This systematic review of reviews provides a synthesis of international evidence regarding interventions in primary care, mental health, and drug treatment settings, for people who are homeless who use drugs. Given the large body of existing evidence available on the topic, a systematic review of reviews was considered to be the most appropriate approach. The review methodology proceeded in accordance with guidelines from the Joanna Briggs Institute [41], and was reported according to the Preferred Reporting Items of Systematic Reviews and Meta-Analyses (PRISMA) guidelines [42] (S1 PRISMA checklist). No protocol was registered with an open-access registry (e.g. PROSPERO) prior to publication.

This review was undertaken as part of a larger piece of research commissioned by the Health Research Board, Ireland, and undertaken by the same authors in 2019–2020 [43]. The larger study combined an analysis of current drug trends and provision of services in Ireland (with contextual mapping) with the systematic review. This current review provides an updated search and new data. The main outcomes of this review focused on: i) substance use; ii) housing; and iii) ‘other’ outcomes. We also extracted and synthesised, where possible, information regarding treatment entry/engagement and retention (engaging the population of interest to enter treatment/engage with a service), and successful completion of treatment (attrition rates throughout treatment duration).

The PICOS framework (population, interventions, comparators, outcomes, and study design) [44] was used to formulate the inclusion/exclusion criteria (see Table 1) and identify appropriate literature search terms.

An information specialist (MM) led the development and application of the search strategies, supported by all members of the research team. The searches were conducted across 10 electronic databases (see Table 2). All searches were run on 30 December 2019, with an updated search conducted on 3 October 2020. We also searched a range of organisational websites from December 2019 to January 2020 to ensure that any relevant reviews situated in the grey literature were identified (S1 Table). Full search strategies can be found in S1 Data. Reference details identified through the literature search were collated and managed using EndNote. Reference lists of included articles were screened for additional reviews. No date or language restrictions were included in order to minimise bias and ensure that all relevant reviews could be captured. Two reviews written in languages other than English (Canadian French and Spanish) were included, translated via Google Translate and deemed of acceptable quality by the research team for the purposes of data extraction.

One reviewer (JM) screened all titles and abstracts, alongside the full-text of articles that were considered relevant. A second reviewer (WM) independently assessed 20% of all titles and abstracts to ensure inter-rater reliability, as deemed to be good practice in rapid systematic review methodology [45]. The relevance of each article was assessed according to the criteria set out in Table 1. Any discrepancies were resolved by consensus or, if necessary, by consulting a third reviewer (HC). As a second reliability check TP, HC, WM, and JM discussed all identified relevant papers in consultation with HS. By consensus, it was agreed that only reviews where at least 40% of all included papers were relevant to substance use and homelessness were to be included, to ensure that the review maintained a firm focus on both topics. Adopting a minimum percentage in this context has also been used in other systematic reviews [46]. Reviews of both quantitative and qualitative studies were included, as were non-systematic reviews. Papers reporting pooled data or meta-analyses without an accompanying systematic review were rejected.

Reviews were not excluded based on quality appraisal scores but evidence quality was noted in accordance with the recommendations proposed by the Centre for Reviews and Dissemination [47]. Two reviewers (JM and HC) independently assessed the quality of the included systematic reviews using the JBI Critical Appraisal Checklist for Systematic Reviews and Research Syntheses [41] (S2 Data); and the quality of the non-systematic reviews using the Scale for the Assessment of Narrative Review Articles (SANRA) [48] (S3 Data). Any disagreement in scores was resolved through consensus and, if necessary, by a third reviewer (WM). Overall, the quality of the included systematic reviews was moderate, with three achieving the highest possible score of 11, and six receiving a score of six or lower. The included non-systematic reviews were apparised to be of moderate to high quality. Quality appraisal allowed for the study strengths and weaknesses to be considered but papers were not excluded based on their scores. The final scores are presented in S2 Table.

Data relating to study design and key characteristics, including populations, interventions, outcomes, and implications for policy and practice, were extracted by one reviewer (JM) into an Excel spreadsheet. Data from the reports identified through the grey literature search were extracted into the same spreadsheet by a second reviewer (WM). The data extraction table (S3 Table) was shared with other team members (HC, TP, HS) to check and ensure accuracy.

As this systematic review of reviews includes both quantitative and qualitative reviews regarding diverse types of interventions and outcomes, pooling of data was not possible, and a narrative synthesis was deemed the most suitable option for data analysis. One author (JM) summarised included studies in a narrative synthesis using textual description of each study included. Thematic summaries were developed based on the type of intervention in the included studies which enabled the synthesis and supported comparisons to be made between each study [49]. Although the search focused on controlled drugs, the team also extracted data on about alcohol, prescription drug and tobacco use, if these were included. One of the reviews previously identified for inclusion [50] only presented an abstract from a conference, with the full review not available/not published. Full data extraction was therefore not possible for this paper.

Included reviews were published between 2004 and 2020, and consisted of: four grey literature reports [39, 52–54]; 18 systematic reviews [2, 31, 46, 50, 55–68], two of which also included a meta-analysis [2, 65]; and three non-systematic reviews [69–71]. Thirteen reviews included quantitative studies only, 11 included any study type/mixed designs, including one realist synthesis [62], two systematic review of reviews [52, 60], one ‘state of the art’ review [61], and one review was a meta-ethnography of qualitative studies [31]. The number of included studies per review ranged from four [2] to 151 [53], with five reviews not reporting how many studies were included in the final synthesis [39, 54, 68, 70, 71].

Eleven of the reviews were undertaken in the United Kingdom (UK), four in the United States of America (USA), six in Canada, three in Europe (Spain, Ireland, and a Dutch/Belgian collaboration), and one was an international collaboration (Switzerland, the UK, and Canada). Nearly all reviews (22/25) were international in focus, with two focusing on the USA and one on the UK only. The majority of primary studies were undertaken in the USA.

The included reviews were diverse in terms of their primary focus and included a range of interventions (Table 3). Two of the included reviews focused on any/all health interventions, rather than on a specific intervention type, thus they included a variety of programmes ranging from harm reduction for people who use drugs to sexual health promotion programmes.

The included reviews varied in terms of their inclusion of populations of interest, with only a few focusing specifically on people who use drugs who reported being homeless [31, 61, 64, 70]. Others focused on people who were homeless and had co-occurring serious mental health problems and alcohol/drug use (COSMHAD) [62, 69] people who were homeless [59], or people who were homeless with mental health problems [56] as the primary population of interest, where substance use was secondary. Full details of the studies are presented in S3 Table.

There were notable differences in the proportion of participants who were homeless between the primary studies in the included reviews. For this reason some adopted minimum percentages for inclusion, for example Barker and Maguire [46] only included reviews when a minimum of 30% of included studies had a focus on homelessness, and Ponka et al. [63] required more than 50% of any study participants to be identified as ‘homeless’. The definition of homelessness also varied between the reviews, and between the included primary studies, which made it difficult to make direct comparisions between reviews.

The included reviews discussed a wide range of outcomes, including: those relating to substance use (reduction in drug and alcohol use (or tobacco); relapse rates; fatal and non-fatal opioid overdose rates; mean injecting frequency; and increase in treatment entry); housing; and ‘other’ outcomes, for example: well-being/quality of life (QoL); mental health; criminal justice system involvement; and societal integration. Four reviews [31, 62, 69, 71] grouped into ‘components of good practice’ focused on the elements of successful treatment rather than, or in addition to, investigating types of specific treatments. These outcomes have been synthesised below.

Four of the included reviews discussed components of good practice. Carver et al. [31] explored the views of people who used services and found that both harm reduction and abstinence-based treatments were considered effective but, in several studies, harm reduction-oriented services were preferred. However, clients also reported that abstinence-based treatments should be made available for when people are ready, highlighting that people who are homeless and experience problem substance use often desire an integrated approach to treatment. The review suggested that five components were important for effective treatment: i) the provision of a facilitative service environment; ii) compassionate and non-judgemental support; iii) adequate time in treatment; iv) choices regarding treatment; and opportunities to (re)learn how to live; and v) with these being delivered within the context of good relationships, person-centred care, and an understanding of the complexity of people’s lives. Longer treatment duration and stability was also valued, particularly by women [31].

Sun [71] reported four components of successful strategies for helping people who are homeless with COSMHAD: i) ensuring an effective transition for individuals with COSMHAD from an institution (e.g. hospital, foster care, prison, or a residential programme) into the community; ii) increasing the resources of people who are homeless with COSMHAD (e.g. helping them apply for government entitlements or supported employment); iii) linking individuals to supportive housing, including HF options, and being flexible in meeting housing needs; and iv) engaging individuals in treatment for COSMHAD. This includes incorporating modified ACT, motivational interviewing (MI), cognitive behavioural therapy, contingency management, and COSMHAD-specialised self-help groups.

Motivation for, and maintenance of, behaviour change was reported as a central factor for success in community-based services for people experiencing homelessness and COSMHAD [62]. Called ‘client choice’ in some programmes [62], this concept facilitated respect for the client’s treatment preference, even if this was not in line with what was considered the optimum treatment approach. Clients having input into staffing and programme elements resulted in a programme that was maximally tailored to their own needs, with data suggesting that both sense of mastery and perceived level of choice were mediators in the causal pathway between housing and a person’s psychiatric symptoms.

Provision of a more supportive, less intensive approach in residential programmes for people with COSMHAD was found to be a key to success [69]. Programmes rated by participants as being high in ‘support’, ‘involvement’, and ‘task orientation’, were associated with better outcomes, although it is not clear how these characteristics translated into specific programme components. In addition, specific modifications over the different stages of recovery, with a focus on slower, more concrete substance use counselling, flexibility in treatment, and greater support and guidance from staff, were also highlighted.

Collectively, these reviews suggest that flexibility is needed in treatment approaches, and that support should be tailored to the person. If possible, a combination of approaches should be used to offer choices to people who may not be ready for/do not want complete abstinence. Service providers need to be supportive and the treatment needs to be integrated, comprehensive, holistic, and person-centred, in order to increase effectiveness. Optimal duration also needs to be considered, with evidence suggesting that longer treatment leads to better outcomes, as well as being preferred by clients.

Twelve of the included reviews mentioned treatment engagement and/or retention [31, 39, 52, 54, 57–60, 67, 69–71] and six mentioned completion rates [46, 58, 59, 68–70], however, only one presented data as completion percentages [70], and one only provided completion percentages from one of the included studies [58].

There was some evidence of HF participants having higher rates of retention in a methadone treatment programme, compared with TF clients, and of increased engagement with medical treatment and mental health services. However, this was not the case for all clients, with identified barriers including boredom and isolation [57]. HF programmes were criticised in another review for a lack of engagement with services among those with very high levels of problem substance use, suggesting that TF could achieve better substance use outcomes, since they actively pursue abstinence from drugs and alcohol [54]. However, TF models have been reported to achieve relatively low rates of success, often losing between 40% and 70% of participants due to strict regimes, participants becoming ‘stuck’, or participants being evicted from services due to not meeting the abstention criteria [54]. One TF approach, called the ‘Birmingham model’, was found to lead to higher than average completion rates, with reports of 65% of participants completing a programme lasting 24 weeks [70].

Regarding case management approaches, de Vet et al. [58] noted participants not adhering to treatment and a lack of service use between groups in their included ICM studies. For example, 71% of participants assigned to shelter-based ICM services for men experiencing both substance use and homelessness did not complete the programme. On the other hand, Penzenstadler et al. [67] highlighted higher rates of treatment engagement and retention for ACT, as well as evidence of greater medication compliance, with significantly higher contact with patients in the ACT and integrated assertive community treatment (IACT) groups compared with controls. Overall, the authors concluded that ACT could be a promising approach that may be useful for promoting treatment engagement for people experiencing problem substance use.

Regarding harm reduction, findings on OST retention in treatment were mixed [60]. There does not appear to be any effect on treatment retention rates whether buprenorphine was administered under supervised or unsupervised criteria. However, methadone maintenance therapy was found to be more effective than non-pharmacological approaches in retaining heroin dependent patients in treatment, with no statistically significant difference in dropout rate between participants in slow release morphine versus methadone [60]. This suggests that the relative superiority of one pharmacological agent over another on retention outcomes remains unclear. Naltrexone implants showed significantly better treatment retention than placebo implants or oral naltrexone, and extended-release naltrexone led to significantly greater retention in treatment compared to TAU. However, successful completion of treatment rates did not differ when comparing oral naltrexone versus placebo [60].

Two studies included in Hwang et al.’s review [59] focusing on the treatment of latent tuberculosis (TB) for people who are homeless reported that, compared with TAU, a cash incentive increased attendance at an appointment for initial assessment of a positive tuberculin skin test. For people experiencing homelessness with latent TB, receiving directly observed preventive therapy, cash incentives, and non-cash vouchers at each visit were equally effective in increasing completion rates [59]. In other studies, there was some evidence that MI and motivational enhancement therapy (MET) increased treatment engagement in the short term for those experiencing homelessness and COSMHAD, and some evidence of benefits from the MI group in terms of increased attendance with aftercare [71]. Regarding engagement in treatment for people with HIV, Bates et al. [52] reported that adherence to highly active antiretroviral therapy (HAART) among people who used drugs was comparable to that among people who did not use drugs. However, people who used drugs and engaged in OST had increased adherence to HAART and better treatment outcomes, compared with people who used drugs who engaged in HAART alone.

For people with HIV, there was also evidence in support of the use of directly administered antiretroviral therapy, both alone and integrated in medication-assisted therapy, to improve treatment and outcomes related to blood-borne virus (BBV) infections. In terms of people with chronic HCV, there were no significant differences in BBV treatment dropout between people who inject drugs and those who do not who received combination treatment for HCV (ribavirin plus recombinant, or pegylated interferon-α). Lastly, for people experiencing homelessness who also injected drugs, an accelerated Hepatitis B immunisation schedule (with doses administered at 0, 7, and 21 days, and a booster at 12 months) resulted in superior completion rates, compared with traditional schedules with similar seroconversion rates [68].

Regarding peer support interventions, Barker and Maguire’s [46] review reported that their included IPS studies showed baseline data for 1,829 participants and completed data for 1,341 participants, with a loss to follow-up of 488 or 27% of participants. The authors [46] reported that one of the included studies suffered such extreme attrition from its control group that they excluded those data from the analysis, although the percentage dropout was not reported. This highlights challenges in retention in research studies for this group.

Overall, the evidence suggests that engaging and retaining people who are homeless and have substance use problems in treatment can be difficult, regardless of intervention type. There is evidence that ACT can lead to increased engagement rates for people who are homeless and use drugs, and that integrated services for people with COSMHAD lead to better engagement and retention than segregated treatments. Results regarding HF suggest that engagement can be difficult and that social isolation may be a problem for those using the service. Completion rates for the various treatment interventions are rarely reported, but tend to be low for case management interventions, especially for ICM.

Our findings point to the need for a range of harm reduction oriented services to be available to those experiencing homelessness and problem drug use, including OST, NSP, SCFs, and peer distribution of THN. ‘Full’ harm reduction should therefore be made available to ensure people can access support without the expectation of abstinence. Additional work is also required to support those with BBVs through increased public health surveillance and research [65].

It is clear that the housing situation of individuals has a notable effect on their lives and should not be dictated by their substance use. Flexible and choice-led approaches to housing like HF may be beneficial, with more research required to identify the key components of HF and other approaches [54, 70]. Setting clear and realistic goals, particularly within the context of HF, is important, and services should recognise that achievable goals will differ between individuals [54]. This review has highlighted the potential of ACT, SCM, and CTI, and more research is required to compare these and other case management models in order to identify which models or specific components are most effective. Current treatment duration is often relatively short and there is evidence that extended treatment is associated with improved outcomes and perceived as beneficial [31, 75]. Therefore, further research is also required to identify the optimal length of treatment duration. Additionally, treatment requires suitable funding to ensure that it can continue for as long as necessary, so secure funding sources are also recommended. This is particularly important, but increasingly challenging, in the context of the COVID-19 pandemic, with already vulnerable services closing or restricting access [76, 77]. More research is also required regarding optimal policies on discharge planning for statutory agencies, which impact on continuity of care [78].

It is apparent that integrated care and partnership working are important aspects of providing services to people who are homeless [25]. Integrated mental health and problem substance use services appear to be particularly important for those experiencing homelessness and COSMHAD, with secure funding also required for such services [53]. However, more research is needed regarding such services in order to establish effective components of integrated programmes of support.

The way in which services are delivered appears to be vitally important, with compassionate and non-judgemental staff. It is therefore essential that services prioritise staff training to support them to gain an understanding of people’s complex lives, and the need for person-centered approaches, empathy and compassion. The context in which services are delivered is also crucial. For example, Pleace [39] noted the need for existing networks and support for joint working, and also to recognise the potential impact of: the availability and extent of welfare systems; social care and healthcare systems; general economic conditions; housing and labour markets; and waiting lists for social rented housing, on the effectiveness of interventions. Relatedly, involving peers in the delivery of services can be beneficial and more research is required to fully understand the effect of such individuals at the intersection of homelessness and problem drug use, as well as the impact of such services on peer workers themselves.

More qualitative research is required to understand people’s experiences of the various approaches, particularly from the viewpoint of sub-groups of people who are homeless with more complex needs due to their age, gender, ethnicity or sexual orientation/identity [31]. The heterogeneity of the populations and interventions included in this review point to the need for more research at the intersection between homelessness and problem drug use specifically, to ensure that the interventions for this group of individuals does meet their specific needs. While we can make suggestions regarding effectiveness, it would be misleading or inaccurate to base policy and service recommendations on evidence that is not specific to those experiencing homelessness and problem drug use.

Steps were taken throughout this review to enhance methodological rigour, including involvement of at least two people in literature searching, screening, quality appraisal, data extraction, and analysis. Including quantitative and qualitative reviews provided a more detailed understanding regarding the effectiveness of interventions, with insight into clients’ perspectives. We also included a range of international reviews, including two non-English reviews, to provide a detailed investigation of the topic.

Several limitations should be noted. Firstly, some of the included reviews were not systematic and were limited in their reporting on included studies, thus their findings should be interpreted with caution. Secondly, some of the reviews are relatively old, so the included studies are even older. The findings of these studies may be limited in terms of their relevance today, especially if no newer reviews have been conducted (e.g. [66]). Thirdly, while most of the reviews were international in focus, most primary studies were conducted in the USA or Canada, which may limit the transferability of the findings to countries where there are clear differences in terms of homelessness, healthcare, substance use and other related systems [79].