Homelessness compounds the difficulties of persons with severe mental illness, both exacerbating their psychiatric symptoms and precluding effective treatment. Moving individuals with severe mental illness off the streets, out of emergency shelters and into housing has thus been a top priority for mental health service systems and those who advocate for clients. Yet uncertainty about the best housing options to provide and resistance by some homeless persons to the housing options that are available has made it difficult to design appropriate policies and effective programs.

The Pathways Program in New York City was designed to lessen this resistance by offering independent apartments to persons with severe mental illness who have been living on the streets and rejecting offers of services or service-oriented housing (Tsemberis and Eisenberg, 2000). Pathways offers this housing with no prerequisite transitional residential programs and only minimal ancillary service requirements, and yet has achieved a retention rate of 80%. In 2002-2003, the Metro Boston Region of the Massachusetts Department of Mental Health (DMH) adapted this program for Boston’s service system in a program termed Housing First. If this program model succeeds in Metro Boston, it could lessen the need for extensive outreach programs, reduce expenditures for staffed transitional residences, and, most importantly, improve the quality of life and ultimately the treatment outcomes of DMH clients who are homeless.

The 2003 Survey was designed to increase understanding of the types of persons potentially eligible for Housing First services. It built on the knowledge developed in the 1990-1995 Boston McKinney Project housing study and reexamined some of the dilemmas for effective housing policy that the McKinney Project findings highlighted. In 1990 baseline McKinney surveys, most persons in the shelters funded by the Department of Mental Health, Metro Boston Region, desired to live independently, but the strength of this desire varied inversely with their ability to do so. In addition, clinician ratings of readiness for independent living (Schutt and Goldfinger, 1996) tended to predict successful outcomes, while those clients who desired to live alone but who clinicians rated as needing support were at high risk of failure (Goldfinger et al., 1999). Substance abuse was also a key predictor of poor housing outcomes, even though substance abusers were more eager than others to live independently.

This new survey describes the residential preferences of homeless mentally ill persons who use either generic shelters or a DMH transitional shelter in Boston, clinicians’ assessments of these same persons’ needs, and the correspondence between homeless persons’ preferences and the clinician assessments. It also indicates some of the predictors of homeless persons’ needs and identifies the extent to which these needs changes over time in the DMH shelter.

Methods

The population for this research was individuals staying at one of Metro Boston’s four transitional shelters for persons identified as having severe and persistent mental illness, as well as persons using any of the three large generic shelters and one day program for adult individuals in Boston who were identified by Department of Mental Health outreach psychiatrists or shelter staff as being seriously and persistently mentally ill. Interviewing occurred over several days in the four generic shelters and over two days in the baseline interviewing in the DMH shelter. All shelter users who were considered by clinicians to be seriously and persistently mentally ill and were available at the time of the interviews were invited to participate. There was no regular schedule for appointments with the outreach psychiatrists that would have allowed a more systematic sampling procedure. A consent form approved by the Harvard Medical School Institutional Review Board was read to each person who assented to the interview. The form included a separate consent to allow a designated clinician to complete the clinician rating forms for the subject.

At the Metro Boston DMH transitional shelter, one-third of the approximately 60 residents available on two different nights were selected for the study. At the three generic shelters and one day program, the sample consisted of 20 persons who were on the caseload of a DMH outreach psychiatrist or were considered to be eligible for psychiatric outreach by a clinician on the shelter’s staff. Of 24 persons in these four locations who were read the project consent form, two declined to sign. Both were young men.

Client interviews were conducted by trained research staff at a time and place agreed to by the client. These interviews continued for two months (December 24 2003 – February 20 2004). Data were also gathered about the persons who were interviewed from their DMH outreach psychiatrist or the referring shelter clinician. Outreach psychiatrists and shelter staff were not asked to complete forms about the interviewees unless and until the interviews had been completed and the interviewees gave their written consent. (Only one person who had consented to the interview subsequently refused to consent to having a clinician report on his needs. He was not included in the study.) Consent to collect the clinician information was given on a form that conformed to HIPAA requirements.

After four months, the clinicians at the DMH shelter completed the three clinician rating forms for all subjects initially interviewed at that shelter. No such followup was conducted at the generic shelters due to changes in outreach staff.

The sample and methods were shaped by practical constraints. The homeless outreach clients were all approached in generic shelters, where the two outreach psychiatrists had regularly scheduled visits. It was not possible to conduct interviews with potential outreach clients on the streets. Shelter clinicians helped in recruitment of generic shelter clients to be interviewed. For this pilot study, just one of the four DMH shelters was chosen for the interviewing.

Homeless subjects were interviewed with a revised version of the original McKinney residential preferences instrument. Staff completed four forms on each subject, all adapted from the McKinney project: a housing recommendation form, a residential safety form, and the Life Skills Inventory (Rosen et al., 1989) at baseline and three months. Two research assistants from the Graduate Program in Applied Sociology at the University of Massachusetts Boston and the first author conducted all interviews.

Multiple indexes were constructed from the preference interview and the clinician rating forms using procedures developed in the Boston McKinney Project. All indexes used in this report met standard criteria for inter-item reliability (see Appendix). Comparisons of average index scores between shelters and changes in their value over time were tested for statistical significance with t-tests and analysis of variance, and only those differences meeting accepted criteria are discussed. Comparisons were also made in clinician ratings between consumer groups defined in terms of the correspondence between consumer residential preferences and clinician residential recommendations. Since the study was designed only to test instruments and data collection procedures, no additional subject characteristics were measured and hence no multivariate analyses are conducted.

Findings

Shelter Satisfaction

Overall, satisfaction with living in the shelters was moderately high, with almost half rating themselves as “satisfied” and a total of 60% indicating they were either “satisfied” or “very satisfied” (Table 1).

Table 1

Satisfaction with living in the shelter

	Percent
Very satisfied	15%
Satisfied	45
Dissatisfied	25
Very dissatisfied	15
Total	100% (40)

The overall satisfaction level was slightly higher at the DMH shelter (Figure 1).

Satisfaction was highest with shelter staff and shelter security, while it was lowest with the available space, privacy, and the other people using the shelter (Figure 2). Respondents at the DMH shelter and the generic shelters were similar in terms of satisfaction with staff, privacy, security, comfort, help with benefits and the number of people, but respondents at the DMH shelter were more satisfied with “the kinds of people living here” and less satisfied with “your freedom” at the shelter.

Residential Preferences

Most respondents were eager to move out of the shelter (Table 2), even if taking medication or participating in substance abuse treatment were pre-conditions for this move (Table 3).

Table 2

Feeling about leaving this shelter

	Percent
Very excited	69.0%
Somewhat excited	19.0
Somewhat unsure	7.1
Very unsure	4.8
Total	100% (42)

Table 3

Want to move into housing even if…

	medication was a condition	participation in s/a treatment was a condition
	Percent	Percent
Move	83.3%	81.0%
Not sure	7.1	7.1
Stay	9.5	11.9
Total	100% (42)	100% (42)

Feelings were mixed about “keeping in touch” after leaving the shelter, with about half of the respondents reporting they would want to keep in touch with any of the other people at the shelter and two-thirds expressing an interest in keeping in touch with any of the service staff (Table 4)

Table 4

Will keep in touch after move

	with staff	with people
	Percent	Percent
No	28.6%	45.2%
Unsure	4.8	4.8
Yes	66.7	50.0
Total	100% (42)	100% (42)

These attitudes did not vary appreciably between the DMH and generic shelters (Figure 2).

When offered alternative living options, respondents expressed a clear preference for living alone and without staff (Table 5). The preference for living without roommates was strongest, expressed by about four in five respondents; even when the alternative was living with just one or two others, “where you had your own bedroom,” 86% preferred instead to live alone in a small two-room apartment. The same marked preference for independent living was expressed when the alternative was living in a home that was managed by the residents themselves (Table 5).

The preference to live in a place without staff support was less strong than the desire to live without roommates. Just over half the respondents preferred having no staff to having full-time staff “to help you manage in your new place” (and 10% were neutral on this issue), but only 28% preferred to live in a home managed by the residents as compared to a home managed by staff (Table 5). These preferences did not differ appreciably between the DMH shelter and the generic shelters.

Table 5

Respondents’ preferences for the alternative living options.

Option 1	6 or 7 others	1 or 2 others	6 or 7 others	Full-time staff	Resident managed	Resident managed
Option 2	Alone in small apt	Alone in small apt	1 or 2 others	No staff	Staff managed	Alone in small apt
Prefer option 1	14.3%	14.3%	2.7%	38.1%	27.5%	11.9%
Neutral	7.1	0	10.8	9.5	10	4.8
Prefer option 2	78.6	85.7	86.5	52.4	62.5	83.3
Total	100% (42)	100% (42)	100% (37)	100% (42)	100% (40)	100% (42)

Perceived Readiness for Independent Living

In spite of their marked preference for independent living, many respondents liked the idea of having someone to help them with the things they have a hard time managing alone, after they moved into their own place (Table 6). Almost three-quarters liked the idea of having such help at least somewhat, although 17% disliked the idea a lot. This preference did not vary between the DMH and generic shelters.

Table 6

Feeling about having someone to help with the things hard to manage alone

	Percent
Like the idea a lot	57.1%
Like the idea somewhat	16.7
do not really care/not sure	2.4
Dislike the idea somewhat	7.1
Dislike the idea a lot	16.7
Total	100% (42)

When asked about specific things “people may have to do when they live in their own place,” almost all respondents rated themselves as able to use public transportation by themselves, while many had doubts about their ability to fill out forms and budget money by themselves (Figure 3). Other activities involved in living independently were rated as, on average, “OK” by most. DMH shelter residents felt somewhat less confident in their ability to shop and clean house than generic shelter residents, but they were more confident in their ability to deal with neighbors on their own.

Clinician Ratings

Clinicians rated each homeless person in the survey in terms of most appropriate residential placement, life skills and risk level.

Residential Recommendation

The overall clinician residential recommendations did not differ between the DMH shelter and the generic shelters; shelter residents in both settings were rated as, on average, about equally likely to succeed in independent and group living arrangements (Figure 4).

Across the specific residential characteristics, clinicians rated generic shelter residents as less likely to be able to manage on their own without substance abuse treatment compared to DMH shelter residents, but as less likely than DMH shelter residents to need staff visits or staff designed activities (table not shown).

Life Skills

Problems with life skills were rated in terms of ability to get along with others (“prickliness”), ability to care for oneself, and sociability. On average, respondents were rated between the levels of not having the problems of prickliness and inability to self-care and having these problems to a slight degree, but were rated as unsociable between a slight and moderate amount (Figure 5). Sociability was seen as a greater problem for respondents using the generic shelters than for those using the DMH shelter.

Risk Level

The clinician raters reported only a moderate level of risk for most of the nine risk factors. The greatest risk was reported for assaultive/destructive behavior, substance abuse, non-compliance with psychotropic medication, and financial problems (Figure 6). Clinicians perceived a low level of risk, on average, in terms of parasuicidal and suicidal behavior, victimization, fire setting and medical problems, and rated few sample members as responding to command hallucinations to harm themselves or others. These risks were seen as comparable for the DMH and generic shelter samples, with the exception of medical problems, which were seen as a greater risk in the DMH shelter, and medication non-compliance, which was seen as a greater risk in the generic shelters.

Substance Abuse

Clinicians rated about half the sample members as being moderate to severe substance abusers. The substance abuse index indicates substantial variability in the sample (Figure 7).

Figure 7

Consumers who received high scores on the substance abuse index at baseline were more likely to be rated by clinicians as being more intrusive (Figure 8). They did not differ in sociability or self care.

Consumer Preference and Clinician Recommendation

Consumers’ residential preferences did not correspond to clinicians’ residential recommendations: Those consumers who desired greater independence were no more likely to be recommended as ready for independent living by clinicians than were consumers who were seeking more support. However, the correspondence between consumer preferences and clinician recommendations was related to consumer functioning. Compared to the other groups, the group of consumers who sought more independence but who were rated by clinicians as needing more support were judged as having poorer life skills and to be at greater risk in four areas: intrusiveness, assaultiveness (except when compared to the consumers who sought less independence than their clinician recommended), substance abuse, and non-psychiatric medical problems (Figure 9). No differences were detected between the groups defined jointly by consumer and clinician preference in terms of the other risk factors.

Consumers who sought more independence but who were rated by clinicians as needing more support were also more likely to be rated as improving in risk due to substance abuse after four months in the DMH shelter (table not shown). There were no other differences between these four groups in terms of change in other risks or life skills.

Change over Time

The clinician raters at the DMH shelter provided a follow-up assessment four months after their first assessment. On average, the 19 DMH shelter residents had improved during this period on the intrusiveness dimension of the Life Skills scale, but not on the dimensions of self-care or sociability (Figure 10).

There was no change in clinician residential recommendations over this period, but among the eight dimensions of risk assessed, clinicians identified an improvement in risk due to substance abuse (scores on the composite substance abuse index also declined) (Figure 11).

Conclusions

In summary, a majority of shelter users in this sample reported that they were at least somewhat satisfied with the shelter, rather than dissatisfied. When asked about specific shelter features, shelter users reported the highest levels of satisfaction with shelter staff and the safety of their shelter. DMH shelter users were slightly more satisfied than generic shelter users with their shelter overall, and particularly so in terms of the other people staying there, but they were no more likely to want to keep in touch with other shelter guests after leaving the shelter. Although the DMH shelter users were less satisfied with the amount of freedom they had in the shelter than those in the generic shelters, the relative satisfaction with co-residents may indicate the relative success of DMH shelter management in maintaining a supportive environment. The relative social satisfaction of the DMH shelter users was also reflected in their greater confidence than the generic shelter users about their ability to deal with neighbors in any future home.

In spite of their relatively positive feelings about each of the shelters in which they stayed (including the one day program studied), the homeless persons in this study were eager to move out of the shelters they were using into independent apartments. However, although most rejected the idea of living in a group home, about half indicated that some level of staff support would be helpful in the new residence.

Clinicians evaluated the individuals in the sample as needing even more support than did the homeless persons themselves, and the clinicians’ evaluations did not correlate with variation in the homeless persons residential preferences. Homeless respondents who were more eager to live on their own, without roommates or staff, were not more likely to be judged by their clinician as able to live on their own without support.

Each of these empirical patterns replicates those found with the Boston McKinney Project in 1990, which also studied homeless persons who were identified as severely mentally ill but sampled exclusively from the three DMH shelters not included in this study. The replication of these patterns with a different sample after the passage of 15 years indicates that they reflect relatively stable orientations of both persons who are homeless and severely mentally ill and of the clinicians who work with them. This stability is also reflected in the lack of variability between the two shelter types in either homeless persons’ residential preferences or their clinicians’ residential recommendations.

The extent of correspondence between homeless persons’ residential preferences and their clinicians’ residential recommendations had a strong relationship with the functioning and needs of the homeless persons, as had also been the case in the Boston McKinney Project sample. Those homeless persons who were most interested in living independently but were judged by their clinicians as most in need of support were more likely to be assaultive and to have substance abuse and other medical problems.

The DMH shelter users seemed to be more treatment oriented than their generic shelter counterparts. The DMH shelter had more procedures in place to support medication compliance than the generic shelters, and thus it is not surprising that the DMH shelter users were viewed by clinicians as more medication compliant, although they also were rated as having more medical problems than the generic shelter users. In addition to being less compliant with psychiatric medications, the generic shelter users were also rated as having more problems socializing. The Housing First model, which bypasses the transitional shelters, was designed to serve this less treatment compliant and less sociable subset of homeless persons with mental illness.

Based on the evaluations by their own clinicians, the DMH shelter seems to have had some success in improving the functioning of the persons who were staying there at the start of the study. After four months at the DMH shelter, the shelter users were judged to have reduced somewhat their levels of intrusive behavior and their substance abuse. In particular, it was the persons who rated their need for independence so higher than reflected in their clinician’s residential recommendations whose initially high levels of substance abuse declined over the next four months.

The limited availability sample used in this pilot study precludes generalization to the larger population of homeless persons with mental illness, but the replication of empirical patterns previously identified in larger studies suggests that the respondents in this limited sample were not markedly different from other such shelter users. This stability also reflects the failure of the numbers of homeless persons with mental illness to have declined in the 15 years between the Boston McKinney Project and this pilot study. We hope that the findings from this study will help to stimulate more attempts to provide these persons with the housing that they are so eager to obtain. A programmatic focus on the discrepancy we have identified in the orientations of shelter users and their clinicians, as well as learning from relatively successful shelter practices we have identified may help to improve efforts to move homeless persons with mental illness into housing.

References

Goldfinger, Stephen M., Russell K. Schutt, George S. Tolomiczenko, Larry J. Seidman, Walter Penk, Winston Turner, and Brina Caplan. 1999. “Housing Placement and Subsequent Days Homeless Among Formerly Homeless Adults with Mental Illness.” Psychiatric Services, 50:674-679.

Rosen, Alan, Dusan Hadzi-Pavlovic, and Gordon Parker, G. 1989. “The Life Skills Profile: A Measure Assessing Function and Disability in Schizophrenia.” Schizophrenia Bulletin 15:325-337.

Schutt, Russell K. and Stephen M. Goldfinger. 1996. “Housing Preferences and Perceptions of Health and Functioning Among Homeless Mentally Ill Persons.” Psychiatric Services, 47:381-386.

Tsemberis, Sam and Ronda Eisenberg. 2000. “Pathways to Housing: Supported Housing for Street-Dwelling Homeless Individuals with Psychiatric Disabilities.” Psychiatric Services, 51:487-493.

Appendix

Index Statistics

Descriptive Statistics
	Mean	Std. Deviation	Cronbach’s α
CLINICIAN RATING SCALES
Life Skills Prickliness*	1.63	0.46	0.79
Life Skills SelfCare*	1.51	0.54	0.84
Life Skills Sociability*	2.19	0.58	0.84
Overall Life Skills Rating*	1.87	0.41	0.89
Baseline Recommendation for Support	2.60	0.65	0.76
Baseline Risk Score	2.14	0.74	0.68
SHELTER GUEST SCALES
Life Satisfaction	2.23	0.95	0.85
Preference for Independent Living	4.30	1.07	0.78
Ability to Manage Daily Tasks	1.13	0.24	0.75
Preference for Staff Support	2.79	1.20	0.49
Shelter Satisfaction	2.21	0.63	0.85
Severity of Substance Abuse	2.70	0.91	0.91

*Higher scores indicate poorer skills.

James Feldman, MPH, MD1,3, Eileen Reilly, MD1,3, Martha Schinagle, MD1