Introduction

There is evidence that past incarceration is a major risk factor for homelessness (Greenberg and Rosenheck 2008) and increasing concern about the needs of homeless people with a history of incarceration. Homeless adults with a history of incarceration may be particularly disadvantaged in exiting homelessness because of cultural stigmatization, formal barriers to using public housing subsidies and prejudices that make it difficult to obtain employment and income benefits (Breakey et al. 1992). For example, housing applicants with certain criminal offenses are excluded for rental subsidies through the federal government. An estimated 3.5 million Americans are thus ineligible for most federal housing assistance as a result of legislation passed in 1996 (McNiel et al. 2005).

Several studies, however, have compared the effectiveness of programs for homeless adults with and without a history of incarceration, and found similar levels of effectiveness. For example, at Seattle’s Downtown Emergency Service Center, a study of 347 homeless adults who moved into supportive housing found that those with a criminal history were able to retain housing continuously for two years at the same rate as those with no criminal history (Malone 2009). Supported housing in which case management is linked with housing subsidies has emerged as the leading service model for helping homeless adults to exit homelessness (Fakhoury et al. 2002; Rog 2004; Rosenheck et al. 2003).

Homelessness and incarceration are particularly important problems for veterans as some veteran cohorts, especially those who served during the early years (1973–1980) of the All Volunteer Force (members of the U.S. Military after the draft ended in 1973) are overrepresented in both the homeless population (Gamache et al. 2001) and the prison population (Greenberg and Rosenheck 2007; McGuire 2007).

In 1992, in response to evidence of extensive veteran homelessness, a collaboration between the Department of Housing and Urban Development and the VA established a supported housing program for veterans called the Housing and Urban Development-Veterans Affairs Supportive Housing (HUD-VASH) program. Veterans who are admitted into HUD-VASH are assigned a HUD-VASH case manager who assists clients in obtaining a housing voucher, locating and moving into an apartment, and providing intensive case management support. HUD-VASH has been found to improve housing outcomes, levels of social support (Rosenheck et al. 2003) and reduce substance abuse (Cheng et al. 2007). However, HUD-VASH does not specifically target homeless veterans with histories of incarceration and it is not clear whether a history of incarceration is a barrier to obtaining housing in this program. Since 2007, as part of a major initiative to end homelessness among veterans (National Alliance to End Homelessness 2008), the VA has expanded the HUD-VASH program to try to serve thousands of additional veterans.

In this paper, we examined data on veterans who entered the HUD-VASH program in 2008 and 2009 to (a) identify distinctive characteristics of those with a history of incarceration and (b) to evaluate whether, after adjusting for those characteristics, those with a history of incarceration are less likely to get housed and/or experience greater delays in the housing attainment process.

Methods

Source of Data and Study Design

This study used VA administrative evaluation data on veterans admitted to the HUD-VASH program during fiscal years (October 1–September 30) 2008 and 2009. The program was targeted broadly at homeless veterans without restriction, but with an emphasis on serving chronically homeless adults, those with mental illness and addiction disorders, families, female veterans, and veterans who had served in recent Middle East Conflicts. There were no explicit restrictions to entry based on past criminal offenses, except sexual offenses. Veterans were referred to HUD-VASH by both VA service providers and clinicians in the community. Based on measured housing acquisition times (with an average total of 103 days in this sample), we believe that studying veterans admitted to the program in 2008 and 2009 allowed ample time for veterans to complete the housing acquisition process. Given that data collection stopped in April 2011, subjects had a full fourteen months to obtain housing. Referring clinicians, including VA homeless outreach staff, completed formal admission forms and HUD-VASH case managers completed structured progress reports on the process of housing acquisition. The research design used was an observational cohort design that compared clients with histories of incarceration with other clients on socio-demographic and clinical characteristics at program entry and on housing acquisition times for various phases of the housing acquisition process after adjusting for potentially confounding baseline differences. Past involvement in the criminal justice system, as measured by reported lifetime incarceration at the baseline interview, was used to divide the sample into three groups: no incarceration history (NI); short-term incarceration history, defined as one year or less of incarceration (including any time spent in jail or prison); and longer-term incarceration history, defined as a total of more than one year of incarceration. A dichotomous comparison of one year or less versus one year or more was used to provide an, albeit imperfect, comparison of incarceration for misdemeanor as contrasted with felony offenses. All procedures were approved by the Institutional Review Boards at the parent site and at each participating site.

Measures

Background Characteristics and Mental Health Diagnoses

Information on age, gender, race, marital status, military service war era, lifetime episodes of homelessness and lifetime years incarcerated were based on participant self-report and confirmed by case managers.

Housing

Participants were asked where they spent the past 30 days in each of seven living arrangements that were collapsed into 6 categories: nights in own place (in own apartment, room, or house) nights in someone else’s place (in the apartment, room, or house of a family member or friend), nights in transitional/temporary housing (in transitional housing, transient hotels, and boarding homes) nights in residential treatment (in VA and non-VA residential treatment programs) nights in institution (in hospitals, nursing homes, prisons, and jails) and nights homeless (in shelters, outdoors, or in vehicles).

Employment and Income

Participants were asked the number of days they worked in the past month and the amount of income they received from employment (including compensated work therapy and supported employment) disability/public assistance (VA service-connection, welfare, etc.) and all other sources.

Clinical Status

Mental health status was assessed with both self-report and clinician-rated measures. Self-report measures included a mental health symptoms score, which was a summed score of 8 items from the Psychiatric subscale of the Addiction Severity Index (McLellan et al. 1980). A social quality of life score was rated by clinicians, which was collected using 3 items from the Heinrich–Carpenter Quality of Life Scale (Heinrichs et al. 1984). Clinician-rated measures included an observed psychosis rating scale (Dohrenwend 1982) the Global Assessment of Functioning (American Psychiatric Association 2000) and the Clinician Alcohol and Drug Use Scales (Drake et al. 1996).

To assess their working alliance with clients, case managers rated their alliance on 5 items using a 6-point Likert rating scale (Neale and Rosenheck 1995).

Housing Acquisition Times

Case managers documented critical dates during the housing acquisition process as participants achieved milestones through the process. These dates were used to calculate the number of days between each milestone. Specifically, we examined how many days it took for participants to progress from admission into the HUD-VASH program to visiting the Public Housing Authority (P.H.A) from visiting the P.H.A to obtaining a housing voucher, from obtaining the voucher to first looking at an apartment, from first looking at apartment to signing an apartment lease and from signing the lease to moving-in to the apartment. The total days it took from admission to each milestone was also calculated by summing each of the process times.

Data Analysis

First, the three groups (no incarceration (N = 5,023) incarceration less or equal to one year (N = 6,324) and incarceration more than one year (N = 3,210) were compared on sociodemographic, clinical and community adjustment characteristics. Chi square tests and analysis of variance were used to compare means as frequencies across the three groups, but the sample sizes were so large that virtually all comparisons were significant at p < 001. We thus determined that differences of greater than 20 % between either incarceration group and the non-incarceration group were substantial enough to require adjustment in multivariable analysis of housing attainment processes. In addition, because variation across sites has been shown to be the strongest predictor of housing process times (Tsai et al. 2011) we included as adjustment for dichotomous dummy coded covariates representing sites. Finally, since the measures of the housing attainment process were non-normally distributed, we conducted multivariate analyses on log-transformed measures.

Results

Characteristics at Program Entry

A majority of participants reported a past incarceration history with N = 6,324 (43.4 %) reporting a shorter history (≤1 year) and N = 3,210 (22.1 %) reporting a longer history (>1 year) of incarceration.

Before enrollment into the supported housing program, participants with incarceration histories were substantially more likely to be chronically homeless. This was defined as three or more episodes of homelessness or greater than one year of continuous homelessness [NI = 2,367 (47.2 %) SI = 3,581 (56.7 %) LI = 2,060 (64.3 %)] with more homeless episodes in the 3 years prior to admission, as well (NI = 1.69 ± 1.18, SI = 1.97 ± 1.31, LI = 2.16 ± 1.41 for the three groups respectively).

As the extent of incarceration history increased, so did the percentage of participants who were male, older, and who were separated or divorced. The proportion of White, Latino and other ethnicities all decreased with more extensive incarceration history while the proportion of African American participants increased [NI = 2,373 (47.6 %) SI = 3,057 (48.6 %) LI = 1,947 (61 %)]. Additionally, more incarceration was associated with having fewer children in legal custody, and fewer children residing with the participant.

Participants with incarceration histories were significantly more likely to be diagnosed with substance abuse, alcohol abuse/dependence and drug abuse/dependence. Incarceration history was also associated with higher rates of psychotic disorder diagnosis, bipolar diagnosis and dual diagnosis. Rates of depression, however were broadly similar across the groups [NI = 2,424 (48.9%) SI = 3,262 (52.3 %) LI = 1,550 (49 %)]. Rates of PTSD were similar across groups, but participants with incarceration histories were more likely to suffer from non-combat trauma, and less likely to suffer from combat trauma than the no incarceration group. The number of medical problems was 9 % higher in those with short incarceration histories and 22 % greater in those with long incarceration history when compared to no incarceration.

Military service in the Vietnam Theater of operations was associated with greater history of incarceration whereas service in Iraq/Afghanistan was associated with a reduced percentage of participants with incarceration histories. Incarceration was not associated with combat exposure as a higher percentage of participants with NI reported having received fire in a combat zone than those with short or long histories [NI = 1,164 (23.5 %) SI = 1,204 (19.3 %) LI = 458 (14.4 %) respectively].

Participants with short and long incarceration histories reported working more days (NI = 4.80 ± 9.03, SI = 5.29 ± 9.28, LI = 5.82 ± 9.73) but earned less total income (NI = 842.58 ± 600.37, SI = 813.13 ± 579.02, LI = 765.00 ± 37.81) than those with no incarceration history. They tended to receive less Service-Connected income, less welfare but more employment income and more non-Service-Connected pension income. Housing of participants in the 30 days prior to admission also varied; those with more incarceration histories spent more days in residential treatment (35–46 % more respectively) and fewer nights in someone else’s place (in the apartment, room, or house of a family member or friend) (23–34 % less respectively).

Quality of life measures such as Social Activity (NI = 3.75 ± 1.76, SI = 3.79 ± 1.69, SI = 3.92 ± 1.67) Social Withdrawal (NI = 4.08 ± 1.63, SI = 4.14 ± 1.60, LI = 4.21 ± 1.59) and sense of purpose (NI = 4.10 ± 1.45, SI = 4.03 ± 1.40, LI = 4.07 ± 1.40) assessed at time of entry into the HUD-VASH program, were very similar across the three groups.

Housing Attainment Process

Participants in each category achieved similar rates of successful housing, defined as moving into their own apartment, through the program. The no incarceration group’s rate of housing success was slightly lower at [N = 2,868 (57.1 %)] when compared to the short incarceration [N = 3,745 (59.2 %)] and long incarceration groups [N = 1,873 (58.3 %)]. Our analysis of veteran’s progress through each step of the housing attainment process demonstrated only modest differences between the three groups. We analyzed the total number of days participants took to reach each milestone from the date of admission to HUD-VASH and found modest differences varying by 6–18 %. To better understand the progression of each group to housing attainment, we also analyzed the number of days between each milestone. Except for one exception discussed below, we found only modest differences with most differences in days varying between 2 and 26 %. Those with incarceration histories, both short and long, progressed to many milestones faster than the group with no incarceration history. Examples include: time to visiting the P.H.A time to obtaining a housing voucher, and time to taking their first look at possible apartments.

In order to address the skewed, non-normal distribution of data, we first conducted log-transformations of the data (housing acquisition times). We then conducted a multivariate analysis adjusting for baseline measures where between-group differences of greater than 20 % were detected, as well as adjusting for the site. With these adjustments, we found that there were few significant differences in the average of the log-transform of the outcome variable between groups after admission into the HUD-VASH program.

One exception was the number of days between participants signing a lease and moving into housing (Lease To Move In: NI = 4.86 ± 14.41, SI = 3.77 ± 12.4, LI = 17.35 ± 590.80) where participants with a longer incarceration history took almost two weeks (257 %) longer to move into housing than the other two groups. However, the mean total days from admission to move-in was similar across groups (NI = 109.50 ± 90.44, SI = 97.74 ± 73.74, LI = 101.58 ± 78.64) with the no incarceration group taking longer than those with short or long incarceration histories to complete the process. Case managers across all three groups felt that a positive clinical relationship was established at similar rates [NI = 2,865 (57.8 %), 3,533 (56.5 %), 1,809 (57 %)] and the perceived alliance rating between case managers and participants did not vary significantly (NI = 4.85 ± 0.85, SI = 4.76 ± 0.86, LI = 4.79 ± 0.86).

Discussion

This study found that most homeless veterans enrolled in this large supported housing program have been involved in the criminal justice system. Over 65 % of HUD-VASH clients had been incarcerated prior to admission. Prior studies of the general homeless population found rates of incarceration as high as 50 % in 1996 (Burt et al. 1999) but more recent studies estimate the rate at 25 % (Metraux and Culhane 2006). Our study concurs with these findings but shows a higher prevalence of incarceration among homeless veterans; possibly because it is a clinical sample and veterans are overwhelming male.

Before entry into the HUD-VASH program, homeless veterans with incarceration histories showed greater psychiatric symptoms and more substance abuse diagnoses than those veterans with no incarceration history. This is consistent with previous studies on the general homeless population (Kushel et al. 2005; McGuire and Rosenheck 2004). PTSD increased with incarceration history. Some studies suggest that incarcerated veterans are more likely to have been involved in combat than non-incarcerated veterans (Boivin 1987); however we found a higher percentage of those without an incarceration history reported combat experience (21.9 %) than those with short incarceration (19 %) or long incarceration history (13.7 %) and that this particularly pertains to veterans who reported service in recent Middle East conflicts.

Despite many differences in demographics, mental and physical health, veterans with a history of incarceration were not statistically any less likely to obtain housing, suggesting that an incarceration history does not have a negative effect on housing acquisition in HUD-VASH. Prior studies of housing programs for homeless people with serious mental illness have reported discrimination based on criminal history, where “public entities and individual housing operators use criminal background information to assess for risk and screen out undesirable people” (Malone 2009). Recent studies show that, in service settings, criminal history does not provide good predictive information about housing placement or retention (Malone 2009). While we did not study housing retention, a history of incarceration did not appear to be a substantial factor in terms of assessed therapeutic alliance or progression through the housing process. This finding is consistent with a previous study that used a non-VA sample of chronically homeless adults and found incarceration histories do not hinder attainment of supported housing (Tsai and Rosenheck 2012). The HUD-VASH program does not impose entry restrictions based on past criminal offenses and also provides extensive case management, two factors which might mitigate the historic barriers that a criminal justice history creates for access to public housing. The results demonstrate that the HUD-VASH program is a fair housing assistance program and helps homeless veterans with histories of incarceration, mental illness and/or substance abuse just as well as homeless veterans without these histories. Therefore, we conclude that the HUD-VASH program represents an effective model for housing assistance programs as reflected in its ability to successfully assist individuals whose histories might otherwise impede them from exiting homelessness.

Several limitations require comment. The main limitation of these data is that they were collected during a period of rapid program implementation when the primary focus was on timely program implementation rather than program evaluation. Thus there is an identifiable, but potentially important, amount of missing data. Nevertheless, we adjust in our key analyses for site and have a large national sample. This study was also limited in that all measures were based on self-report, which may be susceptible to various reporting biases and inaccuracies. Participants were grouped based on their reported history of incarceration because administrative information about the nature of their criminal convictions and sentences served was not available. Although the cohort classified as having extensive incarceration had similar outcomes in HUD-VASH, more research is needed on the pathways in which this group became involved in the criminal justice system. Finally, although clients in HUD-VASH showed large improvements in the housing domain, we did not present outcome data on work status, clinical status, or quality of life, regardless of criminal history. Additional efforts are needed to evaluate and improve the social reintegration of formerly homeless veterans after they obtain housing. Nevertheless, this study, like others, is reassuring in that veterans with different incarceration histories appear to have experienced few differences in progression through the housing process after enrollment in HUD-VASH.