Developing programme theories to address unmet needs of people experiencing homelessness and severe mental illness: protocol for a realist review

Patient and Public Involvement

Patient and Public Involvement, defined as research being carried out ‘with’ or ‘by’ members of the public, rather than ‘to’, ‘about’ or ‘for’ them³³, is a key feature of realist approaches. Extensive and interactive engagement with relevant stakeholders helps to ensure authentic, contextually relevant evidence, clarity, and the transferability of findings^34–36. The HOPE project, in which the current research protocol was nested, has been working with a multi-stakeholder partnership of researchers, policy makers, practitioners, and people with lived experience of mental illness (Lived Experience Advisory Group-LEAG) starting from the HOPE project proposal development. Each HOPE country team (Ethiopia, Kenya, and Ghana) established a multi-sectoral Community Advisory Board (CAB) that meets at key stages to guide and inform the project³². Similarly, during the development of this realist review protocol, key stakeholders including multidisciplinary professionals and people with lived experiences of mental illness (n=7) involved throughout the protocol development process and included as co-authors. We will draw on patient and public involvement as part of our active stakeholder engagement throughout the actual study phases.

Accordingly, we will convene a separate Research Advisory Group (RAG). The RAG members will comprise multidisciplinary professionals and researchers with combined relevant expertise in the area of people who are homeless and have SMI, complex intervention design, and people with experiential knowledge of homelessness and SMI. The members will be selected from the HOPE multi-stakeholder consortium members, including investigators from each HOPE project country, all members of the LEAG, and other relevant professionals. The RAG will provide methodological guidance and feedback throughout the study phases. In addition to the RAG, we will actively involve and work with the LEAG throughout the review process. The LEAG and RAG will meet online at least three times (along with continuous e-mail exchanges) during the review process. Key findings of each phase of the review will be presented, and feedback will be obtained from the LEAG and RAG. Notes of the meetings will be kept on programme theory to maintain an audit trail of the decisions made and improvement of the findings.

Rationale for a realist review

Homelessness is often described as a “wicked problem”, which refers to a complex problem resulting from multilevel and interacting individual, interpersonal, organizational, and systemic factors, which require a coordinated solution from many stakeholders, including social and political actors^41–43. Co-occurrence of homelessness and SMI requires approaches for evidence synthesis and intervention development which are able to account for complexity^1,26,44. Theory-driven approaches such as realist review is appropriate for a deeper understanding of complex social programmes within heterogeneous social contexts^26,35,45.

Unlike traditional systematic reviews that focus on whether interventions work, realist reviews focus not only on “what works” or the effectiveness evaluation of interventions but also on the underlying theories and mechanisms. This offers an opportunity for generating context-sensitive causal explanations about how, why, for whom, and in what contexts interventions work or do not work^24,29. Realist reviews bring together evidence and insights from multiple sources, allowing synthesis and triangulation of theoretical and empirical evidence. This approach promotes stakeholder engagement, optimizes learning and generates accessible recommendations for policy and practice that can be applied in different settings^28,40,46.

Furthermore, realist reviews aim to recognize and account for the complexity and non-linear causal relationships of social systems at multiple levels: macro-level (systemic external situations that influence the function of the programme), meso-level (organizational culture, rules, regulations, structure and functioning of services), and micro-level (knowledge, attitudes, skills, values, health, and behaviour of individuals involved in the programme)^29,40,47. In realist approaches, the social world is understood by identifying underlying causal mechanisms using generative causation through realist logic. Realist logic includes inductive, deductive, abductive, and retroduction reasoning (See Table 1 for definitions of key realist concepts)^48–50.

As a theory-driven approach, realist review begins with IPTs and ends with a refined programme theory that explains how a complex intervention or programme works^29,34,40. Programme theories are typically expressed as context-mechanism-outcome (CMOs) configurations^29,38,51. For the proposed review, we will use the expanded model of intervention-context-actors-mechanisms-outcome configurations (ICAMOs)^37,49,52. The ICAMO approach postulates that “outcome(O) is produced by mechanism(M) activated in context(C) through actors(A) when interventions(I) are implemented”⁵³. Realist approaches recognize that interventions and actors interact within their social reality, which also influences how the intervention is implemented and how the actors respond (or not) to the resources offered by the intervention^39,54. The ICAMO configurations make programme resources more explicit (i.e. differentiate intervention from context), and seek to clarify actors' roles, perspectives, and reasoning in designing and implementing interventions^37,49.

Realist review stages

There are several ways to conduct a realist review, each with a different number of iterative steps^35,55. This review will be informed by the iterative stages set out by Pawson et al.^29,56, and recent realist review methodological papers that provide further depth and information regarding searches, data extraction, analysis, and synthesis approaches^48,55,57,58. The findings will be reported according to established realist quality and publication standards: Realist and Meta-narrative evidence syntheses–Evolving standards (RAMESES)^24,59. Each stage will incorporate stakeholder consultations. The review protocol for this study was preregistered with the International Prospective Register of Systematic Reviews (PROSPERO, CRD420251067136); and the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Protocols (PRISMA-P)⁶⁰ checklist can be found online on Open Science Framework, https://doi.org/10.17605/OSF.IO/G432U⁶¹. The stages of the review are illustrated in Figure 1, and detailed below.

Figure 1. Overview of realist review steps, adapted from Wong G, et al.⁶⁵.

1. Step 1. Clarifying the scope of the review

Different approaches can be used to clarify the scope of the study and identify/develop IPTs; for example, informal scoping reviews, programme documents, systematic reviews of scholarly articles, and interviews or focus groups with stakeholder groups⁶². In this study, we will use multiple data sources. We will conduct a preliminary informal literature search to identify existing literature about interventions for people who are homeless and have SMI.

More importantly, we will synthesise findings from formative work conducted in the HOPE project³². This includes a systematic review⁶³, a global expert consultation exercise, and ethnographies, surveys, theory of change workshops and intervention mapping within the study sites in Ethiopia, Ghana and Kenya. The HOPE systematic review aims to identify interventions for people who are homeless and have SMI in LMICs, as described elsewhere³². Findings will provide information on the scope of existing evidence, types of interventions, and implementation strategies in LMICs, which can inform the scope of the proposed review. In our study, we will deductively analyse data from the formative works and scoping review to identify the intervention, context, actors, mechanisms and outcomes (ICAMO). We will create a codebook for the elements of ICAMO. The codebook will guide the deductive coding of relevant documents and serve as starting point for stakeholder consultations and the literature review^58,64.

The ICAMO configurations will then be mapped using the realist logic of abductive and retroductive theorizing to develop the IPTs⁵⁰. The IPTs will be formulated as early working hypotheses structured in the format “if…then…because” statements, which make tentative links between ICAMO elements and explanations of which interventions address the unmet needs of people who are homeless and have SMI, for whom, how, why and in which contexts^59,65.

We will conduct a scoping review of literature to locate existing formal theories (also called substantive theories) to provide a framework to inform the development of the IPTs³⁹. Formal theories are pre-existing theories within a range of disciplines, such as behavioural sciences, social sciences, at a higher level of abstraction than programme theories, which provide a bridge to a wealth of existing research and knowledge about a topic^39,49,66. Formal theories help interrogate IPTs about how an intervention brings about change at multiple levels (individual, interpersonal, institutional, infrastructural, and structural)^24,39,67. We will conduct a scoping review using the BeHEMoTh approach (behaviour of interest, health context, exclusions, models or theories) to locate existing formal theories in studies for people who are homeless and SMI⁶⁸. The identified formal theories will be presented for stakeholder groups (described below) to select and prioritize relevant theories which inform IPT development.

2. Search for evidence

Relevant academic literature will be used to refine the IPTs and test the hypotheses formulated in Stage 1^24,58. The process of searching, screening and analyzing relevant material in a realist review is not a single-point exercise and has no one prescribed method. Rather, the process requires ongoing multiple iterations, and different search methods to support the different components of the realist synthesis process (See Figure 1)^24,55,58. In the proposed review, formal and additional searches will be conducted. Details of the search process and record of any subsequent alterations will be documented in a research logbook to provide transparency and enhance the quality of the review^24,55.

Formal search

A formal search for published documents will be conducted in databases, which include MEDLINE, PsycINFO, EMBASE, the Cochrane Central Register of Controlled Trials (CENTRAL), Global Health, Cumulative Index to Nursing and Allied Health Literature (CINAHL), PsycARTICLES, Web of Science, Global Index Medicus; Applied Social Sciences Index and Abstracts (ASSIA), and International Bibliography of the Social Sciences (IBSS). Many of these sources are chosen due to their use in the HOPE systematic review related to this study⁶⁹. The search terms will be adapted from the HOPE systematic review, but increasing the scope by not restricting the search to LMICs.

Additional searches

Where necessary, we will use complimentary search techniques such as ‘CLUSTER’ searching (Citations; Lead author; Unpublished materials; Scholar searches; Theories; Early examples; Related projects), to identify associated additional documents that might add to the conceptual and contextual richness of documents identified in the formal search⁷⁰. This search will be purposive, directed by emerging programme theories using search techniques such as searching for "sibling" (i.e., directly linked outputs from a single document) or "kinship" (i.e., associated papers with a shared contextual or conceptual pedigree) papers associated with included documents^55,70,71. We will also conduct forward and backward citation searches, using Google Scholar to identify further related papers from the wider literature. Searching will continue until sufficient data are found (‘theoretical saturation’) to conclude that the refined programme theories are sufficiently coherent and plausible^24,55.

Inclusion and exclusion criteria

We will apply pre-determined inclusion and exclusion criteria, which will likely be refined in light of the emerging data and theoretical development as the review progresses²⁴. The eligibility criteria will be broad and include the following:

Study designs: In keeping with RAMESES guidelines^24,59, there are no restrictions on the types of study designs. However, given that there is a large body of systematic reviews available on various intervention types for people who are homeless and have SMI, to limit the scope of our review and be pragmatic due to time constraints, a systematic review of reviews (quantitative as well as qualitative) will be considered.

Population: The population of interest for this review is adults aged 16 years or above who are homeless and have SMI. Homelessness is operationalised as “(a) literal (street) homelessness, (b) residence in a homeless shelter, squatting, (c) admission in a psychiatric facility, jail or religious healing facility with no place of residence to go post-discharge, and (d) previous street homelessness. This definition excludes the following groups: (a) individuals described as “unstably housed, which include people living in overcrowded homes, living in substandard places with inadequate amenities)”, (b) “landless” individuals, or those living in accommodation with insecure tenancy, (c) individuals living in slum/shantytowns but in stable accommodation, (d) internally displaced people, (e) child homeless populations”¹⁶.

SMI will be operationalised as “diagnosis of schizophrenia, schizoaffective disorder, bipolar disorder with psychotic features, and major depression with psychotic features according to ICD-10/DSM-V guidelines or clinical assessment. We will also include studies that use the broader terms of SMI, psychosis, psychosocial disability, mental disability, people who are homeless with comorbid SMI and alcohol/substance use disorders”¹⁶. We will exclude people with substance use disorders without SMI; people with SMI who are at risk of homelessness, and people living in substandard housing.

Intervention(s): The main interventions identified from the HOPE intervention mapping exercise and co-produced ToC include, but are not limited to, are community-based interventions intended to address the multifaceted complex unmet needs of people who are homeless and have SMI. These include: housing interventions (including supported living, community-based residential interventions), self-help, peer support, drop-in centres (e.g., day centres), empowerment interventions (i.e. interventions to strengthen exercise of human rights and/or to increase involvement of people who are homeless and have SMI in their own care and/or service improvement and/or system strengthening), recovery or rehabilitation (e.g., community-based rehabilitation), family-based interventions, substance use interventions, physical health interventions, non-governmental organization-based interventions, livelihood/vocational interventions (i.e. any interventions to improve economic status or livelihoods at the individual and/or family level), interventions to increase engagement in education, interventions designed to increase social inclusion of people with lived experience of homelessness and SMI, psychological interventions, other support interventions (such as provision of food/clothing/other basic needs), service models for delivery of care, and studies to support medication adherence or to safely manage discontinuation. However, we will exclude efficacy studies of pharmacological interventions, and studies of long-term institutionalization.

Contexts: We are looking for evidence about interventions in largely community-based settings (excluding humanitarian settings) for people who are homeless and have SMI. Intervention settings may include residential settings (including temporary or long-term shelters/hostels), the health system (primary, secondary and specialist mental health care levels), and traditional or religious facilities. The review will have no restriction by geography. We will thus extend the HOPE systematic review, which focused solely on LMICs.

Outcomes: The outcomes of interest include, but are not limited to those prioritised by people who are homeless and have SMI as identified from the HOPE ethnographies and surveys in Ethiopia, Ghana and Kenya, and identified by the HOPE global expert consultation and systematic review. Domains of relevant outcomes include basic needs (water, food, self-care, clothing, shelter), social needs, clinical outcomes (mental and physical health, mortality), livelihood (economic outcomes at the individual and/or household levels), social inclusion, empowerment (i.e. strengthening of human rights, support or advocacy for people who are homeless and their caregivers), and quality of life/well-being.

Language: English

Publication dates: We will restrict our focus to studies published within the past 10 years (2015–2025) to ensure the findings reflect the relatively contemporary landscape of programmes for people who are homeless and have SMI.

3. Document selection and appraisal

Screening

All data sources will be exported to Endnote X7 reference manager. Selection will be accomplished in stages. First, author SA will screen all potentially relevant documents retrieved by title and abstract for relevance against inclusion and exclusion criteria detailed above. Second, the full-texts of documents deemed to meet the inclusion criteria will be read in detail and assessed for relevance, richness, and rigor as detailed below^24,36,48. A random sample of 10% of documents at the two screening stages will be checked by another reviewer for consistency of application of screening criteria. Any disagreement will be resolved through discussion or if necessary, by consulting with the rest of the team.

Appraisal of the studies will use the criteria of relevance, richness and rigour⁴⁸. The relevance of a document refers to its ability to provide information related to theory building, testing, and the review`s aims and objectives^24,35,48. Rigor in realist approaches refers to the credibility, plausibility, and trustworthiness of the methods used to generate the relevant data or a theory^24,48. Assessment of rigor will focus on the extent to which a document/study provides a detailed description of their methods (e.g., sample size, data collection, data analysis), whether specific inferences drawn have sufficient support to make a methodologically substantial influence when developing/testing IPTs, and how reliable and generalizable their findings are based on those methods^24,35,48. This traditional assessment of rigor should be balanced, as “nuggets of wisdom” for realist review can be found in methodologically weak studies according to the conventional hierarchy of evidence^36,72. Hence, in line with realist review practices, we will not exclude any study based on rigor; rather, studies providing relevant evidence will be included, and results of appraisal of rigor will be included for transparency⁷³. Richness refers to the level of theoretical and conceptual depth and detail provided in a document^24,48,70. This help to assess the extent to which texts could provide in-depth explanation of how and why an intervention worked⁷⁰. Sources can have ‘conceptual richness’ and/or ‘contextual thickness’⁷⁰. Contextual thickness entails sufficient detail that enables the reader (i) to establish what is occurring in the intervention and the wider context and (ii) to infer whether findings can be transferred to other people, places, situations, or environments. Appraising for richness ensures only sources that can meaningfully contribute to the research question are included, making the review manageable and the findings more relevant^48,74.

For all studies that meet our eligibility criteria and are included in the review, data quality appraisal will be undertaken using predefined criteria for assessment of relevance, richness, and rigor adapted from published studies (See Table 2)⁷⁵. Each article will be appraised and classified into high, medium and low ratings for each appraisal category depending on its contribution to IPT development and established methods used in previous studies^35,48,75. No article will be excluded or included based on just one criterion, judgement will be made based on the overall value added to the research question.

4. Data extraction.

SA will design and pilot a bespoke data extraction form in a Microsoft Excel spreadsheet to extract text segments across each full-text study deemed to be relevant, rich, and rigour congruent with realist review guidelines^35,58. First, studies classified as high according to our appraisal will be extracted followed by those rated medium/moderate and finally low richness papers will be extracted for any additional data or insights that could contribute to the refinement of IPTs. Data extraction will stop when theoretical saturation is reached (until the literature no longer contributes novel insights)⁴⁸. SA will independently complete the extraction of data. Information will be captured on each document a characteristic includes: author(s), publication year, methods/design, population/participant demographics, setting, and key findings. In addition, elements of ICAMO will be extracted separately for each included study using the coding framework. For quantitative studies, rather than extracting numerical data, description of key findings and author`s interpretation will be captured. A random 10% sub-sample of data extraction documents will be reviewed independently by one other reviewer for accuracy. Any disagreement between the reviewers will be recorded and resolved through discussion or in consultation with a third reviewer.

5. Data analysis and synthesis.

Data analysis and synthesis are intended to find and align evidence using realist logic to make sense of the different data contributions and refine the IPTs^29,36. For the proposed study, these will be informed by the processes outlined by realist methodological papers, which will involve formulating ICAMO configurations at multiple levels, consolidating recurring patterns or demi-regularities, and conceptual mapping/theory building^35,57,76.

First, we will conduct a deductive thematic analysis of extracted data from each article and categorize them into the ICAMO components using a coding framework developed based on the IPTs in Step 1. The extracted data will be organized into an evidence table of ICAMO elements. The ICAMOs will be linked to the earlier IPTs to iteratively refine them into programme theories (PTs) by testing them against the ICAMO using a retroductive approach to uncover the causal mechanisms behind observed outcomes^50,77,78.

Secondly, we will consolidate recurrent ICAMO configurations into demi-regularities (statements of what works, for whom, and in what circumstances)^57,76,79. We will explore patterns within the ICAMO configurations and apply thematic labels based on similarity (categorizing conceptually similar ICAMOs)^57,80,81. The consolidation process helps manage and reduce large amounts of data through grouping or categorizing conceptually similar IPTs⁵⁷.

Finally, we will use the process of configurational mapping to create a visualization of the findings framed within the ICAMO model to explain the programme theory^37,57,79. We will use a conceptual map, such as feedback loop diagrams/causal loops, as an analytical strategy to map out pathways to identify and represent the elements of ICAMO and to illustrate the links between these constructs. This will help us to develop refined programme theories for people who are homeless and have SMI^57,77. Realist scholars contend that interventions are influenced by various levels of contexts(individuals, interpersonal, institutional settings, infrastructures/structural)⁸². We will inform the refined IPTs with relevant formal theories that support theorizing at these multilevel contexts^39,83 and construct conceptual maps depicting their relationships systemically^57,76,79. This analysis will enable us to understand the most relevant and important mechanisms that work at different levels and build more transferrable IPTs across various settings⁷⁷.

The refined IPTs will be presented and discussed with the RAG and LEAG for further feedback and to ensure an appropriate interpretation of results, contextual sensitivity, rigor, trustworthiness, coherence, and plausibility of the refined IPTs^24,65.

Next steps

The findings from this review, namely the refined programme theories, will be evaluated through realist evaluation of the intervention using a multiple case study design in Ethiopia, Ghana, and Kenya. The realist evaluation will enable further refinement of the programme theories through real-world testing. Then, the findings of the realist evaluation will be supported by relevant formal theories to substantiate the inferences made, make sense of the patterns of findings (demi-regularities) of emerging ICAMO configurations from HOPE pilot evaluation empirical studies, and develop middle-range theory (MRT), which will be transferable to wider contexts^24,34,39,49. This greater clarity on ICAMOs can further enhance transferability of the HOPE programme across different settings and usefully inform efforts to meet the needs of people who are homeless and have SMI.