Cognitive impairment and homelessness: A scoping review

Abstract This paper reports the findings of a scoping review designed to identify research which has explored the relationship between cognitive impairment and homelessness. A systematic search of databases for articles published between 2007 and 2017 was conducted using keywords relating to cognitive impairments and homelessness. Sources were expanded using manual searches of citations and grey literature. Forty studies represented in 45 papers were selected for review using predefined inclusion criteria. Sources were subject to quality appraisal and data were extracted in line with review questions. Prevalence studies were over‐represented in the review, while qualitative data were lacking. Aetiology of impairments was delineated by acquired and developmental causes. A variety of measures were employed by studies which were not validated in homeless populations. Studies did not give sufficient consideration to co‐occurring disorders and overlapping symptoms between aetiologies. Because of these factors, it was difficult to conclude that all studies had accurately measured what they set out to; however, the evidence suggested that cognitive impairment was disproportionately over‐represented in homeless populations. Cognitive impairment was found to be both a risk factor to and perpetuator of homelessness. Risk factors for homelessness were similar to those of the general population, though exaggerated by sequelae of certain cognitive impairments. The results of this review suggest that more attention needs to be paid to the underlying socioeconomic disadvantages, persons with cognitive impairments face which may lead to homelessness. Further research should prioritise the voice of homeless persons with cognitive impairments, to better understand both causes of homelessness and effective methods of rehabilitation.

. It is of concern that these figures only reflect statutory homeless, which is quantifiable due to interaction with government agencies or specific services. In 2011, the charity Crisis reported that 63% of single homeless persons were "hidden" from support and statistics (Reeve, 2011).
The consequences of homelessness for both the individual and society are damaging. A health audit in 2014 found that 80% of homeless persons self-reported a mental health issue and 41% long-term physical health problems (Homeless Link, 2014).
Malnourishment and substance (mis)use are also over-represented in the homeless population (Homeless Link, 2014;Sprake, Russell, & Barker, 2013). From a societal perspective, factors associated with homelessness put a strain on resources. In 2010, the Department of Health reported that homeless hospital usage cost the NHS an estimated £85 million a year (Department of Health, 2010). Overall the cost of preventing homelessness is reported to be far lower than the cost of managing it (Pleace, 2015).
Homelessness is associated with a range of physical and mental health disparities including cognitive and neurological impairments (Backer & Howard, 2007;MacReady, 2009). For this review, cognitive impairment (CI) will be used as a broad term encompassing functional and clinical impairments which affect cognitive ability.
Functional impairment refers to the affects a CI has on specific abilities such as problem-solving, attention, or memory. Clinical conceptualisations concern diagnoses such as dementia, traumatic brain injury (TBI), learning or intellectual disability (learning disability), and autism spectrum disorder (ASD) (Centre for Persons with Disabilities, 2018). Such diagnoses are represented in the general population at moderate levels. The prevalence of learning disability in adults in England is conservatively estimated to be 2.16% (Public Health England, 2015). Dementia affects 1.3% of the entire UK population and 7.1% of those aged 65 and over (Prince et al., 2014).
Regardless of aetiology, many of the functional aspects of CIs and the way they are understood and supported by society result in poor social and economic outcomes. While not all people experience the same challenges, certain CIs are associated with communication, emotional, and adaptive functioning difficulties (Headway, 2018;NAS, 2018a). Furthermore, persons with CIs, as with other disabilities, are often misunderstood, undersupported, and subjected to prejudice at both individual and policy level (Aiden & McCarthy, 2014;EHRC, 2017). Such experiences can inhibit a person's ability to find employment, maintain relationships, apply for welfare, or understand how to pay rent or bills (Backer & Howard, 2007). In turn, this may prevent a person securing or maintaining suitable housing.
Once homeless, it is difficult to identify and appropriately support individuals with CIs. Persons with CI can experience higher rates of mental health problems which are aggravated by homelessness (Eapen, 2014;Homeless Link, 2014), making it hard to distinguish underlying issues.
The limited research in this field tends to approach CI through the lens of brain deficit or diagnosis. Two known reviews have examined the prevalence of generalised CI in homeless populations (Burra, Stergiopoulos, & Rourke, 2009;Depp, Vella, Orff, & Twamley, 2015). It is hoped that by expanding such comparisons, we can better understand how to identify, support, and rehabilitate homeless people with CI. By employing systematic methodology, this review thus sought to (a) summarise existing research on CI and homelessness and identify gaps in knowledge, (b) ascertain how common CIs are in homeless populations and how they are measured, (c) identify risk factors to and perpetuators of homelessness for individuals with CIs. This article begins with a brief explanation of the epistemic framework to this study before setting out methods. Findings of this scoping review are followed by a discussion of context, including ramifications for current practice and policy.

| Epistemological framework
The social model of disability distinguishes between impairment and disability, understanding the latter to be structured by social oppression, inequality, and exclusion (Thomas, 2004). Although the variations of the social model are commonly adopted in disability research, it is often posited as dichotomous to a biomedical understanding of disability and criticised for denying the importance of embodiment (Oliver, 2012). Alternatively, disability can be understood in a social-relational way. This acknowledges that some restrictions of ability are down to impairment effects (Thomas, 2004), but that those which are "disabling" are caused by socially constructed barriers. This epistemic framework underpins this review and is returned to later in the article.
For the purpose of this review, the terminology "cognitive impairment" (CI) has been adopted. This phrasing reflects existing What is known about this topic • Research indicates that cognitive impairments are overrepresented in homeless populations.
• Research focuses on individual cognitive impairments such as brain injuries or learning disabilities.
• There is insufficient recognition and support for cognitive impairments in homeless populations.
What this paper adds • The first systematic scoping review to look comparatively at research examining cognitive impairments with different aetiologies in the homeless.
• Findings from this review question the usefulness of further prevalence studies, which do not consider the overlap of co-occurring disorders and do little to improve the lives of homeless people with cognitive impairments.
• Findings from this review suggest risk factors to, and perpetuators of, homelessness for people with cognitive impairments are exacerbated by inherent socioeconomic disadvantages.
literature, although it is noted such a deficit-focused language oversimplifies the aetiologies discussed.

| ME THODOLOGY
A scoping review is useful for determining the size and nature of a given knowledge base and can identify gaps in the literature for further research (CRD, 2008). While scoping review methodology was employed in a number of ways, for this study, a review was conducted in accordance with Social Care Institute for Excellence (SCIE) criteria (Rutter, Francis, Coren, & Fisher, 2013). This approach involves a detailed and systematic literature search to identify relevant papers followed by a process of data extraction and analysis.

| Search strategy
A systematic search of The Cochrane Library, MedLine, PsycInfo, Scopus, Social Care Online and Web of Science was conducted for literature published in English between 2007 and 2017. To cover a wide range of cognitive impairment, "homeless*" was searched for in combination with the following terms: autis*, asperger*, ASD, brain injury, TBI, ABI, cognitive impairment, cognitive function*, neurodevelopmental, NDD, learning disab*, intellectual disab*, ADHD, attention deficit, down's syndrome, dementia, william* 1 . Searches were limited to abstract, title, and keywords and were modified where appropriate to yield more relevant or wider scoping results. While multiple terms were used to denote CI, pilot searches indicated that papers examining specific types of homelessness such as rough sleeping or staying in shelters included "homeless*" in their title, abstract, or keywords. Synonyms and types of homelessness were therefore removed to avoid erroneous results. Database searches yield a total of 625 sources. Manual searches were performed using reference lists derived from database results. This was accompanied by a search of grey literature and known authors through the knowledge of the field. Internal search tools of homeless charity, government, and media websites were used to locate potential material.
This led to the inclusion of two research studies not published in academic journals (Campbell, 2015;Pritchard, 2010) and also provided supplementary sources which were used to corroborate findings.
Known reviews had included quantitative papers focussed on the prevalence of CI in homeless populations. Given the commitments to a social-relational understanding of disability, and the aim to identify risk factors to homelessness for CI populations, it was decided that qualitative papers are included in the review.

| Study selection
The search began on 1 November 2017 and concluded on 31 November 2017. Identified studies went through three stages of screening. Initially titles were scanned and subjected to the inclusion and exclusion criteria set out in Table 1; duplicates were removed at this stage. Exclusion and inclusion criteria were then applied via an abstract screening, after which texts were read in full. The rational for inclusion or exclusion of a source was founded upon the initial review questions and structural restraints of the review. For example, due to this review's focus on homeless adults, studies which exclusively sampled children or adolescents under # Inclusion/Exclusion Criteria Guidance 1 Include: homeless* in abstract or title AND (autis* or asperger* or ASD or brain injury or TBI or ABI or cognitive impairment or cognitive function* or neurodevelopmental or NDD or learning disab* or intellectual disab* or ADHD or attention deficit or down's syndrome or dementia or William*) in abstract or title.
Modify for databases as appropriate

| Quality appraisal
Included studies went through a process of quality appraisal and were given a rating of poor, fair, good, or very good. Due to methodological differences, studies were categorised by type and rated using distinct criteria. Criteria for appraisal was adapted where appropriate from multiple sources (Burra et al., 2009;CASP, 2017;CRD, 2008;Higgins & Green, 2008;Rutter et al., 2013). Qualitative papers, for example, were rated on research design, consideration of ethical issues, and centrality on user views, among other factors. Papers focussed on prevalence were assessed according to sample size, adjustment for confounding factors, and comparisons to matched control groups, among other factors. Studies were given a point for each matched criterion resulting in a score of up to 12 (systematic reviews) or 14 (prevalence, cohort, and qualitative studies).
All systematic reviews were rated good or very good. A greater range was displayed in empirical studies within which seven studies were poor, 13 fair, nine good, and eight very good (Keyser & Mathiesen, 2010, appraised against both qualitative and prevalence criteria). All qualitative studies presented details of how informed consent was obtained, though this was found to be lacking else- where. Other limitations included small sample sizes, not describing refusals, and not providing sufficient demographic detail. Studies tended to use convenience sampling and did not make efforts to reduce sampling bias.
Studies were assessed on the information available, and unreported information was treated as a negative score. It is acknowledged that reporting limitations may have negatively impacted the quality ratings of some studies. As criteria were adapted from multiple sources to increase their relevance to study type, it may be difficult to draw comparisons between the ratings of studies with different methodological processes. With these points in mind, quality ratings were used as a guide for inclusion in this review, as opposed to an objective measure. In the results section of this review, good and very good papers account for 63.7% of references, fair papers for 24.2%, and poor papers for 12.1%. Limitations of included papers have been noted where appropriate, with particular reference to those of poor quality.

| Description of included studies
A total of 40 studies were included. These consisted of 36 empirical studies and four systematic reviews.
Thirty-two studies were quantitative and primarily concerned with screening for CIs. Three qualitative and one mixed methods paper were also identified. Empirical studies mostly (n = 29) recruited from homeless populations, with definitions of homelessness ranging from absolute homelessness to recently rehoused, or in unsecure or unsuitable accommodation. Additional sample criterion included veteran status (Barnes et al., 2015;Brenner et al., 2017) and older age (Brown et al., 2017;Brown, Kiely, Bharel, & Mitchell, 2012;Hurstak et al., 2017;Joyce & Limbos, 2009;Okamura et al., 2017). Of studies which did not sample from homeless populations, two recruited from prisons (Hennessey, Stein, Rosengard, Rose, & Clarke, 2010;McCarthy et al., 2016), and other sampling frames included professionals who worked with the homeless , mental health service clients (Keyser & Mathiesen, 2010), and records of hospital admissions (McMillan et al., 2015). These studies were included be- RCT studies of the homeless (Nikoo et al., 2017;Stergiopoulos et al., 2015;Van Straaten et al., 2017. Some papers reported on the same project or sample. These papers are at times reported on independently due to differences in sample type (

| Measure and prevalence of cognitive impairment
This section describes the types of CIs examined by included studies. Prevalence rates are presented where available along with the measures or screening tools utilised. A table summarising the measures employed by prevalence studies is found in Table 5.

| Facets of cognitive disfunction
Studies in Category A measured facets of cognitive dysfunction.
Although all studies had higher rates of CI in their samples in comparison to the general population, CI was broadly defined and screened for using a variety of measures. The most common measure used

| Traumatic brain injury
The most prominent of aetiologies in Category B was TBI. As Table 3 illustrates, 13 of 18 studies reported on brain injury in the home- still gave a prevalence rate 5.4 times higher than the general population it was compared to. The highest rates of around 90% were found in the sample of veterans examined by Barnes et al. (2015) and Brenner et al. (2017) and may be partially accounted for by increased risk of TBI associated with serving in the armed forces.
Few studies investigated how head injury occurred; those which found assault to be the most common cause (Barnes et al., 2015;Gargaro, Gerber, & Nir, 2016;Topolovec-Vranic et al., 2014). Cause of injury which was further examined by Topolovec-Vranic et al. problems (HCDP), and low income housed found having a head injury was better predicted by previous head injury, drug dependence, or a seizure disorder than a history of homelessness. They note, however, that among the HCDP group, rates of head injury were 400 times higher than in the general population, and that the time interval between injuries decreased with each successive head injury.
Although TBI is associated with CI, this was not routinely measured. Andersen et al. (2014) found association between TBI and poorer cognitive performance in their homeless sample. Other studies which noted high rates of moderate to severe head injuries (Barnes et al., 2015;Brenner et al., 2017;Hwang et al., 2008;Svoboda & Ramsay, 2014) suggested these related to ongoing neurological deficits.

| Neurodevelopmental disabilities
Category C studies were concerned with developmental disability.
Developmentally caused impairments included learning disability, ASD, and ADHD. Qualitative research was most prominent in this category.
Ten studies concentrated on the prevalence and support of learning disability in the homeless population. These studies used variations of standardised measures to assess prevalence. Core intellectual ability was measured using the WASI alongside adaptive functioning by Oakes and Davies (2008), and Wechsler-based measures (WAIS III) were also adopted by Nishio et al. (2015).
Other measures of learning disability included the LNST (Keyser & Mathiesen, 2010)  The suggestion prevalence of CI in the homeless population may be partially accounted for by learning disability was also supported by one review (Depp et al., 2015), while the authors did not attempt to ascertain cause of CI, they note a below average mean score in studies that estimated premorbid IQ.
Other developmental CIs examined were ASD and ADHD. Using   (Evans, 2015).

| Risk factors and perpetuators
Risk factors and perpetuators of homelessness were often overlooked by studies where emphasis was placed on measuring prevalence. Those which were identified are examined below.
Economic success was negatively associated with all types of CI. The deficits following TBI such as with memory, attention, or planning made it difficult to maintain employment and housing (Anderson et al., 2014;Oddy et al., 2012). Low employment rates following a TBI were suggested as exacerbating circumstances and leading to homelessness (Topolovec-Vranic et al., 2013).
Similarly individuals with developmental disabilities had high levels of job instability (Hennessey et al., 2010;Keyser & Mathiesen, 2010;McCarthy et al., 2016). Low levels of educational attainment and inconsistent employment were also suggested as leading to poverty and subsequent homelessness (Hennessey et al., 2010). Many prevalence studies found that within homeless populations levels of educational attainment were lower for those with CI (Hurstak et al., 2017;Patterson et al., 2012;Rogoz & Burke, 2016;Stergiopoulos et al., 2015;Van Straaten et al., 2017). CI was also linked to contact with the Criminal Justice System. History of TBI was found to be independently and significantly associated with arrest and incarceration Topolovec-Vranic et al., 2014) and persons with NDDs were over-represented in the prison population (Hennessey et al., 2010;McCarthy et al., 2016).
Relationship breakdown was commonly cited as a cause of homelessness. Breakdowns could be the result of the death of caregiver, commonly a parent, with no appropriate support system in place, or as a result of social deficits such as difficulties building and maintaining relationships or vulnerability to abuse (Campbell, 2015;Hebblethwaite et al., 2007). Relationship breakdown was mostly associated with developmental impairments (Campbell, 2015;Gouveia et al., 2017;Mercier & Picard, 2011), though one study suggested emotional and family problems resulting from TBI as a cause of homelessness (Gargaro et al., 2016). Relationship difficulties were also considered to perpetuate homelessness by both individuals with learning disabilities and service providers working in the homelessness sector Hebblethwaite et al., 2007). Lastly, cultural differences were considered significant when a learning disability was seen as "incurable" or a "burden," as homeless individuals were less likely to be reintegrated with their families (Gouveia et al., 2017).
Substance dependence was also associated with those with developmental CIs (Mercier & Picard, 2011;Van Straaten et al., 2014). In Mercier and Picard's (2011) study, a third of their learning disability sample attributed homelessness to substance abuse. The complex interplay between mental health, substance (mis)use, CI, and homelessness was not directly explored in papers, but acknowledged, and seen as exacerbating existing problems (Mercier & Picard, 2011;Pluck, 2011). Overall co-occurring conditions were not seen to be adequately supported. Campbell (2015) noted that though eight of her 12 autistic participants had a co-occurring condition, only one was receiving support for this at the time of homelessness. Tripathi et al. (2013) also highlighted untreated mental health problems as the main cause and perpetuator of homelessness.
Some sequalae associated with CIs were seen as prolonging homelessness or as barriers to rehabilitation. Pritchard (2010) explained how his autistic clients had grown used to the routine of living on the streets, making it difficult for service providers to successfully rehabilitate. learning disability needs were found to be similar to needs of non-learning disability homeless, but were enduring as opposed to temporary (Van Straaten et al., 2017). Furthermore, it was suggested persons with TBI may struggle to interact appropriately with services and find it difficult to commit to program rules (Brenner et al., 2017). These examples suggested that housing difficulties resulted from impairment-associated behaviours or symptoms.
Barriers to rehabilitation were not always internalised, as some studies highlighted the difficulties homeless services had meeting the needs of individuals with CIs. These studies reported a lack of awareness of CIs, a lack of specialised programs to treat CI and co-occurring conditions, and inappropriate and low-quality housing which was unsustainable in the long term Hebblethwaite et al., 2007;Oakes & Davis, 2008;Topolovec-Vranic et al., 2013). Difficulties in accessing support were also highlighted for persons with mild or "high-functioning" ASD who were either undiagnosed, experiencing long delays during the diagnostic process, or did not meet the threshold for care or mental health services (Campbell, 2015;Pritchard, 2010). These findings reflect a social-relational understanding of CIs, in that risk factors to homelessness do not solely arise from individual pathology or symptoms but also from the way such individuals are supported.

| D ISCUSS I ON
The following discussion is split into three sections. Initially, the identification and prevalence of CI in homeless populations is considered alongside current debates on the classification and recognition of specific aetiologies. Second, experiences of homelessness are discussed, examining how risk factors to homelessness for persons with CI compare with the general population. This discussion concludes with a consideration of implications for future research, service delivery, and policy.

| Prevalence
Although the variety was displayed in the methods used and outcomes measured by included studies, CI was found to be over-repre-

| Accuracy of measures
Most studies claimed to use "standardised" or "widely accepted" The MMSE was commonly used to ascertain CI but is too general to identify focal cognitive impairments (Backer & Howard, 2007;Burra et al., 2009). Studies which solely employed the MMSE had little justification for discussion of aetiology and CI affects based on such generalised prevalence rates. Self-report mechanisms were also widely adopted, particularly for studies determining TBI. Although limited capacity should not be assumed in populations with CI, it is acknowledged that the presence of CI or circumstances surrounding homelessness such as substance abuse could affect an individual's ability to accurately self-report. None of the included papers explored this possibility satisfactorily.
In determining the presence of learning disability or ASD, access to medical records or life history is especially useful. This can be difficult when sampling older, homeless populations who may not possess or be able to retrieve such information. In lieu of this, some studies relied upon the author's clinical expertise (Pritchard, 2010) or overly inclusive screening tools (Nishio et al., 2015;Van Straaten et al., 2014). It is possible that such studies overestimated the levels of developmental disability in their samples. However, a history of reduced contact with services may also decrease the likelihood of developmental disabilities being picked up earlier on in life. Campbell (2015) noted in her study of autistic adults only 58% had an official diagnosis at the time of homelessness. This may be partially accounted for by the fact that autism screeners have not been sufficiently adapted for diagnosis in adults (Mukaetova Ladinska, Perry, Baron, & Povey, 2012), and that "higher functioning" adults with learning disabilities and autism have been shown to develop coping mechanisms which mask their disabilities. Given these considerations, it is also possible that learning disability and ASD were under-represented in some studies.

| Diagnostic boundaries and overlapping symptoms
Studies used differing definitions of CI. Some differences were internal, where adoption of screening measures necessitated a particular definition of an aetiology. This was seen most clearly in TBI studies where loss of consciousness or hospitalisation could determine the classification of a head injury but was not consistent between studies. Differences in definition were also contextual and related to temporal understandings of the impairment in question. The updating of the DSM-IV to DSM-V, for example, included diagnostic changes which may limit the generalisability of included studies.
Furthermore, it is likely that conceptualisations of cognitive impairments may differ according to culture. Although it is not within the scope of this review to debate the accuracy of diagnostic criteria, the context of individual studies must be considered when comparing international research over a 10-year period.
A further issue with studies which determined the presence of specific CIs is the recognition that aetiologies are not mutually exclusive. This is exemplified in Bymaster's , Chung, Banke, Choi, and Laird (2017) TBI study in which demographic tables describe high rates of diagnosed learning disability or special education in childhood. It is widely accepted that CIs often occur concurrently. ASD, for example, may be diagnosed alongside learning disability, Down's Syndrome, and ADHD (NAS, 2018b). Furthermore, some of the aetiologies considered in this review share similar symptoms and diagnostic criteria. The effects of TBI on executive functioning, in particular the ability to plan and organise, resonate with known ASD symptoms (Singh et al., 2016). Frontal lobe impairment, as examined by Rogoz and Burke (2016), is associated with learning disability, TBI, and dementia. Despite these considerations, studies tended to examine aetiologies exclusively.
Homelessness is associated with mental health problems and substance use disorders, issues commonly experienced alongside CIs. Although it is thus unsurprising that studies reported very high rates of mental health and substance misuse disorders, it is possible that the interplay between these factors further complicates the identification and classification of specific CIs. Backer and Howard (2007) note that symptoms of diseases can mimic organic brain disorders. They go on to explain that CI in a homeless person could indicate AIDS-related dementia, prolonged depression, side effects from medication, or self-medication with psychoactive substances. Ennis et al. (2015) also note that symptoms associated with mental health problems and substance abuse are common to post-concussive syndrome. The findings from this review echo Burra et al.'s (2009) concerns that research in this field does not do enough to measure the impact of co-occurring factors. As Backer and Howard explain, "For many homeless people, their impairments may come from more than one source, and the pattern of their cognitive problems may shift over time" (Backer & Howard, 2007, p377).

| Risks factors and perpetuators
Studies which reported risk factors to homelessness highlighted the social and economic instability experienced by persons with CI.
In particular, persons with CI were likely to become homeless after

| Risks and perpetuators "same but amplified"
Risk factors to homelessness were heightened by the presence of CI.
However, these factors were no different to those affecting the general population. UK homeless charity Crisis (2018)

| For future research
The majority of included studies were prevalence based and substantiated claims that all types of CI are overrepresented in the homeless population. Due to variety in definition and measures adopted, it is suggested that further prevalence studies will continue to produce differing percentages of CI rates in the homeless, which, nonetheless, show over-representation in this population. Shifting focus to conceptualising CIs and examining how measures are affected by variables such as being homeless or having a co-occurring mental health problems is necessary before types of CI can be more accurately recorded.
The usefulness of prevalence studies is questioned, however, as, unless they are acted upon, they do little to address the issue at hand. Studies which did not focus on prevalence tended to be concerned with the experiences of learning disability homeless, which is unsurprising given the social model of disability's permeation of this field. These studies were more effective at identifying risks factors to homelessness for these populations. In light of this, future research should focus on the experiences and perspectives of those with CI. The adoption of social or social-relational understandings of disability across multiple disciplines could help facilitate this.
It is also of note that only two studies could be considered longitudinal. While it is difficult to follow up with homeless populations, research which attempts to examine the impact of CIs on homeless populations over time would be better positioned to identify barriers to rehabilitation. Types of homeless population should also be carefully considered. The fact that most studies recruited from shelters and the majority of participants were male reduces generalisability. In line with current trends in disability research, it is lastly suggested that more is done to include those with CI in the research about them. When conducted well, inclusive research has the potential for unique contribution and actualisation of social change (Walmsley, Strnadova, & Johnson, 2017). Including cognitively impaired persons in the research process could draw attention to important ethical issues, such as gaining informed consent, which were not explored sufficiently in studies.

| For service delivery and policy
The multifactorial risk factors identified suggest that persons with CI need to be supported by services which take a whole person approach. The high rate of co-occurrence between CIs and other disorders suggests the need for integrated interventions to tackle issues concurrently (Baker & Howard, 2007). In order to do so, CIs and co-occurring conditions need to be identifiable. This requires an increased awareness amongst practitioners accompanied by adaptations to service delivery to maximise accessibility. Such adaptations might include making easy read versions of information or supplying advocates to be present at welfare assessments. Research also suggests that the treatment of the complex and interlinking problems associated with homelessness should occur alongside, as opposed to prior to, rehousing efforts (Cornes, Manthorpe, Joly, & O'Halloran, 2014 Government, 2017). Such austerity measures are inherently disadvantaging to persons with CIs. Policy makers must thus strive to challenge the barriers which restrict opportunity and quality of life. This means looking at the socioeconomic disadvantages experienced by persons with CI as well as the individual events which trigger homelessness. This could include redressing some of the welfare concerns raised here, or examining the relationship between CI and imprisonment, or the risk persons with CI are at when ageing carers pass away and no plan has been made for continued support.
Until the stigmatisations and economic realities which continue to jeopardise their life outcomes are challenged, persons with CI will continue to be over-represented in homeless populations.

| LI M ITATI O N S
While this review focussed on homeless populations, some studies have been included which sampled from prisons, mental health services, and hospitals. These studies reported that people with CI experienced increased instances of homelessness and were picked up in database searches due to their use of "homeless*" in keywords, titles, or abstracts. It is acknowledged that their inclusion reduces the comparability of studies, but to exclude them would have limited access to information on the link between homelessness and CIs, particularly concerning developmental disabilities.
Despite the international scope of this review, some findings have been generalised to the UK policy context. It is acknowledged that differing social and economic climates may affect the extent to which CI is understood and supported internationally. The impact of this is reduced somewhat given the concentration of research in Canada, the United States, and the United Kingdom, and research which suggests the principle causes of homelessness are similar across Westernised countries (Crane et al., 2005) It is up to the reader to consider the relevance of research findings across globalised settings.

ACK N OWLED G EM ENTS
The authors thank Dr Demi Patsios for comments on a previous version of this manuscript.

CO N FLI C T S O F I NTE R E S T
None known.