According to the Annual Homeless Assessment Report (AHAR) put together by the US Department of Housing and Urban Development, in 2024 America saw the largest number of people experiencing homelessness since record-keeping began ([17]). This places an enormous humanitarian and economic burden on communities in the United States, which can be ameliorated by effective programs which support rehousing ([4]). However, many homeless Americans experience health challenges beyond the problem of shelter ([11]; [25]). From psychiatric illness to infectious diseases, there is a wide array of comorbid conditions which often present additional challenges for people already struggling to reintegrate socially ([11]; [15]; [25]).

Current literature supports prioritizing the housing problem, specifically through Housing First programs, but medical comorbidities can complicate a person’s ability to engage with these programs ([33]). While nothing precludes simultaneous treatment of comorbid conditions, obtaining adherence in this po-pulation has historically been very difficult for multiple reasons ([7]; [15]; [45]). Therefore, addressing patient adherence represents a key challenge on the path towards improving homelessness overall.

The most recent AHAR shows that 16.4% of homeless Americans are currently in rural or largely rural areas, but it remains unknown whether this population is represented in current trials ([17]).

The purpose of this review is to identify recent efforts to improve treatment adherence among people experiencing homelessness in America. A scoping review method is well suited for this task because, at present, the literature concerning treatment of homeless patients is not well unified. This scoping review will lay the groundwork for a more systematic approach to this topic in the future.

We conducted a scoping review without the assistance of research software. The review design was based on the Arksey and O’Malley framework described in JBI, and PRISMA-ScR guidelines. Literature searches were performed via the databases PubMed, EMBASE, and Web of Science, as well as ClinicalTrials.gov, the current US government registry for clinical trials.

The search terms used were (homeless AND adherence AND trial). Variations and synonyms of these search terms were attempted, such as with “unhoused”, “unsheltered”, “unstably housed”, etc., but this did not add novel results, as all studies using varied terminology also included the word “homeless”. Variations and synonyms of “adherence” were not considered. The most common, “compliance”, represents an approach to treatment which is opposed to the idea of seeking adherence; this difference is more than semantic, as it reflects the view of the investigators towards their participants. Variations on “trial” were not considered, as this is a central concept to the present review. However, the word “trial” was omitted from the registry search, given that all results were trials by nature of the registry. Search results were date limited to the range of the inclusion criteria.

Search results were screened by title and abstract, then by full text, based on the following inclusion criteria:

1) Original research articles from Jan 1, 2015 to Mar 24, 2025;

2) Published in the English language;

3) Published in peer-reviewed journals;

4) Trialed on American participants;

5) With either all participants described as homeless, or a clearly distinguished subset corresponding to participants described as homeless or any variation on this term;

6) With an intervention targeting treatment adherence.

Figure 1. PRISMA flow diagram for study selection (https://doi.org/10.6084/m9.figshare.33289446).

Records not meeting all these criteria were excluded. This review is directed towards present action in the United States, for which we believe recent (≤ 10 years old) evidence from that region is of particular relevance. A preset form was used to extract data from all studies identified using these criteria. The form, with data as extracted, can be found in Supplemental Figure 1. The form was used to catalog all studies which met inclusion criteria, along with their study designs, target problem, intervention, sample size, follow-up timing, outcomes, and limitations, if any. Further, each study was analyzed to discover whether its participants were in urban, rural, or both types of environments, if known. With respect to outcomes, only statistically significant findings relevant to treatment adherence were recorded. The protocol, as performed, is described completely here; it was not registered with any online database.

This review identified 20 total publications corresponding to 16 trials. The numeric discrepancy stems from three trials reviewed here which were associated with multiple identified publications. A list organizing these details and presenting characteristics relevant to the present review can be found in Table 1. A PRISMA flow diagram illustrating the selection of these publications can be found in Figure 1. This figure was produced using PRISMA2020 software ([10]).

Table 1. Summary of trials assessing adherence improvement in Americans experiencing homelessness.

Abbreviations: (LAI) Long Acting Injectable, (TRQ) Tablets Routine Questionnaire adherence metric, (VA) Veterans Administration, (MI) Motivational Interviewing, (NRT) Nicotine Replacement Therapy, (MAT) Medication Assisted Treatment, (PrEP) Pre-exposure Prophylaxis, (SoC) Standard of Care, (RN/CHW) Registered Nurse with Community Health Workers.

Regarding study participant demographics and geography, 81.25% (13/16) of the studies considered here addressed urban populations explicitly. The remaining three studies did not specify participant location, but each was administered out of major urban centers ([5]; [29]; [32]). None of the studies identified by this review addressed treatment adherence in rural or semi-rural contexts. Most of the studies reviewed here used the US federal definition of homelessness from the Housing and Urban Development Department: “an individual or family living in a supervised publicly or privately operated shelter designated to provide temporary living arrangements” or “an individual or family with a primary nighttime residence that is a public or private place not designed for or ordinarily used as a regular sleeping accommodation for human beings, including a car, park, abandoned building, bus or train station, airport, or camping ground.” Assessing studies for the inclusion of people experiencing homelessness was complicated by participants experiencing frequent moves and changes in housing status. Generally, all studies reviewed here included anyone with unstable housing as “homeless” and made reasonable efforts to follow-up with all participants.

In addressing treatment adherence, the studies under review clustered around interventions for a small number of problems that correspond to key public health challenges. Broadly, these problems were:

Addiction pathology (including alcohol, tobacco, and opiate use disorders) ([23]; [29]; [32]; [41])Type 2 Diabetes Mellitus ([43])Hepatitis C Virus ([21]; [27]; [30])HIV treatment and pre-exposure prophylaxis ([37]; [40])Housing stability ([5]; [8])Psychiatric illness ([3]; [9]; [36])Primary care ([20])Tuberculosis ([31])

In their approach towards these problems, the interventions under review fell into several broad categories worth treating separately because it is here that the trends between studies become clearest. Given the heterogeneity of trial designs, outcomes, and sample groups, quantitative subgroup analysis of the reviewed studies is not feasible. Instead, we have presented here a narrative summary of the categories observed in the currently available studies.

The field of interventions targeting treatment adherence among people experiencing homelessness has been highly active over the past ten years. We can see that the interventions grouped well into a small number of categories. In particular, psychosocial and direct action interventions, together, account for 11 of 16 recent trials. The data extracted in the course of this review provide enough information to make recommendations for future studies targeting treatment adherence among homeless adults in America.

Psychosocial interventions represent the largest group of trials in this review, at 6 of 16, but only two demonstrated statistical significance. This was likely due to a variety of causes. Peer specialist support ([5]), personalized coaching ([43]), and social skills training ([8]) trials were limited by availability of personnel, participants, and space, respectively. Beyond improving those material factors, the studies which did find significant results can perhaps point to further ways to optimize psychosocial strategies. In both Towe et al. ([40]) and the series of CAE-L trials ([3]; [36]), we can see that psychosocial approaches were paired with an additional treatment with known efficacy; a Housing First program in the former, and a Long-Acting Injectables program in the latter. This seems to indicate that psychosocial approaches work best as an adjunct to a larger program. Further, these interventions were proactive. While the studies which failed to demonstrate significance relied on patients to come to them, Towe et al. and the CAE-L trials provided support measures which reached out to patients and helped them keep appointments ([3]; [36]; [40]). This level of proactive effort may be resource-intensive but appears to be necessary based on these trials. Consider also the trial run by Ojo-Fati et al., which did pair a psychosocial approach, motivational interviewing, with another efficacious approach, nicotine replacement therapy, yet still failed to find a significant improvement in the adherence variable of interest ([32]). It is worth considering that what differentiates this from more successful trials is the degree to which their psychosocial intervention was less proactive towards its patients.

The value of proactive approaches is driven home further by the class of interventions referred to here as direct action. In these trials, researchers met patients directly on the streets ([31]), or created pop-up clinics in homeless shelters ([9]; [21]) or heavily trafficked areas ([27]). All trials in this category reported improvements in their respective adherence rates ([9]; [21]; [27]; [31]). An interesting question on this point, however, is raised by Khalili et al. in their paper. They were able to obtain results which compared well to similar studies, but could not reach remission rates like that seen in one major trial which obtained a 97% remission rate among homeless people with HCV ([2]; [21]). Khalili et al. raise the point that the very high remission rate obtained in this prior study was the result of a program which only included participants who made it through multiple organizational hurdles to pre-emptively screen for adherence in their participants ([21]). Is it reasonable to compare such results to a pop-up clinic trying to engage historically nonadherent patients? The answer is clearly no, which provides a stronger basis for the importance of separating traditional clinic-based studies from what is described in this review as direct action. Older studies ([26]) suggest that treatment engagement and adherence rates outside of curated study samples were previously very low, on the order of <1% per year. This paints the numbers achieved by Khalili et al. and others in a very different light.

The trials which utilized contingency management of alcohol use disorder and emergency management of opiate use disorder complete the list of things that appear to work well. Langabeer et al.’s trial of emergency buprenorphine induction for OUD produced results in line with similar trials in Canada ([22]; [23]). These successes highlight the importance of finding opportunities to engage marginalized patients as proactively as possible. The success Novak et al. found with their contingency management program highlights the diversity of potentially effective engagement strategies ([29]).

On the other hand, we can see clear indications of what kinds of treatment strategies are not likely to improve adherence. The failure of text message systems in Kershaw et al. ([20]), and of video DOT in Tsui et al. ([41]), seems to indicate that a remote approach is less effective at reaching this patient demographic. Moreover, while some of the unfinished trials identified in this review are recent enough that data would not be expected yet, a large proportion of unfinished trials related to smartphone-based mobile health interventions are all at least three years old. We can see the oldest two unfinished mHealth trials are now eight years old, with no published data. Both of the completed trials provided study phones and screened for participant literacy and technical ability beforehand ([20]; [41]). A high-quality review from Heaslip et al. highlights this issue of comfort with technology as a possible confound ([12]). They cite data indicating that people experiencing homelessness are more likely than age-matched peers to have a cell phone, though they ultimately conclude that no mHealth studies in this population have yet yielded statistically significant results. They report communication and connectedness as critical to surviving homelessness. However, simply possessing a phone does not necessitate the skills and willingness necessary to engage with mHealth support. There exists a generational divide, seen currently around age 65, where the older generation often lacks the technical literacy to engage with mHealth interventions ([12]). Further, they cite participant skepticism across all age groups and multiple studies regarding the privacy and usage of mHealth data. While these concerns and limitations may shift over time, presently this seems to argue strongly against mHealth as a viable modality for homeless Americans.

Taken altogether, this review represents strong evidence that improving the adherence of homeless patients to a wide variety of treatment modalities is absolutely possible. However, some approaches appear more effective than others. Interventions which seek to maximize personal contact with homeless patients appear to demonstrate better adherence improvements. That may be through helping them get to appointments, ensuring follow-up for long-acting injectables, setting up clinics within homeless shelters, or meeting them out on the street; these approaches have all shown significant results. Conversely, the opposite appears to be true: leveraging technology to provide care remotely does not provide comparable results. In the future, our research efforts can benefit from heeding these lessons. The trials included in this present review can provide valuable data as we seek to establish more rigorous, quantitative baselines. Additionally, future trials can benefit by focusing effort on the modalities highlighted in this review.

The ambiguity of what constitutes a reasonable baseline for comparison is a common theme throughout the trials in this review. Randomization into intervention and control arms is often not appropriate, and we can easily see how using reference data for comparison is often misleading. This ambiguity represents the largest limitation of the present review. Additionally, while most studies follow the definition of homelessness provided by the United States Department of Health and Human Services, all participants may not have met this definition throughout the course of each study, and there are likely also people in unstable living situations not captured by this definition that would benefit from programs supporting the homeless. The ambiguous realities of marginalized life in America present unique challenges to rigorous, quantitative investigation. These challenges will likely remain a limitation both to the present review and to future research. Moreover, it is important to note that the present review is intentionally limited to the American cultural context; these results may not generalize well to all locales. Even within the United States, the present review may not represent all locations well. The participant demographics in the trials reviewed here were all either explicitly or implicitly urban. This may bias evidence in favor of urban-centric solutions. Finally, many of the studies cited here had very low sample sizes, often with 50 or fewer participants. While they offer important evidence to inform the direction of future research, these small sample sizes do limit generalizability of their conclusions.

We recommend future studies address adherence improvement in homeless populations outside large urban centers. We recommend further that the results presented here be compared with similar analyses from other global regions.

This review demonstrates that while treatment adherence among people experiencing homelessness remains an evolving field of research, rough categories have emerged from current studies which can inform our efforts going forward. Three key points stand out as especially relevant to future research. First, interventions appear more likely to improve adherence to whatever extent the intervention proactively engages the patients. Second, psychosocial interventions show significant benefit when paired with additional evidence-based services. Finally, leveraging communications technology (e.g. smartphones) does not currently appear to produce significant results in this patient population, even when phones are provided.

Special thanks to Nell Aronoff, the University at Buffalo Libraries Liaison to the Jacobs School of Medicine and Biomedical Sciences. Special thanks also to Howard Callihan, PATH Case Manager at Bert Nash CMHC, and to Uriah Binegar for editing support. All necessary material and financial support were generously provided by the Jacobs School of Medicine and Biomedical Sciences.