Making healthcare accessible for single adults with complex needs experiencing long-term homelessness: A realist evaluation protocol

Background: Over the last several years, homelessness has increased in Ireland and across Europe. Rates have recently declined since the coronavirus disease 2019 (COVID-19) pandemic, but it is unclear whether emergency housing measures will remain in place permanently. Populations experiencing long-term homelessness face a higher burden of multi-morbidity at an earlier age than housed populations and have poorer health outcomes. However, this population also has more difficulty accessing appropriate health services. A realist review by the authors found that important health system contexts which impact access are resourcing, training, funding cycles, health system fragmentation, health system goals, how care is organised, culture, leadership and flexibility of care delivery. Using a realist evaluation approach, this research will explore and refine key system-level factors, highlighted in our realist review, in a local health care system. Aim: The aim of this study is to understand how funding procedures and health system performance management impact service settings, staff, providers and their ability to make services accessible to populations experiencing homelessness. Methods: A realist evaluation will be undertaken to explain how funding and health system performance management impact healthcare accessibility for populations experiencing homelessness. Data will be collected using qualitative and realist interview techniques and focus group methodology. Secondary data such as policy documents and budgets will utilised. The analysis will follow Pawson and Tilley’s iterative phases starting with building an Initial programme theory, then data collection, data analysis, synthesis and finally building a refined programme theory. Conclusion: Building on a realist review conducted by the same research team, this study will further test and refine findings that explain how health system factors impact healthcare accessibility for populations experiencing homelessness. The study has the potential to inform policy makers, health planners and managers of contextual factors that can be modified to increase healthcare accessibility.


Introduction
Homelessness has been on the rise in Ireland and internationally over the last several years. From March 2015 to March 2020, the total number of people experiencing homelessness in the Republic of Ireland (including single adults and families with children) grew from a total of 4,261 people to 9,335 representing an increase of 119 percent 1 . The total number of people being reported as homeless has declined since the outbreak of coronavirus disease 2019 (COVID-19) to 8,728 people in July 2020 1 . However, it is unclear whether emergency housing measures will be in place permanently 2 .
According to the European Typology of Homelessness and Housing Exclusion (ETHOS) homelessness occurs on a continuum including the following experiences: • Rooflessness (sleeping rough, without any shelter) • Houselessness (having somewhere to sleep but temporary in shelter or institution) • Living in insecure housing (insecure tenancies, threat of eviction, violence) • Living in inadequate housing (overcrowding, unfit housing, caravans on illegal campsites) 3 The homelessness statistics in Ireland cover several different populations with somewhat differing needs, including single adults and families. The people included in the statistics are those relying on emergency accommodation and does not include people who are staying with family members, rough sleeping, camping, etc 1,4 .
In addition to the physical aspects of homelessness there is an important temporal aspect. Researchers use three temporal categories: crisis homelessness, intermittent homelessness or chronic homelessness. Crisis homelessness occurs for people who experience sudden homelessness after an eviction, divorce or job loss. Intermittent homelessness occurs for people who move in and out of homelessness interrupted by access to adequate housing or institutional care stays. Chronic homelessness is defined as experiencing homelessness lasting more than a year continuously or for individuals with a disabling condition, four episodes of homelessness over a period of two years 5 .
The focus of this research is on the subset of the homeless populations who are single adults experiencing chronic homelessness. This population group experiences lengthy periods of homelessness while moving between any or all types of accommodation outlined in the ETHOS typology, sometimes also with time spent in institutional settings and/or in sufficient accommodation 5 . Such people typically have lived lives marked by trauma from a young age and often have complex health and social care needs 6,7 . Many experience tri-morbidity, that is the simultaneous presence of physical and mental ill health and substance use disorders 8 . As a result, populations who experience homelessness and other forms of social exclusion have higher levels of morbidity than other populations and the onset of morbidity and multimorbidity happens at an earlier age 9-11 . Homeless people also tend to die at a much younger age than their housed peers 10,11 . In Dublin the median age of death for homeless people who die while accessing Dublin Region Homeless Executive supported services is 42 years old (44 years old for males and 37 years old for females) 12 . In line with Tudor Hart's inverse care law, this high-need population generally has more difficulty accessing healthcare than other populations, despite their poorer health and low life-expectancy 7,13-16 .
The majority of research examining healthcare access for populations experiencing homelessness is focused on the patient journey from the perspective of the individual seeking to access health services 13-15,17-21 . While it is important to understand the factors that make healthcare accessible for highly vulnerable population groups, there is a risk that this focus leads to inappropriately placing responsibility for accessing healthcare with populations experiencing homelessness and not with the health system. To remedy that imbalance, this research aims to understand access from a systems perspective.
Health systems are also characterised by a high level of complexity as they consist of a number interacting people, places, policies and actions as described in the WHO's definition of a 'health system': '(i) all the activities whose primary purpose is to promote, restore and/or maintain health; (ii) the people, institutions and resources, arranged together in accordance with established policies, to improve the health of the population they serve, while responding to people's legitimate expectations and protecting them against the cost of ill-health through a variety of activities whose primary intent is to improve health' 22 . In other words, a health system is an open system with many parts, at times interacting, and moving along non-linear pathways on diverse timelines. At the intersection with patients experiencing homelessness seeking to access healthcare services, health systems contribute to a number of intended and unintended outcomes which emerge as more than the sum of their parts (eg a patient may nominally access healthcare in the emergency department but leave before being seen even through physical access is possible due to a stigmatising culture which makes the environment unwelcoming) [23][24][25][26][27] Realist approaches (realist review and realist evaluation) are designed for understanding complex phenomena. The realist logic of inquiry is based on an understanding of reality as existing independent of our ability to observe it. It views the world as being one where we cannot observe or measure many of the processes that produce outcomes we are interested in 28 . For example, in seeking healthcare access, a person experiencing homelessness may be met with stigma and feel unwelcomed and as a result decide not to seek care. In this case,

Amendments from Version 1
We have responded to helpful reviewer comments to make the protocol clearer with regards to the justification for using realist evaluation and how this work builds on a previous realist review by the same authors. The title was also revised for clarity.
Any further responses from the reviewers can be found at the end of the article

REVISED
we can measure the lack of realised healthcare access in low healthcare utilisation. However, we cannot independently observe and measure stigmatising attitudes nor can we measure the response arising in the person experiencing stigma. In the realist view of the world, we can theorise about what is happening to cause the outcome of poor healthcare access by understanding the social and psychological processes commonly at play in situations where health services prove accessible or inaccessible for populations experiencing homelessness.
In the school of Pawson and Tilley 29,30 , realist evaluation seeks to understand the underlying mechanisms that generate a given outcome in a particular context using primary data. It is methods neutral and can employ data sources from a variety of study designs and methods. Realist approaches seek to uncover the conditions in which something works and for whom it works and understanding why rather than merely whether something works or not.
To add to the understanding of health system factors that improve access to healthcare for long-term homeless adults, we carried out a realist review 31 . This realist review produced a number of context-mechanism-outcome configurations (CMOCs) and an overarching programme theory (see Figure 1). This programme theory synthesised the full set of findings from the review showing the interlinking set of factors which must all be in place for health systems to successfully provide healthcare access to populations experiencing homelessness. The programme theory explains that important health system contexts which impact access are resourcing, training, funding cycles, health system fragmentation, health system goals, how care is organised, culture, leadership and flexibility of care delivery. Key mechanisms which arise in these contexts are provider attitudes, provider confidence, staff and provider experience of stability and sustainability, organisation-centred thinking, flexibility, transparency, timeliness, demonstration of respect and empathy, trust, adaptability, and anticipation.
Conversely, when those system features are partially in place or not in place at all, the patient experience of health services is impacted negatively. Eg when health services are fragmented, care cannot be delivered in a timely manner because each service or speciality operates on their own schedule and each step on a patient pathway depends on a referral from the last.
This study will focus on a subset of the findings of our realist review to evaluate how funding procedures and health system performance management impact service settings, staff, providers and their ability to make themselves accessible to populations experiencing homelessness. We have chosen to further investigate these two areas of study because they are under-researched and they will contribute to the work of planners and policy makers 32-34 .

Study design
In this study, we will use realist evaluation, a theory-driven approach well-suited for analysing complex topics and interventions as detailed above.
Data collection will take place in Dublin with the goal of learning lessons particular to how health services make themselves accessible or not to populations experiencing homelessness. The study will be limited to Dublin region homeless healthcare services because the majority of homeless adults in Ireland live in Dublin 35 and the majority of healthcare services are provided in Dublin 36 . Transferrable findings that will be useful internationally will be generated.
The study will employ documentary analysis of policy documents (Health Service Executive (HSE), Department of Health, non-government organisation (NGO) sector) and health service utilisation and budgeting statistics, as well as semi-structured and realist interviews with stakeholders in the health and NGO sectors. This study will also include focus groups with individuals with lived experience of homelessness who will aid in challenging, confirming and further refining study findings.
The study will follow the iterative realist evaluation design as set out by Pawson and Tilley 30 in the following phases ( Figure 2): 1. Initial programme theory building 2. Data collection 3. Data analysis 4. Synthesis

Refined programme theory building
The phases will repeat as needed. If it becomes clear that more data or different sources are needed, these will be incorporated into the synthesis in order to produce the most complete refined programme theory possible.

Initial programme theory
Theory building is the start and the end of realist work -the work starts with an initial programme theory and ends with a refined one after iterative rounds of data gathering and theory building.
Building on our realist review, the team has selected two context-mechanism-outcome configurations (CMOCs) to further explore and refine. The review indicates what is known about what works in promoting healthcare access in homeless populations from a health systems perspective -see the full programme theory above in Figure 1. Due to time and resource constraints it is not possible to further test the full set of six detailed CMOCs generated in the review.
For this study we will focus on two particular areas of the full programme theory explained in the two CMOCs described below. These relate to: • How health services are funded • How health systems manage performance We have chosen to further investigate these two areas of study because they are under-researched and they will contribute to the work of planners and policy makers 32-34 .
Building on the below CMOCs from our review, we will, where possible, further explain and add to the relationships between the contexts, mechanisms and outcomes identified in the international literature by expanding the analysis with the use of primary data and additional secondary data sources. CMOC2 (Figure 4) suggests that when the parts of a health system operate in silos with narrowly defined goals 34,38,39,44,45 , performance management is aligned with meeting those goals and they become the priority of staff even if they are not aligned with the needs of their patients 39,44,45 . Performance management practices have tended to inhibit the ability of staff to deliver services in holistic, coordinated and flexible ways. As a result, healthcare is organised around the needs of providers and the system not the person 15,38,39,44-46 .
Building on these CMOCs and the understanding of the underlying causative relationships they describe, the analysis will evaluate how funding procedures and health system performance management impact service settings, staff, providers and their ability to make themselves accessible to populations experiencing homelessness.
We will seek to answer the following questions: • How are health services for homeless populations funded?
• For whom do funding arrangements work and not work, and why?
• How is health system performance management intended to translate into service design and delivery for populations experiencing homelessness?
• In what circumstances are health system performance management and organisational targets useful for creating healthcare accessibility for homeless populations and when are they not?
• Who (patients, staff, providers, the system) does health system performance management work for, to what extent?
Data collection Data will include grey literature and secondary sources identified through internet searching and snowballing as well as primary data collected via qualitative and realist interviews and focus groups. Realist interview/ focus group guides are available as extended data 47 . Approximately ten key informant interviews will take place with professionals with significant experience in preventive, primary and secondary healthcare sectors and NGOs that provide services to homeless, and the HSE National Office for Social Inclusion, as well as relevant experts and academics. These will be identified by members of the research team which has broad networks in health and social services for homeless populations in Ireland. Snowball sampling will also be used to identify further interviewees. Interviews will take place via online video conferencing due to  the current COVID-19 pandemic or in person, if and when possible. Interviews will be recorded and transcribed.
The research team will also conduct at least one focus group with experts by experience (individuals with lived experience of homelessness). An agreement to engage with individuals who are part of the Depaul peer health worker panel is in place but it remains to be seen if COVID-19 restrictions will allow us to proceed. It is likely that several focus groups conducted with subgroups within the target population would be useful.
Women who have experienced homelessness may feel more comfortable participating in a women-only group and the same may be the case for LGBT or migrant people. If possible, given time and COVID-19 constraints, the research team intends to conduct more than one focus group. If holding focus groups is not possible, other methods will be considered. The goal of the focus groups is to add a service user's perspective to the analysis of factors that impact health accessibility, particularly uncovering ones which may not be evident to individuals who have not personally had the experience of homelessness.
Data collection will be iterative and additional data sources may be collected as needed after the first data analysis phase to substantiate findings from interviews and focus groups. Eg if a particular policy or budget or procedure is discussed by a source, the research team will endeavour to access documentation regarding the item being discussed to add it to the analysis beyond the interviewee's opinion.

Data analysis
All collected data will be transcribed and then inputted and analysed in NVivo software. For key informant interviews, a set of initial codes will be generated inductively while coding the first several interviews. The same codes will then be used across the rest of the interviews with more added as needed. Data collected from the focus group will likely have some similar codes and some different codes due to the difference in perspective and will be coded separately and inductively. Policy documents and grey literature sources will also be included in the analysis and will be coded separately from the primary data with a set of initial codes being developed inductively from what the team deem to be the likely most useful sources. Subsequent documents will be coded deductively using those codes while adding more as needed. Sources which may be identified over the course of the data collection from interviews and focus groups will be coded using the same procedure.

Synthesis
Each set of data (interviews, focus groups and secondary literature) will be analysed separately and will then be synthesised at the interpretation stage using triangulation 48 to formally explore the agreement and disagreement between sources and to identify how these contribute to the analysis. The goal is not to only find instances where there is agreement between sources but also to investigate areas of disagreement or where findings only occur in one set of sources. Eg service users may have a perspective on healthcare accessibility that is invisible to planners and providers of services but which is nonetheless important to understanding how services are and are not made accessible.
Following the data triangulation, CMOCs will be built using a combination of data sources. CMOCs will be iteratively refined and further developed as needed based on team discussions. A final programme theory will be constructed from CMOCs at the end of the project.

Ethics
Ethical approval has been granted by the Trinity College Dublin Health Policy and Management and Centre for Global Health Research Ethics Committee (application 22/2019/01).
Given that the population in question is a highly vulnerable one, the team is keenly aware of our responsibility in conducting this research ethically and sensitively. Trusted and known gatekeepers will be used to recruit focus group participants to provide every opportunity for potential participants to say no to participating for any reason. Participants will also be given a thorough oral explanation of the informed consent form along with the written copy of the document to ensure that they understand that participation is voluntary and that they may leave the study at any time. A professional who is known to the individuals in the peer health worker panel and who has significant experience working with the population, will take part in moderating the focus group in order to help provide a familiar and safe environment. Should a need for any support services arise as a result of participation in the focus group, Dr Cliona Ni Cheallaigh, a member of the research team and the clinical lead for the Inclusion Health Team in St. James's Hospital Dublin will personally ensure that the individual or individuals are given the right care and treatment.

Dissemination of information
The main product of this research will be a research article which will be disseminated via a peer-reviewed journal. It will also be disseminated to research, policy making and practice networks, and through the Centre for Health Policy and Management, Trinity College Dublin website.

Study status
Documentary analysis and data collection via interviews has commenced. Three pilot interviews have been completed. Realist interviews are underway. Data analysis and synthesis is yet to be done.

Discussion
To our knowledge this is the first realist evaluation seeking to explain how funding procedures and health system performance management impact service settings, staff, providers and their ability to make themselves accessible to populations experiencing homelessness.

References
With the predominance in the literature of studies examining individual level factors that impact on healthcare access, there is a lack of understanding of the impact of system level decisions and priorities on downstream health service delivery for socially excluded populations. This study will add important knowledge to the field by specifically examining health system contexts and mechanisms.
A strength of this study is the use of a realist approach. The study will uncover modifiable contexts which impact upon healthcare accessibility and will be of interest to policy makers and healthcare planners and managers. A potential impact of the study is to inform health system and policy leaders, with the ultimate aim of making healthcare more accessible for populations experiencing homelessness.

Open Peer Review
Barbara Clyne

Royal College of Surgeons in Ireland, Dublin, Ireland
This paper describes a protocol for a realist evaluation of funding procedures and health system performance management on accessibility of services for adults experiencing chronic homelessness. Research on this vulnerable population represents an important area of health services research and the findings could have important policy impacts. The research builds on a previous realist review by the study team, demonstrating their expertise in this area.
The protocol is clear, well written and the overall study is well described. However, there are a number of clarifications that I feel would improve the manuscript overall. These are as follows: Title: The title and the study aim don't seem to align to me. The title "How accessible is healthcare for single adults experiencing long-term homelessness and complex needs?" implies that this study is about accessing healthcare from the perspective of adults experiencing homelessness. But the study aim is to understand how funding procedures and health system performance management impact service settings, staff, providers and their ability to make services accessible. So providers of homeless services and the system are really the focus of the study. If providers can make the service available, is different to whether or not adults experiencing homelessness then access it. I think this is especially the case for the focus group with adults who previously experienced homelessness who may or may not be part of the study. There is also a little contradiction within the introduction, 4 th paragraph, the focus of this research is homeless populations, and 5 th paragraph, aim to understand system perspective. Suggestion: making healthcare accessible for homeless adults....

○
The introduction gives a good justification as to why realist evaluation is suited to exploring barriers to health care access from the perspective of homeless populations, but as above, the focus of this study is the system so a sentence or two justifying why this approach is suitable would be good.

○
The research builds on a previous realist review by the study team. Figure 2 outlining the full programme theory is a nice visual, but I think a full statement of what the programme theory is would be beneficial.

○
The protocol is focused on 2 of the 6 elements of the overall theory from the realist review. Could the authors provide brief info on the other 4 elements and a more detailed description of why these 2 elements were selected over the other 4. It would also be helpful if the authors reflected on how the focus on 2 elements only impacts the overall theory? ○ For consistency, it would be helpful to describe CMOC 2 consistently with the rest of the manuscript in terms of calling it performance management, as opposed to goals and incentives. Or define what is meant by performance management earlier on.

○
Realist evaluations tend to require good data on outcomes and where possible, data about outcomes should be triangulated (at least using different sources, if not different types, of information). In terms of your CMOCs, where is the outcome data coming from? In terms of data collection overall, I think it would be helpful to provide a table/figure demonstrating the data source for each of the CMOs to ensure you have all the adequate data.

○
For your realist interview approach, you will be interviewing key stakeholders. With this approach, they will be asked to contradict/refine your theory. Could you comment on how you might reduce social desirability bias and get the respondents to really challenge the researcher. How experienced in their area will the participants be and how senior/junior will the researcher be? ○ Will you be using any reporting standards such as RAMESES II? ○ Is the rationale for, and objectives of, the study clearly described? Yes

I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.
Author These are as follows: Title: The title and the study aim don't seem to align to me. The title "How accessible is healthcare for single adults experiencing long-term homelessness and complex needs?" implies that this study is about accessing healthcare from the perspective of adults experiencing homelessness. But the study aim is to understand how funding procedures and health system performance management impact service settings, staff, providers and their ability to make services accessible. So providers of homeless services and the system are really the focus of the study. If providers can make the service available, is different to whether or not adults experiencing homelessness then access it. I think this is especially the case for the focus group with adults who previously experienced homelessness who may or may not be part of the study. There is also a little contradiction within the introduction, 4th paragraph, the focus of this research is homeless populations, and 5th paragraph, aim to understand system perspective. Suggestion: making healthcare accessible for homeless adults.... Yes good point. Thank you for that suggestion. We have amended the title.
The introduction gives a good justification as to why realist evaluation is suited to exploring barriers to health care access from the perspective of homeless populations, but as above, the focus of this study is the system so a sentence or two justifying why this approach is suitable would be good.
Thank you for pointing out the need to have more detail here. We have added more text to the document.
The protocol is focused on 2 of the 6 elements of the overall theory from the realist review. Could the authors provide brief info on the other 4 elements and a more detailed description of why these 2 elements were selected over the other 4. It would also be helpful if the authors reflected on how the focus on 2 elements only impacts the overall theory? Thank you for this suggestion.
We have chosen not to describe each of the six CMOCs in the protocol because they are included in the full set of findings described in the full programme theory and because to do so it would repeat a lot of the realist review in which they are each fully explained and which is available as a preprint (https://www.researchsquare.com/article/rs-79236/v1) (and hopefully soon will be published!) We selected the two which, based on the existing literature, had the greatest potential for impact while also being under-researched and we felt that two provide the right balance of depth and breadth. We are not able to comprehensively build on each of the CMOCs from the review and the two were chosen as a starting point for the realist evaluation because as they are under-researched and policy relevant. It is important to note also that they are just a starting point which will guide the initial data collection and initial coding of data in the evaluation. The end product will move beyond them to produce a number of subsequent CMOCs which will be refined and combined to produce a set of CMOCs and a final programme theory for the realist evaluation.
For consistency, it would be helpful to describe CMOC 2 consistently with the rest of the manuscript in terms of calling it performance management, as opposed to goals and incentives. Or define what is meant by performance management earlier on. Good point. Have made that change in the graphic and replaced and have updated text as well.
Realist evaluations tend to require good data on outcomes and where possible, data about outcomes should be triangulated (at least using different sources, if not different types, of information). In terms of your CMOCs, where is the outcome data coming from? In terms of data collection overall, I think it would be helpful to provide a table/figure demonstrating the data source for each of the CMOs to ensure you have all the adequate data. The two CMOCs carried over will be a starting point for generating many more using the data collected for this study. New CMOCs will be developed to further investigate the topics explored in these two CMOCs in further detail. To this end, the data for the study, including outcomes data, will be generated from policy documents, interviews and focus groups. Additionally, we'll use other data sources such as funding information from HSE Social Inclusion to understand how funding flows to service providers in the voluntary sector, as well as peer reviewed and grey literature which discuss the topics. Each CMOC will always draw on several data sources. At this stage, since the CMOCs have not been constructed we cannot include a table to demonstrate the data source for each of them as that will depend on the data collection to come.
For your realist interview approach, you will be interviewing key stakeholders. With this approach, they will be asked to contradict/refine your theory. Could you comment on how you might reduce social desirability bias and get the respondents to really challenge the researcher. How experienced in their area will the participants be and how senior/junior will the researcher be?
The interview technique is informed by a number of sources (eg (1,2)) as well as practical learning undertaken by the primary researcher in trainings provided by experts in the field (Geoff Wong and Justin Jagosh). As outlined in the RAMESE guidelines for realist interviews(1), theory testing in the interview setting is a process that takes place bit by bit and the researcher's theories are typically not introduced in their full form nor formally as 'here is my theory.': 'Interviews should start with general questions about interviewees' role in, experiences of and views about the programme. Subsequent questions follow up their responses, asking them to tell their stories about specific experiences or issues with the programme, its participants, and constraints. Each of these questions investigates something about the programme theory being tested. Specific elements of programme theory can then be explicitly introduced and tested with the respondent. In order to avoid allegations of 'leading the interview', it is a good idea to test multiple, including contradictory, theories about the same aspect of the programme with the same respondent'. (1) The realist interview process is often described as a 'teacher-learner' process whereby the interviewer presents theories under examination using normal (non-realist) language as the 'teacher' explaining some aspect of their findings to the interviewee. For example, in our study the researcher may ask 'I have heard from other people I have spoken with (or I have read in the published literature) that funding cycles in health services from homeless populations are often short and unstable. Is that your experience? Could you tell me more about that?' to begin to test CMOC1. The teacher and learner roles are not static -the interviewer moves fluidly back and forth. We may then, moving into learner mode, depending on what the interviewee answers, follow up asking about the effects of long or short or mixed funding cycles they've experienced in relation to staffing stability and their own experience of working for a given service. We might then ask about other aspects of the topic such as, for example explaining that in the literature we have read that that people working at the frontline are seen by their managers as naturally caring people who are rewarded by doing caring work and therefore are not always compensated fairly, and get their thoughts on that. And so on and so forth. (1,2) In this manner, the causal relationship between mechanisms and outcomes is investigated by asking about specific aspects of the theory rather than by laying out the whole theory and asking for validation. This should diminish the risk of social desirability bias as the researcher does not present theories as 'my work' but rather as what they've been told by other interviewees or learned from the literature, and often will present contradictory findings and points of view for discussion with the interviewee. Interviewees will be frontline medics, nurses and social workers and service managers with many years' experience as well one individual from HSE Social Inclusion (programme management level). Others will likely be added per snowballing.
The researcher undertaking the interviews is a PhD student.