The effectiveness of substance use interventions for homeless and vulnerably housed persons:  A systematic review of systematic reviews on supervised consumption facilities, managed alcohol programs, and pharmacological agents for opioid use disorder

Summary

Background

Substance use is disproportionately high among people who are homeless or vulnerably housed. We performed a systematic overview of reviews examining the effects of selected harm reduction and pharmacological interventions on the health and social well-being of people who use substances, with a focus on homeless populations.

Methods and findings

We searched MEDLINE, EMBASE, PsycINFO, Joanna Briggs Institute EBP, Cochrane Database of Systematic Reviews and DARE for systematic reviews from inception to August 2019. We conducted a grey literature search and hand searched reference lists. We selected reviews that synthesized evidence on supervised consumption facilities, managed alcohol programs and pharmacological interventions for opioid use disorders. We abstracted data specific to homeless or vulnerably housed populations. We assessed certainty of the evidence using the GRADE approach.

Our search identified 483 citations and 30 systematic reviews met all inclusion criteria, capturing the results from 442 primary studies. This included three reviews on supervised consumption facilities, 24 on pharmacological interventions, and three on managed alcohol programs. Supervised consumption facilities decreased lethal overdoses and other high risk behaviours without any significant harm, and improved access to care. Pharmaceutical interventions reduced mortality, morbidity, and substance use, but the impact on retention in treatment, mental illness and access to care was variable. Managed alcohol programs reduced or stabilized alcohol consumption. Few studies on managed alcohol programs reported deaths.

Conclusions

Substance use is a common chronic condition impacting homeless populations. Supervised consumption facilities reduce overdose and improve access to care, while pharmacological interventions may play a role in reducing harms and addressing other morbidity. High quality evidence on managed alcohol programs is limited.

Keywords:

Death rates – Drug therapy – Alcohol consumption – Systematic reviews – opioids – Housing – morbidity – Heroin

Introduction

Homeless and vulnerably housed populations experience disproportionately high rates of substance use [1–3]. While individuals may use substances as a way to cope with adverse living conditions, stress, and the trauma of homelessness [4–6] it is also the case that without stable housing individuals often experience barriers to accessing and following treatment recommendations for substance use disorders [7]. Childhood traumatic experiences, posttraumatic stress disorder and adult exposure to other traumatic experiences can also influence substance use [8]. Despite the substantial unmet care needs of this population, people struggling with both substance use and homelessness experience overlapping barriers to accessing care, including stigma related to care itself [9], and abstinence-based treatment first housing services [10,11]. Together, these barriers increase perceived loss of control over one’s life, contribute to mistrust of the health system, and perpetuate lower access and adherence to care and treatment [12–14].

There is substantial literature demonstrating that people who are homeless benefit from receiving tailored, patient-centred care within interprofessional primary care teams with an integrated approach to community and social services [15–17]. Specifically, harm reduction and pharmacological interventions represent important approaches to facilitate care and treatment of individuals experiencing homelessness and concomitant substance use disorders. At its core, harm reduction is a pragmatic approach that aims to reduce the adverse consequences of substance use and related socio-cultural marginalization [18–20]. Within a harm reduction approach, specific interventions include opioid overdose response with naloxone, supervised consumption facilities, and managed alcohol programs [21]. Harm reduction approaches are particularly relevant for homeless populations given the extreme health inequities and stigma impacting both substance use and homelessness [1,22,23]. The effectiveness of medications for substance use disorders including methadone, buprenorphine, slow-release morphine, diacetylmorphine, and naltrexone have been documented in the general population [24,25], but the evidence specific to homeless populations is yet to be synthesized. This review is one of a series of reviews to study the effectiveness of interventions for the care of homeless, vulnerably housed, and persons with lived homelessness experience. The objective of this review is to assess the effectiveness of specific harm reduction and pharmacological interventions among homeless or vulnerably housed individuals with concomitant substance use disorders.

Methods

This systematic review adheres to the guidance provided in the Cochrane Handbook for overviews of reviews [26] and is reported according to the Preferred Reporting Items of Systematic Reviews and Meta-Analyses (PRISMA) guidelines [27] (S1 File). Ethical approval was not required for this study.

Selection of priority interventions

We conducted a Delphi consensus process with 84 experienced healthcare practitioners and 76 persons with lived homelessness experience to prioritize person-centered and clinically meaningful community interventions, outcomes, and marginalized subgroups [28]. Among these, addiction care was highly prioritized. We then scoped literature using Google Scholar and PubMed to determine a list of interventions and terms relating to each of the Delphi priority topic categories. Priority topic working groups were formed to arrive at a consensus and inform the final selection of the interventions to be included in this review. Each working group consisted of medical practitioners, allied health professionals, and community scholars (people with lived experience of homelessness or vulnerable housing). Our working group deliberated the value of systematic reviews and evidence based guidelines on various addiction interventions, giving significant weight to the needs and opinions of persons with lived experience of homelessness. The working group agreed that focusing on safe consumption sites (of drugs and/or alcohol, i.e. SCFs and MAPs) and medications to manage opioid use disorder should be the priority.

Search strategy and selection criteria

Our initial search of published literature revealed no experimental trials conducted specifically among homeless populations [29]. We therefore followed the GRADE approach and expanded our search strategy to include grey literature and revised our inclusion criteria to incorporate general population and systematic review evidence for the interventions of interest, and published a protocol reflecting these changes on the Cochrane Equity Methods website [30]. We made this decision fully aware that most studies among “general populations” had a large representation of homeless populations in their samples, and we did not identify any substantial reason to believe that the mechanisms of action of our interventions of interest would differ between homeless populations who use substances and the general population of people who use substances.

We included systematic reviews of quantitative or qualitative studies that focused on at least one of the interventions specified in our protocol among homeless or general populations. Any type of comparator was considered eligible (such as usual services, alternative intervention or no intervention). We included all outcomes of interest including: mortality, morbidity, substance use, mental health, access to care, and retention in treatment. We excluded overviews of reviews and narrative literature reviews (See Table 1 for full inclusion and exclusion criteria).

We developed a systematic search using relevant keywords and MeSH terms for relevant published systematic reviews. Keywords included terms such as “opioid-related disorders”, “supervised consumption”, “supervised injection”, “managed alcohol”, “methadone” and “harm reduction” (See S2 File). We searched MEDLINE, EMBASE, PsycINFO, Joanna Briggs Institute EBP, Cochrane Database of Systematic Reviews and the Database of Abstracts of Reviews of Effects (DARE) for systematic reviews from database inception to August 2019. There were no language restrictions. The reference lists for all articles selected for full-text review were manually searched for relevant citations. These were cross-referenced against our original search results and any additional potentially relevant citations were screened. We conducted a grey literature search (see S2 File) and consulted experts in the field for any additional studies.

Results were uploaded to Rayyan reference manager software to facilitate the study selection process [37]. Two review authors independently assessed each study for inclusion by title/abstract, followed by full-text; disagreements were resolved through discussion or a third reviewer.

Data analysis

We developed a standardized data extraction template which included study design, search parameters, inclusion criteria, number and design of included studies, results and conclusions. We extracted study details specific to homelessness or housing instability status of participants when available. We extracted data in duplicate and resolved disagreements through discussion. We assessed the methodological quality of reviews in duplicate using AMSTAR 2 (See S3 File) [38].

Following guidance from the Cochrane Handbook [39], we assessed primary study overlap within the included systematic reviews using a citation matrix (See S4 File). We synthesized outcome data narratively and organized the presentation of evidence by intervention. A narrative synthesis is defined as a “synthesis of findings from multiple studies that relies primarily on the use of words and text to summarize and explain the findings of the synthesis. Whilst it can involve the manipulation of statistical data, the defining characteristic is that it adopts a textual approach to the process of synthesis to ‘tell the story’ of the findings from the included studies” [40]. Where appropriate, we included existing meta-analyses with pooled effects, and report estimates of effect as reported in the source review. Numerical estimates of association are only reported in instances when these were provided in the source review. We assessed the certainty of the available numerical evidence for critical patient-important outcomes (mortality, morbidity, substance use) using GRADE methodology, considering risk of bias of the primary studies, indirectness, imprecision and inconsistency [26] (See Table 2: GRADE definitions). Where data was unavailable, a GRADE assessment was not done.

Role of the funding source

The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all of the data in the study and had final responsibility for the decision to submit for publication.

Results

Our systematic search identified 486 citations. We identified an additional 11 articles through grey literature and reference searching. After removal of duplicates, we screened 355 unique articles on title and abstract, excluded 283 articles and screened 72 articles by full text. Of these, we excluded 42 articles (See S5 File: Table of excluded studies) and included 30 systematic reviews in our final synthesis (See Fig 1: PRISMA).

Of the 30 included systematic reviews, 24 reviews which included 352 unique primary studies reported on pharmacological interventions for opioid use disorder [41–64]. Two systematic reviews which included 90 unique observational studies and one qualitative meta-synthesis reported on supervised consumption facilities (SCFs) [65–67]. One Cochrane review with no included studies [68] and two grey literature reviews reported on 51 studies of managed alcohol programs (MAPs) [69,70].

Fifty-eight percent (14/24) of systematic reviews on pharmacological interventions included studies which reported housing instability or homelessness [41,42,44,46,50,51,53,55–60,64]. All systematic reviews on SCFs included observational studies reporting the rate of homelessness ranging from 10% to as high as 75% among SCF clients, the majority of studies were from Sydney, Australia and Vancouver, Canada [65–67]. Evidence identified on MAPs was specific to homeless populations [69,70]. See Table 3 for additional study characteristics.

Pharmacological interventions for opioid use disorder

Reviews on pharmacological interventions reported on the use of methadone, buprenorphine, diacetylmorphine (heroin), levo-α-acetylmethadol (LAAM), slow release oral morphine and hydromorphone for treatment of opioid use disorder. Certainty of evidence from these reviews ranged from very low to moderate (see Table 4). One review [57], including one observational study among homeless populations, reported pooled all-cause mortality rates of 36.1 and 11.3 per 1000 person-years for participants out of and in methadone maintenance therapy respectively (rate ratio 3.20, 95%CI 2.65 to 3.86), and mortality rates of 9.5 per 1000 person years for those not receiving buprenorphine maintenance therapy compared to 4.3 per 1000 person-years among those receiving the therapy (rate ratio 2.20, 95%CI 1.34 to 3.61; GRADE certainty of evidence: very low). Higher rates of mortality were reported among heroin-dependent persons using LAAM (relative risk 2.28 (95%CI 0.59–8.9; GRADE certainty of evidence: low)) [42] and people chronically injecting opioids receiving diacetylmorphine (11 vs 3, p<0.05, no further data reported) [41] compared to methadone, however there were not enough deaths in these samples to draw any meaningful conclusions.

Overdose-specific mortality rates were similarly affected, with pooled overdose mortality rates of 12.7 and 2.6 per 1000 person years for participants out of and in methadone maintenance therapy (GRADE certainty of evidence: very low), and rates of 4.6 and 1.4 per 1000 person years out of and in buprenorphine maintenance therapy (GRADE certainty of evidence: low) [57] (See Figs 2 and 3). Conversely, in an older review among heroin-dependent persons including fewer studies, compared to non-pharmacological approaches, methadone maintenance therapy had no statistically significant effect on mortality (relative risk = 0.48; 95%CI: 0.10–2.39; GRADE certainty of evidence: moderate) [51].

With respect to morbidity, pharmacological interventions for opioid use disorder reduced the risk of Hepatitis C (HCV) acquisition (risk ratio 0.50, 95%CI 0.40–0.63; GRADE certainty of evidence: low) [53] and HIV infection [44]. Because of the high risk of bias and variability in several aspects of studies examining the risk of HIV infection, combined totals were not calculated [44]. Methadone was associated with more adverse events compared to buprenorphine, including sedation [50,60], but not compared to slow release oral morphine (81% SROM vs 79% methadone, p = 0.62 [no further data reported])[63]. Buprenorphine patients had a significantly lower prevalence of fatigue (15.5% vs 25.1%; risk ratio 0.62; 95% CI 0.41–0.95; GRADE certainty of evidence [as reported in [64]: moderate) but no difference in insomnia (risk ratio1.12 95%CI 0.78–1.62) compared to methadone patients [64].

Therapies for opioid use disorder among general populations were associated with statistically significant reductions in non-prescribed opioid use [41,44,45,50,51,58,60], injection drug use and sharing of injection equipment [44,46]. Studies conducted among homeless populations demonstrated an association between prescribed use of diacetylmorphine or methadone and reductions in non-prescribed drug use or more general illegal activity [71]. Several studies also reported reduced risk behaviours such as having multiple sexual partners, unprotected sex or participation in drug-related crime [46]. Buprenorphine and methadone further showed improvements in social functioning, physical health, and psychological morbidity [58].

Findings on access to care and retention in treatment were mixed. A review on buprenorphine treatment found that it was associated with expanded access to treatment for patients who may not enroll in methadone clinics, and facilitated earlier access to treatment for patients who have more recently initiated opioid use [58]; however there did not appear to be any effect on retention whether buprenorphine was dosed under supervised or unsupervised settings (risk ratio 0.99, 95% CI 0.88–1.12) [55]. A systematic review highlighted that methadone maintenance therapy was more effective than non-pharmacological approaches in retaining heroin-dependent patients in treatment (risk ratio 4.44, 95% CI:3.26–6.04) [51]. Of note, methadone maintenance therapy use was negatively associated with homelessness in a separate observational study (odds ratio 0.29, 95% CI 0.17–0.49) [72]. One review reported that difference in dropouts was not statistically significant between participants in slow release morphine vs methadone (risk ratio 0.98; 95% CI: 0.94–1.02, p = 0.34) [63]. The relative superiority of one pharmacological agent over another on retention outcomes remains unclear.

Finally, three reviews reported on naltrexone treatment. No findings were specific to homeless populations. Naltrexone was associated with few opioid overdose events [41,48]. Naltrexone implants were significantly better than placebo (risk ratio 0.57; 95% CI 0.48–0.68) or oral naltrexone (risk ratio 0.57; 95% CI 0.47–0.70) at reducing opioid use [48]. Extended-release naltrexone led to significantly greater retention in treatment, significantly greater proportions of opioid-negative urine drug tests, and significantly greater reductions in illicit opioid use compared to treatment as usual [41]. Results of a meta-analysis were not statistically significant for the successful completion of treatment (odds ratio 0.78; 95% CI 0.34–1.75) when comparing oral naltrexone versus placebo [47]. Naltrexone implants showed significantly better treatment retention than placebo implants (risk ratio 3.20; 95% CI 2.17–4.72) and oral naltrexone (risk ratio 3.38; 95% CI 2.08–5.49) [48].

Pharmacologic agents for reversal of opioid overdose

We identified one review reporting the effectiveness of naloxone among populations experiencing opioid use disorder, [41] which identified four trials that investigated naloxone in the treatment of suspected or established opioid overdose. Three trials involved prehospital treatment of suspected opioid overdose [73–75]. One trial [76] compared intravenous naloxone to physostigmine for in-hospital treatment of heroin overdose, finding no significant difference in efficacy. Two trials [74,75] compared intranasal to intramuscular naloxone, finding no significant difference in efficacy. However, Kerr et al. [75] reported that supplementary naloxone was administered to fewer patients who received intramuscular naloxone (IN 18.1%; IM 4.5%; difference: 13.6%, 95% CI 4.2–22.9). Neither of these trials reported on housing stability of the patients, however, Kelly et al. [74] reported that more patients were treated in public places such as a park or street than in private residences. One trial [73] compared intravenous naloxone to nalmefene (1 or 2mg), finding no significant difference in efficacy. Adverse events occurred in 30.9% (2mg nalmefene), 15.9% (1 mg nalmefene), and 15.5% (naloxone) of patients; group differences were not statistically significant (p>0.08; [no further data reported]). No reviews on naloxone reported on substance use, mental health, or access to care outcomes.

Supervised consumption facilities

Our search identified three reviews providing very low to low certainty evidence (see Table 5) that examined supervised consumption facilities [65–67]. Vancouver’s INSITE, the first legal supervised drug injection site in North America, and the Sydney Medically Supervised Injection Centre (MSIC) are two SCFs which have been extensively evaluated in these reviews. Vancouver’s INSITE was established in 2003, and it is estimated that 20% of its clients are absolutely homeless, with many more living in single-residence rooms [77]. In Sydney, the MSIC facility has been operating since 2001 with an estimated 11% of its clients in unstable accommodations [78]. In both sites, no deaths from opioid overdose were reported [66]. As well, there was a significant decrease in the number of deaths in the immediate vicinity of the Sydney SCF, from an average of 4 to 1 deaths per month (p<0.001) [65].

The establishment of INSITE in Vancouver led to a 35% decrease in the number of fatal opioid overdoses within 500 meters of the SCF (from 253.8 to 165.1 deaths per 100,000 person-years; p = 0.048), compared to 9% in the rest of the city [65]. There were 336 reported opioid overdose reversals in 90 different individuals within the Vancouver SCF over a four-year (2004–2008) period [66]. Simulation studies estimate that, over the same four-year period, up to 1004 opioid overdoses may have been avoided in Vancouver, including 453 life-threatening opioid overdoses [66], suggesting that between 2 and 12 cases of lethal opioid overdoses might have been avoided each year (GRADE certainty of evidence: very low) [65]. Similarly, there were 409 opioid overdose reversals associated with the establishment of MSIC in Sydney over 17 months, and similar protective effects of SCFs have been reported in Germany [65]. There were no substantial changes in rates of relapse into injection drug use, ceasing injection, ceasing binge drug use, or participation in methadone maintenance therapy after the Vancouver SCF opened [65]. The rate of recent initiation into injection drug use among SCF clients was markedly lower than the estimated background community-level rate of injection initiation [65]. Frequent SCF use in Sydney was positively associated with experiencing a non-fatal opioid overdose within the SCF (AOR = 6.1; 95% CI 4.3–8.6) [65]. There was also a significant decrease in opioid overdose emergency department presentations, from an average of 11 to 7 per month (35% reduction, p<0.001) [65].

Observational studies conducted in Vancouver and Sydney showed that regular use of SCFs was associated with decreased syringe sharing (aOR = 0.30, 95%CI = 0.11–0.82), syringe reuse (aOR = 2.04, 95%CI = 1.38–3.01), and public-space injection (aOR = 2.79, 95%CI = 1.93–3.87) [66]. The SCF in Vancouver was associated with reductions in the number of people injecting in public [65]. It has been noted that factors associated with public injection in Vancouver include homelessness (aOR = 3.1, 95%CI = 1.46–6.58) [66]. One review suggested that SCFs prevented incident HIV and HCV infections (GRADE certainty of evidence: very low) [65], but the exact number of cases averted depends on varying mathematical model assumptions.

Reviews noted that SCFs may have improved access to care for vulnerable populations; their advantages included competent staff and non-judgmental transfer to a hospital if necessary, education on safer injection, and transfer to other medical and social structures [66]. SCFs mediated access to ancillary services (e.g. food and shelter) and fostered access to broader health supports [67]. SCF attendance was associated with an increase in referrals to an addiction treatment center and initiation of methadone maintenance therapy (aHR = 1.57, 95%CI = 1.02–2.40) [66]. A study in Sydney found that frequent SCF use was positively associated with referral to addiction treatment, but not addiction treatment uptake [65]. In a Danish study where 40% of participants were unstably housed, being advised to seek treatment for a medical condition by SCF staff was associated with an increased likelihood of receiving treatment (51.3 vs. 25.7%, p = 0.003 [no further data reported]) [65]. On average, 21% of individuals attending the SCF in Vancouver wanted, but were unable to access, substance use treatment. The main obstacle in access was the waiting list, and homelessness was significantly associated with lack of access (OR = 1.47, 95%CI = 1.09–1.98) [66]. No systematic reviews reported on the effects of SCFs on mental health outcomes.

Managed alcohol programs

We identified one Cochrane systematic review [68] and two grey literature systematic reviews [69,70] examining the effects of managed alcohol programs (MAPs) for homeless populations. There was a lack of high-quality evidence in the peer reviewed literature for this intervention [68]. Existing evidence consisted largely of uncontrolled evaluations of small scale pilot programs. Few studies reported on deaths among MAP clients [70]. In a case study of a Canadian MAP, “out of few dozen men who cycled through the program, eight clients of the MAP had passed away since its opening in 2006” [70]. Additionally, a MAP within a shelter-based palliative care program had 28 consecutive patients die between 2001 and 2003 [70]. Impacts of MAPs on hepatic function are mixed, with some studies reporting improvement in hepatic laboratory markers over time, and others showing increases in alcohol-related hepatic damage [70], however this may have occurred regardless of entry into a MAP. This evidence suggested that MAPs result in stabilized alcohol consumption and can facilitate engagement with medical and social services [69]. MAP clients also reduced their consumption of non-beverage alcohol products, liquids containing a form of comparatively harmful forms of alcohol that were not intended for human consumption (e.g., mouthwash, hand sanitizer, etc.) and were consumed instead of beverage alcohol for the purposes of intoxication [70]. MAP clients experienced significantly fewer social, health, safety, and legal harms related to alcohol consumption [70]. Individuals participating in MAPs demonstrated fewer hospital admissions and a 93% reduction in emergency service contacts [69]. MAPs also promoted improved or stabilized mental health [69] and medication adherence [70]. No systematic reviews reported on retention in treatment outcomes.

Discussion

We conducted a systematic review of systematic reviews to understand the effectiveness of substance use interventions of relevance to homeless and vulnerably housed populations with concomitant substance use disorder. To our knowledge, this is the first systematic review on harm reduction approaches and pharmacological interventions for this population. While our review did not focus exclusively on studies among populations experiencing homelessness, people experiencing homelessness and people who use substances both face considerable social marginalization, multiple physical and/or mental health concerns, and high rates of premature morbidity [79]. Our review found that these interventions reduce harms associated with substance use and mitigate morbidity and mortality [41,44,45,50,51,53,57,58,60,65,66]. Since homelessness and unstable housing are factors that continue to be associated with public injection, infection, and reduced access to care [9,80,81], this further highlights the importance of offering low threshold services in close proximity to homeless populations and integrating housing supportive services for individuals who are homeless and have concomitant substance use disorders [82].

Several studies on pharmacological interventions demonstrated improved outcomes for mortality, HCV and HIV acquisition, psychological morbidity and non-prescribed opioid use [44,53]. Results suggest that buprenorphine and methadone are the most effective pharmaceutical agents to address mortality and morbidity among people who use substances [50,57,60]. However, these options may not be feasible for many people experiencing homelessness as they may face additional barriers when accessing opioid agonist therapy (i.e. accessing pharmacy daily, attending regular appointments) compared to those who use substances and are stably housed [83–86]. Much of the included evidence in this review took place prior to the current opioid epidemic, thereby missing important contextual factors related to treatment need and acceptability for homeless populations. Emerging evidence suggests that injectable heroin and injectable hydromorphone are both acceptable and associated with improved outcomes for people who are treatment-refractory [87]. Furthermore, the prevalence of stigma among health care providers towards people who use substances is well-documented and can result in barriers to healthcare access and poor health outcomes [88]. Individual patient characteristics and preferences should be considered when choosing a first-line opioid agonist treatment, including risk-factors for treatment drop-out such as unstable housing, lack of social support, and concurrent mental illness [89]. Other low-threshold pharmaceutical options are being adopted, including slow release oral morphine, diacetylmorphine, and injectable hydromorphone, which warrant ongoing pragmatic study, particularly with homeless and vulnerably housed populations. Traditionally, opioid use disorder pharmacotherapy was prescribed by specialists, however as with other chronic conditions, opioid use disorders can be managed in a primary care setting where primary care teams and clinician support and training can facilitate improved outcomes [89,90].

Non-prescription use of opioids and the resulting increase in overdose-related deaths has become a global issue over the past decade. Recent literature continues to highlight the importance of naloxone in averting overdose-related deaths [91,92]. Though we included naloxone as a pharmacological agent in our review, we did not search specifically for take-home naloxone programs as an intervention, and there are existing reviews documenting the value of these programs [93,94]. There is yet to be a systematic review considering their use among homeless and vulnerably housed populations, despite their disproportionate representation in the overdose epidemic [95,96]. However, recent studies indicate that take-home naloxone education programs are feasible within existing settings offering services to homeless and vulnerably housed populations [97] and the wider community [98,99].

Homeless and vulnerably housed populations are at risk of public injection and other high-risk consumption behaviours [100]. Our findings highlight that SCFs had positive impacts on fatal overdose prevention, public injecting and other high-risk consumption, which constitute an important mechanism for drug use-associated morbidity and mortality [101]. Additionally, SCFs can act as a bridge to other health and social services for those who seek them, although clients are often faced with long wait-lists to obtain services and treatment [65]. While public perception persists that SCFs may increase non-prescribed substance use and local crime and discourage individuals from seeking addiction treatment, such concerns are not supported by existing evidence [65,66]. The effective implementation of SCFs is unfortunately influenced by mixed public perceptions [102–106]. Finally, most formal evaluations, including those captured by this review, are of stand-alone SCF sites. In light of the current overdose epidemic, current initiatives may include integrated service models, mobile sites and virtual sites [107,108]. Future research examining the experiences of homeless or vulnerably housed persons engaging with these sites, with considerations of their acceptability and barriers/facilitators in accessing them, is needed to guide best practice.

Though a number of publications document their feasibility and acceptability, we could not identify any systematic reviews of high-quality comparative evidence (e.g. randomized trials) on MAPs. However, a national Canadian evaluation of MAPs is currently underway [109]. Previously, studies examining MAPs were largely conducted as case studies and small sample pilot projects [68], however, these studies demonstrated that such programs facilitate stabilization by providing a safe space that decreases the risks associated with substance use, promotes social and cultural reconnections, and provides an environment that fosters self-change [110–112]. Recent systematic reviews from grey literature conclude that few studies report on deaths of MAP clients, however many clients of MAPs have advanced, life-limiting illness prior to entering the program, and their health may deteriorate despite the best efforts of the health care providers [70], therefore rendering studies on mortality difficult to conduct and interpret. Moreover, reviews are often unable to provide high quality evidence on MAPs because of ethical and logistical barriers which need to be overcome in studying such vulnerable heterogeneous populations with complex needs [69].

Strengths and limitations

We conducted a structured systematic search, used high quality methods to synthesize systematic reviews, included results in meta-analyses where appropriate, and used GRADE methods to assess the certainty of the effects. One of the strengths of this review is its ability to capture a wide body of evidence regarding our selected interventions. Several limitations should also be acknowledged. Due to the lack of existing trials studying these interventions among homeless and vulnerably housed populations (indeed, randomized trials of harm reduction interventions have particular ethical concerns), we expanded our search to the general population. The GRADE certainty of the evidence for health system and public health interventions is often low or very low [113], and this phenomena is replicated in our review. While we extracted homelessness and housing stability patient characteristics when possible, population-specific factors must be considered when interpreting our findings. Additionally, this review only included reviews that followed a systematic methodology, and excluded primary studies. A weakness inherent to secondary analysis is that our review may be limited by the decisions and potential biases of the authors of the included studies on the reporting of results. This may have impacted the variety of studies found and excluded the most recent primary studies in the field. We also recognize that grey literature searches have limitations, including the possibility that existing evidence is missed. While many homeless individuals were included in certain primary studies of pharmacological programs, homeless-specific results may have been omitted from the original reviews as a result of selective reporting. In addition, many of our included reviews are of low quality, since we did not exclude any reviews on the basis of methodological quality. The low scoring quality of these reviews indicates that they have one or more critical flaws in their methodology, as defined by AMSTAR 2 [38], and should not be relied on to provide an accurate and comprehensive summary of the available studies. For the majority of the included reviews these critical weaknesses were due to a lack of explicit reference to a review protocol and/or omission of a list of excluded studies. Adherence to a well-developed protocol reduces the risk of bias in the review, while excluded studies should be accounted for fully by review authors, otherwise there is a risk that they remain invisible and the impact of their exclusion from the review is unknown [38]. This lack of transparency means that these systematic review findings should be considered with reservations. Therefore, future reviews should aim to explicitly mention and adhere to a protocol and report all studies under consideration to improve their overall reporting quality. Finally, it is important to note limitations in the generalizability of our findings from the range of social and geographical settings reviewed. For example, data from Canada, Australia, the USA and UK may reflect weather extremes, universal healthcare or availability of low-threshold services not relevant to other countries, and vice versa.

Implications for future research

While this review gathers a robust body of evidence of the effects of harm reduction and pharmacologic interventions on the health and social outcomes of people who use drugs who are homeless or vulnerably housed, more high quality research is required on MAPs, the scaling up of SCFs in a variety of settings such as supportive housing environments or as part of healthcare centres, as well as the types and ideal prescribing strategies of pharmacological interventions. Longer follow-up periods are also required to determine which interventions maintain their improvements over time. In order to inform decisions on resource allocation by public health and governmental bodies, future studies should continue to assess the cost-effectiveness of these interventions by evaluating and comparing the cost of their implementation against their benefits at the individual and societal levels [114–118]. In addition, more research looking at the effects of these interventions on specific subgroups of homeless populations that were not captured by this review, such as youth, women and Indigenous populations. As the value of peer-support and lived experience in intervention design and implementation is increasingly recognized [119], their role should be evaluated in future harm reduction interventions. Furthermore, additional research among homeless populations is needed to contextualize our findings and address the unanswered questions regarding the prescribing strategies of pharmacological interventions and implementation of harm reduction programs for homeless and vulnerably housed populations.

Conclusions

Homeless and vulnerably housed populations experience high rates of substance use. Our review found that harm reduction strategies and pharmacologic interventions for opioid use disorder limit or prevent infectious diseases, other chronic conditions and opioid overdose while improving retention and access to care. Policy and civil society debates on the role of harm reduction may constitute barriers to implementation. The disproportionate levels of harm experienced by the most at-risk and highly stigmatized populations, including people who are homeless, heralds a call to action. Policy makers need to address growing calls by a number of organizations and people who use drugs for the decriminalization of drugs, as this reduces the harm caused by prohibition and may encourage more people to access health services while reducing the burden on the criminal justice system, and improve morbidity, mortality and substance use outcomes [120]. Finally, allocating resources to structural conditions that homeless people who use substances face, such as housing, trauma and poverty, has the potential to further reduce harms associated with the risk environment [121–123]. Primary care practitioners may benefit from incorporating harm reduction interventions into their practice to reduce morbidity in homeless populations. However, stigma, lack of available services and not knowing what services are available to homeless populations are often barriers to providing care, and instill bias in the care that is given [124]. Concurrently integrating harm reduction programs in primary healthcare services, building the capacity of staff working in these settings, and changing the culture within organizations has the potential to lead to sustained improvements of health and social outcomes [125].

Supporting information

S1 File [pdf]
PRISMA checklist.

S2 File [pdf]
Search strategy.

S3 File [pdf]
AMSTAR II assessments.

S4 File [pdf]
Citation matrix.

S5 File [pdf]
List of excluded studies.