In 2015, the United Nations (UN) adopted The 2030 Agenda for Sustainable Development, which includes 17 Sustainable Development Goals intended to promote universal peace through social, economic, and environmental action (UN, 2015). The first of these goals – to end poverty in all its forms everywhere – involves several targets, including making sure that those living in poverty and those vulnerable to poverty and its consequences have equitable access to economic resources and basic services. A key indicator of this target is increasing the proportion of the population living in households with access to basic services (UN, 2022), including medical and mental health services. For individuals experiencing homelessness, major gaps exist in the availability of medical, mental health, and other basic services (Barile et al., 2020), especially among individuals living with mental illness (Stergiopoulos et al., 2010).

Associations between mental health problems and homelessness are complex and likely bidirectional in that mental health problems can increase the likelihood of experiencing homelessness, which can, in turn, exacerbate mental health problems (Padgett, 2020). Findings from a recent meta-analysis of studies looking at the prevalence of psychiatric disorders among individuals experiencing homelessness in high-income countries suggest that as many as 76% of individuals experiencing homelessness meet criteria for a current psychiatric disorder, with the most common conditions being substance use disorders, schizophrenia spectrum and psychotic disorders, and depressive disorders (Gutwinski et al., 2021). Moreover, homelessness is associated with increased risk for exposure to traumatic events and associated trauma-related mental health problems, including posttraumatic stress disorder, or PTSD (e.g., Whitbeck et al., 2015). Despite high rates of mental health problems among individuals experiencing homelessness, utilization of outpatient behavioral health and primary care services are low (e.g., Folsom et al., 2005; Khandor et al., 2011), in part because these services may be given lower priority than accessing basic services that ensure survival, such as food and shelter (Pruitt et al., 2018). Indeed, individuals experiencing homelessness who are also living with mental illness are more likely to utilize crisis care services, including inpatient psychiatric hospitalizations and emergency room (ER) services, for medical and psychiatric care (Laliberté et al., 2020; Lombardi et al., 2020).

To ensure that individuals experiencing homelessness are provided with basic resources such as food and shelter that are essential to both accessing health care and recovering from illness, including mental illness, clinicians and policymakers have turned their attention to housing as a health care intervention. For example, a number of health care systems and community agencies have developed permanent supportive housing (PSH) programs that combine non-time-limited housing with voluntary supportive services to promote housing stability and retention (National Academy of Sciences, Engineering, and Medicine, 2018). Early research investigating the impact of PSH programs on individuals living with mental illness found that PSH programs were associated with reductions in time spent hospitalized for psychiatric care and per day costs, although these reductions were greatest for individuals recruited directly from psychiatric hospitals compared to those recruited directly from the community (Gulcur et al., 2003). In one of the largest studies to date looking at the impact of a PSH program on health service utilization outcomes among individuals experiencing both homelessness and mental illness, researchers examined health care utilization outcomes in a sample of 527 adults with a serious mental disorder who were housed as part of the At Home/Chez Soi randomized controlled trial of a Housing First intervention in Toronto, Canada, and followed for seven years post-housing (Lachaud et al., 2021; Mejia-Lancheros et al., 2021). As part of this trial, individuals living with a serious mental disorder who were housed through the Housing First intervention received either assertive community treatment support or intensive case management based on level of need, with assertive community treatment support being used for those with the highest level of need for more intensive services. For study purposes, high need was determined by several criteria, including having a current psychotic or bipolar disorder and associated impairment as determined by one of several possible factors, which could include two or more psychiatric hospitalizations in any one of the five years prior to housing. Individuals who were not assigned to the Housing First intervention received treatment as usual (TAU), which consisted of access to local support services (Hwang et al, 2012). Interestingly, the Housing First intervention was not associated with reductions in all-cause inpatient hospitalizations or psychiatric hospitalizations but was associated with reductions in the number of emergency room visits for individuals with the highest level of need. The intervention was, however, associated with an increase in the number of all-cause and psychiatric inpatient hospitalizations and emergency room visits for individuals with moderate level of need for services compared to TAU (Lachaud et al., 2021). A separate study from the At Home/Chez Soi project found that the Housing First intervention was also associated with a reduction in the number of outpatient primary care and specialty visits among individuals with the highest level of need, whereas those with moderate need saw an increase in outpatient primary care visits compared to TAU (Mejia-Lancheros et al., 2021).

Overall, PSH is generally associated with reductions in crisis and acute care services (i.e., emergency room visits, inpatient hospitalizations) and more frequent use of outpatient care, although findings are somewhat mixed and appear to vary as a function of individual characteristics, including the extent to which individuals were already utilizing different services prior to housing. In other words, PSH, and the availability of support services associated with PSH, may help redirect individuals who rely primarily on crisis care services for their health care needs away from those services and toward less restrictive, outpatient care, while increasing access to all health care services for individuals with no established care. As a result, some tenants may be less likely to utilize certain health care services after being housed, while, for others, utilization rates, including utilization of crisis care, may increase. That said, any differential patterns of engagement in health care services that might vary as a function of prior care are typically obscured when PSH tenants are studied collectively as a single population. Therefore, we sought to build on the extant literature on PSH by conducting an exploratory evaluation of health service utilization patterns and associated charges among individuals living with a chronic mental illness who were and were not utilizing health care services in the year prior to being housed through a PSH program embedded in an urban Community Mental Health Center (CMHC). Given the relatively small sample of tenants included in this evaluation, our approach was primarily exploratory in nature rather than hypothesis-driven, although we specifically sought to explore (1) changes in health care utilization following placement in PSH across all service types, (2) whether using health care services in the year prior to being housed was associated with using those same services again in the year post-housing, and (3) whether the amount of services used across all service categories differed among individuals who did and did not use various health care services in the year prior to being housed.

Methods

Setting and Sample Description

Records were reviewed for 116 supportive housing tenants housed through University Health Behavioral Health’s 500 in Five program. University Health Behavioral Health (UHBH), formerly known as Truman Medical Center Behavioral Health, is a CMHC embedded in University Health (UH), a trauma-one, inner city, safety net hospital located in the Midwestern U.S. In 2017, UHBH, launched the 500 in Five initiative that committed to developing and/or securing 500 units of housing over five years. To support the initiative, UHBH received intensive, multi-month training on designing and operating quality supportive housing conducted by Corporation for Supportive Housing (CSH), a national nonprofit that provides technical assistance and training. The training informed how the 500 in Five program delivered and operated supportive housing for these tenants. UHBH also consulted with CSH on program evaluation and outcome reports. Although the initiative was cut short due to the COVID-19 pandemic, 320 additional housing units were secured in just over 2.5 years.

Referrals for housing through the 500 in Five initiative came through the city’s coordinated entry system. Referrals were also received internally from within the CMHC when it was determined that clients needed housing due to the severity of their mental health problems. Support services, including case management and housing and benefit assistance, for tenants were funded primarily by Medicaid, and non-Medicaid reimbursed support services for individuals pending Medicaid were supported primarily through Department of Housing and Urban Development (HUD) funds. Rental assistance was supported through a variety of mechanisms, including HUD, Missouri Department of Mental Health, and private foundation funds. Data on behavioral health and medical services, including visit-related charges, within the UH system during the year pre- and/or year post-housing were reviewed for 116 supportive housing tenants housed as part of this initiative. Tenants were housed by UHBH between November 1, 2017, and February 25, 2020, a timeframe that allowed for one-year post-housing analysis of health care utilization outcomes for all tenants with valid data. Of the 116 tenants, billing data with information on visits and associated professional charges for health care services received in the UH system during the year pre- and/or year post-housing were available for 80 (69.0%) tenants, who were subsequently included in the current evaluation. Demographically, these tenants were primarily male (n = 48; 60.0%) and predominantly Black/African American (n = 41; 51.3%) and White (n = 32; 40.0%), with a mean age of 39 years (SD = 15.21). All 80 tenants were diagnosed with at least one psychiatric condition, with the most common primary psychiatric diagnoses including posttraumatic stress disorder, or PTSD, (n = 32; 40.0%), schizophrenia or other psychotic disorders (n = 23; 28.8%), and major depressive disorder (n = 19; 23.8%). Twenty-two tenants (27.5%) met criteria for more than one psychiatric condition, and 31 (38.8%) also met criteria for a co-occurring SUD. The 500 in Five initiative had excellent retention in housing, with 78 of the 80 tenants (97.5%) included in the current evaluation remaining in permanent housing after one year. Additional demographic information is presented in Table 1.

Table 1 Sample demographics
Full size table

Archival data reflecting key health care utilization outcomes was extracted from UH’s electronic medical record system and billing data for individuals housed through UHBH’s 500 in Five initiative. Key outcomes included use of services, visits, and associated professional charges for utilization of crisis care services (i.e., emergency room visits, all-cause and psychiatric inpatient hospitalizations) and outpatient services (i.e., all-cause outpatient and outpatient behavioral health visits). To account for instances where multiple procedures were billed for the same visit, a visit was defined as one encounter to any given provider per day (Tyree et al., 2006). For inpatient services, this effectively translated into visit days. Outcomes were included only for health care services rendered up to one year before and after the date the tenant was housed. This project was reviewed and approved as a quality improvement project by the University of Missouri – Kansas City’s Institutional Review Board.

Data Analysis Plan

First, medical records and billing data were used to construct a series of dichotomous variables for different crisis care and outpatient health care services indicating whether the tenant did or did not receive that type of service in the year pre- and year-post housing, as well as whether the tenant received any health services in the year prior to being housed. McNemar’s tests were used to examine changes in health service utilization rates pre- and post-housing, and Goodman and Kruskal’s tau statistic was used to explore how much variation in post-housing health service utilization was associated with pre-housing use across all service categories. Bivariate analyses were used to explore differences between tenants who were and were not actively utilizing any health care services in the year pre-housing on demographic and clinical variables, including primary diagnosis. Pre- and post-trends in the amount of services used, as well as pre- and post-housing differences in professional service charges, were analyzed using paired-samples t-tests. To examine whether housing is associated with different patterns of health care utilization among tenants who were and were not actively utilizing health care services in the year pre-housing, separate analyses were conducted among individuals that utilized UH hospital services (i.e., emergency, inpatient, and outpatient) in the year prior to being housed, and those that did not use UH for health care during the previous year. Missing data were handled using pairwise deletion with tests using all cases with valid data for each tested variable. All analyses were conducted using SPSS v.26 software.

Results

The proportion of tenants who used different health care services in the years pre- and post-housing for the full sample is presented in Table 2. Across all service types for the full sample of 80 tenants, 87.5% (n = 70) utilized health care services at least once through the hospital’s system in the year pre-housing, and 97.5% (n = 78) utilized health care services through the hospital’s system at least once in the year post-housing (p = .039). In terms of crisis care services, results indicated that the proportion of tenants overall who utilized emergency room and inpatient services before and after being housed was not statistically different (all p’s > 0.10). Approximately 58.8% (n = 47) of the sample visited the emergency room at least once in the year pre-housing, with 55.0% (n = 44) of the sample utilizing emergency room services at least once in the year post-housing. Moreover, approximately 23.8% (n = 19) of the full sample had at least one inpatient hospitalization in the year pre-housing, with a similar rate observed in the year post-housing (n = 16; 20.0%). When specifically examining psychiatric hospitalization rates, 20% of the sample had at least one psychiatric hospitalization in the year pre-housing compared to 12.5% (n = 10) in the year post-housing. Pre-housing utilization of specific crisis care services did not account for a statistically significant amount of variance in post-housing crisis care utilization.

Table 2 Pre- and post-housing health service utilization rates by past year use across all service categories (n = 80)
Full size table

Results did, however, reveal statistically significant differences in the proportion of tenants accessing outpatient care after being housed. For example, approximately 67.5% (n = 54) of the sample used outpatient services at least once in the year pre-housing, increasing to 92.5% (n = 74) of the sample in the year post-housing (p < .001). A similar trend emerged when looking specifically at outpatient behavioral health services such that 50.0% (n = 40) of the sample used outpatient behavioral health services at least once in the year pre-housing, compared to 81.3% (n = 65) in the year post-housing (p < .001). While pre-housing, all-cause outpatient service utilization did not account for a statistically significant amount of variance in post-housing all-cause outpatient service utilization, pre-housing use of outpatient behavioral health services did account for a statistically significant amount of variance in post-housing outpatient behavioral health service utilization, accounting for approximately 12.4% of the variance in post-housing outpatient behavioral health service use (p = .002). More specifically, tenants who did not receive outpatient behavioral health services in the year pre-housing were disproportionately represented among tenants who did not receive any outpatient behavioral health services in the year post-housing (standardized residual = 2.0).

Outcomes Among Tenants Utilizing Health Care Services in the Year Pre-Housing

Next, we sought to explore whether the degree to which individuals who were already using various health care services pre-housing changed in the year post-housing using all available visit and professional service charge data for the subset of tenants who used those particular services in the year pre-housing. The mean number of visits and associated professional charges across both crisis care and outpatient service categories for the subset of tenants who used those particular services in the year pre-housing are presented in Table 3. Of the 70 tenants who received some form of health care service in the year prior to being housed, billing data used to derive visit and associated professional charge variables were only available for 63 of the 70 tenants. So, seven tenants were excluded from this subset of analyses. Across all service types for tenants who received some form of health care service in the year prior to being housed, tenants averaged almost nine unique provider visits in the year pre-housing and approximately seven visits in the year post-housing, t(62) = 1.40, p = .168. No statistically significant difference was observed in total professional charges from pre- to post-housing, t(62) = − 0.19, p = .852.

Table 3 Pre- and post-housing health care utilization outcomes for tenants utilizing health care services in the year pre-housing (n = 63)
Full size table

In terms of crisis care services, a statistically significant decrease in emergency room visits was observed from pre- to post-housing among tenants who utilized emergency room services in the year prior to being housed, t(39) = 3.42, p = .001, although professional charges associated with emergency room visits were not statistically different from pre- to post-housing, t(39) = 1.30, p = .200. When utilization and costs associated with inpatient psychiatric hospitalizations were examined, statistically significant reductions were observed in both the mean number of inpatient psychiatric hospitalization visit days, t(12) = 2.71, p = .019, and professional charges associated with psychiatric hospitalizations, t(12) = 2.84, p = .015, in the years pre- and post-housing among tenants with a prior psychiatric hospitalization. In terms of outpatient care services, the mean number of outpatient visits and associated professional charges did not significantly change from pre- to post-housing among tenants who were already utilizing outpatient services prior to housing. Similarly, no statistically significant differences were observed in terms of the number of outpatient behavioral health visits and associated professional charges from pre- to post-housing.

Bivariate Analyses Comparing Tenants with and Without any Health Care Utilization in the Year Pre-Housing

In all, ten individuals (14.3% of the total sample) were identified using both medical record and billing data that had no record of crisis care and/or outpatient care services through the hospital’s system in the year pre-housing. Bivariate analyses were conducted to compare tenants that did and did not utilize any health care services in the past year on demographic and mental health outcomes. Tenants who did and did not utilize any health care services from the hospital system in the year pre-housing did not differ in terms of gender as reported in the medical record, race, ethnicity, having a primary diagnosis of a serious, persistent mental illness (i.e., bipolar disorder, schizophrenia, or another psychotic disorder), or having a co-occurring substance use disorder (all p’s > 0.05). Tenants did, however, differ with regard to age in that tenants who utilized health care services in the year pre-housing were, on average, older (M = 40.43 years; SD = 15.03) than tenants who did not utilize health care services in the year pre-housing (M = 28.50 years; SD = 12.64), t(78) = -2.39, p = .019.

Outcomes Among Tenants with No Health Care Utilization in the Year Pre-Housing

To explore the impact of housing on the degree of health care utilization among individuals who were not already receiving health care services through the hospital’s system, we examined health care utilization outcomes among tenants who did not receive any crisis care and/or outpatient services at the hospital in the year pre-housing. Of the 10 tenants with no record of any prior service use in the year pre-housing, 80.0% (n = 8) visited the emergency room at least once, 30.0% (n = 3) had at least one inpatient hospitalization, 10.0% (n = 1) had at least one psychiatric hospitalization, 80.0% (n = 8) had at least one outpatient visit, and 50.0% (n = 5) had at least one outpatient behavioral health visit in the year post-housing. Pre-housing use of any health care services did not account for a statistically significant amount of variance in post-housing use of any category of health care services except for outpatient behavioral health care, accounting for approximately 9.2% of the variance in post-housing outpatient behavioral health service use (p = .007). More specifically, tenants with no record of any prior service use in the year pre-housing were disproportionately represented among tenants who did not receive any outpatient behavioral health services in the year post-housing (standardized residual = 2.3).

The mean number of visits and associated professional charges across both crisis care and outpatient service categories for the subset of tenants who did not receive any health care services in the year pre-housing are presented in Table 4. Across all service types, tenants with no prior health care utilization averaged approximately two visits in the year post-housing, t(9) = -4.71, p = .001. No statistically significant difference was observed in total professional charges from pre- to post-housing, t(9) = -2.12, p = .063.

Table 4 Pre- and post-housing health care utilization outcomes for tenants with no health care utilization in the year pre-housing (n = 10)
Full size table

In terms of crisis care services, a statistically significant increase in both emergency room visits, t(9) = -3.67, p = .005, and associated professional charges, t(9) = -2.52, p = .033, was observed from pre- to post-housing for tenants with no past year health care utilization. No statistically significant differences were observed in terms of all-cause inpatient hospitalization or psychiatric hospitalization visit days and associated professional charges in the years pre- and post-housing for tenants with no past year health care utilization. In terms of outpatient care services, statistically significant increases in mean outpatient visits, t(9) = -2.86, p = .019, and associated professional charges, t(9) = -2.45, p = .037, were observed in the years pre- and post-housing, although no statistically significant differences in mean visits or professional charges were observed in terms of outpatient behavioral health services.

Lastly, we conducted exploratory analyses to examine whether post-housing health care utilization outcomes differed between tenants who did and did not utilize any health care services at the hospital in the year pre-housing. In the year post-housing, there were no statistically significant differences between tenants who were and were not utilizing health care services prior to being housed in terms of mean number of visits, t(71) = -1.00, p = .322, or total professional service charges, t(71) = -1.10, p = .275.

Discussion

Homelessness and poverty are inextricably linked in that poverty increases risk for homelessness (Giano et al., 2020), and housing individuals experiencing homelessness is a critical part of the solution to ending poverty (Children’s Defense Fund, 2019). Moreover, homelessness often introduces significant barriers to accessing medical and mental health care and other basic services (Barile et al., 2020), especially among individuals living with mental illness who are also experiencing homelessness (Stergiopoulos et al., 2010). Results from the current evaluation suggested that being housed through a PSH program embedded in an urban CMHC may increase access to outpatient health care and contribute to changes in the frequency with which different health services are utilized post-housing.

Most notably, there was a statistically significant increase in utilization of outpatient services for the full sample after being housed. This was also the case for outpatient behavioral health services. This suggests that the combination of stable housing and support services, including case management, characteristic of PSH may play an important role in helping link individuals to needed health care services. While the proportion of tenants using crisis care services, including emergency room and inpatient services, was similar pre- and post-housing for the full sample, those tenants who used crisis care services in the year prior to being housed tended to use those services less often after being housed, with statistically significant reductions observed in the number of emergency room visits and inpatient psychiatric hospitalization visit days from pre- to post-housing for utilizers of those services, respectively. These findings are consistent with several studies of PSH programs showing that PSH is typically associated with decreased use of crisis care services, including psychiatric emergency room visits (Raven et al., 2020) and psychiatric inpatient hospitalizations (Aidala et al., 2013). Furthermore, results revealed that tenants with no record of service utilization from the hospital system in the year pre-housing were more likely to utilize both emergency room and outpatient services from the hospital system in the year post-housing as evidenced by increased visits and charges. These findings are similar to those found in a sample of adults with a serious mental disorder who were housed as part of the At Home/Chez Soi project, which found that individuals living with a serious mental disorder and a moderate level of need saw an increase in emergency room visits, inpatient psychiatric hospitalizations, and outpatient primary care visits after being housed (Lachaud et al., 2021; Mejia-Lancheros et al., 2021). In other words, while housing may reduce utilization of costly crisis care services for individuals who are actively using those services as their primary source of health care, housing may also enable individuals who avoided such care while homeless to shift priorities from survival and other basic needs to health care for chronic conditions requiring both acute and outpatient services.

When comparing tenants who utilized health care services in the year pre-housing to those who did not utilize services in the year pre-housing, no differences were found between tenants who did and did not utilize health care services in terms of gender as reported in the medical record, race, ethnicity, having a primary diagnosis of a serious, persistent mental illness (i.e., bipolar disorder, schizophrenia, or another psychotic disorder), or having a co-occurring substance use disorder. Statistically significant differences were found with regard to tenant age, however, in that tenants who utilized any health care services from the hospital system in the year pre-housing tended to be older than those with no prior record of service utilization. The small proportion of tenants in the overall sample with no record of pre-housing service utilization precludes our ability to draw meaningful conclusions about this finding. Even so, these findings are consistent with previous research suggesting that young adults as a whole are less likely than other age groups to utilize health care services (Fortuna et al., 2009). Because these findings also suggest that individuals with no prior use of health services in the year pre-housing were disproportionately represented among tenants who received no outpatient behavioral health services in the year post-housing, these findings further speak to the need to better understand and address common barriers to mental health care experienced by young adults. For example, previous research has shown that attitudes toward mental health services tend to be especially negative among young people with low socioeconomic status (e.g., Jagdeo et al., 2009). For many young adults experiencing both mental illness and homelessness, negative attitudes toward mental health care may be reaffirmed when they do seek care in that stigma and discrimination experienced within the health care system itself is commonly identified as a barrier to care by many people experiencing homelessness, especially if individuals fear that disclosing relapses or symptoms might lead to loss of housing or support (Patterson et al., 2015). In order to successfully retain tenants in housing and promote greater utilization of health care services, it is critical that support services provided to tenants in PSH programs like the one described in this paper, along with associated health care services, are provided in an inclusive environment by staff knowledgeable in culturally responsive practices and challenges commonly faced by people experiencing and exiting homelessness.

Clinicians and policymakers can also use data on health care costs associated with housing to improve the development and implementation of housing interventions. Pre- and post-housing trends in health care costs associated with crisis care and outpatient care were noticeably different for tenants who did and did not utilize health care services in the year prior to housing. Specifically, statistically significant reductions in costs associated with crisis care were observed among tenants utilizing health care services in the year prior to housing, although these costs overall may be offset by increased costs of outpatient care for tenants who did not utilize health care services in the year prior to housing. To the extent that information on health care costs can be used to make future medical costs more predictable (Wright et al., 2016), these findings suggest that health care systems implementing PSH programs may be able to more accurately predict future medical costs associated with these programs by using algorithms that incorporate factors like pre-housing utilization of health care services when modeling cost projections. Research has clearly shown, however, that health care costs are not a proxy for health care needs, and using cost algorithms to project future health care needs introduces significant racial bias into health care decision making (Obermeyer et al., 2019). Because utilization of health care services among individuals experiencing homelessness and living with mental illness is intertwined with a number of biopsychosocial and environmental factors, including the availability of services within a community and racial discrimination, among other factors, an important question to consider is whether different algorithms for predicting future service utilization might be needed for different groups of consumers. So, future research is needed to determine whether variables such as pre-housing patterns of health care utilization can improve prediction of actual costs and, importantly, do so in a way that does not perpetuate racial biases in the delivery of health care services post-housing.

Several limitations must be considered when interpreting these findings. First, information on the number of visits and associated professional charges pre- and post-housing was derived from billing data, which was incomplete for tenants in this sample. That is, rates of utilization drawn from both medical records and billing data revealed slightly higher rates of service utilization than were reflected in our billing data. Because incomplete data is a common problem associated with billing and claims data, visit and cost estimates derived from these data are likely underestimates (Tyree et al., 2006). Second, health care utilization outcomes were only available for services received in the UH system, and information regarding health care services obtained from other hospitals and/or clinics in the area is unavailable. UH is located in a large city in the Midwestern U.S. with three other major hospitals located within a 10-mile radius; however, UH is the only safety net hospital in the metropolitan area. So, while it is likely that some of the tenants in this sample may have sought and received services from other medical centers during the two-year evaluation window, it is also likely that UH was the primary source of medical care for many of these tenants. Third, the onset of the COVID-19 pandemic during the evaluation period may have impacted our findings given that the post-housing period overlapped with the onset of the pandemic and associated lockdown and other public health precautions that took place starting in March 2020. It is unknown whether the pandemic had any effect on how these patients used health care. If anything, the results presented here could underestimate what the true impact of housing and services would be in the absence of the pandemic, given that people were generally thought to be avoiding “unnecessary” visits to doctors and hospitals. Fourth, methodological limitations, including the small sample size, short follow-up period, and lack of a comparison group, limit our ability to draw firm conclusions about the extent to which changes in health care utilization outcomes were due to the 500 in Five initiative or other factors. Moreover, the lack of a comparison group may have led to confounding based on age differences between groups that did and did not utilize health care services in the year pre-housing. Fifth, because of the exploratory nature of this evaluation, we conducted multiple comparisons that may have statistically increased our chances of finding some pre- and post-housing changes on key outcomes simply by chance alone. Future studies with more rigorous designs and corrections for multiple comparisons will be essential in drawing more firm conclusions about the impact of PSH interventions on health care outcomes. Despite these limitations, this evaluation joins a growing body of literature documenting the positive impact of housing as a health care intervention and key strategy in ensuring that those living in poverty and those vulnerable to poverty and its consequences have equal rights to economic resources and access to basic services, including health care.