Loading...
HomeMy WebLinkAboutfinal-prk-reportE v a lu a t i n g P r o je c t R o o mk e y i n A l a me d a C o u n t y Lessons from a Pand em ic R es pons e t o Homel essnes s C ODY Z E G E R MAY 2 0 2 1 A r e p o r t for t h e A l a m e da Co u n ty O f f i c e o f H om e l e s s C a r e a n d C o o r d i n a t i o n Executive Summary - 1 - Executive Summary In response to the COVID-19 pandemic, California was the first state to implement a program specifically designed to protect people experiencing homelessness. Dubbed Project Roomkey (PRK), the program began as a way to utilize vacant hotel rooms as non-congregate shelter to prevent the spread of COVID-19 among the community. What followed was one of the fastest expansions of the state’s shelter capacity in its history and an opportunity to reinvent emergency shelter provision. With support from state and federal funding, the state tasked communities across California with Project Roomkey’s implementation. Alameda County jumped at the opportunity to participate in the program and with the help of city, state, and nonprofit partners, opened 13 hotel and trailer sites, doubling its shelter capacity in less than six months. Between March 2020 and March 2021, these sites served over 1,700 participants. Given the federal requirement to prioritize people at higher risk of complications from COVID-19, these participants were significantly older and sicker than the population of people experiencing homelessness overall. To effectively serve these participants, a new form of shelter was needed: one that could not only bring as many people inside as quickly as possible, but also meet their medical needs and put them on a path toward more permanent housing. The county responded by adding health care and housing navigation services to the non-congregate shelter environment. This combination formed the backbone of a new “PRK model” of shelter provision. This report seeks to understand the benefits of the PRK model in order to inform post-pandemic homeless services. Through quantitative analysis of administrative data and interviews with service providers and government staff, this is one of the first in-depth studies of Project Roomkey outcomes anywhere in the state. Evaluating Project Roomkey in Alameda County - 2 - The initial lessons from Alameda County are clear: 1. The low-barrier, non-congregate shelter model was universally preferred by service providers and made shelter more appealing to many people living outside. 2. The health care services and other amenities provided at PRK sites helped participants stabilize and address long- standing medical issues. 3. The focus on housing navigation and the creation of new housing subsidies for people at PRK sites led to 65% of exiting participants entering housing—nearly double that of traditional congregate shelters in the county. 4. The speed at which the program was implemented required government and nonprofits to collaborate in new, beneficial ways that could be maintained beyond this program. This report demonstrates that with a substantial infusion of state and federal funding as well as coordination between government and nonprofit partners, long-term progress can be achieved toward addressing the homelessness crisis. While the success of Project Roomkey should be celebrated, it also came at a significant financial cost. Project Roomkey is estimated to have cost about $260 per participant per night. This is many times higher than congregate shelter and in line with other service-intensive environments such as medical respite centers. Therefore, while the PRK model may be more effective than congregate shelter at addressing homelessness, its cost could make it most viable as a short-term intervention, not a long-term solution for people experiencing homelessness. In order to understand how to build on the success of the PRK model, this report delves into which components helped improve participant outcomes and suggests ways to bring those pieces forward into future county shelter operations. While shelter alone is not the answer to homelessness, it is an important part of meeting people’s immediate needs. This report shows that shelter can and should be improved in ways that put people on a path toward stability and long-term housing Table of Contents - 3 - Table of Contents Executive Summary ............................................................................................................. - 1 - Table of Contents ................................................................................................................ - 3 - Acknowledgements ............................................................................................................ - 4 - Introduction ........................................................................................................................... - 5 - The Pandemic-Fueled Creation of Project Roomkey ............................. - 6 - Getting Off the Ground ........................................................................... - 7 - Methods ................................................................................................................................. - 9 - Project Scope .......................................................................................... - 9 - Quantitative Analysis .............................................................................. - 9 - Qualitative Analysis ............................................................................... - 11 - Limitations and Future Research Needs ................................................ - 12 - Project Roomkey in Alameda County .......................................................................... - 14 - Initial Program Implementation ............................................................. - 14 - Shelter Model Comparisons .................................................................. - 16 - Participant Demographics ..................................................................... - 20 - Program Cost Estimates ........................................................................ - 23 - Lessons Learned ............................................................................................................... - 26 - Bringing People Inside .......................................................................... - 27 - Addressing Health Needs ..................................................................... - 32 - Increasing Access to Housing ............................................................... - 38 - Improving Collaboration ........................................................................ - 45 - Bringing the PRK Model Forward: Conclusion and Recommendations .............. - 49 - Conclusion ............................................................................................. - 49 - Recommendation 1 ................................................................................ - 50 - Recommendation 2 ................................................................................ - 53 - Recommendation 3 ................................................................................ - 56 - Appendix A: Data Definitions ........................................................................................ - 59 - References .......................................................................................................................... - 61 - Evaluating Project Roomkey in Alameda County - 4 - Acknowledgements This paper would not be possible without the advice, guidance, and feedback I received from so many individuals. From Alameda County, Kerry Abbott was immediately willing to let me delve into this program and is truly invested in understanding how to best serve people experiencing homelessness. Dr. Kathleen Clanon led me through what it meant to make big change and was willing take time out of pressing pandemic work to give me feedback. Rachel Metz jumped into the details with me, always providing incisive comments, asking the right questions, and guiding me toward making this report as useful as possible. Mona Palacios was the linchpin in understanding program costs, collecting and sharing every stray invoice and helping me decode budgets and contracts. Last, but most certainly not least, Martha Elias was the best data wrangling partner I could have asked for. Her consistent curiosity in digging for answers and willingness to lend a hand at any time made all of this data analysis possible. From UCSF, Cynthia Nagendra helped me design this report from the outset. She shared invaluable knowledge about shelter provision, guidance on what questions I should ask, and always helped me keep the big picture in mind. Dr. Gabriel L. Schwartz also generously volunteered to be my Stata guide, talking me through both strategic and philosophical data analysis questions. From Berkeley, Dan Lindheim coached me through this process and consistently posed questions I hadn’t considered. My APA section provided much needed support and made my weeks so much brighter. Of course, this report could not have happened without all those who participated in interviews. From service providers to county and state staff, their willingness to take time out of their days doing demanding, vital work and share honest thoughts with me made this report what it is. I hope reading this makes them appreciate all they have helped accomplish. It goes without saying that there are countless others involved in making our homeless response system function each day, including the people who it is intended to serve. I hope this project helps move us toward a world where people are served more effectively to end homelessness as we know it. The author completed this study as part of a Master of Public Policy Program at the Goldman School of Public Policy, University of California at Berkeley. This study fulfills the program’s thesis requirement. The analysis and opinions contained herein belong solely to the author and are not endorsed by the Goldman School of Public Policy, or any other organization. Introduction - 5 - Introduction On April 3, 2020, California Governor Gavin Newsom stood outside a Sacramento hotel to announce the official launch of Project Roomkey: a statewide initiative to provide emergency shelter for tens of thousands of people experiencing homelessness to prevent the spread of COVID-19 in the community.1 The Governor celebrated Project Roomkey (PRK) as a “first-in-the-nation initiative” to protect people experiencing homelessness from COVID-19 and other states soon followed suit. 2 Bolstered by an infusion of state and federal funding for this effort, Newsom set a goal of securing 15,000 rooms across the state that could be used as non-congregate shelter.3 In order to jumpstart the initiative, Newsom allocated $150 million to support local communities in protecting the health and safety of people experiencing homelessness. By the end of 2020, the state had made a total of $512 million available that could be used to expand and sustain Project Roomkey.4 While the cities, counties, and tribes participating in PRK designed and implemented their programs differently, many larger urban communities chose to create both Isolation & Quarantine (I&Q) rooms where people who had been exposed to COVID-19 could stay for the duration of their illness, as well as Shelter-in-Place (SiP) rooms for people experiencing homelessness who needed a non-congregate shelter to follow social distancing guidelines. It appears to have worked: in just a matter of months, approximately 35,000 people participated in Project Roomkey.5 Additionally, the widespread creation of the SiP rooms represents one of the fastest expansions of emergency shelter capacity as well as one of the first major investments in service-intensive, non-congregate shelter in the state. Given the program’s success providing shelter for so many people so quickly, it is imperative to try and understand what lessons can be learned from the design and implementation of this shelter model. However, with over a year having gone by since its inception and communities being overwhelmed with getting these programs up and running so quickly, Project Roomkey has not been studied in depth. Understanding the impact of the program is crucial to illuminate the strengths and limitations of California’s various emergency shelter models in achieving positive outcomes for people experiencing homelessness. This report represents one of the first studies of Project Roomkey outcomes through an in-depth analysis of Alameda County’s program design and implementation. It aims to break down the various components of Alameda County’s Project Roomkey model in order to identify which components were key to improving health and housing outcomes for people experiencing homelessness. Using both quantitative and qualitative research, this report shows that Alameda County’s Project Roomkey was successful at bringing people experiencing Evaluating Project Roomkey in Alameda County - 6 - homelessness into shelter quickly, connecting them with necessary services that helped them stabilize, and moving a large proportion of people from this temporary intervention to permanent housing. However, since the program is tied to the pandemic, which opened access to vacant hotels and new funding streams, it is likely to end this year. Federal reimbursement is set to run out on September 30, 2021 and with travel beginning again it is likely many hotels will want to use their rooms again. Without the availability of federal funding and vacant hotel rooms, questions remain as to how to bring the successful aspects of Project Roomkey’s non-congregate and low-barrier shelter model into existing shelter programs to continue improving outcomes for people experiencing homelessness. The Pandemic-Fueled Creation of Project Roomkey California’s first case of COVID-19 was documented on January 26, 2020.6 Just over a month later, on March 4, Governor Newsom declared a state of emergency and directed state agencies to identify hotels and other facilities that could be used to isolate and treat individuals who have tested positive for or are high-risk of complications from COVID-19.7 The size and vulnerability of California’s homeless communities quickly became cause for concern during the pandemic. As of the most recent estimate, over 151,000 Californians experience homelessness on any given night. Of these, around 72% were unsheltered while 27% had a disability and faced chronic housing challenges.8 Research has shown that unsheltered adults experience higher rates of physical and mental health conditions9 as well as increased vulnerability to communicable diseases.10 So while homelessness was already a health crisis in California, the addition of a pandemic threatened to exacerbate the problem. Further, California has traditionally relied on congregate shelters to temporarily house people experiencing homelessness. Since congregate shelters are designed for groups of people to live in one or a few large rooms, they tend to be less expensive and easier to find space for. The state’s extensive use of congregate shelter, however, posed yet another risk during COVID-19 due to the airborne transmission of the disease and the inability to isolate in a congregate shelter setting. In March 2020, the CA Department of Public Health recommended that communities decrease the density of congregate shelters and create isolation units for people exposed to COVID-19.11 Communities across California followed these recommendations.12 However, this meant that suddenly many fewer shelter beds were available, leaving more people experiencing homelessness with no option but to sleep outside. San Francisco, for example, lost 75% of its shelter capacity after implementing the recommendations.13 Therefore, the pandemic increased the urgency to create additional non-congregate shelter space across the state to mitigate the transmission of COVID-19 among people experiencing homelessness and front-line staff. Introduction - 7 - Soon after the emergency declaration, a report confirmed the danger posed by these overlapping risks. Due to the underlying health conditions of the population, state modeling estimated that over 60,000 unsheltered individuals could contract COVID-19 within a two-month period—quickly overwhelming the state’s remaining hospital capacity.14 Days later, on March 18, Newsom signed an executive order and a senate bill that together gave communities access to $150 million to expand emergency shelter and isolation capacity for people experiencing homelessness.15 Of this, $50 million was specifically tied to the creation of non-congregate shelter. In November 2020, the state released another $62 million in one-time funding to continue Project Roomkey operations while transitioning participants to permanent housing. This funding was distributed to the state’s 13 largest cities, all 58 counties, all 44 Continuums of Care, and tribal governments through multiple state agencies.16 While spending was largely left up to recipients, the state recommended certain uses: ● Acquisition and lease of hotels, motels, and trailers for isolation and shelter ● Furnishings, medical supplies and services, and sanitary supplies for these facilities ● Street outreach to people experiencing homelessness ● Transportation to and from shelters and medical care ● Staffing support and case management for clients While this funding was an important start, once allocated across the state, it represented a small portion of overall program costs for some participating localities. While the state provided guidance to communities on how to use other funding streams for their Project Roomkey operations, it needed a larger infusion of resources to meet the goal of securing 15,000 non-congregate shelter rooms. Seeking federal support, California was the first state that requested FEMA reimburse communities for the costs of the program. On March 27, 2020 FEMA approved this request and agreed to reimburse the state for 75% of the costs of leasing and operating non-congregate shelter. This funding was exclusively meant to serve people experiencing homeless who had tested positive for COVID-19, been exposed to COVID-19, or were deemed “high risk” of complications from COVID-19 (such as people over 65 or those with certain underlying health conditions). On January 21, 2021 the Biden Administration expanded that reimbursement to cover 100% of the costs through September 30, 2021.17 Getting Off the Ground With this commitment from the federal government as well as additional funds from the CARES Act of 2020, participating communities were able to begin implementation of Project Roomkey. However, due to the complexities involved in billing FEMA for these costs and the risk of not receiving the full reimbursements, it was difficult for many areas to provide the upfront capital necessary to initiate Project Roomkey.18 Alameda County, Evaluating Project Roomkey in Alameda County - 8 - for example, did not start billing FEMA until January 2021, because of the significant staff time it requires to do so and the uncertainty involved in the process. Additionally, FEMA reimbursement did not cover any supportive services for the sites (e.g. case management or mental health counseling)19 so the extra expense could be prohibitive for many local governments without funding already available for homeless services. Those that did start Project Roomkey programs were tasked with identifying individuals that met FEMA the qualifications, negotiating with hotel owners to lease out rooms or entire buildings, and contracting with service providers to operate the sites. According to county staff and providers, these communities had just a matter of days or weeks to start up a program for which they had no previous model. Finding sites in which to house these individuals, while not without its challenges, happened fairly quickly. First, the state allowed the use of over 1,300 trailers it had distributed across California as one option to provide non-congregate shelter. The majority of rooms, however, were secured through agreements with hotels. Due to the travel restrictions caused by the pandemic, hotels were facing single digit occupancy during the spring and summer months, when typical rates would have been closer to 70%.20 Therefore, many hotel owners initially welcomed the offer to lease their vacant rooms to government agencies. The state signed contracts with the first two hotels in Oakland to begin the program, which Alameda County then assumed payment for in August 2020. When the county ran a request for proposals process to identify more hotels interested in leasing out their rooms, they received dozens of responses in just a few days. Project Roomkey provided an opportunity to help hotels, retain some of their staff, and create more non- congregate shelter opportunities at the same time. However, according to county staff, some of these hotels changed their minds when they learned that the rooms were going to be used to shelter people experiencing homelessness and not first responders, health care workers, or other medical professionals. This was true across the state as well where concerns about the image of hotels that chose to serve people experiencing homelessness turned out to be obstacles to their participation.21 Nevertheless, the communities participating in Project Roomkey met the state’s goal of securing 15,000 rooms in less than three months22 and are estimated to have sheltered about 35,000 people since the program’s inception.23 Given the size of Alameda County’s homeless community and their consistent participation in the program from the beginning, analyzing the county’s implementation of and outcomes from Project Roomkey can provide insight into what this unprecedented infusion of resources into the homelessness response system has accomplished and direction on how to continue that success moving forward. The next section will describe the methods used in this study to do so. Methods - 9 - Methods This report utilizes both quantitative and qualitative methods to investigate the benefits that the Project Roomkey SiP rooms had for people experiencing homelessness during the pandemic. While administrative county data provides the basis for the quantitative analysis, qualitative interviews with shelter providers were crucial in contextualizing the findings and elucidating what aspects of Project Roomkey were particularly beneficial or challenging. The study period for all analyses was March 2020 through March 2021 in order to capture the opening and closing of all PRK sites so far. Project Scope As described above, Alameda County created two parallel programs under Project Roomkey. However, this project focuses exclusively on analyzing outcomes from the SiP sites, which the county named “Operation Safer Ground.” Herein, any discussion of SiP or Safer Ground is a reference to this program and all data reported excludes the I&Q program, unless otherwise noted. The I&Q site model has been studied elsewhere and is shown to have effectively served people experiencing homelessness outside a hospital setting and reduced hospital admissions due to COVID-19.24 On the other hand, Alameda County’s combination of non-congregate shelter with comprehensive onsite health care and housing navigation services is not common for shelter provision in California, and has therefore not been widely studied in the state. This report will herein refer to this combination as the “PRK model.” The PRK model is novel because it combines health care and housing services in a way other shelter models do not. Therefore it is valuable to use the relatively fast proliferation of this strategy due to COVID-19 to understand how its outcomes and cost compare to these other shelter models. Quantitative Analysis Participant Demographics and Exits For the analysis of participant demographic information and exit outcomes, I used data from Alameda County’s Homeless Management Information System (HMIS). HMIS is a system mandated by the federal Department of Housing and Urban Development (HUD) that helps communities to track people experiencing homelessness across multiple service providers and programs. It assigns unique identifiers to each individual and updates their record as they move into new programs or complete new surveys. I ran descriptive analyses on participant demographics (race, ethnicity, age, disability status, and more), length of stay in Project Roomkey sites, and exit destination for those who had left the program. I also used HMIS to analyze housing outcomes for participants in Evaluating Project Roomkey in Alameda County - 10 - the county’s shelters between April 2019 and March 2020 in order to compare them with the housing outcomes from PRK. While this database contains the most comprehensive information available on Project Roomkey participants, there are some possible issues with the accuracy of its data. First, since HMIS is a complex database and service providers input data independently, there can be differences in the way demographics or exit destinations are coded. This is especially true with Project Roomkey since programs were created quickly and service providers had to continually hire and train new staff. Therefore, while HMIS was central to this analysis, its fidelity should be checked in future studies with surveys of participants or comparisons to other databases. For example, looking at participants that entered the same rapid rehousing program after leaving Project Roomkey, only 63% have the “rapid rehousing” option chosen in their exit destinations in HMIS. Another 15% of these participants have the “other ongoing subsidy” option chosen. This could be due to the caseloads of the service providers entering the data or their depth of knowledge about the programs participants are entering into after leaving their sites. In order to resolve these discrepancies I aggregated some of the most granular data into higher-level groups. Continuing with this example, I created broader exit destination categories (e.g. “housing,” “shelter,” “medical or treatment facility”) that grouped the related exit options together. After doing so, nearly 95% of participants who had utilized this rapid rehousing program were categorized as being “housed.” These categories were created in partnership with county staff to ensure they matched how the county defines these statistics in other reports. While there isn’t evidence to suggest these data discrepancies would bias the results in any particular direction, it points to the need to check data accuracy in the future by following up with program participants. See Appendix A for a full data dictionary. Additionally, because participants can have multiple stays at different shelters and are often given the same survey multiple times, most participants have many duplicate records. To address this, I created separate multi-step deduplication processes for demographic information and for exit information. However, information on self- reported health conditions could change for a participant between one survey and another. Given the uncertainty around these surveys, I chose to report that a participant had one of these conditions if they ever marked yes in any of their surveys. While this is a standard practice in other research, it demonstrates that health data is not as reliable coming from HMIS. The county Community Health Record was not available for use in this report, but should be used going forward as it is a more accurate repository of participant health conditions. Last, this report utilizes two-sample t-tests to analyze whether the composition of certain groups was proportional (e.g. the race and ethnicity of PRK participants versus those exiting the program to housing). Any mention of statistical significance of specific data refers to this process. While t-tests may not be the perfect method due to the Methods - 11 - characteristics of the available data, I believe it provides a conservative estimate of significance and is therefore useful in identifying the larger statistical trends. All analysis of HMIS data was conducted using Stata Statistical Software, Version 16.1. Point-In-Time Count For comparisons to the broader population of people experiencing homelessness, I used Alameda County’s 2019 point-in-time count.25 The count estimates how many people are experiencing sheltered and unsheltered homelessness on a given night. This methodology is also mandated by HUD and tied to federal funding for homeless services. However, there are numerous critiques of this strategy, projecting that it likely undercounts the unhoused population overall.26,27 While imperfect, the point-in-time count was necessary to understand how proportionately the county was serving people experiencing homelessness across multiple demographic factors. Program Cost I estimated program costs by collecting and reviewing invoices of actual expenditures provided by Alameda County staff. Using these documents, I was able to estimate the costs of running the seven largest SiP hotel sites. While I was not able to receive invoices for all costs, I attempted to get as many as possible in order to estimate program costs more closely than the initial budgeted amounts, which were generated early on in the program. Actual invoices were available for the hotel leases, service provider contracts, nursing staff, caregiver services, transportation, and security during the study period. When invoices were not available for each month within the study period, I used average expenditures for each site from similar invoices to estimate spending. This happened infrequently, but points to the need to access internal county financial systems to generate more precise spending estimates for the program. I was not able to get actual expenditures for miscellaneous costs (e.g. office equipment, technology, administrative oversight, personal protective equipment), but was instructed by county staff to estimate those at roughly 10% of overall program costs, based on miscellaneous costs from other programs. This means that cost estimates in this report are not exact, but are more likely to overestimate total costs of the program rather than underestimate them. Therefore the data related to cost is more conservative and may decrease once more precise data is available. Qualitative Analysis For this report I interviewed 25 individuals from 14 nonprofit organizations and government agencies that participated in Project Roomkey. All interviews occurred between February and May 2021. These interviews remained anonymous in order to Evaluating Project Roomkey in Alameda County - 12 - allow staff to speak freely about benefits and challenges of the program to the author— who is conducting this project on behalf of Alameda County. Additionally, this anonymity allowed providers to discuss anecdotes about program participants without revealing any information about their location. To provide a deeper understanding of how the PRK model differed from other forms of shelter and add nuance to the results of the quantitative analysis, I made it a priority to interview at least one person from each service provider that operated a Project Roomkey site in Alameda County. I was able to speak with staff from every site, including those that had closed. Collectively these staff had decades of experience running congregate shelter programs and therefore could speak to the differences of the PRK model. The interviews consisted of asking a set of standard questions about the organization’s role in Project Roomkey, their perceived advantages and challenges of the model, and what lessons they wanted to bring forward into future emergency shelter provision. From these interviews, I identified four common themes, which I have used to create the Lessons Learned section of this report. For background on the creation of Project Roomkey statewide, I spoke to staff at the California Department of Social Services, conducted an exhaustive review of publicly accessible state documentation, and reviewed as many news articles about the program as I could find through February 2021. Background information on Project Roomkey in Alameda County specifically came from conversations with staff from various county agencies; the cities of Oakland, Berkeley, and Fremont; and internal county documents that recorded details on program design and implementation. Last, I spoke with service providers that participated in Project Roomkey in two other counties in order to see how their experiences compared to that of the providers in Alameda County. While these interviews were informative, more work needs to be done to understand whether the lessons from Alameda County can be extrapolated to other communities in California that implemented a Project Roomkey program. Limitations and Future Research Needs As stated, this report draws on both quantitative data as well as qualitative interviews, and represents one of the most comprehensive analyses of a Project Roomkey program to date. There are, however, limitations to this study based on the time and resources available to the author, which lead to the need for future research. First, and most importantly, while service providers were able to provide insight into the difference between running congregate shelter and a PRK site, they cannot speak to the full experience of their participants. Many of the providers discussed anecdotes and stories about particular participants, but in order to understand what aspects of Project Roomkey were better or more challenging from a participant perspective, they would need to be interviewed directly. Given the short timeline of this project and the lack of Methods - 13 - funding, I was not able to interview participants in a way that would have appropriately valued their time and expertise. I did reach out to multiple advocacy organizations and groups that work directly with participants of Project Roomkey, however I did not hear back in time for the publication of this report. Therefore, future studies should make it a priority to interview both participants who were successful in the program as well as those who were not served well in order to add nuance to our current understanding of the PRK model. Further, while the analyses in this report use the latest available data to provide descriptive statistics on Project Roomkey, they do not provide causal estimates of program outcomes. Future studies could build off of the data in this report by taking an experimental design approach to the outcomes related to the health and housing of PRK participants in order to provide a more robust understanding of the effect of the PRK model compared to other forms of shelter. Additionally, while HMIS data provides the best available view of participant demographics and housing outcomes, it is not the best database to use to assess health outcomes. More accurate participant health information is available via the county’s Community Health Records, which I did not have access to for this report. However, county employees are currently using this database to study the health outcomes from Project Roomkey more deeply and results should be available in the coming months. Last, since Project Roomkey is ongoing in Alameda County, all of this information will change and require ongoing updates. This report bounded analyses to a one-year period in order to take a deeper look at the available data, but participant demographics, health outcomes, housing exits, and program cost will all continue to evolve over time. Also, as the program goes on, staff using HMIS may adjust participant records to more accurately reflect their demographic information or housing outcomes. Therefore, it will be important to do another deep analysis of the program once it ends in order to update the findings of this report. Evaluating Project Roomkey in Alameda County - 14 - Project Roomkey in Alameda County Alameda County is home to a disproportionate share of the state’s population experiencing homelessness: as California’s seventh most populous county, the most recent point-in-time count estimated it had the fourth largest homeless population.28 With just over 8,000 people reported to experience homelessness on a given night, it compares closely with San Francisco and San Diego counties.29 In 2019, the county reported having 1,160 year-round emergency shelter beds to serve around 6,300 unsheltered individuals,30 nearly all of which were in congregate settings.31 These factors motivated the county to respond quickly to the Governor’s offer of funding to house people experiencing homelessness through Project Roomkey. The California Department of Social Services reached out to Alameda County’s Social Services Agency (SSA) and requested they set up emergency hotels. SSA worked with the county Health Care Services Agency and General Services Agency to implement the program, identifying potential sites, creating leasing agreements, contracting service providers, and coordinating all other logistics. By March 27, 2020, only ten days after the Governor’s executive order, the county’s first participant had moved into a Project Roomkey hotel. The speed at which Alameda County was able to respond to this crisis was due to recent work in the county that had been done to connect health care and housing programs. According to county staff, lessons from their Whole Person Care program implementation helped them design the Project Roomkey shelter model so quickly. The county had also recently created an Office of Homeless Care and Coordination, which meant there was a team able to coordinate this type of program. Without this work already completed, the county may not have been able to design and implement a program like Project Roomkey with such short notice. Initial Program Implementation The first site up and running on March 27, 2020 was a large I&Q hotel with the capacity to shelter 100 COVID-positive participants. The first SiP site opened soon after on April 2 with 285 rooms (see Table 1 below). By the end of May, the county and partners had opened eight PRK sites (two I&Q and six SiP) with a total of nearly 800 rooms. However, the height of the county’s Project Roomkey capacity came later, between August and December, during which time they were operating 13 sites as well as a scattered-site voucher program, for a total capacity of nearly 1,100 rooms. This represents a doubling of the county’s shelter capacity in less than six months. Separate from leasing the hotel sites, the state had donated trailers to the City of Oakland. Oakland set up the majority of these as SiP site in an empty parking lot, with two sleeping areas per trailer. They shared the rest with the cities of Alameda and Berkeley to do the same. These three cities had to identify locations for the trailers to sit and coordinate the necessary infrastructure connections (e.g. power, water, and sewer Project Roomkey in Alameda County - 15 - hookups). According to some providers, this setup proved more challenging than anticipated as some of the trailers broke down and connections could take a long time to coordinate. Table 1: Safer Ground Site Information Safer Ground Site City Service Provider Site Type Date Opened Radisson Inn Oakland Abode Services Hotel April 2, 2020 Scattered-Site Vouchers Countywide Various Various May 1, 2020 HomeBase Trailers Oakland Housing Consortium of the East Bay Trailers May 5, 2020 Marina Village Alameda Building Futures Hotel May 7, 2020 Fremont Islander Fremont City of Fremont Hotel May 10, 2020 SpringHill Suites Newark Abode Services Hotel June 5, 2020 Alameda Trailers Alameda Various Trailers June 11, 2020 Berkeley Trailers Berkeley Berkeley Food and Housing Project Trailers June 22, 2020 Days Hotel Oakland Five Keys Schools and Programs Hotel July 1, 2020 Rodeway Inn Berkeley Berkeley Food and Housing Project Hotel July 20, 2020 Quality Inn Berkeley Berkeley Berkeley Food and Housing Project Hotel August 13, 2020 Residence Inn Livermore Abode Services Hotel August 24, 2020 The occupancy agreements used to lease the sites not only paid for the rooms, but also retained hotel workers to run housekeeping, maintenance, and front desk services. Three meals a day were provided for participants at each site, either through the hotel lease or a separate contract. Additionally, the county had contracts with various vendors for shelter operation, security, transportation, nursing and caregiver services, and pharmaceutical services. Since federal reimbursement was tied to serving a specific subset of the homeless community, prioritization of participants was an important step. FEMA would only reimburse expenses related to non-congregate shelter operations for people who tested positive for COVID-19, those who had been exposed, or those who were Evaluating Project Roomkey in Alameda County - 16 - asymptomatic but at “high risk” of complications from COVID-19 and without another way to follow social distancing guidelines.32 Initially, there was an emphasis on identifying those that met FEMA criteria who were also on the county’s “Priority by Name List,” which is used to prioritize people experiencing homelessness for housing. Additionally, the county had to moved people out of existing shelters into the PRK sites in order to reduce crowding per state recommendations. Homeless and health care service providers worked to refer eligible participants from the county’s unsheltered population. The county also utilized a tool that prioritized everyone with an HMIS record by their COVID-19 risk factors and found over 3,000 people eligible for Project Roomkey. According to the assessments participants took upon entry to PRK sites, about 58% were coming from a place not meant for habitation (such as the street, a tent, or an RV), 34% were coming from another emergency shelter, and the remaining 8% were coming from other situations such as staying with friends or family, institutional settings like hospitals or jails, or some form of housing. Shelter Model Comparisons As described above, Alameda County created the SiP sites to provide medically frail people experiencing homelessness with a way to reduce their risk of contracting COVID- 19. Although the primary goal was to prevent the spread of COVID-19 among this community, the county also took the opportunity to include additional services in the provision of emergency shelter, creating the PRK model. The PRK model differs from other shelter models available in Alameda County in a variety of ways (see Table 2 below). While shelters are all run differently, Table 2 collects information from publicly available research as well as interviews with county staff to create a generalized version of various shelter models in Alameda County and the broader Bay Area. Table 2: Shelter Model Comparisons PRK Model Traditional Congregate Shelter Navigation Center33,34,35 Medical Respite36,37,38 Homeless Populations Served High-risk of complications from COVID-19 (e.g. 65+ or underlying health conditions) Varies; some are broad while others have populations of focus (e.g. Transitional-Age Youth, survivors of domestic violence, veterans, etc.) In Alameda County, prioritized for people with an identified path to housing, either through rapid rehousing and employment, or permanent supportive housing Individuals exiting the hospital, or at risk of entering the hospital, with acute medical conditions that make other shelter impractical Room Occupancy Non-Congregate: One household per room + may elect to stay with partners or family Congregate: Multiple households per room Congregate: Multiple households per room + may elect to stay with partners Congregate: Multiple households per room Project Roomkey in Alameda County - 17 - Food Provided Three meals per day One meal per day One meal per day Three meals per day Time Limit None Varies Varies, generally 90 days 90 days Nursing + Caregiver Services Nursing: 1 nurse + 1 medical assistant for every 100 residents Caregiver: 1 caregiver for every 35 participants at participating hotels Nursing: Shelter Health providers support staff in accessing nursing care for residents with acute health needs Nursing: Shelter Health providers support staff in accessing nursing care for residents with acute health needs Nursing + Caregiver: Varies, some have light touch case management with few services, others have more intensive services with 1 nurse for every 20-30 residents Case Management + Housing Navigation 1 housing navigator (or similar staff) for every 25 guests Behavioral health services assigned as assessed Varies, some on-site housing services with undefined ratios Alameda County Behavioral Health Care services provided directly in several sites 1 housing navigator for every 20 participants Varies, as above % of Exiting Participants Who Enter Housing 65% 35% 40% - 90%, varies by model 20% - 30% Approximate Per-Night Cost $260 $50 $100 $200 - $250, varies based on services The PRK Model Project Roomkey’s SiP sites were designed to be low-barrier, non-congregate shelters that provide participants with access to health care and housing resources. Building off the county’s prior work with the Whole Person Care program, which brought health care and housing services together for people experiencing homelessness, county staff knew how difficult it was to meet people’s medical needs while they were living outside. Therefore, they used Project Roomkey as an opportunity to bring these services together again to improve people’s health while they had a consistent place to live. All of the sites were low-barrier, meaning participants could choose to live with partners, bring their pets and other belongings, and were not required to abstain from substance use while on site. The low-barrier nature of the program was intended to provide shelter that would appeal to as many people as possible and make participant stays more comfortable, which in turn would make sheltering in place easier. Evaluating Project Roomkey in Alameda County - 18 - The large majority of participants had their own room or shared with family members. The exceptions to this were the trailer sites where some participants lived with a roommate. The sites were open 24-hours and participants were provided with three meals a day. Most participants had access to transportation to and from the site for important appointments. Since prioritization for these sites was given to people at the highest risk of complications from COVID-19, the participants often came into the program with significant unmet medical needs. Therefore, the county added clinical services (such as nursing and caregiving) with the hope of helping participants access necessary care and creating stability before they moved on, ideally to permanent housing. Specifically, these clinical services were intended to help participants: ● Strengthen relationships with primary care providers in the community ● Facilitate access to the appropriate levels of care ● Foster independence ● Improve health literacy and management of chronic conditions The clinical services available at each site varied, with the highest level concentrated at the largest hotel sites. Nurses were available at most hotels to help participants meet the above goals. The two largest hotels also had staff assisting with caregiving for participants who could not take care of their daily needs alone. As the program progressed, however, the county saw a larger than expected need for caregiver services. While participants could apply for the state’s in-home supportive services (IHSS), service providers faced difficulty helping participants access this program due to long application timelines, uncertainty of approval for all participants, and inconsistent IHSS staff availability. Therefore, the county increased their own caregiver services for these participants as they applied this program. In February 2021, the county concentrated participants with the highest medical needs in one hotel and focused caregiver services there. While health outcomes, including preventing the spread of COVID-19, were the primary goal of the program, connecting people with housing upon exiting the shelter was also a key part of the PRK model. Therefore, housing navigation services to help participants find more permanent housing options were built into each service provider contract. Additionally, the county used funding from the federal Emergency Solutions Grant program within the CARES Act (ESG-CV) to create hundreds of new bridge housing subsidies for Project Roomkey participants. The effects of these investments in health care and housing will be explored more in the next section. Traditional Congregate Shelter Congregate shelters in Alameda County are traditionally focused on providing participants with a place to sleep, with very few, if any, other services. Participants are Project Roomkey in Alameda County - 19 - typically given a shelter bed for themselves and cannot bring their partners, pets, or belongings with them. These beds are in a “congregate” setting where participants sleep in rooms with a number of other participants from other households. Congregate shelters are often only open at night, requiring participants to leave during the day and come back to claim their bed again each night. Some congregate shelters also have limits on participant stays (e.g. 90 days) after which time participants have to leave the site. While service providers have leeway in how they operate these shelters, the average per-night cost for operating shelters is around $50. Given this, many shelter operators are constrained in what services they can provide participants beyond monitoring their safety. Alameda County HMIS data on shelter participants between April 2019 and March 2020 show that around 35% of participants exited to housing. According to the providers interviewed for this project, many have adjusted their shelter operations to include access to more services such as housing navigation or connections to health care. Others have smaller spaces where participants can have more privacy. However, these factors vary by site, so congregate shelters do not systematically include these services. Navigation Center The navigation center model started in San Francisco in 2015 as an attempt to provide congregate shelter in a smaller setting that focused on finding housing for people experiencing homelessness.39 These shelters were designed to be low-barrier and include access to some medical services, case managers, and meals, among other services. They also allowed participants to bring their partners, pets, and belongings with them, unlike traditional congregate shelters. At first the navigation centers in San Francisco had no time limit on a participant’s stay and each bed was tied to a permanent housing placement. A 2015 evaluation of the San Francisco navigation centers estimated that over 75% of participants exited to housing.40 However, due to a lack of resources and conflicts with the city’s typical housing prioritization system, the city changed this model in 2017. Shelter participants were no longer guaranteed a housing placement and participant stays were limited to one month, unless they were in the highest priority group.41 This makes the navigation centers operate more like smaller, low-barrier congregate shelters. Exits to housing have since decreased to around 48%.42 Since then, Alameda County has opened its own navigation centers. While their models are slightly different from San Francisco’s, they also have an emphasis on finding housing for participants. These sites prioritize serving people with a path to housing and have shown to have high proportions of clients exiting to housing. In 2019, the service provider operating these sites reported that of participants who exited, 64% in Berkeley and 88% in Oakland left for housing.43,44 Evaluating Project Roomkey in Alameda County - 20 - Medical Respite The medical respite model is designed for people experiencing homelessness leaving a hospital inpatient setting, or to prevent hospitalizations. It is intended to serve people who have acute health needs that would make them unsuccessful in another shelter environment.45 Medical respite typically provides participants with a bed in a congregate setting and access to high levels of health care services to allow them to recuperate from their hospital stay. Some areas have also created medical respite beds for people who are currently in shelter, but who have medical conditions or functional impairments that prevent them from living on their own in a shelter setting.46 The level of services available varies based on the site, but can range from only providing light case management to providing intensive clinical services. Participants are limited to stays of 90 days and exits to housing are in the range of 20% - 30%, according to county staff. Summary Alameda County’s PRK model was designed to fill a gap in the existing slate of shelter options. In terms of health care services provided, the PRK model falls between those offered at traditional congregate shelters and medical respite centers. In terms of housing navigation services provided, the PRK model falls between those offered at traditional congregate shelters and navigation centers. The PRK model provides more in-home caregiver services than any other shelter model. Providing non-congregate shelter with enhanced health care and housing navigation services was intended not only help prevent the spread of COVID-19, but also as an opportunity to improve participants’ health and housing outcomes while in the program. Following sections will examine the data available on these outcomes. Participant Demographics Between March 2020 and March 2021 a total of 1,708 participants stayed in at least one Safer Ground SiP site. This represents over 20% of Alameda County’s homeless population (see Table 3 below) and around 50% of those the county originally identified as the population eligible for Project Roomkey. Referrals to Safer Ground sites initially stopped in November 2020 due to the program reaching capacity. However, they were reopened in February 2021 and will continue through the end of the program, so these numbers will change going forward. As compared with the 2019 Alameda County point-in-time Count, women and people identifying as White were slightly overrepresented as participants in PRK sites, while those identifying as Multi-Racial were slightly underrepresented. Participants of other races and ethnicities were generally represented proportionately. Additionally, PRK participants were significantly older and reported longer histories of homelessness. PRK also served a higher proportion of people experiencing chronic homelessness as well as people who had ever reported having physical disabilities, chronic health problems, or psychiatric and emotional conditions. Project Roomkey in Alameda County - 21 - Table 3: Participant Demographic Comparisons with Overall Homeless Population Demographic Information 2019 Point-in-Time Count Safer Ground Participants Individuals 8,022 1,708 Gender Male 61% 56% Female 35% 43% Transgender or Gender Non-Binary 4% 1% Race Black or African American 47% 46% White 31% 40% Multi-Racial 14% 7% Asian 2% 3% American Indian or Alaska Native 4% 3% Native Hawaiian or Other Pacific Islander 2% 1% Ethnicity Hispanic/Latinx 17% 17% Age Under 18 4% 7% 18 - 24 9% 2% 25 - 59 73% 56% 60+ 14% 35% Sexual Orientation LGBTQ+ 14% 5% Length of Homelessness More than 1 year 63% 90% Other Conditions Chronically Homeless 28% 67% Physical Disabilities 24% 65% Chronic Health Problems 26% 73% Psychiatric/Emotional Conditions 39% 59% Evaluating Project Roomkey in Alameda County - 22 - The presence of a higher proportion of women in Project Roomkey could be due to the fact that one hotel (as well as the scattered-site voucher program) prioritized women and families in order to better serve survivors of domestic violence and people who were pregnant. Additionally, some of this difference could be due to the fact that the PRK model focused on keeping couples and families together in shelter. According to a few of the providers interviewed, women often do not feel comfortable in congregate shelters, so given the higher proportion of women in the sample, it would be useful for future research to investigate whether the PRK model was able to provide a better environment for those individuals. Participants in Project Roomkey were also generally older with more medical conditions than the homeless population overall. They also had longer histories of homelessness, with 90% of participants having experienced homelessness for more than 1 year and a median length of homelessness just over 4 years. This data is confirmed by the experience of the service providers running the sites who discussed serving an older and sicker population than before. One reason for this difference is that the FEMA eligibility criteria required participants to be “at risk” of complications for COVID-19 as defined by their age and underlying health conditions. It follows, then, that these eligibility requirements selected for participants that were older and had more co- occurring illnesses or disabilities. The overrepresentation of people identifying as White and underrepresentation of people identifying of Multi-Racial could be due to multiple factors. First, county staff reported that they initially had an over-representation of referrals for participants from areas of the county outside Oakland, which are much whiter overall. This may have had to do with the referral response times of providers in the various parts of the county and points to the need to do more targeted outreach to communities that are overrepresented in the homeless population when new programs begin. However, given the low representation of people identifying as Multi-Racial, and the relatively proportionate representation of other racial and ethnic groups, it is also possible the difference has more to do with data entry than participant demographics. As discussed in the Methods section of this report, the quick startup of Project Roomkey could have affected data collection. If staff collecting participant data did not ask participants to self-identify their race, it is possible they identified multi-racial people as the one race they perceived them to be and created an unintentional bias in the available data. When the county re-opened referrals for PRK sites in February 2021, they received a higher portion from Oakland. Looking at participants entering between February and March 2021, it appears the proportion of participants identifying as Black or African American has gone up slightly, White has stayed the same, and Multi-Racial has gone down slightly. It will be important for future research to look into the final demographic breakdown of PRK participants and check the data quality to understand if the program served all racial and ethnic groups proportionately. Project Roomkey in Alameda County - 23 - The data also shows people identifying as transgender or gender non-binary and LGBTQ+ as being underrepresented in SiP sites. While it is important to understand if these groups are being appropriately served by PRK, their underrepresentation may also be a result of the HMIS data collection challenges. Given the speed of program startup and the fact that many participants choose not to answer questions about gender and sexuality, providers may fill out the answers based on what they perceive about a participant. Additionally, if people are asked to select between mutually exclusive gender categories, some individuals may select the option that applies to their gender instead of the one related to their transgender identity. Due to these factors, more work needs to be done to determine whether SiP sites served a proportional number of transgender, gender non-binary, and LGBTQ+ people. Program Cost Estimates As described in the Methods section, the following figures represent the estimated cost of running the seven largest SiP hotel sites in Alameda County: the Radisson Inn, SpringHill Suites, Residence Inn, Marina Village Inn, Rodeway Inn, Quality Inn Berkeley, and the Days Hotel. The restriction to these sites was partially due to the cost information available and partially as a way to ensure cost estimates were inclusive of sites representative of the full PRK model. These sites had varying levels of services, but each of them is a full hotel leased by the county as non-congregate shelter sometime between March 2020 and March 2021. They represent 77% of the total capacity of Project Roomkey sites in Alameda County. These estimates do not include any trailer sites, since the costs of purchasing the trailers was covered by the state. They also do not include the rooms in the Fremont Islander hotel because they dedicated only a portion of their rooms to Project Roomkey. Last, they do not include estimates of the scattered-site voucher program as the services participants had access to did not line up as closely with those in the hotel sites. The below numbers are estimates of program costs, representing all expenditures related to Project Roomkey by Alameda County as well as the state and the city of Berkeley. Since the state covered some PRK costs for the first few months of the program and the city of Berkeley provided nursing services to Berkeley sites, the county’s portion of the total costs is slightly smaller than the program costs overall. While a precise accounting of program costs won’t be available until county financial records are examined, this report uses both actual expenditures and an approximation of miscellaneous costs to create the closest cost estimate as possible, but airs on the side of overestimating total cost to remain conservative. Based on this process, the total estimated program costs of running the seven SiP hotels between March 2020 and March 2021 was approximately $58 million. Of the total program costs, Alameda County spent about $51 million. Prior to December 31, 2020, specifically, the county spent about $36 million, $3 million of which was covered by state funding and the rest by the federal CARES Act. Starting January 1, Alameda County began billing FEMA, which is expected to cover approximately 70% of ongoing program Evaluating Project Roomkey in Alameda County - 24 - costs. The county plans to use state and federal funding sources to pay for the remaining costs. These seven sites opened at various points in the summer and fall of 2020 and two have since closed. While they were open, these sites served 1,456 participants in 775 rooms, for a total of 223,670 nights (or around 5 months each, on average). Therefore, the estimated cost per participant per night was around $259 (herein, the “per-night” cost). The per-night cost estimate can be broken down further to understand which components required the largest investments (see Figure 1 below). The hotel lease average of $176 per night includes costs of a room, laundry, and three meals a day. It falls within the county’s contracted lease rates for the hotels which were between $150 - $200 per night for occupied rooms and $90 - $120 per night for unoccupied rooms. The contracts with service providers added about $27 per night, while transportation and security added another $25. The miscellaneous cost estimate came out to $24 per night. The nursing and caregiver services made up the smallest portion, only adding $7. Figure 1: Safer Ground Per Night Costs Disaggregated by Use Total per-night cost estimated at approximately $259. The $7 estimate for nursing and caregiving is lower than expected by some county staff. This is possibly explained by the fact that many nurses were working at PRK sites on a volunteer basis and only two sites had significant caregiver services. However, even restricting the cost estimates to look only at months when nurses were being paid or only sites that had caregivers, the per-night cost only increases by about $2. Therefore, the overall breakdown of total per-night costs still seems relatively accurate. The per-night cost of these Safer Ground hotels was much higher than other forms of shelter (refer back to Table 2 above). According to county staff, lease rates for congregate shelters in Alameda County are about $50 per person per night, while navigation centers are closer to $100. Hotel Lease, $176 Service Agreement, $27 Transportation + Security, $25 Miscellaneous Estimate, $24 Nurse + Care- giver, $7 Project Roomkey in Alameda County - 25 - The PRK model appears to cost a similar amount to programs designed for populations with high medical needs, like medical respite centers. On average, the medical respite beds in Alameda County are estimated to cost around $200 - $250 per night. The PRK model, then, presents a similar cost to medical respite, with less intensive health care services and more intensive housing navigation. Summary Overall, this analysis confirms the assumption of nearly all county staff and service providers that this program was much more expensive to operate than traditional congregate shelter and similar to the cost of medical respite. At such a high cost, and with federal support for these programs set to end this year, the county is unlikely to be able to sustain this level of expense for leasing hotel rooms to use as shelter indefinitely, especially if these funds could be going to more permanent housing solutions. Simply looking at cost, however, ignores improvements in participant outcomes from this investment in shelter services. If the PRK model is better at helping participants stabilize and exit homelessness, it may result in long-term outcomes that make it worth the investment. Therefore, the next section will explore what aspects of Project Roomkey made the biggest difference in participant health and housing status, in order to incorporate these components into future shelter or temporary housing programs. Evaluating Project Roomkey in Alameda County - 26 - Lessons Learned The following lessons have been compiled from interviews with 25 individuals across 14 organizations representing both service providers operating PRK hotels and government staff overseeing program implementation. The service providers interviewed had decades of experience and could therefore speak to the differences between the PRK model and congregate shelter, in particular. This section attempts to bring together the most common themes from these interviews while also representing the diversity of viewpoints among service providers. All quotes come from interviews conducted between February and May 2021. These interviews remained anonymous in order to allow staff to speak freely about their experience in the program and share anecdotes without revealing participant information. Quantitative data is also embedded within these sections where available in order to validate or add nuance to the experience of the interviewees. The major lessons are that the PRK model: 1. Increased shelter acceptability and engaged people experiencing homelessness who may not have otherwise come into shelter. 2. Facilitated better access to services for participants, but could be better tailored to work effectively for some populations (such as survivors of domestic violence and people with higher health care needs). 3. Nearly doubled the proportion of participants exiting to housing, while also posing challenges for facilitating those exits. 4. Led to the creation of new collaboration between service providers and government agencies, despite some challenges with program logistics. Lessons Learned - 27 - Bringing People Inside The PRK model increased shelter acceptability and engaged those who may not have otherwise used shelter. “I think non-congregate [shelter] is the way to go in the future. To me, congregate seems almost barbaric at this point.” Having private space created autonomy and reduced tension among participants, compared to congregate shelter. Allowing participants to stay with their communities made the prospect of moving inside more appealing. Removing time limits may have given participants time to stabilize. One of the most common advantages that service providers identified about the PRK model was its success in engaging communities that may not have previously wanted to use congregate shelter. Many providers described that some people experiencing homelessness are reasonably hesitant to move into congregate shelter for many reasons, including not wanting to stay in a crowded environment, having to leave their communities behind, and not being sure if it would help resolve their homelessness. These and other reasons have been discussed repeatedly by people experiencing homelessness in various surveys and articles.47,48,49 All of the providers interviewed, however, said that the PRK model was able to better address the needs of these individuals by offering participants a private room, three meals a day, and the ability to stay with partners and pets. These factors were key to moving a large number of people off the streets in such a short time. “If this was congregate shelter,” one provider noted, “we’d have to spend 50 million years convincing people to come inside.” While it is safe to assume that the pandemic was a factor in people’s decision to seek shelter, providers also thought the PRK model had other advantages that helped bring people inside for the first time. In Alameda County, the SiP hotels were designed to be low-barrier in order to encourage as many people to come inside as possible and follow the county’s shelter in place guidelines. This meant they allowed participants to choose who they wanted to stay with, keep their belongings, and didn’t put restrictions on private substance use behaviors. As discussed above, this differs significantly from the setup of most congregate shelters. 1. Evaluating Project Roomkey in Alameda County - 28 - Many Alameda County providers acknowledged that it would be unrealistic to get rid of congregate shelter altogether due to the severe lack of shelter beds to begin with—in 2017 it was estimated that Alameda County had nearly 10 unsheltered individuals for every 1 emergency shelter bed.50 However, most felt that the PRK model created a much better living environment for participants. Since data is not collected in HMIS about a participant’s reason for entering shelter, the best way to confirm this impression would be through interviews with participants of Project Roomkey. However, some of the data that is available in HMIS supports this statement. When a person experiencing homelessness interacts with a provider, they are typically entered into the HMIS database and given a unique identifier. Among Project Roomkey participants, around 23% did not have a unique identifier before their PRK stay. This indicates PRK may have been the first formal interaction with any county services related to homelessness for nearly a quarter of participants. Additionally, the data shows PRK residents had longer histories of homelessness than the population overall. Around 90% of participants reported being homeless for more than one year (as compared with 63% in the overall population) with a median length of homelessness just over 4 years (see Figure 2 below). PRK residents having longer histories of homelessness could indicate that they had more years of experience with the county’s homeless system, but had yet to find a viable solution to their needs. Figure 2: Participant Time Homeless Prior to PRK Entry While these estimates are descriptive of Project Roomkey participants, they would need to be compared with data from congregate shelters to understand the magnitude of any improvement. However, providers overwhelmingly reported that the PRK model made a difference in making shelter a more appealing option for people living outside. They cited having a private room, getting to stay with loved ones, and not having a strict time limit on participant stays as important factors for getting more participants inside than before. Private Rooms Being able to shelter people in private rooms was a first for many service providers and created an increased level of autonomy for participants that was key to keeping many Less than 1 Year, 10% 1 -5 Years, 48% 5 -10 Years, 27% More than 10 Years, 16% 0%25%50%75%100% Lessons Learned - 29 - people in shelter during the pandemic. “The [site] was a huge success because you’re giving people their own room and their own space,” one provider said. “Treating people like adults, letting them come and go with no curfew, not having set meal times. Also, they aren’t 32 inches apart, so naturally you’re going to get better results.” In addition to privacy and autonomy, the individual rooms reduced the tension and potential risks that exist in a congregate setting. Providers explained that staff in congregate shelters have to constantly monitor participant behavior in order to prevent fighting and other interactions that could make the environment unsafe. These actions can result in participants getting removed from congregate shelter altogether and potentially being sent to live back outside. In a non-congregate setting, however, many providers felt they did not have to monitor participants’ every move, which led to a more positive environment overall. One provider described a participant who had not succeeded in congregate shelter before due to conflicts with other participants. “One [participant] in particular has schizophrenia and a developmental disability. People would steal from him and bully him in a congregate shelter setting,” she said. “After he moved into the [PRK site] he gained a lot of weight and lowered his substance use. He started making art and selling it. He seemed much much happier and more stable.” Some providers also discussed the potential dangers for women and transgender individuals in congregate shelter settings. “A lot of people won’t enter congregate shelters for very good reasons. Including sexual violence, trauma, the risk of being robbed,” one explained. And while she clarified that the use of private rooms didn’t completely eliminate the possibilities of those issues, “the risk was much much lower.” At some of the trailer sites, providers tried pairing people with roommates since the larger trailers had two bedrooms. This strategy saw mixed success, however, and created challenges for staff when participants did not get along. While one of these providers said they were usually able to move people around if they started having issues with their roommate, another said they had to give up on the idea of roommates altogether. “We finally realized that it wouldn’t work, even though we spent hours trying to match people, meeting with them, talking to them beforehand,” she said. “Even if people had been getting along in encampments, when they moved into the trailers together that broke down. They are such a small space that it just doesn’t work.” Keeping Communities Together In addition to private space, the ability to keep communities together was central to bringing so many new people inside. Almost all of the providers said that allowing people to stay with their partners and pets helped make shelter more appealing to people who may have otherwise avoided going inside. This held true outside of Alameda County as well. One service provider who participated in LA’s Project Roomkey program described being able to bring two entire Evaluating Project Roomkey in Alameda County - 30 - encampments inside for the first time, after many prior failed attempts at offering them shelter. “Those people wouldn’t go to congregate shelters before,” she explained. “That changed when we offered them spots in a hotel together, though, and they could move into a private room with their partner, dog, and other possessions.” Through Project Roomkey, she was able to house two entire encampments, totaling nearly 200 people, by offering them the option to stay together. A couple of providers in Alameda County said that community building activities and creating relationships among participants was crucial to the success of their sites. One explained that since participants were supposed to be sheltering in place, she had to come up with community activities in order to keep everyone engaged. While she did not feel she had the resources to do this as much as participants needed, it allowed her to keep participants at the site who otherwise may have felt too isolated and left. “They learn about each other and it turns out many people were neighbors or old classmates. Some are friends and family,” she described. “There’s a big sense of community here.” Another provider described a situation in which the community building events on site led to participants finding people they were willing to live with in shared housing after leaving the site. Due to COVID-19 however, some providers were not hosting community activities, which they said could be a challenge for participants. “We aren’t doing group activities like children's groups, parenting workshops, and employment workshops that we’d normally do,” one provider said. “This made it harder for some individuals who struggle with loneliness because there isn’t a place for them to congregate. It was hard that we couldn’t get together and have community events.” In addition, some providers said the physical setup of their site didn’t allow for community events. This indicates that future homeless service programs should include resources for community building and consider the physical space of any new shelters or permanent housing sites to ensure community events are possible. Removing Time Limits One of the other major aspects providers said was helpful in bringing more people inside was the lack of time limits on a participant’s stay. Typically congregate shelters place a limit on how long participants are able to stay in a bed (e.g. 90 days). Due to the ongoing nature of the pandemic, however, the Safer Ground sites allowed participants to stay for as long as they needed. While the lack of time limits also presented some challenges, providers generally agreed that it made shelters more appealing. On average, PRK participants stayed at a site for about 7 months. Since the program is not over, however, this average will likely increase. For those who exited the program during the study period, though, the median length of stay is closer to 5 months. For those still in the program, the median is about 8 months. Only 20% of PRK participants had stays under 90 days (see Figure 3 below). For those who exited the program in less Lessons Learned - 31 - than 90 days, a higher proportion exited to places not meant for habitation, suggesting their overall outcomes may not have been as positive compared to those who were able to remain at the sites longer. Figure 3: Participants by Length of Stay This data shows that the vast majority of participants relied on the lack of time limit and suggests that those with shorter stays may have had less favorable housing outcomes. While this could have again been due partially to the pandemic, providers also think these longer stays helped participants’ stability. One provider said that the lack of time limits also helped build trust between participants and staff on site. “Since we were holding onto people, they learned over time to trust us and access our services,” she said. “Sometimes with other time-limited services you don’t build that trust because you know you’re getting kicked out anyways.” Another provider said that staff had built such good relationships with participants that even after they moved into housing, the participants would seek their help. “A few people keep coming back by to try and talk to their housing navigators because they have built such a strong relationship.” While they couldn’t continue helping participants who had moved on to another program, this demonstrates the strong relationships some providers were able to develop through the PRK model’s flexible timeline. After running a shelter using the PRK model, one provider said she can’t imagine going back to a congregate setting. “I think non-congregate [shelter] is the way to go in the future,” she said. “To me, congregate seems almost barbaric at this point.” She quickly added, however, that congregate shelter may be necessary if we are going to provide shelter for everyone. “I’d rather have congregate than have people sleeping outside.” One of the county staff members who helped design Project Roomkey felt similarly, saying that congregate shelters aren’t healthy for anyone and emphasizing the need to provide people with appropriate services in addition to housing. She also added that the county is working to bring more resources to homeless services by getting creative with health care funding. “There are savings in the health sector when we do this right, so we can’t think of them separately,” she said. Less than 3 Months, 20% 3 -6 Months, 20% 6 -9 Months, 33% 9 -12 Months, 15% More than 12 Months, 11% 0%25%50%75%100% Evaluating Project Roomkey in Alameda County - 32 - Summary Overall, it appears that the low-barrier, non-congregate design of the PRK model helped make shelter more appealing for people who may not have wanted to access it before. The private rooms, ability to stay with community, and lack of timeline were all called out by providers as particularly beneficial aspects of the PRK model. Addressing Health Needs The PRK model facilitated better access to services, but could be better tailored to serve all populations. “Once they got into a room, we finally saw their needs being met. A lot of medical providers would tell us that this was the longest period of time the person had been inside and consistently accessing services.” Participants were able to more easily access necessary health care and stabilize, especially those who traditionally have less success in a congregate setting. The PRK model could be better tailored to work for some populations, such as survivors of domestic violence and people with higher health care needs. Once participants were placed in a room, the PRK model emphasized providing more services to the sites than congregate shelters traditionally do. As described earlier, Alameda County contracted to provide a variety of health care and other services for participants. While not every site had the same services available at all times, providers said the PRK model helped stabilize participants and provide them with much needed support. However, some pointed out that large, low-barrier shelter sites do not effectively serve all populations of people experiencing homelessness and identified the need for more tailored solutions as well. According to county staff, the health care services at PRK sites were designed to help improve specific participant outcomes: health-related quality of life, chronic disease management, service utilization (such as the emergency department or inpatient care), and connection to benefits and services. At a descriptive level, clinical teams comprised of nurses, medical assistants, and the PRK medical directors, completed nearly 700 comprehensive assessments with participants. These led to connections with appropriate health care, health coaching on 2. Lessons Learned - 33 - chronic disease management, and advanced care planning. Additionally, they enrolled 92 residents in Medi-Cal, connected 89 with a medical home to provide comprehensive health care, and prevented around 100 avoidable ER visits, among other activities. Once COVID-19 vaccines were available, these teams administered over 600 vaccinations to PRK participants and staff, with even more happening beyond the study period. Providers also spoke to how they saw participants change throughout the course of their stays, discussing the importance of this care for participant stability as well as the need to tailor the PRK model to serve some specific groups. Accessing Care + Creating Stability While offering health care and other services to participants was not new for providers, the PRK model committed more resources toward stabilizing participants than other shelter models. As a baseline, the county was able to fund nursing, caregiving, housekeeping, transportation, and three meals a day, which are rarely available in a congregate shelter setting. Building these services into the program allowed participants to access resources they may have needed but could not get before. One provider explained that many of the participants at her site had not been receiving adequate health care for a long time, but their entry into PRK helped address that. “We saw people come inside who have been on the streets for decades or haven’t seen doctors or had appropriate serves for so long,” she explained. “Once they got into a room, we finally saw their needs being met. A lot of medical providers would tell us that this was the longest period of time the person had been inside and consistently accessing services.” This impression is confirmed by the data available on participants’ self-reported health conditions. Figure 4 below demonstrates that PRK participants reported higher rates of physical disabilities, chronic conditions, and mental or psychiatric illness than those in the population of people experiencing homelessness overall. Figure 4: Comparing Participant Self-Reported Health Conditions 24%26% 39% 65% 73% 59% Physical Disabilities Chronic Health Problems Psychiatric / Emotional Conditions 2019 PiT Count Safer Ground Participants Evaluating Project Roomkey in Alameda County - 34 - Providers attributed at least part of this success to people having their own space that they could return to for a longer period of time. Another provider described a transformation they saw in the stability of participants at their site overall: “Something about having your own space is really impactful. It gives [participants] time to think what their next steps are and go back to their old selves. When we opened in May many people didn’t have anything. They were just off the street, hadn’t taken a shower, didn’t have any extra clothes. Between May and now there has been a real change: people are definitely more stable than when they started.” Providers also said that giving participants a consistent address and access to reliable transportation made a big difference in helping them receive the services they needed. “I can’t tell you how important having transportation has been for us,” one said. “People can go to their appointments without having to pay or find a bus or a ride.” Another explained that pharmacies and physicians would call their site every day to confirm a participant lived there, since they had never had a consistent address at which to receive their medication before. One site manager described a participant who is one of the highest users of emergency services in the county. This participant’s mental health issues had made it difficult for them to maintain positive relationships with other congregate shelter participants in the past. “We have someone at [our site] who hasn’t stayed in a shelter for more than a few days before, but now he’s been inside since September,” she said. “Just the ability to come back consistently and have his own space has greatly improved his ability to access services.” Summarizing their view of the success of the PRK model in reaching those not typically served well by shelters, one provider noted, “People with severe mental health challenges are not usually successful in congregate [shelter]. The biggest success here is that we took a kinder, more empathetic approach to really work with individuals that had mental health issues and weren’t so quick to exit people.” Another provider echoed the advantage of getting to work with participants for more time, saying, “People are more relaxed and calm, getting three meals a day, the environment itself is much more respectful. With [congregate] shelter you are so busy just trying to work with people in an environment that isn’t fit to live in. Here we have a real opportunity to work with people.” An Alameda County staff member affirmed this perspective, saying the PRK model was designed to give participants space to worry less about their immediate needs and think toward the future. “I think there is something about people being able to have their own space, with a lockable door and a bathroom,” she said. “People get a chance to set down their burdens and not have to focus on keeping themselves together. It allows them to focus on themselves in a new way and reach for a different future.” Lessons Learned - 35 - Additional Services Crucial to Success Aside from health care and housing, many providers called out the importance of having access to seemingly more basic services such as meals and transportation. One provider said specifically that the ability to have reliable transportation to take clients to and from their myriad appointments was invaluable. If a goal of shelter is to help stabilize participants by getting them access to public benefits and medical care, which often require multiple appointments at varying times of day, it may be useful to continue providing transportation options. Additionally, multiple providers said that being able to provide three meals a day meant that participants didn’t have to worry about where they would find money to buy food or wait in line for free meals elsewhere. One staff member who helped oversee health care resources at the PRK sites said that she could see a significant change in participant health even before they had accessed medical services. “It is so powerful to give people food and shelter,” she said. When asked what services they would have wanted more of, most providers noted the need for more mental health care workers. While some sites had access to licensed social workers, others didn’t have consistent behavioral health resources, which they said many participants needed. Additionally, while the larger hotels had regular access to nursing staff, staff at other sites said a more frequent presence of nurses or doctors would have helped participants. A few providers also mentioned that working with volunteers not accustomed to serving people experiencing homelessness was challenging and required staff to help oversee that participant experiences with these services were positive. Last, two providers said that caregiver services were invaluable, but that they were unable to provide them for every participant who needed them. The services provided as part of the PRK model created a space where participants could reliably access the care they needed, however providers explained that the model may need to be better tailored to effectively serve some groups. Communities in Need of More Tailored Support The ability to access reliable services and remain in one place with no time limit was clearly a benefit to participants’ stability. However, providers had varying experiences when it came to actually engaging with all participants. At some sites, providers said their physical environment made it easy to engage: “Here at the site we see people every day. If there is something they need, we can easily call and talk to them. Everyone who passes our office will say hi. We are much more accessible than at previous sites.” Evaluating Project Roomkey in Alameda County - 36 - For others, however, the physical setup of the site, including the private rooms, created barriers to engaging some participants regularly. For sites without a central entrance space, providers said it could be difficult to keep track of participants coming and going, and therefore challenging to encourage them to engage in services. While providers described the negative aspects of congregate shelter, as noted above, some also acknowledged that the non-congregate setting could create added difficulties in engaging with some participants. “Because people got to stay inside their rooms, and there was COVID-19 outside, it was harder to engage people,” a provider explained. “They would just stay in their rooms and by the end some people had not really engaged at all because they either didn’t have to or had mental health challenges that prevented them from engaging.” While no providers stated they would prefer congregate shelters, it does demonstrate the need to identify new ways to engage participants in a non-congregate setting. Additionally, providers explained that the PRK model was not necessarily an improvement over congregate shelter for some groups. The groups mentioned most frequently were people who needed high levels of care (e.g. those with severe physical or mental illnesses) and survivors of domestic violence. While PRK was able to get these people into shelter, it was not necessarily designed to meet their specific needs. Some providers reported that they had participants who needed much higher levels of care than they were able to provide. “Sometimes we joke around and we run a skilled nursing facility, but without the training,” one provider said. “We have people that can’t get out of bed on their own, use the bathroom on their own, or eat on their own. So they need our help, but our staff are not trained or paid to be in-home care providers.” The presence of clinical teams on site was a significant help, but many providers reported their staff had to spend a lot of time taking care of people’s basic needs instead of focusing on long-term goals like their housing search. “Twenty to thirty percent of people require high-touch services,” one provider said. “We don’t have the capacity to be checking in with so many people all the time. It takes so much just to monitor and help with the basic survival stuff.” While this was a significant use of staff time, it also created problems at sites where people were living with a roommate. As a provider at one of these sites stated, “Anyone who needs a higher level of care, who cannot take care of themselves or has an extreme mental health situation, isn’t served well here.” Two providers reported trying to connect participants with in-home supportive services (IHSS) through the state, but being unable to access the resource during COVID-19. “IHSS can be the difference between people making it here or not. It is such an invaluable resource,” another provider explained. “But there is a lot of paperwork and not a lot of flexibility with timelines.” Lessons Learned - 37 - In the meantime, the county did attempt to connect people with these services by contracting with caregivers for participants who were applying for long-term IHSS care. According to county staff, though, there were not enough of these services available at the beginning of the program to meet the needs of the participants. In more recent months, they have reconfigured the sites to focus one hotel on serving those with the highest health care needs and have ramped up caregiver services there. According to this staff member, this kind of support can be crucial for people coming out of homelessness, but is not provided in most shelters. “These people couldn’t have stayed in shelter without this support,” she said. “They would have been kicked out.” Additionally, this staff member added that the PRK model may help fill an important gap in our current slate of shelter services. “Some of the largest gaps we see are individuals who are institutionally frail, but would prefer not to be in an institutional setting,” she said, referring to people who may be better served in a skilled nursing facility or medical respite center, but choose to remain outside because they do not feel comfortable in those places. “They are happy to be [in Project Roomkey] because they have a similar level of freedom, are treated respectfully, and can get the care they need.” While serving people with high medical needs was challenging for providers, the available data does not demonstrate that any particular condition made participants statistically less likely to be successful at a PRK site. If a certain group were particularly unsuccessful in this program, they would likely have higher exit rates to non-housing destinations (such as medical facilities, other shelters or back to homelessness). However, looking at participants who exited to non-housing destinations, there is no significant difference based on self-reported physical or mental health conditions. This speaks to the possibility that, with the proper level of services available, the PRK model may be better poised to effectively stabilize and house people with high medical needs. Another group for whom the PRK model may need to be more tailored was survivors of domestic violence or people with other significant trauma. While there was one hotel dedicated to serving this population, it has since closed. Since it was a smaller site, all of the participants were able to move to another PRK site if they wanted to, but moving to a larger site created new challenges for some participants. “Some of the clients we had stabilized went to another hotel and relapsed,” one provider said. “They are used to living a life where they have a small world that they have created to protect themselves. If we want to have a hotel that is super low threshold, we should have that, but we need to have another space for people that need more protection.” This provider extended this lesson to other groups as well: “When you design a system based on who the most homeless people are—single men—you miss being able to serve other people. We have to stop and think, ‘What about the families, what about domestic violence survivors, what about youth?’ We need to go down the checklist and ask ourselves what we are going to do for those populations.” Summary Overall, providers felt that the PRK model was a significant improvement over traditional shelters in stabilizing participants and improving their health outcomes. They thought Evaluating Project Roomkey in Alameda County - 38 - future shelter provision should include access to these comprehensive services as well. However, some discussed the need to have more tailored options for groups that may not be successful in a large, low-barrier, non-congregate setting. Increasing Access to Housing The PRK model nearly doubled the number of participants exiting to housing, while also posing challenges to facilitating those exits. “People are going to pick apart the data for years to come… But the bottom line is I don’t think there’s ever been a situation where over 400 people have been housed in this fast of a period.” 65% of PRK participants who exited PRK entered housing, as compared with 35% in congregate shelter last year. Some participants were initially hesitant to move out of PRK sites, but most elected to move into housing once the temporary nature of the program became clear. Possibly the most striking outcome from Alameda County’s Project Roomkey is the proportion of participants who moved from the shelters into housing. Even though connection to housing was a secondary goal of the program—given its focus on participant health outcomes—the PRK model nearly doubled the percent of participants leaving to more permanent housing. While some providers reported challenges in motivating participants to move out initially, nearly all of them stated that this model housed more participants than they had seen in past programs. More Participants Exiting to Housing During the period of study, 65% of the 815 participants who left the program went to a housing destination (see Table 4 below). As a comparison, according to HMIS data on exits from emergency shelters in Alameda County from the prior year (April 2019 - March 2020), around 35% of participants exited to housing. For detail on the HMIS destinations included in these categories, see Appendix A. 3. Lessons Learned - 39 - Table 4: PRK Participant Exits by Category Exit Category Participants Percentage Housing 532 65% Place Not Meant for Habitation 104 13% Shelter 71 9% Other/Data Not Collected 50 6% Deceased 25 3% Medical or Treatment Facility 24 3% Jail 9 1% Total 815 100% Around 13% of Project Roomkey participants exited to locations not meant for habitation (e.g. tents, RVs, etc.) and 9% exited to other shelters. While these numbers are much lower than usual for congregate shelters, future research would have to speak with participants to understand the challenges that led them to non-housing destinations. Further, PRK residents who were exiting to housing accessed potentially more stable public subsidies at much higher rates than those in shelters in prior years (See Figure 5 below). Only around 29% of participants from congregate shelters in Alameda County accessed public subsidies in the year prior to PRK, while 43% left to live with friends or family. In comparison, nearly three-quarters of PRK residents used a public subsidy and only 11% stayed with friends or family. Figure 5: PRK Participant Exits to Housing by Destination Public subsidies tend to be a more reliable form of housing as they often come with support from a service provider, so a participant’s outcomes are monitored more closely. Exits to live with friends or family, while possibly beneficial, are harder to 74% 13% 10% 3% Public Subsidies No Ongoing Subsidies Staying with Family or Friends Other Destinations Evaluating Project Roomkey in Alameda County - 40 - monitor long-term. Future studies should follow up with participants and program administrators to understand the true retention rates of participants who left to these various destinations. Exit Demographics HMIS data shows that the racial and ethnic breakdown of participants exiting from PRK appears generally proportional to their representation in the program overall. Further, among participants exiting to housing, no statistical difference could be detected between the composition of the racial and ethnic groups (see columns 1 and 2 in Table 5 below). This indicates that once participants entered PRK, their housing outcomes were relatively equal. Table 5: Exits to Various Categories by Race and Ethnicity combined Race and Ethnicity All Participant Exits (1) Exits to Housing (2) Exits to Places Not Meant for Habitation (3) Exits to Shelter (4) Black or African American 43% 44% 36% 46% White - Non- Hispanic/Non-Latinx 28% 29% 34% 10%** White - Hispanic/Latinx 13% 13% 11% 18% Multi-Racial 8% 7% 7% 14%* Asian 3% 3% 6%* 1% American Indian or Alaska Native 3% 2% 4% 7%* Native Hawaiian or Other Pacific Islander 2% 1% 3% 3% Total 100% 100% 100% 100% Table 5 utilizes two-sample t-tests to test whether the percentage of people from a particular racial or ethnic group who left to a specific exit category (columns 2, 3 and 4) is statistically different from that group’s representation among participants exiting PRK overall (column 1). Those marked with * are statistically significantly different with 90% confidence. Those marked with ** are statistically significantly different with 95% confidence. The racial category “White” is split into two groups based on the participant’s ethnicity. Exits to non-housing destinations like places not meant for habitation and shelter, however, do see statistically significant differences by race. Participants identifying as Asian are overrepresented in exits to places not meant for habitation. Participants identifying as Multi-Racial and American Indian or Alaska Native are overrepresented in exits to shelter, while those identifying as White - Non-Hispanic/Non-Latinx are Lessons Learned - 41 - underrepresented. These differences are all significant at either 90% or 95% confidence and should be investigated further to understand why certain racial and ethnic groups have disparate non-housing outcomes. Multiple Factors to Housing Success One reason providers gave for the high percentage of exits to housing was the addition of funding from the federal ESG-CV program. As previously stated, Alameda County used these funds to create hundreds of new bridge housing subsidies for Project Roomkey participants. Through this funding the county was able to subsidize twelve- month housing placements for participants moving out of a PRK site, with a commitment to continue funding the subsidies beyond the first year. Additionally, the county participated in California’s 100-Day Challenge which provided a structure to collaborate with the cities of Oakland, Berkeley, and Alameda to align funding around housing exits. Using the ESG-CV funds, the county contracted Abode Services to provide a centralized housing navigation team that assisted participants across sites in working with landlords and finding available housing in apartments across the county. To help ensure participant success after exiting to housing, the county also funded 11 housing providers to offer ongoing tenancy support. The data shows that 217 Project Roomkey participants were housed using this specific subsidy between November 2020, when it became available, and March 2021. This represents over 40% of the total exits to housing overall and was therefore clearly a significant factor in the program’s high housing exit rate. Many other participants exiting to housing with public subsidies utilized the county’s standard process to access permanent supportive housing. Some providers affirmed that the resources for these new subsidies as well as the work of their housing navigators was crucial in helping participants find housing. As one provider said, referring to the new subsidies, “We would never have been able to house so many people if not for Abode and the resources they had access to.” Another provider described the addition of these resources as unprecedented and said, “I’ve never seen so many people with housing options. In shelter you’re lucky if 25% of people have real housing options—not just living with relatives. I think the infusion of housing resources has been awesome.” Another provider described what it was like to see so many of her participants given the opportunity to access housing once it was announced that the site was going to close. “It was magical and intense,” she said. Providers, however, were more mixed on whether the PRK model, specifically, was better at preparing participants to move into housing, or if it was the newly available subsidies alone. According to one, the ability to give participants time to stabilize and get to know shelter staff was crucial in creating a successful housing search. “Having people in one place and having time helps,” she said. “The time-limited shelters can be Evaluating Project Roomkey in Alameda County - 42 - hard to get people stabilized and focusing on housing.” Another provider, however, was less sure that the PRK model had an effect on housing outcomes since most of the participants at her site who found housing got access to these new subsidies. One way of understanding the impact of these subsidies on housing exits overall is to examine what the rate of exits to housing would have been without them. Even when the participants that used this subsidy are removed from the sample, the rate at which those who exited PRK sites entered housing was 53%. As stated above, this is still significantly higher than exits to housing from congregate shelters. Therefore it appears the PRK model, with its increased housing navigation and case management, may have had an effect independent from this new subsidy. One Alameda County staff member observed that, at the very least, the PRK model made it easier for participants to access housing simply by bringing them inside. When people are eligible for supportive housing, but are not living in a shelter or other indoor setting, she explained, “it’s really hard to get them through the process.” Since the PRK model was appealing to people who may not have chosen to go into congregate shelter, it made it easier for them to get access to the housing they were eligible for. If they hadn’t been in a PRK site, she said, they may have lost that opportunity. While more research is needed to extricate the effect the PRK model had on participant housing outcomes from the new subsidies, one provider made sure to emphasize the overall success of the program. “People are going to pick apart the data for years to come,” she said. “But the bottom line is I don’t think there’s ever been a situation where over 400 people have been housed in this fast of a period.” All the providers agreed with this sentiment and felt that the infusion of new funding for housing subsidies was central to the program’s success. “A shelter program is only as good as the availability of housing on the other side,” one provider said. Since this study was conducted while Project Roomkey was still in operation, more time is needed to understand the retention of participants in their housing. While county staff had not heard of participants dropping out of the subsidy programs they fund, this is an important area for future research. Understanding retention rates among Project Roomkey participants would help illuminate whether people receiving the PRK model are more prepared to stay in their housing than those coming from congregate shelter. Finally, one important source of permanent housing not captured in this data is Project Homekey. Project Homkey is a follow-up initiative to Project Roomkey that provides state funding for local governments to buy hotels, enabling them to convert the rooms into permanent housing units. While these sites were not ready to be used during the study period, they are likely to be a growing source of permanent housing exits for PRK participants going forward. While nearly all providers affirmed that more participants left PRK to housing than in prior shelter programs, some also discussed challenges it posed in moving people out of the sites and into housing. Lessons Learned - 43 - Challenges Facilitating Exits Despite the significant increase in the number of participants getting access to housing through Project Roomkey, providers noted that the PRK model created some difficulties in facilitating participant exits. Since people had their own rooms, access to numerous services, and three meals a day, without paying any portion of their income, many providers reported that some participants were understandably hesitant to leave, even when they were offered housing. “People don’t always have a huge motivation to leave because it’s comfortable, safe, and free,” one provider explained, emphasizing that subsidized housing typically has participants pay 30% of their income to rent. “They have staff available 24/7, maybe even partners or romance too. They have created a community and family here on site.” Another provider said that moving from a service-rich environment like PRK to living largely on their own, can be daunting for participants. “We see how much support people need and they can get it at our sites because they are designed to do that,” she said. “So to move from here to a place where they aren’t getting that support can be very challenging for people—and rightly so.” An added challenge was that not all housing options participants were offered were for individual units. Some were for shared housing where participants would be matched with others to share a multi-bedroom home or apartment. One Alameda County staff member explained, “When you’re leaving a situation where you have your own room...and you’re going to shared housing, and paying some percent of your income, it is a really hard transition.” Many of the providers noted that while this could be frustrating, it was understandable, acknowledging that they themselves wouldn't necessarily feel excited about moving from a room of their own to living with strangers. This challenge existed in other counties as well. “People liked the space more than congregate shelter,” an LA County provider said. “Some people come into congregate sites and say it’s so awful they’ll do anything to get out. This isn’t true in a non- congregate shelter setting.” However, she explained that those initially hesitant to leave shelter will eventually make the transition. “In the end, people will leave the hotel when it’s time and we will do our damndest to make sure everyone has a place to go.” Most of the Alameda County providers who reported these challenges shared a similar view about participants’ eventual participation in their housing search. One reported that when the county told them the site would close, the temporary nature of PRK became clear to participants and they were more willing to move to housing. “When we thought we were going to close the site in February, suddenly everybody wanted a case management appointment,” she said. “It’s hard when you don't have an end date to get people motivated to have a plan.” Evaluating Project Roomkey in Alameda County - 44 - Another provider elaborated on this point, saying that the focus of the program at first was to get as many people inside as possible, so expectations around participants looking for housing weren’t set up from the start. “Project Roomkey was challenging because residents weren’t starting off with the expectation that they would have to participate actively in their housing plan,” he said. “Shifting the focus from getting everyone into the hotels and being safe, to going out and looking for housing was hard.” He added, however, that the funding for the new bridge housing subsidies wasn’t available at the start of Project Roomkey, so expecting everyone to find housing at that time would have been unrealistic. Nevertheless, providers clearly had a challenging responsibility to create a space that was more comfortable and effective at stabilizing participants than traditional shelter, while also helping motivate them to find housing and move out of the sites. “It feels like two tensions at the same time,” one provider said. “You want to create that respite from the strain of sleeping in bunks in a room, but you also want to make it clearly an interim environment, so people are motivated to find housing.” The timing of PRK exits backs up the idea that participants were eventually willing to move to housing. In November 2020, the county prepared to close all PRK sites due to the lack of long-term program funding. Additionally, this is when the bridge housing subsidies became available and there was a push to find housing for participants. The data shows that after this exits to housing more than doubled for the next two months (see Figure 6 below). While most hotels ended up being able to remain open, this demonstrates that when the temporary nature of the program became clear and more housing resources were added, participants were willing to move out of PRK sites. Figure 6: Exits to Housing by Month 0 3 11 14 20 14 17 33 60 121 101 59 79 0 50 100 150 Ma r c h 2 0 2 0 Ap r i l 2 0 2 0 Ma y 2 0 2 0 Ju n e 2 0 2 0 Ju l y 2 0 2 0 Au g u s t 2 0 2 0 Se p t e m b e r 2 0 2 0 Oc t o b e r 2 0 2 0 No v e m b e r 2 0 2 0 De c e m b e r 2 0 2 0 Ja n u a r y 2 0 2 1 Fe b r u a r y 2 0 2 1 Ma r c h 2 0 2 1 Ex i t s t o H o u s i n g Housing push begins Lessons Learned - 45 - This was further confirmed by a provider at a trailer site where most participants had roommates. “[This site is] different from a hotel because a trailer feels more temporary, so they don’t have the same feeling of just wanting to stay,” she said. “They’re living in a parking lot and they know it’s just a program that could close any time. People feel secure right now, it’s better than being out in the street, but at the same time they know they have to move on. Anytime there is a waitlist opening [for housing] they really come out to apply.” While the providers at trailer sites didn’t necessarily think the trailers were preferable to hotels, this points to a need to find ways to provide shelter environments that both allow participants to feel comfortable and stabilize, and focus on finding permanent housing for everyone. Providers named multiple ways to address these “dual tensions,” including setting better expectations from the start, creating some time limits for participant stays, or offering more appealing and varied housing options. Summary The PRK model nearly doubled the percent of participants exiting to housing as compared with traditional congregate shelter. This increased access to housing was due to added resources, including new subsidies and funding for housing navigation. While some participants were hesitant to leave PRK sites at first, once more housing options were available and the temporary nature of the program became clear, most were willing to move on. Improving Collaboration The PRK model created new partnerships and coordination between homeless service providers, despite challenges with quick program implementation. “We’ve seen this past year that the local, state, and federal government can really change policy fast. We’ve been able to create things that would have taken decades to do otherwise.” PRK created a chance for new collaboration among service providers and county staff. PRK started quickly and was high intensity for staff, but overall was a successful mobilization of resources. The last theme consistently highlighted by service providers taking part in Project Roomkey was that the program created opportunities for providers, government agencies, and other organizations to work together in new, more effective ways. While 4. Evaluating Project Roomkey in Alameda County - 46 - providers faced significant staffing issues due to the uncertainty of the program’s timeline and the significant needs of the participants, most agreed that Project Roomkey was a successful response to the pandemic and presents opportunities for future improvement to the county’s homeless service provision. New Collaborations According to providers, a major benefit of Project Roomkey’s design was that it created new channels of communication between their organization and the other nonprofits or government agencies they work with. Because there was such a fast mobilization to get PRK sites up and running, Alameda County staff were highly involved in meeting with all the service providers individually as well as regularly bringing them together. Specifically, providers found the connection between the health care and housing sectors to be particularly important to participant success. “We’ve had a lot of support from key players in Alameda County,” one provider stated. “The medical directors have been very hands on, which has been integral to getting access to a lot of services and answering important questions.” They went on, giving an example of one of their participants who isn’t technically qualified for a housing subsidy yet, but needs one. “The coordination between these groups helps us fill in all the cracks that this person would otherwise fall through.” Another provider agreed PRK fostered a new level of collaboration, which was a change from the way they had worked before. “The teamwork was incredible,” she said. “The staff that interacted with my staff were so nice and supportive. Even when people were exposed to COVID-19 and had to move to another hotel and come back, it was seamless.” Yet another said that the presence of health care staff was the biggest improvement in the PRK model. “I’ve been doing this for 27 years and we finally got regular access to county departments we didn’t have before,” she said. “Having a relationship with leadership of county health care agencies and the staff there was a game changer.” Multiple providers also reported that having medical professionals and nursing staff present made it faster to get necessary documentation participants need for housing. Clinical teams reported completing 259 Verification of Disability forms for SiP hotel residents. This can be particularly important, as having this documentation can make the difference in whether or not a participant can access housing. “Having health care staff on site made it easier to get a verification of disability, social security cards, and other documents,” one provider said, adding that she hoped this collaboration would remain even after Project Roomkey ends. Providers did report some challenges in accessing critical government services, such as IHSS or the DMV, due to offices being closed during the pandemic. However, some reported that the new collaborations resulted in opportunities to make homeless service programs less rigid. “A lot of organizations have bent over backwards to work with [PRK Lessons Learned - 47 - sites] and be flexible on what documents our clients need,” one provider said. She went on to explain that the Housing Authority of Alameda County allowed participants to receive housing subsidies so long as they were working on getting their proper documentation together, a change from prior requirements that meant many more people could get housed. Another provider said that Project Roomkey forced them to collaborate in new ways to meet the diverse needs of the participants. “Since homelessness affects people of all ages and races, we brought in experts from our aging unit, family support team, substance use disorder team, and street health outreach team, who hadn’t necessarily worked together in this way before,” she said. County staff agreed that Project Roomkey allowed for new forms of collaboration and the ability to make change faster than ever before. “We’ve seen this past year that the local, state, and federal government can really change policy fast,” one staff member said. “We’ve been able to create things that would have taken decades to do otherwise.” While most providers mentioned that they would have liked more consistent communication from the county through the program, they all acknowledged that the county faced immense challenges in starting a program so quickly and with such little certainty from the federal government. Rocky Start Up, but an Overall Success Nearly every provider discussed the challenges they faced in starting their PRK sites with such a short timeline. Many providers had to ramp up staffing and begin work in a matter of weeks or even days. One described the creation of their site saying, “frankly it was quite amazing how fast the program got up and running.” Another stated, “I was basically hiring seven days a week for 4 months straight to get us where we are today.” Since Project Roomkey was a quick reaction to an unexpected pandemic, providers generally felt these challenges were unavoidable, however many described the issues they faced in training and supporting staff appropriately once the program was going. With the need to hire up so quickly, many providers struggled to bring on new staff members that were able to handle the intensity of a program like Project Roomkey. One provider explained, “We were all new. There was so much learning that had to happen in addition to the service provision.” Since most sites were low-barrier and provided participants with more autonomy, newer staff often needed support in learning how to engage with the community. “A lot of shelter monitors we hired had never worked with this population before,” another provider said. “We try to have at least one experienced staff person on every shift, but we had to teach them evidence-based practices fast.” Evaluating Project Roomkey in Alameda County - 48 - One provider explained that the non-congregate setup of their site created some difficulty for new staff. “It is more challenging to adequately train and oversee staff because participants are much more spread out,” she said. “There are so many rooms and a lot of things that can go wrong.” Even with staff in place, some providers said more resources were needed in supporting them through the day-to-day operations of the site. One explained that since staff had to go into people’s rooms to engage them, they could find themselves in many sorts of situations like helping clean up someone’s room, assisting them with necessary tasks like going to the bathroom, and dealing with overdoses. “It’s really traumatizing to know you may find a dead body in a room one day,” she elaborated. “People don’t necessarily understand that. It weighs on our staff and causes a lot of other related issues to be in that high-stress environment all the time.” While she acknowledged that this is part of their work as a service provider, she also said that more resources for training and supporting staff were needed from the county to effectively run a low-barrier environment like Project Roomkey. In addition to the high-intensity situations staff could find themselves in, providers reported that the uncertainty of Project Roomkey’s timeline made it difficult to retain staff. The uncertainty of the pandemic caused by COVID-19 case rates and spread, the term of the federal emergency declaration, and funding availability meant that the county could never be sure how long the program would last. Even though providers understood this, they explained that not knowing when their sites would close made it difficult to keep staff. “The uncertainty was hard - only being open month to month to month,” one provider said. “This made it hard to hire people and keep staffing operations going.” Another provider said they were told a few different times to prepare to close their site, but each time the dates got extended so the site didn’t have to close. “We started losing staff because we kept pushing back the date we were going to close,” she said. “If we had known we were going to stay open, we could have kept people.” Many providers described losing higher-level staff to more permanent positions inside and outside their organization. With the Biden Administration’s commitment to continue reimbursing localities for this work through September 2021,51 this uncertainty is likely less of an issue going forward. Providers did say, however, that continuing their new level of collaboration with others involved in county homeless services as well as receiving the appropriate resources for training staff to work in PRK-style shelters, will be critical to the success of future programs. Summary Project Roomkey provided a chance for service providers and government agencies to work together in new, collaborative ways that improved service provision at the sites. Providers faced some challenges with hiring, training, and retaining staff in a program that included so much uncertainty, but hope to keep up these new levels of collaboration in future programs. Bringing the PRK Model Forward: Conclusion and Recommendations - 49 - Bringing the PRK Model Forward: Conclusion and Recommendations Conclusion Alameda County’s Project Roomkey program was a strikingly fast mobilization of resources to protect the people experiencing homelessness most vulnerable to the COVID-19 pandemic. In only a matter of months, the county more than doubled its available shelter beds and created a new model of non-congregate shelter provision. Based on the data available, the PRK model appears to have been more appealing to those who may not have wanted to use shelter before, helped participants stabilize through connection to appropriate health care services, put many more on a path toward permanent housing, and facilitated new collaborations between homeless service providers. This report demonstrates that with significant state and federal investments, clear outcome goals, and coordination among government and service providers, real progress can be made in addressing the homelessness crisis. At the start of 2020, Gavin Newsom proposed the idea of using hotel conversion to augment the state’s pace of permanent supportive housing production. Just one year later the state had tripled that production through hotel conversions, due to the funding and coordination propelled by the pandemic.52 The PRK model was able to quickly address many of the barriers to using congregate shelter that people experiencing homelessness have long identified. One Alameda County staff member described how seeing Project Roomkey unfold so quickly has changed the way she thinks about the possibility of meaningfully addressing homelessness, saying, “Normally how I think of my work is relentless incrementalism. If we have the right infrastructure, though, it’s clear we can make changes in leaps and bounds.” The PRK model, however, comes at a significant financial expense. At a nightly cost per person that is multiple times higher than other forms of shelter, continuing the PRK model at the current scale may not be the most sustainable or appropriate use of county funds without continued federal reimbursement. Additionally, as COVID-19 cases decrease and travel reopens, many hotels will likely want their rooms back, leaving the county with fewer spaces available for non-congregate shelter. In an attempt not to lose ground on addressing homelessness once funding and space run out, the state has allocated $800 million to buy hotels and convert them to permanent supportive housing. This follow-up program to Project Roomkey, called Project Homekey, has already led to the purchase of 94 hotels with over 6,000 units across California at a much lower cost than building new supportive housing.53 Using Evaluating Project Roomkey in Alameda County - 50 - these funds, Alameda County has already purchased two of the PRK sites to convert into permanent housing. While it is critical to create as many permanent housing opportunities as possible, the county will continue to operate shelter in some capacity to address the immediate needs of people experiencing homelessness. Therefore, this report has attempted to shed light on which pieces of the PRK model can be replicated in other programs and at what cost—either financially or to the success of participants in the program. The following section attempts to synthesize these lessons into three complementary recommendations. While this is not an exhaustive list, it is intended to begin a conversation about how the county can improve upon current shelter operations in order to best serve people experiencing homelessness. While creating a more financially sustainable program is likely necessary for the county, it is important to do so without sacrificing the quality of shelter services and leaving participants with interventions that do not serve them well. Therefore, it is important to be clear about the potential tradeoffs involved in adjusting the PRK model. In order to do so, discussion of these recommendations will be organized around the following criteria: ● Effectiveness: Does this recommendation help stabilize people experiencing homelessness and put them on a path toward long-term housing? What are some key components to consider in implementing the recommendation? ● Reach: Who does this recommendation serve and who does it leave out? ● Cost: What is the relative cost of this recommendation? At what scale could it be implemented? Recommendation 1 Maintain the PRK model at a smaller scale focused on people with high medical needs who would not be able to stabilize in congregate shelter but do not feel comfortable in other institutional settings. Key Components Non-congregate shelter offering increased privacy for participants. Low-barrier design allowing participants to bring their partners, pets, and belongings with them, with opportunities for community building Nursing and caregiver services focused on stabilizing participants’ immediate medical needs and connecting them with long-term care. Housing navigation from an organization other than the hotel operator, dedicated to finding appropriate long-term housing options. Longer participant stays to give time for stabilization, with an emphasis on housing search and temporary nature of the shelter. Access to regular transportation and meals. Bringing the PRK Model Forward: Conclusion and Recommendations - 51 - Even with the end of Project Roomkey, it is possible that there are some hotel owners interested in continuing to lease their rooms to the county as shelter. If this is the case, the county could continue providing non-congregate shelter and services in the PRK model while adjusting the program to focus on a higher-needs population that is not currently served well in a traditional congregate shelter setting. Effectiveness As this report demonstrates, the PRK model was able to fill a gap in the slate of shelter models currently provided in Alameda County. Through its mix of low-barrier, non- congregate and on-site health care and housing services, it was able to make shelter more acceptable to people experiencing homelessness, help them stabilize, and put them on a path toward permanent housing. Since many providers and county staff noted that people with high medical needs are often not served well in congregate shelter, but do not want to be in an institutional setting like a medical respite center, the PRK model should be maintained to serve this group. The increased privacy and service level of the PRK model may create a more effective setting in which people with significant medical needs can stabilize and search for appropriate housing, without the added stress of staying in a congregate setting. However, even though Project Roomkey served an older and sicker population, if the PRK model is shifted to focus exclusively on this group, participant outcomes may change and would need to be monitored. By definition, this group would have significant unmet medical needs and therefore would likely show more significant benefits from the increased service level of the PRK model. If participants are sicker, though, they may need more time to stabilize and prepare to look for housing, which could decrease the number of participants served and exiting to housing each year. However, if a clear expectation is set from the start that the goal of this shelter is to help participants stabilize and find appropriate housing, people may be primed to move on more quickly. Additionally, if congregate shelters do not work well for this community, then the alternative is likely leaving them to live outside where it will be difficult to meet their health care and housing needs. If we want to make more people comfortable using shelter, help them stabilize, and prepare them to move into housing, the PRK model may be the best interim step for this group. Reach More research needs to be done as to what conditions to prioritize for people receiving this intervention. However, based on provider interviews it seems that people with a number of physical disabilities and/or mental or behavioral health diagnoses may need more support than they can currently get in congregate shelter, but may not feel Evaluating Project Roomkey in Alameda County - 52 - comfortable in an institutional setting. The PRK model could be continued in order to serve this population specifically. The tradeoff of this option is that it would serve a much smaller population of people experiencing homelessness in Alameda County. Those without significant medical needs who have been homeless for shorter periods, would likely not be prioritized into this setting. It also may not be able to keep larger families or communities together, as people’s needs within these groups may differ. Last, some people may find their needs are so serious that they do need to be in a more institutional setting in order to receive appropriate care. Cost At a rate of around $260 per person per night, the PRK model may be one of the more expensive shelter models the county utilizes. Since this recommendation focuses the PRK model on a smaller population, the costs will be somewhat contained. Based on the estimate of the PRK program costs, if the county was able to lease 50 beds for this program they could expect to serve about 65 people at a time (given that the number of individuals served was about 1.3 times that of the number of rooms available if participants can stay with their partners). At a per-night cost of $260, this would come out to an annual cost of about $6.1 million for the hotel lease, service providers, nursing and health care services, transportation and security, and miscellaneous costs. While these costs are high compared to congregate shelter, it is estimated that leaving chronically homeless individuals outside costs taxpayers about $35,578 annually, and putting them in permanent supportive housing can lower that by about 49.5% due to reduced utilization of high-cost institutions (e.g. emergency departments, inpatient hospital stays, jails, etc.).54 By this estimate, providing space for at least 65 people with high medical needs to leave homelessness, stabilize, and find permanent housing each year, this program could help save at least $1.1 million annually. Additionally, since this would focus on a population with high medical needs, it may be possible to utilize health care-related funding streams to expand the potential funding available for this program In order to decrease costs and serve more people, the county could consider putting individuals in these rooms with a roommate (someone outside their existing household). This would allow the sites to serve more people at a time and may make the environment feel more temporary, encouraging people to actively seek other housing opportunities. However, some providers warned that roommate relationships can cause tension for participants and take significant staff time to manage. Even when staff had done work to carefully match people with roommates, these relationships could deteriorate after spending too much time in such a small space together. Additionally, Bringing the PRK Model Forward: Conclusion and Recommendations - 53 - this may not be successful for people who have mental or behavioral health challenges that make living with another person difficult. Another option would be to shorten the timeline people are able to stay at the sites. A strict timeline for participant stays might work against this goal and reduce the proportion of people exiting to housing, as providers identified it as being important in improving participant stability. However, if participants were required to actively participate in a housing search during their stay, this could help increase turnover within the sites while helping a large proportion of people access more permanent housing. The potential tradeoff of this strategy is that the lack of timeline for PRK participants seemed to help them stabilize before moving to housing, so this may not work as well for participants that need extra time to stabilize. Recommendation 2 Address drawbacks of congregate shelter by adjusting service design, including creating or expanding centralized teams to provide clinical care and housing navigation. Key Components Nursing services focused on stabilizing participants immediate medical needs and connecting them with long-term care. Mental health services to help address issues related to living in a congregate setting and prepare participants to live inside long-term. Housing navigation to find appropriate long-term housing options that will work based on participants’ medical and mental health needs. Added privacy within congregate shelters and keeping communities together when possible. Access to transportation and food. Since non-congregate shelter may be difficult to come by and the cost may be prohibitive to serve large groups of people, the county should focus on bringing the services that improved Project Roomkey participant outcomes to existing shelter operations. While none of the providers interviewed said they prefer a congregate shelter environment, many acknowledged the need for it and discussed how to add components of the PRK model to congregate shelter. Without a significant sustained investment in homeless services from the state or federal government, congregate shelters are likely to be the predominant mode of serving the immediate needs of people experiencing homelessness. However, providers reported that the bare-bones nature of a lot of those settings can be detrimental for participants and discourage them from utilizing the resource. In order to make shelter more acceptable to people experiencing homelessness, help them stabilize, and create more opportunities to find permanent housing, the county should create or expand existing centralized service teams to address participant health and housing needs. Evaluating Project Roomkey in Alameda County - 54 - Effectiveness Providers acknowledged the fact that many people have negative experiences with congregate shelters either due to conflict with other participants or the belief that their homelessness will not be resolved by moving into shelter. This is confirmed by the provider stories of participants who were asked to leave congregate settings and the data showing exits to housing are around 35%, with nearly half of those going to stay with friends or family. Therefore, investing in improvements to existing congregate shelters may be crucial to making them a more acceptable option for many people living outside and could help increase their ability to put people on a path toward long-term housing. In order to make congregate shelter an environment that people want to utilize, the county could try and create more privacy within congregate shelter, either by finding smaller spaces or by arranging large rooms differently to help give people a sense of having their own space. Additionally, they could designate some shelters for larger communities of people who want to stay together, or allow people to move in with their partners. These solutions do not fully address the drawbacks of a congregate environment as compared to non-congregate shelter, but could help to mitigate them. Additionally, since providers discussed the challenges congregate settings pose to participants’ mental and physical health, increasing the level of clinical services available could help provide people with support in addressing their own health challenges or dealing with conflict when it arises. Some of these health care services already exist in Alameda County congregate shelters. The county’s Shelter Health providers support shelter staff in accessing nursing care for residents with higher needs and behavioral health care services are provided at some sites. However, it could be useful to do an assessment of how evenly these services are provided across different congregate shelter sites in order to understand where they could be added or what specific services are missing (such as chronic disease management or connections to community-based care). This could help shift congregate shelters to be a healthier environment for participants. To address concerns about shelter not being effective at resolving their homelessness, teams of housing navigators could be utilized to assist congregate shelter participants. The housing navigation services in Project Roomkey were frequently identified as being important in helping participants understand the options available to them and help exit the sites successfully. While housing navigation was generally the responsibility of the service provider operating a site, many providers said this was a piece they could have used more help with. Since providers were focused on day-to-day needs of participants, some discussed the need for a group whose sole focus is helping with housing search. These teams could help increase exits to housing from congregate shelter and focus on getting people into more stable housing destinations. Bringing the PRK Model Forward: Conclusion and Recommendations - 55 - Last, it is clear from Project Roomkey that there are certain smaller pieces of the program that may have had an outsized impact on participant outcomes. Access to transportation and food were consistently referenced as being helpful to participant stability. Therefore, maintaining some level of transportation available for shelter residents to get to and from important appointments as well as providing more meals each day could make participants more comfortable in a congregate environment. Reach Since congregate shelters are less expensive and not focused on one population in particular, they have the potential to address the immediate needs of a large swath of people experiencing homelessness or focus in on one particular group. Additionally, since shelter spaces are large and potentially flexible, they could be used to serve communities or families that want to stay together. As discussed, congregate shelters may not be appropriate for people with significant physical or mental health needs and therefore the PRK model or other forms of shelter may be needed to better serve those individuals. Congregate shelters also may not be appropriate for serving people with significant trauma, so a non-congregate shelter focused on the needs of that population may be more effective. Cost While each of the components of this recommendation add to the cost of congregate shelter, they are relatively small in comparison to the cost of the PRK model and may be able to recreate many of its benefits for participants. In total the costs of the PRK model not related to the hotel operation (nursing and caregiver services, service agreements including housing navigation, transportation and security) added up to about $52 per participant per night. However, the actual costs of these services are likely lower. Since service agreements with providers included housing navigation as one piece of their work, it is difficult to estimate exactly how much that housing navigation cost, but likely less than the total reported here. Additionally, if people with higher medical needs are being served at separate sites using the PRK model, this would require fewer nurses and caregivers in congregate shelters, reducing that cost as well. Further, while transportation was reported as being useful for participants, congregate shelters likely do not need security in the same way, which would reduce that estimate. Finally, if these services were centralized and able to serve shelters across the county, this might result in an even further reduction to the per-night cost. Increasing the services provided at congregate shelters would no doubt increase the costs of service provision. However, if congregate shelters are intended to effectively serve people experiencing homelessness by helping them stabilize and resolve their Evaluating Project Roomkey in Alameda County - 56 - homelessness, a financial investment along with coordination may be needed to see improvements in those outcomes for participants. Recommendation 3 Continue utilizing state and federal grant funding while it remains available to purchase hotels and create long-term housing subsidies for people experiencing homelessness. Key Components Take advantage of the opportunity to use new funding streams and procure cheaper permanent housing. Connect new rooms or subsidies to shelter stays to allow for a more seamless transition into housing. Ensure long-term funding is available to maintain subsidies if state or federal funding runs out. While the majority of service providers running Project Roomkey sites found that the PRK model was important to helping stabilize participants and find them housing opportunities, most acknowledged that having housing subsidies available for every participant that wants one is critical. Without the subsidies the county created for PRK residents, many participants may have had to wait longer for housing opportunities or eventually leave the site back into homelessness. Even though maintaining some PRK-style shelter beds and improving congregate shelters are important steps in meeting people’s immediate needs, they are not long- term strategies to resolve homelessness. Therefore, the county should continue to take advantage of opportunities to use state and federal funds to purchase hotels or create additional long-term housing subsidies for people experiencing homelessness. Effectiveness Given the recent federal court order requiring Los Angeles to provide shelter to all unhoused residents of skid row55 and a similar proposal in San Francisco,56 an important conversation is occurring about the appropriate amount of funding to dedicate to shelter programs alone. While shelter is important in addressing people’s immediate needs, many advocates and service providers alike have worked to clarify that it is not a replacement for permanent housing.57 Purchasing hotels would allow the county to convert buildings into permanent supportive housing, creating many more long-term housing options for people living outside. Since building new supportive housing can be prohibitively expensive and take years, this is an important opportunity to add to the county’s stock of housing for people experiencing homelessness. Bringing the PRK Model Forward: Conclusion and Recommendations - 57 - If hotels are not available, the county can put funds toward creating scattered-site supportive housing, where participants are placed in apartments or homes across the county with long-term subsidies and services are provided to them in place. This is the method the county utilized in creating its bridge housing subsidies for PRK participants, and resulted in 217 participants finding housing in the first five months. Given the well-document efficacy of permanent supportive housing,58 this recommendation is possibly the most effective at creating stability and putting people on a path toward housing. The one caveat to using state or federal funds for this purpose is that they are often one-time grants. Therefore, it is important for the county to identify long-term sources of funding for any new subsidies created through one-time funds. Reach While permanent supportive housing can be an effective intervention for many people experiencing homelessness, it is going to take time and many more resources than are currently available to serve all those who need it. The county already has a system in place to prioritize people experiencing homelessness for permanent supportive housing opportunities, so access to these subsidies would likely continue through that system. This could be adjusted slightly, however, to support the success of the first recommendation to maintain the PRK model at a smaller scale. Since the population of people prioritized for the PRK model beds would have higher medical needs and likely longer histories of homelessness than the population overall, they may overlap with the group prioritized for permanent supportive housing. If new housing subsidies are tied to the PRK model site, not only would it ensure those participants have a place to go when they stabilize, but it would also help open up those beds more quickly for more people. Additional research is needed to figure out how closely those two groups overlap, but this could facilitate more movement through the shelter system in a way that makes it more of a temporary stop, instead of someone’s only option. Cost The impetus for this recommendation is the potential to purchase hotels that cost less than building permanent supportive housing as well as the availability of state and federal funds to do so. Since these hotels and funds may not be available for long, it is important that the county act now with the resources that are available and seize upon any future funding. Based on the hotels they’ve purchased so far, county staff estimate the cost to purchase and renovate the hotel rooms comes to under $200,000 per unit. This is significantly less expensive than building similar units in the Bay Area, which can cost Evaluating Project Roomkey in Alameda County - 58 - upwards of $500,000 per unit.59 Therefore, this is an important opportunity to make limited funds go further in creating permanent supportive housing. Also, permanent supportive housing can be relatively inexpensive to operate. County staff estimate the per-night cost of one unit of permanent supportive housing is around $100. Separately, Alameda County is spending about $15,000 per household per year for the bridge housing subsidies they created for PRK participants. This amounts to about $41 per household per night for the housing subsidy. This cost estimate is a fraction of the corresponding $176 lease cost for the PRK hotels. Therefore, while building permanent supportive housing can be one of the most expensive activities for continuums of care, maintaining it or creating long-term housing subsidies can be a more financially sustainable investment even compared to some shelter models. It is crucial, then, that the county take advantage of any opportunities available now to secure these spaces at a lower cost and utilize state or federal funding that may not be available going forward. Recommendations Summary In order to serve the short- and long-term needs of people experiencing homelessness, Alameda County should maintain the PRK model at a smaller scale to serve those with high medical needs that would not be served well in either congregate shelter or institutional settings. Additionally, they make congregate shelters more effective at resolving someone’s homelessness by adjusting their design to include components that made the PRK model successful—such as creating or expanding teams of health care and housing navigators. Last, the county should utilize the hotels and external funding available due to COVID-19 in order to add to its supply of long-term housing opportunities for people experiencing homelessness. Together, these recommendations can build off the success of Project Roomkey to improve the slate of services available to people experiencing homelessness in Alameda County. Appendix A: Data Definitions - 59 - Appendix A: Data Definitions The following appendix reports the detailed exit destinations selected in HMIS that were rolled up into broader categories for the sake of this report. These categories were defined in partnership with county staff to closely match how they already report these statistics. Table 4: Participant Exits by Category Housing • Foster care home or foster care group home • Host Home (non-crisis) • Owned by client, with ongoing housing subsidy • Owned by client, no ongoing housing subsidy • Permanent housing (other than RRH) for formerly homeless persons • Rental by client, with VASH housing subsidy • Rental by client, with RRH or equivalent subsidy • Rental by client, with HCV voucher (tenant or project based) • Rental by client in a public housing unit • Rental by client, no ongoing housing subsidy • Rental by client, with other ongoing housing subsidy • Residential project or halfway house with no homeless criteria • Safe Haven • Staying or living with friends, temporary tenure (e.g. room, apartment or house) • Staying or living with family, temporary tenure (e.g. room, apartment or house • Staying or living with family, permanent tenure • Staying or living with friends, permanent tenure • Transitional housing for homeless persons (including homeless youth) Place Not Meant for Habitation • Place not meant for habitation (e.g., a vehicle, an abandoned building, bus/train/subway station/airport or anywhere outside) Shelter • Emergency shelter, including hotel or motel paid for with emergency shelter voucher, or RHY-funded Host Home shelter • Hotel or motel paid for without emergency shelter voucher Other/Data Not Collected • Client refused • Data not collected • No exit interview completed • Client doesn't know • Other Deceased • Deceased Medical or Treatment Facility • Hospital or other residential non-psychiatric medical facility • Long-term care facility or nursing home • Psychiatric hospital or other psychiatric facility Evaluating Project Roomkey in Alameda County - 60 - • Substance abuse treatment facility or detox center Jail • Jail, prison or juvenile detention facility Figure 5: Exits to Housing by Destination Public Subsidies • Permanent housing (other than RRH) for formerly homeless persons • Rental by client in a public housing unit • Rental by client, with HCV voucher (tenant or project based) • Rental by client, with RRH or equivalent subsidy • Rental by client, with VASH housing subsidy • Owned by client, with ongoing housing subsidy No Ongoing Subsidies • Owned by client, no ongoing housing subsidy • Rental by client, no ongoing housing subsidy Staying with Family or Friends • Staying or living with family, permanent tenure • Staying or living with family, temporary tenure (e.g. room, apartment or house) • Staying or living with friends, permanent tenure • Staying or living with friends, temporary tenure (e.g. room, apartment or house) Other Destinations • Foster care home or foster care group home • Transitional housing for homeless persons (including homeless youth) • Host Home (non-crisis) • Residential project or halfway house with no homeless criteria • Safe Haven References - 61 - References 1 Office of Governor Gavin Newsom. “At Newly Converted Motel, Governor Newsom Launches Project Roomkey: A First- in-the-Nation Initiative to Secure Hotel & Motel Rooms to Protect Homeless Individuals from COVID-19.” Press Release, April 3, 2020. https://www.gov.ca.gov/2020/04/03/at-newly-converted-motel-governor-newsom-launches-project- roomkey-a-first-in-the-nation-initiative-to-secure-hotel-motel-rooms-to-protect-homeless-individuals-from-covid-19/. 2 Twitter. “Office of the Governor of California on Twitter,” April 3, 2020. https://twitter.com/CAgovernor/status/1246154220685815809. 3 Office of Governor Gavin Newsom. “At Newly Converted Motel, Governor Newsom Launches Project Roomkey: A First- in-the-Nation Initiative to Secure Hotel & Motel Rooms to Protect Homeless Individuals from COVID-19.” Press Release, April 3, 2020. https://www.gov.ca.gov/2020/04/03/at-newly-converted-motel-governor-newsom-launches-project- roomkey-a-first-in-the-nation-initiative-to-secure-hotel-motel-rooms-to-protect-homeless-individuals-from-covid-19/. 4 Office of the Governor. “Letter to Project Roomkey Partners Re: Ongoing FEMA Approval,” December 18, 2020. https://www.cdss.ca.gov/Portals/9/FEMA/Gov-Ongoing-FEMA-Approval-Letter.pdf. 5 Conor Dougherty. “One Way to Get People Off the Streets: Buy Hotels.” The New York Times, April 17, 2021. https://www.nytimes.com/2021/04/17/business/california-homeless-hotels.html. 6 Kate Eby. “Coronavirus Timeline: Tracking Major Moments of COVID-19 Pandemic in San Francisco Bay Area.” ABC7 San Francisco, March 27, 2020. https://abc7news.com/6047519/. 7 Jennifer Hernandez. “Letter to All County Welfare Directors and Federally Recognized Tribal Governments Re: Project Roomkey Initiative,” June 1, 2020. https://www.cdss.ca.gov/Portals/9/Additional-Resources/Letters-and- Notices/ACWDL/2020/ACWDL_Project_Roomkey_Initiative.pdf. 8 National Alliance to End Homelessness. “Homelessness Is a Problem in California.” Fact Sheet, May 2020. https://endhomelessness.org/wp-content/uploads/2020/05/CA-fact-sheet-2019-PIT-Count-Data.pdf. 9 Janey Rountree, Nathan Hess, and Austin Lyke. “Health Conditions Among Unsheltered Adults in the U.S.” Policy Brief. California Policy Lab, October 2019. https://www.capolicylab.org/wp-content/uploads/2019/10/Health-Conditions-Among- Unsheltered-Adults-in-the-U.S.pdf. 10 Sékéné Badiaga, Didier Raoult, and Philippe Brouqui. “Preventing and Controlling Emerging and Reemerging Transmissible Diseases in the Homeless.” Emerging Infectious Diseases 14, no. 9 (September 2008): 1353– 59. https://doi.org/10.3201/eid1409.082042. 11 “Approaches for COVID-19 Response for Individuals Experiencing Homelessness.” State of California, March 2020. https://www.cdph.ca.gov/Programs/CID/DCDC/CDPH%20Document%20Library/COVID-19/Protocols-Homeless- Pop.pdf. 12 Adam Brinklow. “SF Halts Homeless Shelter Admissions over COVID-19.” Curbed SF, March 30, 2020. https://sf.curbed.com/2020/3/30/21199094/sf-homeless-shelters-coronavirus-covid-19-admission. 13 Mallory Moench. “S.F. Debates Controversial Homeless Proposal to Make City Provide Shelter to All.” San Francisco Chronicle, April 21, 2021. https://www.sfchronicle.com/local-politics/article/S-F-debates-controversial-homeless-proposal- to-16119058.php. 14 Anita Chabria. “Coronavirus Hitting California’s Homeless Population Could Be What Finally Breaks Hospitals.” Los Angeles Times, March 18, 2020. https://www.latimes.com/homeless-housing/story/2020-03-18/coronavirus-homelelss- money-california-risk-150-million. 15 Jennifer Hernandez. “Letter to All County Welfare Directors and Federally Recognized Tribal Governments Re: Project Roomkey Initiative,” June 1, 2020. https://www.cdss.ca.gov/Portals/9/Additional-Resources/Letters-and- Notices/ACWDL/2020/ACWDL_Project_Roomkey_Initiative.pdf. 16 California Homeless Coordinating and Financing Council. “COVID-19 Frequently Asked Questions,” April 3, 2020. https://www.bcsh.ca.gov/hcfc/documents/covid19_funding_faqs.pdf. 17 “Memorandum to Extend Federal Support to Governors’ Use of the National Guard to Respond to COVID-19 and to Increase Reimbursement and Other Assistance Provided to States.” Press Release. The White House, January 21, 2021. https://www.whitehouse.gov/briefing-room/presidential-actions/2021/01/21/extend-federal-support-to-governors- use-of-national-guard-to-respond-to-covid-19-and-to-increase-reimbursement-and-other-assistance-provided-to-states/. Evaluating Project Roomkey in Alameda County - 62 - 18 Nicole Hayden. “Project Roomkey Funding Ends Soon. Over 11,000 Californians Could Become Homeless, Again.” Palm Springs Desert Sun, October 30, 2020. https://www.desertsun.com/in-depth/news/health/2020/10/30/over-11-000- californians-could-become-homeless-again/5875241002/. 19 Robert J. Fenton. “Request for Approval of Non-Congregate Sheltering.” US Department of Homeland Security, March 27, 2020. https://www.cdss.ca.gov/Portals/9/FEMA/202005-DR-4482-CA-Non-Congregate-Sheltering-Request- Response-Letter-03272020.pdf 20 Matt Levin. “Here’s How Putting California’s Homeless in Hotels Actually Works.” CalMatters, April 11, 2020. https://calmatters.org/housing/2020/04/california-coronavirus-homeless-hotels-motels-newsom/. 21 Doug Smith and Benjamin Oreskes. “Program to House Homeless People in Hotels Is Ending after Falling Short of Goal.” Los Angeles Times, September 22, 2020. https://www.latimes.com/california/story/2020-09-22/homeless-people-hotels- project-roomkey-phasing-out. 22 Matt Levin. “Can Newsom Use the Pandemic to Beat Back Homelessness?” CalMatters, December 4, 2020. https://calmatters.org/housing/2020/11/newsom-pandemic-beat-homelessness/. 23 Hannah Wiley. “California Took 35,000 Homeless People off the Street for 1 Year. Did the Program Work?” The Sacramento Bee, April 2, 2021. https://www.sacbee.com/news/politics-government/capitol-alert/article250213025.html. 24 Jonathan D. Fuchs, Henry Clay Carter, Jennifer Evans, Dave Graham-Squire, Elizabeth Imbert, Jessica Bloome, Charles Fann, et al. “Assessment of a Hotel-Based COVID-19 Isolation and Quarantine Strategy for Persons Experiencing Homelessness.” JAMA Network Open 4, no. 3 (March 2, 2021): e210490. https://doi.org/10.1001/jamanetworkopen.2021.0490. 25 Applied Survey Research. “EveryOne Counts! 2019 Homeless Count and Survey,” July 2019. https://everyonehome.org/wp-content/uploads/2019/07/2019_HIRDReport_Alameda_FinalDraft_8.15.19.pdf. 26 Alastair Boone. “Is There a Better Way to Count the Homeless?” Bloomberg CityLab, March 4, 2019. https://www.bloomberg.com/news/articles/2019-03-04/the-problem-with-hud-s-point-in-time-homeless-count. 27 Doug Smith. “A New Database Gives a Sharper Picture of Homelessness across California.” Los Angeles Times, April 8, 2021. https://www.latimes.com/homeless-housing/story/2021-04-08/a-new-database-gives-a-picture-of-homelessness- across-the-state. 28 Joe Colletti. “Recent Homeless Count Reports from California Continuums of Care Reveal a Significant Increase in the Number of Homeless Persons Counted in 2019.” Homeless and Housing Strategies for California (blog). August 4, 2019. https://homelessstrategy.com/recent-homeless-count-reports-from-california-continuums-of-care-reveal-a- significant-increase-in-the-number-of-homeless-persons-counted-in-2019/. 29 Applied Survey Research. “EveryOne Counts! 2019 Homeless Count and Survey,” July 2019. https://everyonehome.org/wp-content/uploads/2019/07/2019_HIRDReport_Alameda_FinalDraft_8.15.19.pdf. 30 US Department of Housing and Urban Development. “HUD 2019 Continuum of Care Homeless Assistance Programs Housing Inventory Count Report,” October 16, 2019. https://files.hudexchange.info/reports/published/CoC_HIC_State_CA_2019.pdf. 31 Elhalaby, Rawan. “Housing Oakland’s Unhoused.” Advanced Policy Analysis. UC Berkeley, Goldman School of Public Policy, May 2018. https://static1.squarespace.com/static/5c09723c55b02cf724c3d340/t/5d5179b42b874d0001b4114c/1565620680179/Fina l+Elhalaby%2C+Rawan+APA+5-11-2018+for+DISJ.pdf. 32 Robert J. Fenton. “Request for Approval of Non-Congregate Sheltering.” US Department of Homeland Security, March 27, 2020. https://www.cdss.ca.gov/Portals/9/FEMA/202005-DR-4482-CA-Non-Congregate-Sheltering-Request-Response-Letter-03272020.pdf 33 Kyle Patterson, Laura Marshall, Ryan Hunter, Peter Radu, and Peg Stevenson. “More than a Shelter: An Assessment of the Navigation Center’s First Six Months.” San Francisco Office of the Controller, December 10, 2015. https://sfcontroller.org/sites/default/files/FileCenter/Documents/6994- An%20Assessment%20of%20the%20Navigation%20Center%27s%20First%20Six%20Months%20-%20Final.pdf. 34 Mayor Jesse Arreguín. “East Bay’s First Navigation Center Shows Success in First Year,” September 25, 2019. https://www.jessearreguin.com/press-releases/2019/9/25/east-bays-first-navigation-center-shows-success-in-first- year. 35 City of Fremont. “Understanding a Housing Navigation Center for Fremont.” Fact Sheet, September 5, 2019. https://www.fremont.gov/DocumentCenter/View/41463/Navigation-Center-Questions. 36 Department of Health Care Services. “California Advancing & Innovating Medi-Cal (CalAIM) Proposal.” State of California - Health and Human Services Agency, January 2021. https://www.dhcs.ca.gov/provgovpart/Documents/CalAIM-Proposal-03-23-2021.pdf. References - 63 - 37 San Francisco Department of Public Health. “Medical Respite Fact Sheet.” Fact Sheet, December 18, 2017. sfdph.org/dph/files/newsMediadocs/2017PR/Medical-Respite-Fact-Sheet-12-18-17.pdf. 38 Bianca Freda. “Housing Is a Health Intervention: Transitional Respite Care Program in Spokane.” Center for Health Care Strategies, 2017. http://www.chcs.org/media/Respite-Program-Case-Study_101217.pdf. 39 Kyle Patterson, Laura Marshall, Ryan Hunter, Peter Radu, and Peg Stevenson. “More than a Shelter: An Assessment of the Navigation Center’s First Six Months.” San Francisco Office of the Controller, December 10, 2015. https://sfcontroller.org/sites/default/files/FileCenter/Documents/6994- An%20Assessment%20of%20the%20Navigation%20Center%27s%20First%20Six%20Months%20-%20Final.pdf. 40 Ibid. 41 Laura Waxman. “Navigation Center for the Homeless Switches Focus from Housing to Triage.” Mission Local, January 19, 2017. https://missionlocal.org/2017/01/navigation-center-for-the-homeless-switches-focus-from-housing-to-triage/ 42 Kate Eby. “San Francisco's homeless navigation centers: 7 questions answered.” ABC7 San Francisco, January 15, 2020. https://abc7news.com/sf-navigation-center-homeless-shelter-vs-nav-centers/5853129/ 43 Mayor Jesse Arreguín. “East Bay’s First Navigation Center Shows Success in First Year,” September 25, 2019. https://www.jessearreguin.com/press-releases/2019/9/25/east-bays-first-navigation-center-shows-success-in-first- year. 44 City of Fremont. “Understanding a Housing Navigation Center for Fremont.” Fact Sheet, September 5, 2019. https://www.fremont.gov/DocumentCenter/View/41463/Navigation-Center-Questions. 45 Bianca Freda. “Housing Is a Health Intervention: Transitional Respite Care Program in Spokane.” Center for Health Care Strategies, 2017. http://www.chcs.org/media/Respite-Program-Case-Study_101217.pdf. 46 San Francisco Department of Public Health. “Medical Respite Fact Sheet.” Fact Sheet, December 18, 2017. sfdph.org/dph/files/newsMediadocs/2017PR/Medical-Respite-Fact-Sheet-12-18-17.pdf. 47 Greg B. Smith. “Sleeping Behind the Bronx Zoo: Why Some New Yorkers Choose Streets Over Shelters.” THE CITY, April 5, 2021. https://www.thecity.nyc/2021/4/5/22366758/new-yorkers-choose-streets-over-homeless-shelters 48 Robert Davis. “Why People Experiencing Homelessness Avoid Shelters.” Invisible People, May 20, 2021. https://invisiblepeople.tv/why-people-experiencing-homelessness-avoid-shelters/ 49 “The View From the Outside: A Survey of 300 People Experiencing Homelessness in San Francisco.” Tipping Point Community, March 2019. https://tippingpoint.org/wp-content/uploads/2019/03/The-View-from-Outside.pdf 50 Steven Brown, Samantha Batko, Josh Leopold, and Aaron Shroyer. “Final Report and Recommendations on Homelessness in Alameda County, California.” Urban Institute, January 2018. https://www.urban.org/sites/default/files/publication/96506/final_report_and_recommendations_on_homelessness_in_al ameda_county_california_0.pdf 51 “Memorandum to Extend Federal Support to Governors’ Use of the National Guard to Respond to COVID-19 and to Increase Reimbursement and Other Assistance Provided to States.” Press Release. The White House, January 21, 2021. https://www.whitehouse.gov/briefing-room/presidential-actions/2021/01/21/extend-federal-support-to-governors- use-of-national-guard-to-respond-to-covid-19-and-to-increase-reimbursement-and-other-assistance-provided-to-states/. 52 Conor Dougherty. “One Way to Get People Off the Streets: Buy Hotels.” The New York Times, April 17, 2021. https://www.nytimes.com/2021/04/17/business/california-homeless-hotels.html. 53 Ibid. 54 National Alliance to End Homelessness. “Ending Chronic Homelessness Saves Taxpayers Money.” Fact Sheet, 2015. http://endhomelessness.org/wp-content/uploads/2017/06/Cost-Savings-from-PSH.pdf. 55 Benjamin Oreskes, Emily Alpert Reyes, and Doug Smith. “Judge Orders L.A. to Provide Shelter for Skid Row Residents.” Los Angeles Times, April 20, 2021. https://www.latimes.com/homeless-housing/story/2021-04-20/judge- carter-la-city-county-shelter-skid-row-homeless-fall. 56 Mallory Moench. “S.F. Debates Controversial Homeless Proposal to Make City Provide Shelter to All.” San Francisco Chronicle, April 21, 2021. https://www.sfchronicle.com/local-politics/article/S-F-debates-controversial-homeless-proposal- to-16119058.php. 57 Ibid. 58 Sarah Hunter, Melody Harvey, Brian Briscombe, and Matthew Cefalu. “Evaluation of Housing for Health Permanent Supportive Housing Program.” RAND Corporation, 2017. https://doi.org/10.7249/RR1694 Evaluating Project Roomkey in Alameda County - 64 - 59 Joshua Sabatini. “Plans in Works to Build 1,700 Supportive Housing Units for Formerly Homeless.” The San Francisco Examiner, December 2, 2019, sec. News. https://www.sfexaminer.com/news/plans-in-works-to-build-1700-supportive- housing-units-for-formerly-homeless/.