Returns to Homelessness: Key Considerations for Using This Metric to Improve System Performance

In many health care and social service fields, considerable attention is paid to primary prevention (i.e., preventing a condition before it occurs) and secondary prevention (i.e., identifying and treating a condition as soon as possible after it occurs). However, tertiary prevention—defined as managing a condition after it has occurred or preventing recurrence—is a crucial component of long-term prevention, particularly when it focuses on preventing recurrence of a condition.

In mental health and addiction treatment, a client may manage their symptoms or maintain sobriety, but clinical work is needed to focus on preventing relapse. In the criminal justice system, individuals convicted of a criminal offense may reoffend—an occurrence commonly referred to as “recidivism.” In the field of homeless services, research, and policy, researchers, advocates, and other stakeholders have long been interested in “recidivism,” “relapse,” or “return” to homelessness, but this phenomenon has not been explicitly addressed in past federal policy responses. This dynamic has shifted in the United States in recent years. For example, the US Interagency Council on Homelessness—the federal entity with primary responsibility for efforts to address homelessness—has stated its desire to make homelessness “a rare, brief and one-time experience,” thus implicitly articulating a goal of preventing repeated episodes of homelessness.¹

Moreover, the US Department of Housing and Urban Development (HUD) now requires communities to track returns to homelessness as part of HUD’s System Performance Measure no. 2 and considers performance on this measure in decisions about how federal homeless assistance dollars are allocated.² Similarly, the Department of Veterans Affairs (VA) has annually reported on returns to homelessness among veterans receiving services from the Supportive Services for Veteran Families program, its nationwide homelessness prevention and rapid rehousing program.³ In this article, we focus on HUD and VA, which operate the two largest US homeless service systems, but we acknowledge that the topic applies to other homeless service systems domestically and internationally.

The increasing focus on returns to homelessness is a positive development given that a considerable proportion of individuals experiencing homelessness who successfully move to permanent housing experience homelessness again after some time (so “permanent” housing is aspirational). This is important to the field because repeated homelessness can lead, by definition, to chronic homelessness, which is associated with an array of adverse health and social consequences for individuals and can necessitate resource-intensive interventions.⁴

Yet, several key issues require greater consideration to employ and operationalize the concept of returning to homelessness for effective policy and program efforts to prevent repeated experiences of homelessness. Broadly speaking, these issues fall under two overarching questions: Over what time horizon should returns to homelessness be measured? What should count as a return to homelessness? In the remainder of this article, we engage with each of these questions with the ultimate aim of helping ensure that returns to homelessness are defined and measured in a way that makes them useful for driving improvements in the performance of homeless assistance systems. First, we draw on previous research as well as recent data from HUD and VA to provide context about the frequency and dynamics of returns to homelessness.

DATA ON RETURNS TO HOMELESSNESS

Researchers have taken interest in the phenomenon of returning to homelessness for several decades.^5–7 Studies conducted in the 1990s using administrative databases from emergency shelter systems in New York City and Philadelphia, Pennsylvania, found that returns to emergency shelter were common: more than 40% of men and more than a third of women who exited the single adult emergency shelter system in both cities reentered shelters within two years.⁶ Related work found lower rates of returns to shelters for families.⁷ A key finding from these studies was that risk of returns to homelessness was highest in the initial first few months immediately following the initial exit from a shelter, a finding corroborated in more recent research focused on rapid rehousing as well.⁸

More recent focus in the past two decades on returns to homelessness as measures of performance, such as that of HUD and VA, considers only those who exit the homeless service system to permanent housing—a group that earlier research found to face a lower risk of returns to shelter than those exiting to other destinations (e.g., transitional housing).⁷ The focus solely on those who exit to permanent housing may reflect an assumption that those who leave the homeless service system to destinations other than permanent housing cannot be presumed to have definitively “exited” homelessness. Alternatively, the focus on those exiting to permanent housing may be appropriate for separating tertiary prevention cases from those who exit to destinations such as carceral facilities or hospitals. Moreover, many existing performance measures examine not just those exiting emergency shelter but also those who exit other programs to permanent housing, including those who exit from street outreach services, safe havens, and transitional housing programs.²

Available data suggest that most who exit one of these programs to permanent housing do not return to homelessness in the near term. For example, data from HUD System Performance Measure no. 2 track returns to homelessness at six months, one year, and two years.⁹ In 2021, across 388 continuums of care (i.e., regional bodies that coordinate housing and services), the rate of returning to homelessness after exiting the homeless assistance system to a permanent housing destination was 9% within six months (range = 0%–35%), 13% within one year (range = 0%–39%), and 18% within two years (range = 0%–45%). In 2019 and 2020, the rates of returning to homelessness were highly similar at all follow-up periods.

The wide variability in return rates across continuums of care could be a function of a number of factors, including capacity and types of services available, the client populations served, documentation practices, and availability of staff and resources. VA data tell a similar story. In our analysis conducted across all VA homeless programs, we found that among veterans who exited to permanent housing destinations in fiscal year 2020, 5%, 9%, and 16% returned to homelessness at 6, 12, and 24 months, respectively. We defined a return as when a veteran returning to VA homeless programs was identified as experiencing homelessness at assessment for VA homeless services or upon entry into VA’s rapid rehousing program.

Not only do rates of return vary by service area, but rates of return vary across program types, which is important to consider in performance monitoring. For example, HUD data indicate that the overall return rates at 6, 12, and 24 months among those exiting emergency shelters are 12%, 17%, and 22%, respectively, compared with 5%, 8%, and 12%, respectively, for those exiting permanent housing programs. Variation across program types may be attributable to differences in the populations served or the services provided by a particular program type. Although we value HUD’s approach to tracking returns and consider it practical and useful, this variation underscores the point that, from an applied perspective, a refined approach to evaluating and using this measure may enhance the measure’s effectiveness in driving performance improvement.

TIME HORIZON TO CAPTURE RETURNS

How long must an individual who formerly experienced homelessness be housed for the onset of a homeless episode to be considered a return to homelessness? At one extreme, if a long time horizon is employed, a person who experienced homelessness as a younger child but not again until decades later might be considered to have returned to homelessness. At the other extreme, if a short time horizon is employed, one could argue that a person who has an apartment leased in their name and then enters an emergency shelter after six months of stable housing is not returning to homelessness but experiencing a new, discrete episode of homelessness. Neither extreme approach is likely satisfactory.

HUD guidance on how to operationalize a return to homelessness uses a two-year time frame: it requires continuums of care to report on the proportion of individuals exiting homelessness to permanent housing who return within a period of up to 24 months from the date of their exit and counts individuals as “newly” experiencing homelessness if they have no record of entry into the homeless assistance system within the previous 24 months. There is not a concrete rationale of which we are aware for the use of this 24-month time frame, although it is consistent with the time frame used in early research examining returns to homelessness.^6,7 On the other hand, VA’s Supportive Services for Veteran Families program tracks returns to homelessness for those who exit the rapid rehousing component of its program to permanent housing only for a follow-up of up to 12 months.³

HUD guidance indicates that its return to homelessness metric is intended to be used in conjunction with the number of persons newly experiencing homelessness who enter the system to assess the number of people experiencing homelessness in a given time frame. In this context, the time frame used to assess what is a return to homelessness versus a new episode has no effect on the number of people that a community identifies as experiencing homelessness. However, from the perspective of improving systems performance to reduce returns to homelessness, a more data-driven approach could yield more actionable information. For example, HUD guidance indicates that continuums of care should set targets for return rates that vary by program type.¹⁰

A potential extension of this guidance is to adjust follow-up times (e.g., three months, six months) according to how return rates vary by program type as baseline benchmarks for performance improvement. To be clear, HUD’s guidance that communities track returns at 6-, 12-, and 24-month intervals is a straightforward and practical approach for setting targets. But an even more detailed program-specific approach could be useful. For example, if the majority of persons exiting a particular program who subsequently return do so within 6 months, the 12- and 24-month time points are less useful, and it may make sense to examine returns for that program at one-, three-, and six-month intervals instead. From a clinical practice and program improvement perspective, setting program-specific return time benchmarks could help inform critical periods after clients exit these programs in which it is especially important to bolster prevention efforts.

WHAT COUNTS AS A RETURN TO HOMELESSNESS

In considering measuring returns to homelessness, there are important data limitations to acknowledge and questions to answer about the types of homeless services that should be included and the duration of homelessness that qualifies as a return to homelessness.

Limitations

Current measures of returns to homelessness that HUD and VA use rely on administrative data from the homeless assistance system. However, homeless service systems will generally know a client has begun experiencing homelessness again only if they use a homeless service that is captured in the same data collection system. Therefore, a return to homeless services is an inexact proxy indicator for a return to homelessness because some individuals may begin to experience homelessness and not access homeless services. Depending on how broad of a definition of homelessness is used, other forms of housing instability (e.g., couch surfing) may be missed. In addition, individuals may begin to experience homelessness and access alternative service systems (e.g., residential substance use or mental health treatment) or have other touch points (e.g., domestic violence or religious shelters) that are not captured in centralized, linked data systems and thus may not be identified as experiencing homelessness.

If homeless service systems have the opportunity to integrate data sources from other service systems with their local Homeless Management Information System, it would likely yield a more comprehensive accounting of all client returns to homelessness. And to the extent possible, structured interviews and other corroborating data methods should be considered to validate samples of the data to ensure that data systems are being inclusive and accurate in capturing returns to homelessness. Doing so may also have clinical value by providing clients with a comprehensive safety net.

It may also be worth noting that the current performance measures capture returns to homelessness only after placement in permanent housing. Thus, they do not include individuals who exit a homeless assistance program before being placed in permanent housing (i.e., premature negative exits). Perhaps premature negative exits should be considered in a different category and tracked separately from returns to homelessness, but it may be important to recognize them as potential intervention points in clients’ journeys in permanently exiting homelessness.

Types of Homeless Services That Count

Many communities offer a continuum of homeless programs ranging from brief housing assistance to permanent supported housing services. Thus, there is a question of whether a return to homelessness should be defined as a return to any homeless service or as including returns to only some broader or more narrow set of programs. For its part, HUD has operationalized a return to include only programs for which homelessness (per the statutory definition HUD uses) is an eligibility criterion.¹¹

This is a reasonable approach, but including additional programs or tracking them separately may provide a more complete picture of how individuals who previously exited the homeless assistance system in a given community continue to use resources in that system. For example, use of homelessness prevention services does not typically require a person to experience homelessness, so including such services in a return measure would provide more information about the extent to which those who exit a system to permanent housing continue to rely on services provided by the system to remain stably housed. Similarly, including or separately tracking returns among those who complete intake for a community’s coordinated entry system but are not literally homeless would also provide additional information about continued contacts of those who exited to permanent housing with the homeless assistance system. In both cases, such information would be useful for helping systems understand the scope of need for continued support for persons who have previously exited the system.

Length of Time Experiencing Homelessness

With some caveats, HUD guidance on returns to homelessness suggests that a person who spent one night in a shelter after exiting to permanent housing would be counted as having returned to homelessness. Again, this is practical to implement. But from a conceptual standpoint, it is reasonable to question the appropriateness of counting a person with a history of homelessness who spends only one night in a shelter as having returned to homelessness. This example may be extreme, but it does point to the fact that there is variation in what returning to homelessness means in terms of the longer-term trajectories of individuals’ housing stability, and these differences have implications for interventions. Some may experience a return that is just a brief “blip” on an otherwise highly stable residential trajectory; others may experience a return episode of extended duration. From an intervention perspective, there is a big difference between these two types of return: the former may require little or no intervention, whereas the latter might be an indication of an individual in need of more intensive support.

Thus, at the systems level there may be great value in treating a return not simply as a yes or no indicator but by differentiating between different types of returns to determine the level of need in a functioning safety net. Doing so would provide more actionable information about the extent of need for intervention to assist those who have returned to homelessness to regain stable housing.

CONCLUSIONS

Helping individuals who exit homelessness remain stably housed and avoid repeated episodes of homelessness should be an important component of any strategic approach to end homelessness. Thus, understanding the extent to which this occurs has been of longstanding interest and has recently been incorporated into HUD and VA efforts to use data to drive system-level improvements. We agree with the logic in tracking returns and see the overall value of the approach HUD has adopted in tracking returns. However, additional considerations are needed of how this is defined and measured to be most useful to different homeless service systems.

We have highlighted some key decision points for possible refinements to this measure for programs and how they might be instructive for monitoring and improving system performance. We acknowledge that tracking returns to homelessness by different types and durations and using multiple data sources, as we suggest, will require data management resources and expertise. Having a metric that is too complicated or having too many metrics may also render them of limited utility. Nonetheless, we believe there is value in thinking more strategically about attending to returns to homelessness to support vulnerable populations, allocate resources, and have a clear understanding of causes of returns to homelessness.