219 West Burnside
Portland, Oregon 97209
Contact: Neal Rendleman, M.D.
Phone: (503) 241-3836
Key words: housing status; utilization
of clinic and welfare services
The purpose of this
study was to determine factors which predict
improvement in the housing status of homeless
patients at the Old Town Clinic (OTC).
METHODS
This project was designed as a preliminary
test of two major theses. Foremost was
the impression that patients with a high
degree of engagement and cooperation with
the OTC program had an improved outcome
in many spheres of their lives including
health status, employment, social interaction,
and hygiene. Second was the impression
that various subsets, including women,
younger homeless people, and those with
primary psychiatric problems had a greater
likelihood of improvement.
The original design was that of a prospective
study of all new patients who presented
between October 1, 1996, and March 31,
1997, who identified themselves as being
homeless and who were seen by a health
care provider. Established housed clinic
patients who became homeless over the
course of the entry period were also included.
These patients were then monitored for
housing status through the end of July
1997. Patients were excluded for the following
reasons: persons who left the office unseen
by a clinician after they filled out a
questionnaire; persons who reported that
they were housed during the entry period
and remained so for the length of the
study; and persons who had multiple charts
under the same name.
Each new patient who attended the clinic
during the course of the study filled
out a brief questionnaire which described
their housing status, medical insurance
status, and whether or not they received
certain forms of government assistance.
Those who reported a category of housing
such as sleeping out, in shelter or car,
doubled up, or not paying full rent, were
monitored as homeless. After enrollment,
housing status was tracked through outreach.
Data on housing status which are routinely
collected on every patient on a monthly
basis, were recorded, and retrieved in
a variety of modes. On each visit, all
patients were routinely asked about their
housing situation.
The overall health status of subjects
at their first visit to the clinic was
assessed retrospectively using an illness
severity chart (on a five point scale)
in conjunction with clients charts.
The outreach done to follow housing status
included mailing posters to local shelters
asking patients to come in for a follow-up
and posting notices under bridges and
in other places homeless people were known
to gather. Following this original protocol
resulted in complete housing follow-up
data on only 47 of the 319 patients entered
into the study. The follow-up protocol
was therefore modified to include a review
of the chart for housing status information
that had been recorded after October 1,
1996 and was complete enough to be scored
on the established housing scale. Further
reports of housing status were retrieved
from data on Medicaid cards or from supplemental
clinicians notes. This expanded protocol
netted housing data on 40 additional persons,
resulting in complete follow-up data on
87 persons in this study. Previous studies
attempting to track homeless people have
shown that even a multifaceted outreach
program on a defined homeless population
is stressed to find 10 percent of its
subjects (Gelberg L, Linn L, 1992).
Technicians abstracted information from
medical records, scoring each visit to
the clinic, using a chart abstraction
tool designed by program staff. Detailed
instructions and close support from the
investigators were provided to the technicians
during scoring. This protocol was designed
to measure intensity of engagement with
clinic programs and to reflect the intensity
of the patients commitment to recovery
through the clinic.
The abstracted chart data and the initial
housing questionnaire were entered into
a computer database and coded for numerical
analysis. Frequency distributions for
race, age, gender, severity of illness,
and change in housing were generated.
Changes in housing status were also dichotomized
such that an improvement in housing scored
one, and no improvement scored zero. This
dichotomization allowed the use of multiple
logistic regression if two or more housing
statuses were available. The difference
was taken between the first and the last
housing status. Pearson R correlation
coefficients were calculated among all
of the numerical and dichotomous variables.
Variables that met the original hypothesis
were put into several models for the multiple
logistic regression. Length of follow-up,
gender, number of visits, severity of
illness and insurance status were forced
into a multiple regression model. They
were each tested individually in logistic
regression models. Chi squared, Mann-Whitney
U, and one-way Analysis of Variance tests
were performed between the persons with
complete follow-up and those with incomplete
follow-up for several variables to detect
differences in these groups.
The final sample contained 312 persons.
Complete follow-up data was available
for 87 of the 312. Comparisons between
those with incomplete follow-up and those
with complete followup data were performed.
There were no significant differences
in age, race, or food stamp program participation.
Severity of illness was significantly
different between those with and without
complete follow-up using a Mann-Whitney
U (p=0.006). Total number of clinic visits
were significantly (p<0.000) different
when tested by one-way ANOVA. These groups
also had significant differences between
rates of insurance coverage (p<0.000).
The average age of subjects was 40.4 years.
The distribution of races is shown in
Table 1.
Table 1. Frequency Distribution
of Race of Persons with Complete Follow-up
| |
Frequency |
Percent |
Valid
Percent |
Cumulative
Percent |
| Valid
Unknown |
6 |
6.9 |
6.9 |
6.9 |
| Black |
10 |
0.5 |
1.5 |
8.4 |
| Hispanic |
3 |
3.4 |
3.4 |
21.8 |
| White |
64 |
73.6 |
73.6 |
95.4 |
| Native American |
3 |
3.4 |
3.4 |
98.9 |
| Other |
1 |
1.1 |
1.1 |
100 |
| Total |
|
87 |
100 |
100 |
Not surprisingly, differences were found
between those with and without complete
follow-up. The difference in the number
of encounters is due to the method of
data collection used. More visits made
it more likely that housing status information
would be gathered. Patients receiving
all of their care in one calendar month
or for one acute illness would not receive
follow-up demographic data collection.
The differences between groups in the
average severity of illness are more difficult
to explain. One would expect that sicker
people would require more care and hence
have more visits. There was a nonsignificant
negative correlation between severity
and the total number of visits. Sixty-two
percent of patients whose health was rated
poor were ones with untreated psychosis
or drug use. Clinic visits are a low priority
for these groups (Macnee C, Forrest L,
1997). It may be that this study had inadequate
power to see the hypothesized effect.
This seems less likely as the sign of
the correlation coefficient is the opposite
of what was expected.
In the group with complete follow-up,
engagement score was highly correlated
(Pearson R = .982, p<0.01) with the
number of clinic visits. This meant that
the complex, elaborate, and time consuming
scale used for data collection provided
no additional prognostic information above
that provided by counting the number of
visits. Visit count was the only predictor
variable that had a significant effect.
The odds ratio for any improvement in
housing was 1.22 (Cl 1.05-1.41) per clinic
visit compared to someone who did not
visit the clinic, controlling for both
age and gender effects. That is, a person
has approximately a 20 percent chance
of improving their housing status per
clinic visit. This association supports
the thesis that engagement at the clinic
would be correlated with improved housing
status. Length of follow-up and severity
of illness did not yield odds ratios significantly
different from 1. This may be an artifact
of the relatively short span of follow-up.
A social service facility in Portland
reported excellent results from intense
intervention to raise housing status during
this time.
Insurance status was positively correlated
with engagement, reflecting the fact that
clinicians are able to order outside tests
and procedures when necessary for patients
who are covered by the Oregon Health Plan.
This trend did not extend to increased
numbers of encounters or improved housing
status. This might mean that the multi
factorial approach of the HCH program
did not have sufficient time to reach
this group to improve their housing. It
may also imply that access to health care
alone without the multifaceted, holistic
care provided by the HCH program is inadequate
to improve housing status.
Neither age, gender, nor race predicted
a change in housing status as expected;
this could be because it is genuinely
absent, the time course was too short,
or the changes in each group were too
small to be measured individually. Several
variables yielded predicted improvements
in housing, but these were not statistically
significant results. This study probably
had inadequate power to detect any effects
due to these variables, if they were present.
So no conclusions can be drawn from these
data regarding these factors in either
direction. The charts of 12 patients,
who became homeless during the study,
were qualitatively reviewed in detail.
They included several patients with very
high patterns of utilization, both during
their periods of homelessness and while
they were successfully housed. This group,
which had a heterogeneous mix of different
factors leading to homelessness, were
uniformly a set of people who were acculturated
to homeless lifestyles. The chronic relapsing
nature of their conditions, including
schizophrenia, manic-depressive illness,
alcoholism, injection drug use, and compulsive
behaviors, puts them at risk for recurrent
homelessness. All of them had severe ongoing
medical conditions that are exacerbated
by homelessness (e.g., diabetes, emphysema,
osteomyelitis, and epilepsy). A health
care for the homeless program may simply
have to accept that acculturation to homelessness
is an inevitable adaptive quality of these
patients and accommodate their needs while
making an effort to proceed in a holistic,
multi disciplinary fashion to ameliorate
all their conditions.
The use of a seven-point housing score
proved cumbersome and did not yield additional
information above a simple improveddeclined-
unchanged delta score. Although clinicians
may perceive that there is a gradation
of homelessness (i.e., from sleeping outside
to shelter to treatment programs), the
heterogeneous nature of the causes of
homelessness as well as society_s responses
to it result in a series of tracks which
are followed by members of the individual
groups. Thus, a battered spouse may leave
permanent housing, sleep with her children
in a car, get access to welfare benefits
through the clinic, and move directly
into a new permanent home. Several schizophrenics
tracked during the study moved directly
from sleeping outside to permanent housing.
Thus, the points scored on the housing
status scale proved to be a coarse measure
of success or failure of the program.
In conducting the study, program staff
attempted to obtain an accurate sample
of homeless patients seen in the clinic
during the 9 months of the study. A convenience
sample of other charts inclusive of known
blind, deaf, illiterate and other high-risk
patients were reviewed. This review turned
up no person who was homeless and had
been seen during the study period. Given
that 2,400 visits were generated for 1,200
patients during the interval, some small
number may have been missed because they
concealed their homelessness or because
of the clinic_s failure, but whatever
the case, the number was small. The study
group was not by any means a random sample
of the homeless population in Portland.
Factors which prompt patients to seek
care include: shelters_ requirement that
they get screening for tuberculosis; referrals
by welfare caseworkers for possible disability
certification; desire to improve job status
through access to free hygiene supplies;
attempts at drug detoxification; requests
for refuge by battered spouses; availability
of meal tickets; and the usual complaints
of sickness which prompt visits to ordinary
medical facilities. Persons who are intoxicated,
hostile, or act out sexually are excluded
from the clinic, so these patients are
not part of the clinic population. Hence,
these results can only be applied to the
homeless population in Portland which
seeks care for some reason at the HCH
program and cooperates with routine registration
and consultation procedures. These data
may be comparable to clinic populations
in other cities, but other facilities
with other cultures and populations will
have different patient pools.
This study method of monitoring housing
status, engagement or number of visits
at a homeless clinic, and basic demographic
factors of patients, seems to be an effective
way to monitor change in housing in a
homeless clinic population over time.
The complicated measures devised during
the course of this study appear to have
no real advantage over simpler measures,
such as counting visits, in predicting
outcomes. Age, gender, and race were not
found to correlate with changes in housing
status, although limitations in the study
may have obscured an association. A number
of differences between those with complete
follow-up and with incomplete follow-up
were detected. A more complete understanding
of these differences could lead to ways
of engaging these persons more effectively
in health care and other support programs.
The number of clinic visits was significantly
associated with an improvement in housing
status in a motivated sub-sample of the
patient population. As this is a short
duration observational study of modest
size, this result cannot be taken as definitive,
but further investigation of the factors
predicting housing improvement is clearly
warranted.
