Metropolitan Government of Nashville
and Davidson County
311 23rd Avenue, North
Nashville, Tennessee 37203
Contact: Celia Larson, Ph.D.
Phone: (615) 862-7926
Key words: SF-12 validity; homeless
patients
The purpose of this study was to explore
the validity of the SF-12 in assessing
and monitoring health status among homeless
persons by testing its applicability at
a day shelter.
METHODS
A convenience sample of 145 homeless persons,
who seek services at the Campus for Human
Development (CHD) a day shelter located
in Nashville, Tennessee, participated
in the study. The CHD provides supportive
services (e.g., skills training, showers,
identification, transportation) to remove
barriers that prevent people from being
housed. The CHD partners with the Metropolitan
Health Department to provide primary health
care as well as mental health and substance
abuse services.
Five trained interviewers verbally administered
two questionnaires: 1) the Dartmouth Improve
Your Medical Care Survey and 2) the Short
Form 12 item Survey. Interviews occurred
between 7:00 a.m. and 12:00 p.m. weekdays
for a 4-week period. Each respondent was
offered a snack for participation. The
mornings were chosen because the greatest
numbers of homeless people that use the
day shelter are encountered during that
interval of time.
Initially, an attempt was made to conduct
the surveys by way of respondent self-administration.
However, both fatigue and frustration
with completing an obviously unfamiliar
task and low levels of literacy made self-administration
too difficult. Thus, surveys were completed
via face-to-face interviews.
The SF-12 was scored using the Medical
Outcomes Study (MOS) software program
that creates two summary scores, mental
health (MCS12) and physical health (PCS12).
The scores are represented as t-scores
that are linear transformations with a
mean of 50 and a standard deviation of
10 in the general United States population.
In addition, the eight sub-scales--role
emotional, role physical, physical function,
social function, mental health, vitality,
pain, and general health--were derived.
All scores were transformed to a 0-100
scale with the higher score indicating
less dysfunction, impairment, or pain.
These questions were designed to prompt
respondents to think about their health
and functioning during the past week.
The six functional health status Dartmouth
Primary Cooperative Research Network (COOP)
items from the Dartmouth Improve Your
Health Survey_ were converted to a 0-100
scale. The six questions measured physical
function; role function; social function;
pain; emotional function; and social support.
However, the social support question was
not used because it was deemed beyond
the scope of this study. Respondents were
asked to think about their health and
functioning during the past month when
answering these questions. The survey
also contained a series of questions that
asked respondents to indicate how often
they experienced specific symptoms (e.g.,
headaches, backaches) and whether or not
they had ever been told by a health care
professional that they had any specific
chronic medical conditions, (e.g., high
blood pressure, asthma).
For the symptoms experienced, the response
categories of never, seldom, often and
always were converted to dichotomous response
categories such that never and seldom
were categorized as a negative response
and sometimes, often and always were categorized
as an affirmative response. The questions
that inquired about known medical conditions
merely asked respondents to indicate yes
if they had previous been told that they
had the condition and no if they had not.
Three methods were used to test the construct
validity of the SF-12 with the homeless
population. First, a test for significant
differences in SF-12 scores by medical
conditions or reported clinical symptoms
was performed. Since people differ in
reported conditions such as hypertension
or asthma, one would expect that the SF-12
scores should discriminate between people
who have been told by a health professional
that they have those conditions. That
is, those individuals who report that
they have been told they have these conditions
should have lower health scores than those
who have not. The SF-12 should also differentiate
between those who experience health related
symptoms and those who do not.
Second, a correlative test was performed
on the SF-12 subscales and the physical
and mental summary scores. The subscales
that purport to measure physical health
should correlate more highly with the
physical health summary score than with
the mental health summary scores. The
same pattern should be observed for the
relationship between the mental health
related subscales and the mental health
summary score.
Third, a correlative test was performed
on the SF-12 and another indicator of
functional health status, the adult COOP
functional health status questions. Both
indicators provide an index of five domains
of functional health: physical health,
role function, mental health, social function
and bodily pain. Although the COOP questions
asked about the level of functioning during
the past month and the SF-12 questions
asked about the level of functioning during
the past week, it would be expected that
these indicators would vary together;
both overall and by the severity or presence
of different conditions. Overall mean
differences in the ratings may be expected
because of wording differences and the
referents for answering are anchored at
two different points in time.
Seventy-seven percent of the sample was
male and 85 percent were younger than
45 years of age. The average age was 37
years. Approximately half of the sample
complained of headaches, dizziness or
fatigue, joint pain, backaches and trouble
sleeping either sometimes, often or always.
When all 13 symptoms were summed, 91 percent
experienced one or more health symptoms.
Sixty-five percent of the sample reported
having been previously told by a health
professional that s/he had high blood
pressure. Thirty-seven percent indicated
the presence of asthma/bronchitis/emphysema,
while 20 percent indicated that they had
arthritis. When totaled, 57 percent indicated
the presence of one or more medical conditions.
Respondents scores on the MCS12, PCS12,
and the eight subscales were compared
to the Medical Outcomes Study results
for four severity levels of medical conditions:
minor medical, serious physical, mental
only and serious physical and mental.
The mental health summary score of the
homeless was most similar to the mental
only group and the serious physical and
mental comparison groups.
On all comparisons, the homeless group
was functioning at a substantially lower
level than the minor medical group. The
values obtained for physical functioning
and role functioning, pain and vitality
were closest to the serious physical health
comparison group. For mental health and
role emotional health, the scores were
closest in value to the mental only and
serious physical and mental problems comparison
groups, respectively. The level of functioning
for general health and social functioning
of the homeless were lower than the MOS
scores for the serious physical and mental
comparison groups. It is interesting to
note that the variability for the homeless
sample was greater than the minor medical
comparison group on all indicators.
A series of multi-variate analysis of
variance were conducted for each of the
32 symptoms and conditions on the MCS12
and PCS12. Twenty-six of the 32 univariate
comparisons were found to be significant.
In almost all cases, those with a condition
or who experienced a symptom exhibited
a significantly lower mental or physical
health score then those who did not have
the condition or did not experience the
symptom.
When the SF-12_ scores were analyzed by
categories of the total number of symptoms
or conditions reported, the scores appeared
to differentiate well between the groups.
In general, as the number of symptoms
or conditions increased, the health functioning
scores decreased. Collectively, these
results would suggest that these indicators
are able to differentiate between diagnostic
groups or levels of severity of illness.
Some differences were noted by gender.
T-tests revealed that females had significantly
lower mental health summary scores than
did males, and females reported significantly
more symptoms than did males. The two
groups were not different on physical
health scores or number of conditions
reported.
Results by age did not reveal any significant
or substantially different levels of functioning.
There was a slight decline in physical
scores by age, however, it is interesting
to note that the mental health scores
are substantially lower for the younger
group.
The subscales correlated well with the
summary scores in the manner that would
be expected of internally valid indicators.
Physical function and role physical function
correlated more highly with PCS12 than
with MCS12 scores. Also, role emotional
function and mental health subscales correlated
more highly with the MCS12 than with PCS12.
Consistent with other findings, the general
health and pain subscales correlated more
strongly with PCS12, while social functioning
correlated more strongly with MCS12. The
vitality sub-scale was found to be equally
correlated with both PCS12 and MCS12.
The PCS12 and MCS12 were not correlated
with each other at all (r=0.008). This
supports the concept of independence of
these two scales that have been previously
documented.
The PCS12 and MCS12 scores correlated
well with the COOP items in the manner
that one would expect with the assumption
that they measure similar constructs.
Similarly, they did not correlate as well
with the COOP items that measure dissimilar
constructs. This is evidence of support
for convergent (correlation between similar
constructions) and divergent (correlation
between dissimilar constructs) validity.
Each COOP chart correlated with the MCS12
and PCS12 in the same pattern observed
with the internal sub-scales and summary
scores, except the coefficients were not
quite as high. The COOP Role Function
and Physical Function correlated more
highly with physical health than with
mental health. The COOP Emotional Status
and Social Function correlated more highly
with mental health than with physical
health. The COOP Pain question correlated
more highly with physical health than
with mental health. Social Support, that
was not of major interest in this context,
was modestly correlated with both physical
and mental health.
The diagonal correlation coefficients,
ranging from 0.46 to 0.63, reflected fairly
good convergent validity. The divergent
validity was evidenced by the off-diagonal
correlation coefficients that ranged from
0.41 to 0.51. The only challenges to this
type of validity were the inter-correlations
between the SF12_ and the COOP physical
function and role function scores. The
SF12_ physical function score correlates
somewhat higher with the COOP role function
than with the COOP physical function score.
The results of this study indicate that
the SF-12_ has potential to be used as
an outcome indicator for the health status
of homeless people. Not surprisingly,
the results clearly demonstrated that
the health needs of homeless people are
great, as evidenced by the high percentage
of respondents who reported the presence
of various symptoms and conditions, in
addition to the low physical and mental
health scores. It should be noted that
the variability was fairly high for the
entire sample and for the various sub-groups
that were evaluated.
Although these results were based on a
single small convenience sample of users
of a day shelter, the three approaches
used to assess the appropriateness of
the tool revealed consistent results.
The survey tool has potential to differentiate
between levels of severity of physical
and mental health conditions; it appears
to be internally consistent; and it correlates
fairly well with another similar health
status rating tool, the adult COOP health
status questionnaire. Factors that would
help allow the SF-12 to be used for homeless
patients include the following: 1) the
ease with which it is administered; 2)
with the establishment of appropriate
referral procedures, it may be implemented
in a day shelter or other setting for
assessment of clients for referral at
intake; 3) it can be used to track and
monitor health status; and 4) norms are
available for comparison purposes.
The major limitation of the tool is the
need to verbally administer the survey.
Because of the brevity, however, it may
easily be implemented within existing
processes in many settings that serve
homeless clients.
This study had several limitations that
should lead to continued evaluation of
this instrument with the homeless. The
reliability of the instrument was not
assessed; thus, further research should
examine the test-retest reliability and
the stability and consistency of the scores
over time.
In addition, specific self-reported symptoms
or conditions indicative of mental health
were not represented in the Dartmouth
Improve Your Medical Care questionnaire
used for validation of the mental health
summary scores and the subscales. However,
the SF-12 mental health summary score
did correlate well with the COOP emotional
health item. Nevertheless, more validation
work needs to be done in regard to mental
health status.
A further limitation of this study is
that the sample used in this study was
shelter-based. The use of this instrument
should be explored with homeless persons
in other settings such as clinics or those
in other non-shelter settings. The SF-12_
scores should also be validated with homeless
persons with whom clinical information
is known to determine the extent to which
the scores can differentiate between diagnostic
groups, both physical and mental.
Collecting information on the functional
health of homeless persons based on self-reported
information can be useful in understanding
the health needs of this often hidden
population.
It is necessary that valid and reliable
tools be developed or identified that
can easily be utilized in a variety of
settings in which homeless clients are
served. In spite of questions related
to face validity, the SF-12_ shows promise
as a tool to provide understanding about
the level of physical and mental health
functioning of homeless patients. Potential
beneficial applications include using
the SF-12 for diagnostic purposes (e.g.,
for triage) and the monitoring of health
outcomes.
