ArticlePDF Available

The Effect of Hip Fracture on Mortality, Hospitalization, and Functional Status: A Prospective Study

Authors:

Abstract and Figures

The purpose of this study was to prospectively assess the independent effect of hip fracture on mortalìty, hospitalization, and functional status. Among 7527 members of the Longitudinal Study of Aging who were over age 70 at baseline, 368 persons with hip fracture occurring between 1984 and 1991 were identified. Median length of follow-up was 831 days. Hip fracture was significantly related to mortality (adjusted hazards ratio [AHR] = 1.83; 95% confidence interval [CI] = 1.55, 2.16) when treated as a time-dependent covariate. This effect was concentrated in the first 6 months postfracture (AHR = 38.93, 95% CI = 29.58, 51.23, vs AHR = 1.17; 95% CI = 0.95, 1.44). Hip fracture significantly increased the likelihood of subsequent hospitalization (adjusted odds ratio = 3.31, 95% CI = 2.64, 4.15) and increased the number of subsequent episodes by 9.4%, the number of hospital days by 21.3%, and total charges by 16.3%. Hip fracture also increased the number of functional status dependencies. The health of older adults deteriorates after hip fracture, and efforts to reduce the incidence of hip fracture could lower subsequent mortality, morbidity, and health services use.
Content may be subject to copyright.
The
Effect
of
Hip
Fracture
on
Mortality,
Hospitalization,
and
Functional
Status:
A
Prospective
Study
Fredric
D.
Wolinsky,
PhD,
John
F
Fitzgerald,
MD,
MBA,
and
limothy
E.
Stump,
MA
Introduction
In
1994
the
Office
of
Technology
Assessment
issued
a
landmark
report
that
comprehensively
reviewed
the
literature
on
the
mortality,
in-hospital
and
posthospi-
tal
service
use,
and
long-term
functional
*..
improvement
associated
with
hip
frac-
ture.1
It
concluded
that
94%
of
the
nearly
300
000
hip
fractures
that
occurred
in
the
United
States
in
1990
involved
adults
aged
50
years
or
older.
Nearly
all
of
these
required
hospitalization,
and
the
estimated
mean
hospital
charges
were
$9322.
Inpa-
tient
mortality
was
approximately
4%,
and
about
39%
of
patients
were
dis-
charged
to
a
long-term
care
facility.
Estimated
mean
posthospital
and
other
expenditures
(including
nursing
home
charges)
were
$9852.
The
1-year
postfrac-
ture
mortality
rate
was
approximately
24%.
Among
those
who
survived,
most
did
not
return
to
their
prefracture
func-
tional
status,
and
their
recovery
trajectory
peaked
by
6
months
postfracture.
Accord-
ingly,
a
key
target
in
federal
health
care
policy
is
the
reduction
of
the
incidence
of
hip
fracture
among
older
adults.2
Despite
the
critical
importance
of
hip
fracture
and
the
comprehensive
review
of
the
literature
in
the
Office
of
Technology
Assessment
report,'
the
current
knowl-
..
....
edge
base
remains
inadequate.
Four
limi-
tations
of
previous
studies
account
for
this.
First,
no
national
studies
have
been
conducted. Second,
there
have
been
few
prospective
studies.
Third,
appropriate
comparison
groups
have
seldom
been
used.
Finally,
most
reports
have
been
descriptive,
and
they
have
not
involved
multivariable
analyses.
The
purpose
of
this
paper
is
to
prospectively
assess
the
independent
ef-
fect
of
hip
fracture
on
mortality,
hospital-
ization,
and
functional
status
among
the
7527
older
adults
included
in
the
nation-
ally
representative
Longitudinal
Study
of
Aging
(LSOA).
To
control
for
potential
confounders,
a
variety
of
covariates
are
included
on
the
first
step
of
the
hierarchi-
cal
models.
Proportional
hazards
analysis
is
used
to
estimate
the
effect
of
hip
fracture
on
mortality.
Multivariable
logis-
tic
regression
is
used
to
estimate
the
effects
of
hip
fracture
on
whether
any
subsequent
hospitalization
occurred.
Mul-
tivariable
linear
regression
is
used
to
estimate,
among
those
with
subsequent
hospitalizations,
the
effect
of
hip
fracture
on
the
number
of
subsequent
hospitaliza-
tions,
total
number
of
hospital
days,
and
total
charges.
Multivariable
linear
regres-
sion
is
also
used
to
estimate
the
effect
of
hip
fracture
on
changes
in
functional
status
between
1984
and
1990
among
survivors.
Methods
Sample
The
LSOA
is
the
follow-up
to
the
Supplement
on
Aging
(96.0%
participa-
At
the
time
of
the
study,
Fredric
D.
Wolinsky
was
with
the
Indiana
University
School
of
Medicine
and
the
Regenstrief
Institute
for
Health
Care,
Indianapolis,
Ind.
John
F.
Fitzgerald
and
Timothy
E.
Stump
are
with
the
Regenstrief
Institute
for
Health
Care,
and
Dr
Fitzgerald
is
also
with
the
Indiana
University
School
of
Medicine,
India-
napolis.
Requests
for
reprints
should
be
sent
to
Fredric
D.
Wolinsky,
PhD,
School
of
Public
Health,
Saint
Louis
University
Health
Sciences
Center,
3663
Lindell
Blvd,
St
Louis,
MO
63108-3342.
This
paper
was
accepted
July
19,
1996.
Note.
The
views
expressed
here
are
the
authors'
and
do
not
necessarily
reflect
those
of
the
academic,
governmental,
or
research
institutions
involved.
See
related
comment
by
Kasl
(p
333)
in
this
issue.
March
1997,
Vol.
87,
No.
3
C.
7,
Effects
of
Hip
Fracture
tion
rate)
appended
to
the
1984
National
Health
Interview
Survey
(96.4%
participa-
tion
rate).
It
consists
of
the
7527
Supple-
ment
on
Aging
respondents
aged
70
years
or
older
in
1984
(96.6%
of
all
National
Health
Interview
Survey
respondents
in
that
age
range)
who
were
selected
for
follow-up
interviews
in
1986,
1988,
and
1990
and
to
have
their
Medicare
Auto-
mated
Data
Retrieval
System
(MADRS)
hospitalization
records
and
National
Death
Index
data
abstracted
for
calendar
years
1984
through
1991.
After
these
files
were
linked,
hospitalization
episodes
that
con-
tained the
Intemnational
Classification
of
Diseases,
Ninth
Revision,
Clinical
Modifi-
cation
(ICD9-CM)3
codes
for
hip
fracture
(i.e.,
820.0
through
820.9)
in
any
of
the
five
discharge
diagnosis
fields
were
used
to
identify
the
368
hip
fracture
patients.
For
those
with
multiple
hip
fracture
episodes,
the
first
was
chosen
as
the
index
case.
Consistent
with
previous
reports
on
the
incidence4
and
recurrence5
of
hip
fracture
using
this
data
set,
the
un-
weighted
LSOA
data
are
used
without
adjustments
for
design
effects.6'0
Covariates
To
obtain
the
independent
effect
of
hip
fracture
on
mortality,
hospitalization,
and
functional
status,
it
is
necessary
to
statistically
control
for
potentially
con-
founding
factors.
Previous
studies
of
the
LSOA10-3
have
shown
these
outcomes
to
be
related
in
part
to
factors
identified
in
the
behavioral
model
of
health
services
utilization,'4
which
is
the
most
widely
used
multivariable
framework
for
study-
ing
health
and
health
behavior.15
Basi-
cally,
it
views
the
use
of
health
services
as
a
function
of
the
predisposing,
enabling,
and
need
characteristics
of
the
individual,
as
well
as
the
individual's
prior
utilization
history.
The
measures
of
the
predisposing
characteristics
used
here
include
age,
sex,
race,
education,
two
variables
reflecting
living
arrangements
(living
alone
or
in
multigenerational
households),
and
sepa-
rate
kin
and
nonkin
support
scales.
The
enabling
characteristics
were
represented
by
having
private
health
insurance
for
both
physician
and
hospital
expenses,
being
on
Medicaid,
residential
stability
(had
not
moved
in
5
years),
population
density
(the
10-point
county
adjacency
scale),
financial
dependence
on
Social
Security,
and
three
variables
reflecting
geographic
region.
There
were
25
measures
of
need
characteristics.
Four
decomposed
the
tra-
ditional
measure
of
perceived
health
status.
The
next
11
were
dichotomous
TABLE
1-Means
(or
Proportions)
of
the
Covariates
for
Hip
Fracture
Patients
(n
=
368)
and
the
Control
Subjects
(n
=
7159)
Hip
Fracture
Control
Patients
Subjects
pa
Predisposing
characteristics
Age,
y
79.66
Sex,
%
female
77.99
Race,
%
Black
2.99
Education,
y
10.15
Living
alone,
%
47.01
Living
in
multigenerational
household,
%
16.85
Kin
supports,
no.
sources
1.62
Nonkin
supports,
no.
sources
2.23
76.68
61.17
8.90
9.99
35.90
17.69
1.61
2.39
.0001
.0001
.0001
.4148
.0001
.6786
.9529
.2426
Enabling
characteristics
Private
insurance,
%
Medicaid,
%
Residentially
stable,
%
Population
densityb
Social
Security
dependence,
%
Living
in
Northeast,
%
Living
in
North
Central,
%
Living
in
West,
%
Need
4
Very
good
health,
%
Good
health,
%
Fair
health,
%
Poor
health,
%
Osteoporosis,
%
Broken
hip,
%
Atherosclerosis,
%
Hypertension,
%
Coronary
heart
disease,
%
Angina,
%
Myocardial
infarction,
%
Other
heart
attack,
%
Stroke
or
cerebrovascular
accident,
%
Alzheimer's
disease,
%
Cancer,
%
Arthrtis,
%
Diabetes,
%
Aneurysm,
%
Blood
clot,
%
Varicose
veins,
%
Basic
ADLs,
no.
difficulties
Household
ADLs,
no.
difficulties
Advanced
ADLs,
no.
difficulties
Lower-body
limitations,
no.
Upper-body
limitations,
no.
64.67
4.62
83.15
2.61
66.85
21.20
27.99
16.85
characteristics
20.92
29.62
22.28
13.04
4.62
8.70
12.77
44.29
4.08
6.79
0.82
7.34
6.25
0.27
12.50
57.07
6.52
0.54
2.17
9.78
0.78
0.69
0.18
2.13
0.49
66.77
5.64
84.25
2.60
64.18
23.15
25.66
17.64
20.14
31.52
21.26
11.76
3.61
4.22
12.91
44.98
4.49
6.94
1.97
6.91
7.47
0.57
12.17
53.93
10.00
0.55
1.52
9.90
0.68
0.59
0.15
1.84
0.42
.4057
.4075
.5728
.9125
.2978
.3861
.3201
.6985
.7148
.4427
.6400
.4579
.3136
.0001
.9389
.7965
.7082
.9158
.1144
.7526
.3841
.4489
.8493
.2392
.0288
.9960
.3263
.9403
.0719
.0628
.0120
.0126
.5287
Baseline
health
services
utilization
Hospital
contact,
%
Physician
visits,
no.
Nursing
home
placement,
%
28.26
4.63
3.26
20.88
4.21
2.60
.0007
.0543
.4410
Note.
ADLs
=
activities
of
daily
living.
aBased
on
one-way
analysis
of
variance
for
interval
variables
and
chi-square
test
for
nominal
variables.
bO
=
large
standard
metropolitan
statistical
area;
9
=
rural
county.
variables
indicating
whether
or
not
the
respondent
had
ever
had
osteoporosis,
a
broken
hip,
atherosclerosis,
hypertension,
coronary
heart
disease,
angina,
myocar-
dial
infarction,
any
other
heart
attack,
stroke
or
a
cerebrovascular
accident,
Alzheimer's
disease,
or
cancer.
There
were
also
five
dichotomous
variables
indicating
whether
or
not
the
respondent
had
had
arthritis,
diabetes,
an
aneurysm,
a
American
Journal
of
Public
Health
399
March
1997,
Vol.
87,
No.
3
Wolinsky
et
al.
blood
clot,
or
varicose
veins
during
the
12
months
prior
to
baseline.
The
five
remaining
measures
of
need
characteristics
were
previously
vali-
dated7"1"'16"17
multiple-item
scales
that
consist
of
various
measures
of
disability
in
activities
of
daily
living.
The
basic
activities-of-daily-living
scale
(al-
pha
=
.860)
reflects
difficulties
with
six
items
taken
from
the
Katz
scale
(i.e.,
bathing,
dressing,
getting
out
of
bed,
toileting,
getting
outside,
and
walking).'8
The
household
activities-of-daily-living
scale
(alpha
=
.821)
reflects
difficulties
with
four
items
taken
from
Duke
Univer-
sity's
scale
(i.e.,
meal
preparation,
shop-
ping,
and
light
and
heavy
housework).'9
The
advanced
(or
cognitive)
activities-of-
daily-living
scale
(alpha
=
.638)
reflects
difficulties
with
managing
money,
using
the
telephone,
and
eating.'8"19
The
two
remaining
scales
were
drawn
from
stan-
dard
disability
items.20
One
taps
lower-
body
limitations
(alpha
=
.862),
such
as
difficulties
in
walking
a
quarter
of
a
mile;
walking
up
10
steps
without
rest;
standing
or
being
on
one's
feet
for
2
hours;
stooping,
crouching,
or
kneeling;
and
lifting
or
carrying
25
pounds.
The
other
taps
upper-body
limitations
(alpha
=
.577),
such
as
difficulties
in
sitting
for
2
hours,
reaching
up
over
the
head,
reaching
out
as
if
to
shake
hands,
and
using
fingers
to
grasp
objects.
Baseline
health
services
utilization
was
represented
by
three
measures.
One
was
a
dichotomous
indicator
of
whether
the
respondent
had
been
hospitalized
during
the
12
months
prior
to
baseline.
The
second
was
the
number
of
physician
visits
that
occurred
during
the
12
months
prior
to
baseline.
The
third
was
a
dichoto-
mous
indicator
of
whether
the
respondent
had
ever
been
placed
in
a
nursing
home
prior
to
baseline.
The
Simulated
Hip
Fracture
Date
To
estimate
the
effect
on
subsequent
hospitalizations
it
was
necessary
to
simu-
late
a
hip
fracture
date
for
the
control
subjects.
This
simulation
involved
several
steps.
First,
all
hospital
discharge
dates
were
converted
to
a
day
count
with
the
base
of
January
1,
1984,
set
to
1.
Second,
the
distribution
of
those
day
counts
for
hospital
episodes
involving
a
hip
fracture
was
examined
and
found
to
have
a
mean
of
1600
and
a
standard
deviation
of
800.
Third,
a
random
variable
(the
simulated
hip
fracture
date)
constrained
to
have
the
same
mean
and
standard
deviation
was
generated
for
all
hospital
discharges.
Fourth,
a
paired
t
test
comparing
the
simulated
hip
fracture
date
with
the
actual
hip
fracture
date
among
the
368
hip
fracture
patients
was
performed.
The
mean
difference
was
8
days
(SE
=
49,
P
=
.87),
indicating
no
difference
be-
tween
the
observed
and
simulated
dates.
Accordingly,
the
observed
hip
fracture
discharge
date
is
used
as
the
index
discharge
date
for
the
hip
fracture
pa-
tients,
and
the
simulated
hip
fracture
date
is
used
as
the
index
discharge
date
for
the
control
subjects.
Outcomes
Mortality
status
and
date
of
death
were
taken
from
the
National
Death
Index.
Subsequent
hospitalization
data
were
obtained
by
aggregating
the
MADRS
episode
records
forward
from
the
actual
or
simulated
hip
fracture
discharge
date
through
December
31,
1991.
These
data
were
then
used
to
construct
a
dichoto-
mous
indicator
of
whether
any
subsequent
hospitalizations
had
occurred
and,
among
those
with
subsequent
hospitalizations,
the
total
number
of
episodes,
total
number
of
hospital
days,
and
total
charges.
Be-
cause
of
their
positively
skewed
distribu-
tions,
the
natural
logarithms
(+
1)
are
used
in
the
multivariable
analysis.
Changes
in
the
number
of
limitations
on
the
five
functional
health
status
scales
were
ob-
tained
by
subtracting
the
1984
from
the
1990
counts
among
those
successfully
reinterviewed
in
1990.
The
potential
for
ceiling
and
floor
effects
in
assessing
changes
in
these
functional
status
mea-
sures
has
been
examined
elsewhere'0
and
found
not
to
be
a
problem.
Similarly,
alternative
methods
for
assessing
changes
in
these
functional
status
measures
have
been
examined
elsewhere'0
and
found
to
yield
equivalent
results.
Statistical
Analysis
Two-step
hierarchical
models
were
used
to
obtain
the
independent
contribu-
tion
of
having
suffered
a
hip
fracture.
On
the
first
step
the
44
predisposing,
en-
abling,
need,
and
prior
utilization
covari-
ates
were
entered.
A
dichotomous
marker
for
having
suffered
a
hip
fracture
was
introduced
on
the
second
step.
All
statisti-
cal
analyses
were
performed
with
SPSS
for
Windows,
version
6.1.2S.2'
Results
Descriptive
Data
To
characterize
the
sample
at
base-
line,
Table
1
contains
the
means
(or
proportions)
of
the
44
covariates
for
the
hip
fracture
patients
and
the
control
subjects.
In
general
agreement
with
previ-
ous
reports,'
the
hip
fracture
patients
were
significantly
(P
'
.05)
more
likely
than
the
control
subjects
to
be
older,
female,
and
White
and
to
live
alone,
to
have
fractured
a
hip
previously,
to
not
have
had
diabetes,
to
have
more
difficulties
with
advanced
activities
of
daily
living,
to
have
more
lower
body
limitations,
and
to
have
been
hospitalized
in
the
year
prior
to
baseline.
Table
2
contains
crude
rates
for
mortality
and
hospitalization
for
the
hip
fracture
patients
and
the
control
subjects.
Initial
analyses
did
not
indicate
a
signifi-
cant
difference
in
mortality
over
the
8-year
period,
even
though
43%
of
the
hip
fracture
patients
died
vs
38%
of
the
control
subjects,
yielding
a
mortality
rate
400
American
Journal
of
Public
Health
TABLE
2-Mortality
and
Hospitalization
Rates
for
Hip
Fracture
Patients
and
TABLE
2Mortality
and
Hospitalization
Rates
for
Hip
Fracture
Patients
and
the
Control
Subjects
Outcome
Hip
Fracture
Patients
Control
Subjects
P
Mortality
Events/persons
(%)
157/368
(43)
2706/7159
(38)
.0635a
Events/person-years
(rate)
157/2570
(.061)
2706/47416
(.057)
.4394b
Hospitalization
Any
episodes,
events/
239/368
(65)
2752/7159
(38)
.0001a
persons
(%)
Mean
no.
episodesc
2.3
2.0
.0016d
Mean
no.
daysc
22.7
18.1
.0187d
Mean
charges,c
$
18
105
16422
.2545d
aBased
on
chi-square
test.
bBased
on
univariable
proportional
hazards
model.
CAmong
the
2993
persons
with
one
or
more
episodes.
dBased
on
one-way
analysis
of
vanance.
March
1997,
Vol.
87,
No.
3
Effects
of
Hip
Fracture
of
.061
vs
.057.
Visual
inspection
of
the
cumulative
survival
distribution
among
the
hip
fracture
patients
(data
not
shown),
however,
revealed
a
substantial
short-term
mortality
risk,
such
that
about
20%
of
these
patients
were
dead
within
1
year
postfracture.
This
suggested
reanalysis
using
a
proportional
hazards
model
in
which
hip
fracture
was
treated
as
a
time-dependent
covariate,
along
with
sub-
sequent
stratification
based
on
postfrac-
ture
survival
time.
The
initial
time-
dependent
covariate
analysis
yielded
a
crude
relative
risk
ratio
(RR)
of
2.42
(95%
confidence
interval
[CI]
=
2.06,
2.85),
and
the
stratified
analyses
further
revealed
the
concentration
of
that
mortality
hazard
in
the
first
6
months
postfracture
(RR
=
57.36;
95%
CI
=
43.67,
75.33)
rather
than
subsequently
(RR
=
1.57;
95%
CI
=
1.28,
1.92).
Median
length
of
postfracture
follow-up
among
hip
fracture
patients
was
831
days
(415
days
for
decedents
and
1099
days
for
survivors).
The
hip
fracture
patients
were
signifi-
cantly
more
likely
than
the
control
sub-
jects
to
have
subsequent
MADRS
hospital-
izations
(65%
vs
38%).
When
the
volume
of
hospital
resource
consumption
is
exam-
ined
for
the
2993
persons
with
one
or
more
MADRS
hospitalizations
after
their
actual
or
simulated
hip
fracture
date,
two
significant
differences
emerge.
The
hip
fracture
patients
experienced
about
2.3
hospital
episodes
vs
2.0
for
control
subjects,
and
they
had
about
4.6
more
hospital
days
(22.7
vs
18.1).
Although
mean
total
hospital
charges
were
also
higher
for
hip
fracture
patients
than
for
control
subjects
($18
105
vs
$16422),
this
difference
was
not
significant.
Table
3
shows
mean
increases
in
the
number
of
functional
health
status
limita-
tions
between
1984
and
1990
for
the
hip
fracture
patients
and
control
subjects.
Significant
mean
increases
on
all
five
scales
are
associated
with
having
suffered
a
hip
fracture
between
1984
and
1989.
These
mean
increases
represent
the
aver-
age
number
of
new
functional
limitations
incurred
over
the
6-year
period.
The
differences
in
mean
increases
are
1.29
on
basic
activities
of
daily
living,
.44
on
household
activities
of
daily
living,
.18
on
advanced
activities
of
daily
living,
1.00
on
lower-body
limitations,
and
.23
on
upper-
body
limitations.
Modeling
Mortality
and
Subsequent
Hospitalization
When
the
hip
fracture
marker
was
entered
as
a
time-dependent
covariate
on
the
second
step
of
the
proportional
haz-
ards
analysis
of
mortality
over
the
8-year
period,
it
had
a
highly
significant
effect
(adjusted
hazards
ratio
[AHR]
=
1.83;
95%
CI
=
1.55,
2.16).
The
stratified
analyses
once
again
revealed
that
the
effect
was
concentrated
in
the
first
6
months
postfracture
(AHR
=
38.93,
95%
CI
=
29.58,
51.23,
vs
AHR
=
1.17,
95%
CI
=
0.95,
1.44).
When
the
hip
fracture
marker
was
entered
on
the
second
step
of
the
logistic
regression
model
predicting
having
any
hospital
episodes
after
the
actual
or
simulated
hip
fracture
date,
its
effect
was
also
quite
significant,
yielding
an
adjusted
odds
ratio
of
3.31
(95%
CI
=
2.64,
4.15).
Multiple
linear
regres-
sion
was
used
to
assess
the
independent
effect
of
hip
fracture
on
the
three
mea-
sures
of
volume
of
subsequent
hospitaliza-
tion
among
the
2993
persons
who
had
one
or
more
episodes
after
their
actual
or
simulated
hip
fracture
date.
The
percent-
age
increase
in
these
volume
measures
was
calculated
by
subtracting
1
from
the
exponent
of
the
partial
unstandardized
regression
coefficients
(data
not
shown)
for
hip
fracture
and
multiplying
by
100.22
Having
suffered
a
hip
fracture
signifi-
cantly
increased
the
number
of
subse-
quent
episodes
(an
increase
of
9.4%;
P
=
.0007),
the
total
number
of
hospital
days
(an
increase
of
21.3%;
P
=
.0016),
and
the
total
charges
(an
increase
of
16.3%;
P
=
.0366).
Modeling
Functional
Status
Table
4
shows
the
partial,
unstandard-
ized
linear
regression
coefficients
(b's)
for
the
hip
fracture
marker
and
their
P
values
on
the
changes
in
each
of
the
functional
status
scales.
Here,
the
b's
may
be
interpreted
as
the
regression-adjusted
mean
changes
in
the
number
of
functional
limitations
independently
associated
with
having
suffered
a
hip
fracture.23
Among
the
4138
persons
reinterviewed
in
1990,
having
suffered
a
hip
fracture
significantly
increased
the
number
of
difficulties
in
basic
activities
of
daily
living
(by
1.12),
the
number
of
difficulties
in
household
activities
of
daily
living
(by
.35),
the
number
of
difficulties
in
advanced
activi-
ties
of
daily
living
(by
.21),
the
number
of
lower-body
limitations
(by
.93),
and
the
number
of
upper-body
limitations
(by
.26).
Discussion
This
study
assessed
the
effect
of
hip
fracture
on
mortality,
hospitalization,
and
functional
status.
Several
features
distin-
guish
this
study
from
previous
efforts:
(1)
American
Journal
of
Public
Health
401
TABLE
3-Mean
Increases
in
Number
of
Functional
Health
Status
Limitations
between
1984
and
1990
among
the
4138
Persons
Reinterviewed
in
1990