], and other cocreation tools [89,90].