Table 2.
Summary of included studies.
| Study title, authors, year, country | Design study period | Population/setting, sample size (n) | Intervention | Comparison I outcomes (O) | Methodology | Main conclusions and key results | Overall assessment of quality limitations as reported and methodological appraisal |
|---|---|---|---|---|---|---|---|
| Early identification of a COVID-19 outbreak detected by wastewater surveillance at a large homeless shelter in Toronto, Ontario (26) Akingbola et al., 2022, Canada |
Quasi-experimental January 2021 |
Men’s homeless shelter N = 169 | Wastewater surveillance for COVID-19 | C:with other homeless shelters/previous wastewater surveillance but not clear O:detection of COVID-19 from wastewater surveillance |
Wastewater samples taken over 1 h retrieved twice weekly from the site. The surveillance team were made aware of any COVID-19 symptoms/cases before commencement of surveillance | Wastewater surveillance enabled prompt dissemination of COVID-19 testing in asymptomatic patients, therefore facilitating effective outbreak management | Low no comparison, no follow-up, no mention of inter-rater reliability, n relatively small of only 169 |
| Implementation of Rapid and Frequent SARS-CoV2 Antigen Testing and Response in Congregate Homeless Shelters (27) Aranda-Diaz et al. 2021, United States |
Prevalence study January–February 2021 |
Homeless shelters n = 828 | Testing strategy and isolation provision | C: between stages of intervention, demographics O: detecting COVID-19 infection, isolation, identify outbreaks |
Programme of regular COVID-19 antigen testing in 10 congregate living shelters. Implemented for residents and staff. Positive individuals were referred to isolation and contact-tracing done. Used RE-AIM framework to guide implementation and evaluation. | Testing and isolation strategy was able to be implemented effectively, detect COVID-19 infections, isolate individuals and identify outbreaks. 47.5% eligible residents participated in testing at least once. Identified 10 positive cases, 8 successfully isolated. |
Very low Low participation rate, low adherence to twice-weekly testing, does not have control, short study period. Some limitations in transferability to UK given setting, may be applied to other congregate settings. |
| Clinical Outcomes, Costs, and Cost-effectiveness of Strategies for People Experiencing Sheltered Homelessness During the COVID-19 Pandemic (28) Baggett et al., 2020, United States |
Modelling study April–August 2020 |
Homeless shelters n = 2,258 | Symptom screening, regular testing, alternative care sites (ACSs), temporary housing | C: No intervention O: cumulative infections and hospital days, costs to healthcare sector, cost effectiveness |
Decision analytic model using a simulated cohort residing in homeless shelters, based on literature and national databases. Looked at disease progression, transmission, and outcomes. | From the model: Daily symptom screening with alternative care sites for pending or confirmed cases of COVID-19 was associated with fewer severe COVID-19 infections (37%) and decreased healthcare costs (46%) in the homeless population. Fortnightly PCR testing and temporary housing most effective (81% fewer infections) but much higher costs (542%) % increase/decrease compared with no intervention |
Low Findings were specific to individual adults, homeless families and rough sleepers excluded. Assumed homogenous mixing in shelters which may alter infections projected in model. Did not factor in higher rates of comorbidities in homeless population. Focussed on one location with different cost of living to other areas. Limited transferability given based on US setting and costs but demonstrates times where prevention cheaper than healthcare treatment costs. |
| Comparison of infection control strategies to reduce COVID-19 outbreaks in homeless shelters in the United States: a simulation study (29) Chapman et al., 2021, United States |
Modelling study March–April 2020 |
Homeless shelters n = not stated | Daily symptom screening, PCR testing, universal masking, relocation of possibly infected individuals, staff testing | C: no intervention O: probability of averting an outbreak |
Developed individual-level microsimulation model of COVID-19 transmission in homeless shelters and calibrated to data from PCR surveys across 5 shelters and 3 cities. Assessed risk (low, medium, high) of shelter characteristics, e.g., distancing, volume density. | Combination of strategies (symptom screening, regular testing, relocation, mask wearing) most effective. High risk settings (i.e., high density, high rates in background population) showed little improvement with any strategy. Daily symptom screening ineffective at all levels of transmission (probability of preventing outbreak = 0.04). Combining this with relocating of individuals with high-risk clinical symptoms combined did not improve outcomes. PCR testing 2x weekly for all individuals and universal mask wearing improved probability of averting outbreak better than symptom screening. |
Low Limited data availability meant study only calibrated model to small number of shelter outbreaks, cross-sectional nature, assume equal transmissibility in model, simplifying assumptions, short study period. Some transferability to UK as the interventions studied are relevant to UK populations. |
| Assessment of contact tracing for COVID-19 among people experiencing homelessness, Salt Lake County Health Department, March–May 2020 (30) Fields et al., 2021, United States |
Cross-sectional survey (prevalence) March–May 2020 |
Homeless shelters (majority), n = 169 | Contact tracing | C: general population O: follow-up, number of contacts identified |
Homeless people with laboratory confirmed positive COVID-19 cases documented in surveillance system included in analysis. A general population comparison group was systematically selected from all confirmed cases identified during same period. Person-based contact tracing through interviews asking about contacts, living place, businesses | Challenges in identifying, locating, and reaching cases among homeless population and their contacts. 55% of homeless with positive COVID-19 cases were interviewed (73 uncontactable, 3 refusals) compared to 100% general population. 81% of homeless reported no contacts. Homeless were more likely to be lost to follow-up compared to general population (14.2% vs. 0%, p < 0.0001), contacts of homeless were more often unreachable (13% vs. 7% p < 0.0001). COVID-19 testing completed for 62% (31) of the homeless contacts (42.5%, 322 general population). 16% of homeless contacts compared to 22% of gen pop contacts tested positive p = 0.3 |
Moderate Contact tracing findings from this district may not be transferrable to other areas, also contact tracing done early in pandemic. Does not actually identify reasons for homeless having fewer contacts, more difficult follow up. Low number of women in sample. May be cases of COVID-19 undocumented in homeless. Some transferability given transient nature of most homeless population regardless of country of origin. |
| Assessment of a Hotel-Based COVID-19 Isolation and Quarantine Strategy for Persons Experiencing Homelessness (31) Fuchs et al., 2021, United States |
Prevalence study March–May 2020 |
Multiple homeless categories, n = 1,009 | Hotel-based COVID-19 isolation with some healthcare | C: Between Subgroups of homeless population, demographics O: programme retention/premature discontinuation of quarantine |
Hotel-based care system: individuals unable to safely isolate at home (with mild–moderate COVID-19 infection, pending test, close contact), were referred from other settings (hospitals, outpatients, public health surveillance) Physician-supervised team of nurses and others offered care and monitoring. | Hotel-based isolation strategy that delivered integrated health support for homeless people was implemented safely outside of hospital, adherence was fairly high (81%), although significant association of premature discontinuation with unsheltered (aOR 4.5, 95% CI, 2.3–8.6). Other risk factors were: being a close contact (aOR 2.6), age < 40 (2.5), female (1.8), black ethnicity (1.7). Used sensitivity analysis and regression models. |
Low Missing data on homelessness, results on hospital stay are limited due to times of pandemic, may not be generalisable to all settings due to reliance on other workers outside of public-health. Design listed as retrospective cohort, but no non-exposure group. Some transferability to UK, evidence of successful implementation of hotel isolation which incorporates care, evidence of issues with housing rough sleepers. |
| Implementation of a Recuperation Unit and Hospitalisation Rates among People Experiencing Homelessness with COVID-19 (32) Gai et al., 2021, United States |
Pre/post-intervention March–June 2020 |
Unspecified homelessness, n = 226 | COVID-19 recuperation unit (CRU) | C: pre-intervention O: hospitalisation rates |
Analysis of COVID-19 hospitalisation census from a single hospital. COVID-19 recuperation unit (CRU) opened midway through study period, provided isolation and quarantine for homeless and treatment for substance use. | An alternative care site for homeless with COVID-19 infection was associated with a reduction in hospitalisations in the homeless population. There was a 28% reduction in hospitalisations pre/post-intervention (risk ratio 0.72, 95% CI, 0.63–0.82) |
Low May have missed hospitalisations elsewhere as only one hospital. Some transferability to UK, implementation of hotel isolation, however, does not specify homeless type. |
| Comparison of COVID-19 mitigation and decompression strategies among homeless shelters: a prospective cohort study (33) Hsu et al., 2021, United States |
Prospective cohort study March–May 2020 |
Homeless shelters, n = 381 | Depopulation strategies: provision of lodging in temporary tents in car park, gym, and hotel spaces | C: between interventions O: rates of COVID-19 infection |
Study looks at residents in two homeless shelters which adopted different strategies to reduce density of shelters. Guests from one shelter were distributed to recreational centre space, hotel, while the other in temporary tents in car park. COVID-19 testing and pre + post-test survey | Depopulation strategies to multiple different locations of stable accommodation was better at preventing COVID-19 infection compared with outdoor tent set-up. Tent intervention participants had 6.21x higher odds of positive COVID-19 tests on follow-up compared with stable indoor locations (adjustments for loss to follow-up, age, gender, race, 95% CI 1.86–20.77) |
Low of note: weather conditions impacted outdoor group, incidents of residents going inside and unable to socially distance during storm. Limitations: high loss to follow-up, varying sample collection methods used due to test shortages at some points, wide confidence interval ranges. Study population of two urban shelters in same state—limited transferability. |
| Assessment of a Hotel-Based Protective Housing Program for Incidence of SARS-CoV-2 Infection and Management of Chronic Illness Among Persons Experiencing Homelessness (34) Huggett et al., 2021, United States |
Retrospective cohort study April–September 2020 |
Varied homelessness (shelter, encampment, street), n = 259 | Individual hotel rooms, healthcare workers available | C: homeless in shelters O: rates of COVID-19 infection |
Retrospective analysis of people who were provided protective housing in individual hotel rooms. Participants were homeless who were deemed at risk of severe COVID-19 outcomes if they were infected (age, underlying health conditions). Healthcare workers on-site provided care, testing. | Homeless people in protective housing had lower risk of COVID-19 infection compared to shelter residents. 259 homeless people admitted to hotel, 201 included in protective housing cohort. 11 tested positive, 7 of these were within 5 days of admission. Overall incidence in hotel cohort was 54.7/1000 compared to 137.1/1000 among shelter residents in the same city (95% CI 125.1–149.1 per 1,000 people; p = 0.001). 11 of hotel cohort were transferred to hospitals for severe illness, no deaths. Improvements in chronic disease management, 51% housed after departure. |
Low Estimation of non-intervention COVID-19 incidence may be inaccurate. Large portion of COVID-19 cases in hotel cohort were within 5 days of admission so possible overestimation of risk of infection in hotel. Selection bias risk high – sample recruited targeting their risk factors, voluntary. No unsheltered homeless who were approached to be recruited agreed to participate. Single site study. Some transferability to UK, implementation of hotel isolation. However, comparison is shelter rates of infection. |
| Of not passing: homelessness, addiction, mental health and care during COVID-19 Lenhard et al 2022 (35), UK |
Qualitative study May 2020–April 2021 |
Across homeless support shelters N = 37 (30 service workers, 7 people experiencing homelessness) | Provision of accommodation to support homelessness, telemedicine to provide alternative access to healthcare during the pandemic | C: pre-pandemic O: experience of accommodation, COVID-19 on wellbeing and mental health |
Semi-structured interviews conducted with both homeless service workers and member residents | Those suffering substance misuse and mental health issues found that as a result of the pandemic support was restricted. Digital options were not always suitable for those with challenging needs. One unexpected positive outcome was that some people had a better chance of securing more permanent housing as a result of having been provided housing at the start of the pandemic. | Moderate Though no mention of reflexivity or statement locating the researcher culturally, overall solid methodology, several quotes provided to support themes, qualitative methodology clearly explained and justified |
| COVID-19 among people experiencing homelessness in England: a modelling study (36) Lewer et al., 2020, England |
Modelling study scenarios: 1st wave February–May 2020 2nd wave June 2020–January 2021 | Temporary hostels, rough sleeping, night shelters, n = 46,565 | Hotel accommodation (housing or isolation), reduced mixing with general population, infection control in settings, e.g., distancing, hand hygiene | C: no intervention, second wave scenarios O: rates of COVID-19 infection, hospitalisation, and mortality |
Used a discrete-time Markov chain model, simulated under different scenarios varying the incidence of COVID-19 in the general population and use of prevention measures. First wave and future wave scenarios ran, each 200 times. Prevention measures including COVID-PROTECT (single room + bathroom) COVID-CARE (testing + medically supported accommodation for symptomatic individuals). | Prevention measures including hotel accommodation and medical care with COVID-19, reduced mixing with general population through lockdowns, and infection control strategies, successfully reduce adverse outcomes of COVID-19 in model. Model suggests 21,092 infections, 1,164 hospitalisations, 338 ICU admissions, and 266 deaths among homeless population prevented in the first wave. Even with no second wave in general population, if preventative measures are not continued, estimated additional 11,168 infections, 653 hospitalisations, 189 ICU admissions, and 165 deaths. If second wave but prevention measures continued, 1754 infections and 31 deaths estimated. If hotel accommodation and isolation rooms (PROTECT and CARE) only, rates are high but lower than without 3,654 infections and 54 deaths avoided |
Moderate Uncertainty about COVID-19 rates and severity and homeless population, issues of modelling immunity, unknown actual size of homeless population, assumed no mixing between subgroups, assumed no changes in infectiousness. Based model on population of homeless from surveillance data in London only (rates, hospitalisation, mortality). While UK-based, numbers may not be fully representative of UK homeless populations. |
| “You Have a Place to Rest Your Head in Peace”: Use of Hotels for Adults Experiencing Homelessness During the COVID-19 Pandemic (37) Robinson et al., 2022, United State |
Qualitative study March–May 2021 |
Two Hotels for those experiencing homelessness in New Haven N = 18 | Hotels for those living in congregate shelter/unsheltered settings | C: Pre-pandemic O: account of people’s experiences of the hotels |
Those living in shelters in New Haven were moved to single room ensuite hotel to contain transmission of COVID-19 | On the whole participants stated that access to their own room and facilities (such as bathrooms) offered security, a greater sense of control, and empowered them to make positive changes for their health and wellbeing | Moderate Philosophical perspective unclear, no statement on reflexivity or one to local cultural and theoretical perspective of researchers, limited generalisability since only two hotels in one area focus of the study |
| Implementation of Baltimore City’s COVID-19 Isolation Hotel. Rosecrans et al. 2022, United States | Quasi-experimental study May 2020 |
Baltimore, isolation hotel for those experiencing homelessness N = 93 residents at peak of study | Isolation hotel for those experiencing homelessness-services for those suffering substance misuse | C: other isolation sites O: detection of COVID-19 among homeless population |
Mode of recruitment to centre not clear, but 300 bed facility opened up to homeless in Baltimore, following collaboration between university of Maryland medical system and Lord Baltimore Hotel and Baltimore City Health Department | 78% of residents did full isolation and quarantine routine, and just 6% of residents required transfer to hospital or higher intensity care-projections suggest hotel responsible for prevention of thousands of cases of COVID-19 | Low Little outcome data, follow-up unclear, hard to determine who measured outcome data, no control group method of recruitment of participants also unclear or how demographic information on participants was obtained |
| Lessons Learned through Implementing SARS-CoV-2 Testing and Isolation for People Experiencing Homelessness in Congregate Shelters (38) Scott et al 2022, United States |
Quasi experimental March to May 2020 |
Congregate shelters N = 52 | COVID-19 testing | C: general population O: COVID-19 positivity rates among those experiencing homelessness |
Community partners came together for the create and carry out a pilot testing alongside isolation in a homeless shelter, in order to review the viability of adopting such testing, in other homelessness facilities. | 14 out of 52 residents tested positive, 13 residents with positive tests were moved to isolation hotels, 9 out of 13 were moved with 72 h of the test having been conducted. | Low Pilot study so small n, follow up unclear, no control group, comparison with general population not clear sample of the general population did not get COVID-19 testing in the same way |
| Comparing the initial Everyone In COVID-19 London response to the resurgence of Dec 2020–Feb 2021 (39) Story and Hayward, 2021, England |
Observational report April 2020–February 2021 |
London-based homeless in hostels or hotel accommodation | Hotel accommodation, specifically Everyone In | C: between subgroups hostels and hotels O: rates of COVID-19 |
Limited methodology: Report on rates of COVID-19 collected in London homeless, some in hostels, and some in hotels connected to the everyone in initiative. | Hotel accommodation had a lower risk of COVID-19 infection than hostels for homeless population. Rise in cases in those living in hostel accommodation compared to emergency hotel accommodation and no fixed abode. Hostel group 5.6x increased risk of positive COVID-19 test compared to hotel accommodation. Likely connected to reduced capacity of hotels due to ending of service, leading to crowding of hostels which had worse infection measures |
Low Absence/inaccuracy of available data on the size and characteristics of the accommodation and support offered to this population. Transferability n/a, UK based |