Protecting
Adolescents
From
Harm
Findings
From
the
National
Longitudinal
Study
on
Adolescent
Health
Michael
D.
Resnick,
PhD;
Peter
S.
Bearman,
PhD;
Robert
Wm.
Blum,
MD,
PhD;
Karl
E.
Bauman,
PhD;
Kathleen
M.
Harris,
PhD;
Jo
Jones,
PhD;
Joyce
Tabor;
Trish
Beuhring,
PhD;
Renee
E.
Sieving,
PhD;
Marcia
Shew,
MD,
MPH;
Marjorie
Ireland,
PhD;
Linda
H.
Bearinger,
PhD,
MS;
J.
Richard
Udry,
PhD
Context.\p=m-\The
main
threats
to
adolescents'
health
are
the
risk
behaviors
they
choose.
How
their
social
context
shapes
their
behaviors
is
poorly
understood.
Objective.\p=m-\To
identify
risk
and
protective
factors
at
the
family,
school,
and
in-
dividual
levels
as
they
relate
to
4
domains
of
adolescent
health
and
morbidity:
emotional
health,
violence,
substance
use,
and
sexuality.
Design.\p=m-\Cross-sectional
analysis
of
interview
data
from
the
National
Longitu-
dinal
Study
of
Adolescent
Health.
Participants.\p=m-\A
total
of
12118
adolescents
in
grades
7
through
12
drawn
from
an
initial
national
school
survey
of
90118
adolescents
from
80
high
schools
plus
their
feeder
middle
schools.
Setting.\p=m-\The
interview
was
completed
in
the
subject's
home.
Main
Outcome
Measures.\p=m-\Eight
areas
were
assessed:
emotional
distress;
suicidal
thoughts
and
behaviors;
violence;
use
of
3
substances
(cigarettes,
alcohol,
marijuana);
and
2
types
of
sexual
behaviors
(age
of
sexual
debut
and
pregnancy
history).
Independent
variables
included
measures
of
family
context,
school
con-
text,
and
individual
characteristics.
Results.\p=m-\Parent-family
connectedness
and
perceived
school
connectedness
were
protective
against
every
health
risk
behavior
measure
except
history
of
preg-
nancy.
Conversely,
ease
of
access
to
guns
at
home
was
associated
with
suicidal-
ity
(grades
9-12:
P<.001)
and
violence
(grades
7-8:
P<.001
;
grades
9-12:
P<.001).
Access
to
substances
in
the
home
was
associated
with
use
of
cigarettes
(P<.001),
alcohol
(P<.001),
and
marijuana
(P<.001)
among
all
students.
Working
20
or
more
hours
a
week
was
associated
with
emotional
distress
of
high
school
students
(P<.01),
cigarette
use
(P<.001),
alcohol
use
(P<.001),
and
marijuana
use
(P<.001).
Appearing
"older
than
most"
in
class
was
associated
with
emotional
dis-
tress
and
suicidal
thoughts
and
behaviors
among
high
school
students
(P<.001);
it
was
also
associated
with
substance
use
and
an
earlier
age
of
sexual
debut
among
both
junior
and
senior
high
students.
Repeating
a
grade
in
school
was
associated
with
emotional
distress
among
students
in
junior
high
(P<.001)
and
high
school
(P<.01)
and
with
tobacco
use
among
junior
high
students
(P<.001).
On
the
other
hand,
parental
expectations
regarding
school
achievement
were
associated
with
lower
levels
of
health
risk
behaviors;
parental
disapproval
of
early
sexual
debut
was
associated
with
a
later
age
of
onset
of
intercourse
(P<.001).
Conclusions.\p=m-\Family
and
school
contexts
as
well
as
individual
characteristics
are
associated
with
health
and
risky
behaviors
in
adolescents.
The
results
should
assist
health
and
social
service
providers,
educators,
and
others
in
taking
the
first
steps
to
diminish
risk
factors
and
enhance
protective
factors
for
our
young
people.
JAMA.
1997;278:823-832
NUMEROUS
REPORTS
have
docu¬
mented
the
health
status
of
youth
in
the
United
States,
concluding
that
the
main
threats
to
adolescents'
health
are
pre¬
dominantly
the
health-risk
behaviors
and
choices
they
make.118
Data
indicate
that
more
than
3
of
every
4
deaths
in
the
second
decade
of
life
are
caused
by
social
morbidities:
unintentional
injuries,
ho¬
micides,
and
suicides.
Juvenile
homicide
rates
have
continued
to
escalate
until
re¬
cently,17
and
suicide
rates
among
adoles¬
cents
aged
14
years
or
younger
have
in¬
creased
by
75%
over
the
past
decade.3
Cigarette
smoking
among
teenagers
has
increased
by
as
much
as
2%
per
year
since
1992,
when
19%
of
high
school
seniors
reported
smoking.
Marijuana
use
has
increased
in
each
of
the
last
3
years
among
8th-,
10th-,
and
12th-grade
students.19
For
editorial
comment
see
864.
Some
children
who
are
at
high
risk
for
health-compromising
behaviors
suc¬
cessfully
negotiate
adolescence,
avoid¬
ing
the
behaviors
that
predispose
them
to
negative
health
outcomes;
while
oth¬
ers,
relatively
advantaged
socially
and
economically,
sustain
significant
mor¬
bidity
as
a
consequence
of
their
behav¬
iors.
These
issues
of
vulnerability
and
resilience
have
stimulated
an
interest
in
the
identification
of
protective
factors
in
the
lives
of
young
people—factors
that,
if
present,
diminish
the
likelihood
of
negative
health
and
social
outcomes.20"26
Of
the
constellation
of
forces
that
influ¬
ence
adolescent
health-risk
behavior,
the
most
fundamental
are
the
social
con¬
texts
in
which
adolescents
are
embed¬
ded20;
the
family
and
school
contexts
are
among
the
most
critical.
Yet,
how
ado¬
lescents'
connections
to
these
contexts
shape
their
health-risk
behaviors
is
poorly
understood.
In
the
present
analysis
we
seek
to
iden¬
tify
particular
risk
and
protective
factors
at
the
school,
family,
and
individual
levels
as
they
relate
to
4
broad
domains
critical
to
adolescent
health
and
morbidity
(emo¬
tional
health,
violence,
substance
use,
and
sexuality),
using
data
collected
as
part
of
the
National
Longitudinal
Study
of
Ado¬
lescent
Health
(Add
Health).
From
the
Adolescent
Health
Program,
University
of
Minnesota,
Minneapolis
(Drs
Resnick,
Blum,
Beuhring,
Sieving,
Shew,
Ireland,
and
Bearinger),
and
the
Carolina
Population
Center,
University
of
North
Carolina
at
Chapel
Hill
(Drs
Bearman,
Bauman,
Harris,
Jones,
and
Udry
and
Ms
Tabor).
Reprints:
Michael
D.
Resnick,
PhD,
Adolescent
Health
Program,
Box
721,
420
Delaware
St
SE,
Minne-
apolis,
MN
55455
(e-mail:
resni001@tc.umn.edu).
Downloaded From: https://jamanetwork.com/ by a Columbia University Libraries User on 03/22/2019
Table
1.—Dependent
Variables
Variables
Select
Descriptors
of
Variables
No.
of
Items
Constituting
Variable
(Reliability
Coefficient)
Emotional
distress
In
the
past
week
or
past
year:
felt
depressed,
lonely,
sad,
or
fearful,
moody,
cried,
or
had
a
poor
appetite
Suicidality
In
the
past
year:
seriously
thought
about
committing
suicide
or
attempted
1,
2,
or
more
times
17(ot=.87)*
Violence
In
the
past
year:
had
a
physical
fight,
injured
someone,
was
in
a
group
fight,
threatened
someone
with
a
weapon,
used
a
weapon
In
a
fight,
or
shot
or
stabbed
someone
8
(a=.82)*
Substance
use
Cigarette
use
A
7-category
composite
variable
from
never
smoked
to
smoked
>
1
pack/d
Alcohol
use
Frequency:
an
8-category
variable
from
never/almost
never
to
daily/almost
daily
used
alcohol
Marijuana
use
A
7-category
composite
variable
from
never
used
to
used
marijuana
=:
6
times
in
past
month
Sexual
behaviors
Age
of
sexual
debut
Age
at
first
intercourse:
a
continuous
variable,
with
nonsexually
active
youth
handled
as
event
not
having
occurred
Pregnancy
history
Among
sexually
experienced
females
s15
y,
those
who
ever
reported
a
history
of
pregnancy;
dichotomous
yes/no
variable
*For
most
measures
including
3
or
more
items,
Cronbach29
coefficient
was
used
to
assess
internal
consistency.
METHODS
The
Add
Health
Design
Add
Health
is
a
longitudinal
study
of
adolescents
in
grades
7
through
12
and
the
multiple
social
contexts
in
which
they
live.
The
primary
sampling
frame
included
all
high
schools
in
the
United
States
that
had
an
11th
grade
and
at
least
30
enrollees
in
the
school
(N=26666).
From
this
a
systematic
random
sample
of
80
high
schools
was
selected
propor¬
tional
to
enrollment
size,
stratified
by
region,
urbanicity,
school
type,
and
per¬
centage
white.
For
each
high
school,
the
largest
feeder
school
(typically
a
middle
school)
was
also
recruited
when
avail¬
able.
Overall,
79%
of
the
schools
con¬
tacted
agreed
to
participate,
for
a
final
sample
of
134
schools.
Schools
varied
in
size
from
fewer
than
100
to
more
than
3000
students.
The
schools
provided
a
roster
of
all
en¬
rolled
students
and
96%
(n=
129)
hosted
a
confidential
in-school
survey
from
Sep¬
tember
1994
to
April
1995.
The
survey
was
completed
by
90118
of
119233
eli¬
gible
students
in
grades
7
through
12.
The
in-school
survey
was
administered
only
once,
in
year
1.
Survey
data
will
be
the
subject
of
future
reports.
School
administrators
also
completed
a
half-hour
self-administered
questionnaire
yielding
information
on
the
provision
of
health
services,
school
policies,
school
en¬
vironments,
and
characteristics.
Two
phases
of
school
administrator
data
were
collected
1
y
ear
apart,
beginning
in
y
ear
1.
A
total
of
130
administrator
question¬
naires
were
completed
in
year
1
and
are
included
in
this
analysis.
The
Main
In-Home
Sample
From
students
on
the
school
rosters
as
well
as
students
who
were
not
on an
en¬
rolled
roster
but
who
completed
an
in-
school
questionnaire,
a
random
sample
of
15
243
adolescents
stratified
by
grade
and
sex
was
selected
for
in-home
interviews;
12118
(79.5%)
completed
the
90-minute
interviews.
Of
these,
75%
had
completed
an
in-school
questionnaire.
The
first
phase
of
in-home
interviews
was
conducted
between
April
and
Decem¬
ber
1995
and
is
the
focus
of
this
report.
A
second
phase
was
collected
a
year
later.
Data
collected
during
the
in-home
phase
of
Add
Health
provide
information
on
sen¬
sitive
health-risk
behaviors
such
as
drug
and
alcohol
use,
sexual
behavior,
and
criminal
activities
in
addition
to
detailed
information
on
health
status,
health
ser¬
vice
utilization,
family
dynamics,
peer
net¬
works,
romantic
relationships,
decision
making,
aspirations,
and
attitudes.
Dur¬
ing
the
more
sensitive
portions
of
the
in¬
terview,
adolescents
listened
to
questions
through
earphones
and
directly
entered
their
responses
into
a
laptop
computer,
thereby
greatly
reducing
any
potential
for
interviewer
or
parental
influences
on
their
responses.
For
85.6%
of
the
participating
adoles¬
cents,
a
parent
(in
most
instances
a
mother)
also
completed
a
half-hour
inter¬
view
in
year
1.
Parent
interview
data
are
not
included
in
this
article.
Through
a
set
of
linked
identifiers—
the
in-school
and
in-home
data
sets
and
the
school
administrator
and
parent
sur¬
veys—school
administrator
and
parent
surveys
were
merged.
Extensive
pre-
cautions
were
taken
to
maintain
confi¬
dentiality
and
to
guard
against
deduc¬
tive
disclosure
of
participants'
identities.
All
protocols
received
institutional
re¬
view
board
approval.
More
detailed
méthodologie
information
is
available
in
a
separate
article.27
Analysis
and
Reporting
A
series
of
checks
for
invalid
and
incon¬
sistent
responses
resulted
in
deletion
of
546
(4.5%)
of
the
core
sample
of
12118
ado¬
lescents.
Each
case
in
the
core
sample
was
assigned
a
weight
based
on
the
sampling
design
so
that
the
sample
is
nationally
rep¬
resentative
of
US
adolescents
in
grades
7
through
12.
These
sample
weights
were
used
in
every
statistical
procedure
with
the
exception
of
Cox
regression
(which
does
not
permit
weighting
in
SAS).
The
final
sample
of
11572
adolescents
was
randomly
partitioned
into
explor¬
atory
and
validation
samples
of
approxi¬
mately
equal
size.
Investigators
identi¬
fied
theoretically
relevant
and
empirically
significant
independent
variables
with
the
exploratory
sample;
confirmatory
analyses
were
completed
and
results
are
reported
for
the
validation
sample.
Sepa¬
rate
analyses
were
performed
for
grades
7
and
8
and
9
through
12
except
for
preg¬
nancy
history,
for
which
questions
were
restricted
to
females
aged
15
years
and
older
regardless
of
grade
and
age
of
first
intercourse,
which
latter
category
in¬
cluded
both
sexes
and
all
grades
regard¬
less
of
sexual
experience.
An
analysis
modeling
age
of
first
intercourse
excluded
sexually
experienced
youth
who
reported
having
intercourse
before
age
11
years
(2.0%
of
the
sexually
experienced
sub-
sample)
on
the
assumption
they
repre¬
sented
a
distinct
subgroup
of
youth
who
had
been
sexually
abused
or
had
partici¬
pated
in
nonconsensual
sex.28
Items
used
in
the
measurement
of
the
dependent
and
independent
variables
were
identified
from
a
variety
of
stan¬
dardized,
validated
instruments
used
in
national
and
state
surveys
of
adolescents.
Dependent
variables
were
selected
to
capture
the
major
indexes
of
adolescent
health
and
risk
behaviors
(Table
l).29
In¬
dependent
variables
were
derived
from
a
resiliency
framework,
which
posits
that
young
people's
vulnerability
to
health-
compromising
outcomes
is
affected
by
both
the
nature
and
number
of
Stressors
as
well
as
the
presence
of
protective
fac¬
tors
that
buffer
the
impact
of
those
Stress¬
ors
(Tables
2
and
3).
Adverse
or
success¬
ful
outcomes
are
described
as
emanating
from
the
interplay
of
environmental
fac¬
tors,
familial
factors,
and
individual
char¬
acteristics.30"38
Individual
characteristics
reflect
both
genetic
predispositions
(eg,
the
timing
and
tempo
of
puberty)
and
so¬
cial
and
cognitive
development
variables
Downloaded From: https://jamanetwork.com/ by a Columbia University Libraries User on 03/22/2019
Table
2.—Generic
Independent
Variables
Variables
Select
Descriptors
of
Variables
No.
of
Items
Constituting
Variable
(Reliability
Coefficient)
Family
context
Parent-family
connectedness
Closeness
to
mother
and/or
father,
perceived
caring
by
mother
and/or
father,
satisfaction
with
relationship
to
mother
and/or
father,
feeling
loved
and
wanted
by
family
members
13(a=.83)*
Parent-adolescent
activities
No.
of
different
activities
engaged
In
with
mother
and/or
father
In
past
4
wk
(summed)
10
for
mother
10
for
father
Parental
presence
A
parent
present:
before
school,
after
school,
at
bedtime,
or
at
dinner
(summed)
Parental
school
expectations
Mother's
and/or
father's
expectations
for
you
to
complete
high
school
and
college
2
(r=0.45)t
Family
suicide
attempts
and/or
completions
Suicidal
attempts
and/or
completions
by
family
members
in
the
past
12
mo
School
context
School
connectedness
Feel
that
teachers
treat
students
fairly;
close
to
people
at
school;
feel
part
of
your
school
6
(a=.75)*
Student
prejudice
On
a
5-point
scale,
agreement
that
students
in
school
are
prejudiced
Attendance^
Quasi-continuous
variable
(average
daily
attendance)
Dropout
ratei§
Estimated
dropout
rate
by
grade
in
school
School
typei§
Five
categories:
comprehensive
public,
magnet
public,
parochial,
technical,
other
Classroom
sizet
Average
size
of
class
from
<20
to
s35
Master's
degree:):
%
of
teachers
with
master's
degree
from
<10%
to
>90%
College:):
Proportion
of
students
who
are
college
bound
Parent-teacher
organization:);
%
of
parents
involved
with
a
parent-teacher
organization,
ranging
from
does
not
exist
to
==90%
Individual
characteristics
Self-esteem
On
a
5-point
scale
(agree
to
disagree):
good
personal
qualities,
a
lot
to
be
proud
of,
like
yourself,
feel
loved
and
wanted,
as
good
as
other
people
10(a=.86)*
Religious
identity
Pray
frequently,
view
self
as
religious,
affiliate
with
a
religion
Same-sex
attraction
or
behavior§
Ever
had
same-sex
romantic
attraction
or
same-sex
intercourse
Perceived
risk
of
untimely
death
Perceive
self
at
risk
for
untimely
death
Paid
work
a20
h/wk§
No.
of
hours
per
week
worked
for
pay
during
school
year
1
Self-report
of
physical
appearance§
Appear
older
or
younger
than
most
age
mates
1
each
Repeated
a
grade§
Repeated
1
or
more
grades
1
Grade
point
average
Available
grades
in
English,
math,
history/social
studies,
and
science
In
most
recent
reporting
period
*For
most
measures
including
3
or
more
items,
Cronbach29
coefficient
was
used
to
assess
internal
consistency.
tPearson
correlation
coefficient
was
used
to
assess
reliability
of
2-item
measures
where
appropriate.
^Derived
from
school
administrator
questionnaire.
§ltem
coded
dichotomously,
eg,
yes/no,
any/none.
(eg,
self-image,
future
perspective).
Lon¬
gitudinal
studies
by
both
Werner
and
Smith25
and
Quinton
and
Rutter39
have
identified
the
role
of
environmental
and
familial
contexts
as
well
as
individual
characteristics
in
promoting
heightened
or
diminished
well-being
among
children
who
have
experienced
multiple
life
Stressors.
In
the
present
analysis,
school
char¬
acteristics
(ie,
school
type,
dropout
rate,
attendance
rate,
classroom
size,
teacher
training,
characteristics
of
student
body),
including
"school
connected¬
ness"—a
concept
that
emerges
from
the
interactions
of
the
individual
with
the
school
environment40·41—are
used
to
rep¬
resent
a
key
environmental
force
in
the
lives
of
in-school
youth.
Familial
factors
incorporate
4
components:
parent-fam¬
ily
relationships
(connectedness,
shared
activities,
parental
presence);
norms
and
expectations
for
adolescent
behavior
(school
achievement,
sexual
behaviors);
parental
modeling
(family
suicide
in¬
volvement);
and
household
features
(ac¬
cess
to
weapons,
substances).30'31,37
Indi¬
vidual
characteristics
include
such
fac-
tors
as
employment,
academic
perfor¬
mance,
and
sexual
orientation
as
well
as
self-belief
components
including
reli¬
gious
identity
and
self-esteem.26·39
Independent
variables
within
each
context
were
divided
into
2
sets:
generic
(those
that
were
expected
to
be
associ¬
ated
with
every
dependent
variable,
such
as
parent-family
connectedness,
school
connectedness,
and
self-esteem)
and
domain-specific
variables
(those
that
applied
to
specific
dependent
vari¬
ables
such
as
household
access
to
alco¬
hol,
school
policies
on
fighting,
and
knowledge
of
condom
use).
In
the
pre¬
sent
analysis,
the
selection
of
risk
and
protective
factors
was
guided
by
an
em¬
phasis
on
variables
that
can
be
used
for
assessment
or
are
amenable
to
preven¬
tion
and
intervention
efforts.
All
dependent
and
independent
vari¬
ables
were
standardized
separately
for
each
grade
category
to
a
mean
of
0
and
an
SD
of
1
before
conducting
the
multi¬
variate
analyses,
except
for
dichotomous
variables
and
age
at
first
intercourse.
In
the
case
of
multi-item
scales,
individual
items
were
standardized
before
sum-
ming
items
to
form
scales;
summed-scale
scores
were
restandardized
to
a
mean
of
0
and
SD
of
1.
Consequently,
parameter
estimates
can
be
interpreted
as
stan¬
dardized
ß
(with
the
exception
of
di¬
chotomous
variables);
within
any
par¬
ticular
analysis,
odds
ratios
and
relative
risks
can
be
compared
with
each
other
for
effect
size.
Multivariate
Analysis
Our
analytic
strategy
was
to
highlight
relevant
variables,
their
measurement,
and
the
interrelationships
of
variables
within
domains.
This
broad
approach
provides
a
foundation
for
future,
more
focused
analyses.
The
impact
of
each
of
the
3
contexts
(family,
school,
and
indi¬
vidual
characteristics)
on
each
of
the
adolescent
health
and
risk
behaviors
was
assessed
using
multiple
linear
regres¬
sion
for
the
continuous
and
quasi-con¬
tinuous
outcome
variables,
logistic
re¬
gression
for
pregnancy
history,
and
Cox
regression
for
age
of
sexual
debut.
Each
of
these
analyses
controlled
for
the
ef¬
fects
of
key
demographic
variables:
sex,
race,
ethnicity,
family
structure,
and
Downloaded From: https://jamanetwork.com/ by a Columbia University Libraries User on 03/22/2019
Table
3.—Domain-Specific
Independent
Variables
Variables
Select
Descriptors
of
Variables
No.
of
Items
Constituting
Variable
Sexual
behavior
domain
Perceived
parental
disapproval
of
adolescent
sex
On
a
5-point
scale,
perceived
mother's
and/or
father's
disapproval
of
their
adolescent
having
sex
at
this
time
with
anyone
or
a
special
person
2
( =0.82)*
Perceived
parental
disapproval
of
adolescent
contraception
On
a
5-point
scale,
perceived
mother's
and/or
father's
disapproval
of
their
adolescent
using
contraception
at
this
time
Length
of
time
since
sexual
debut
Interval,
In
months,
between
first
Intercourse
and
the
current
date
Effective
contraceptive
use
with
first/last
sex
Use
of
oral
contraceptive
pills,
Norplant,
Depo-Provera,
intrauterine
device,
condoms,
or
condoms
plus
female
barrier
method
with
first/last
sex
(response
categories:
neither,
1,
or
both
occasions)
Substance
use
in
connection
with
sex
Level
of
alcohol
and
other
drug
use
involved
with
first/last
sex
6
(o=.65)t
Sex
in
exchange
for
drugs
or
moneyt
Ever
given
sex
in
exchange
for
drugs
or
money
1
Virginity
pledget
Made
public
or
written
pledge
to
remain
a
virgin
until
marriage
1
Perceived
benefits
of
sexual
activity
On
a
5-point
scale
(strongly
agree
to
strongly
disagree),
having
sex
would
relax
you,
give
you
physical
pleasure,
make
you
more
attractive,
make
you
less
lonely
5
(a=.70)t
Perceived
obstacles
to
contraceptive
use
On
a
5-point
scale
(strongly
agree
to
strongly
disagree),
birth
control
is
a
hassle
to
use,
too
expensive,
interferes
with
pleasure,
requires
too
much
planning
ahead,
conveys
that
you
are
looking
for
sex
7
(<»=.82)t
Perceived
susceptibility
to
pregnancy
On
a
5-polnt
scale
(strongly
agree
to
strongly
disagree),
perceived
chance
of
getting
pregnant
after
having
unprotected
sex
on
a
single
occasion
in
the
near
future
Perceived
consequences
of
pregnancy
Pregnancy:
one
of
the
worst
things
that
could
happen
at
this
time,
would
be
embarrassing,
would
force
growing
up
too
fast
I
(a=.70)t
Condom
use
knowledge
Knowledge
regarding
correct
use
of
condoms
(summed)
No.
correct
of
5
Contraceptive
use
self-efficacy
Confidence
in
ability
to
use
contraception
or
to
refuse
sex
in
various
situations
3
(a=.65)t
School-based
reproductive
health
services
on
premisest§
Family
planning
counseling
services,
sexually
transmitted
disease
treat¬
ment,
or
prenatal
or
postnatal
services
Violence,
emotional
distress,
and
suicidality
domains
Household
access
to
gunst
Reported
easy
availability
of
a
gun
in
the
home
1
History
of
victimization
and/or
witnessing
violence
Within
the
past
12
mo,
witnessed
or
been
a
victim
of
a
shooting
or
stabbing
5
(a=.66)t
Weapon
carrying
Weapon
carrying
at
school,
in
connection
with
substance
use
4
(a=.74)t
Sale
of
illicit
drugs
Any
sale
of
illicit
drugs
within
the
past
12
mo
1
Involvement
with
deviant/antisocial
behaviors
Destruction
of
property,
theft,
skipping
school
in
past
year;
ever
suspended
or
expelled
from
school
10(<x=.78)t
Body
image
Perceived
weight,
from
very
underweight
to
very
overweight
School
policies
on
flghting§
Warning/minor
action,
suspension,
or
expulsion
for
fighting
with
or
injuring
a
student
or
teacher
or
carrying
a
weapon
at
school
Mental
health
services
at
school
Emotional
counseling,
rape
counseling,
or
programs
for
dealing
with
effects
of
violence
provided
on
school
premises
Substance
abuse
domains
Household
access
to
cigarettes:);
Reported
easy
availability
of
cigarettes
In
the
home
Household
access
to
alcohol^
Reported
easy
availability
of
alcohol
in
the
home
Household
access
to
illicit
substances:):
Reported
easy
availability
of
illicit
drugs
in
the
home
School
policies
on
smoking§
Warning/minor
action,
suspension,
or
expulsion
for
smoking
at
school
School
policies
on
alcohol§
Warning/minor
action,
suspension,
expulsion
for
possessing
or
drinking
alcohol
at
school
School
policies
on
illicit
drugs§
Warning/minor
action,
suspension,
expulsion
for
possessing
or
using
drugs
at
school
Substance
use
programs
at
schoolt§
Drug
education,
drug
abuse,
or
alcohol
abuse
program
*Pearson
correlation
coefficient
was
used
to
assess
reliability
of
2-item
measures
where
appropriate.
tFor
most
variables
including
3
or
more
items,
Cronbach29
coefficient
was
used
to
assess
internal
consistency.
dlchotomously
categorized
variable,
eg,
yes/no,
any/none.
¿Derived
from
a
school
administrator
questionnaire.
poverty
status.
In
these
analyses,
race
was
categorized
as
black
vs
non-His¬
panic
white
as
the
reference
group;
eth¬
nicity
as
"other"
ethnic
group,
which
in¬
cluded
subcategories
of
Hispanic
(98%
white,
2%
black),
Asian/Pacific
Islander,
American
Indian,
and
"other"
(1%
des¬
ignated
2
or
more
ethnic
identities)
vs
non-Hispanic
white
as
the
reference
group;
family
structure
as
2
parents
in
the
home
vs
2
parents
not
in
the
home;
and
poverty
status
as
1
or
more
parents
on
welfare
vs
neither
parent
on
welfare.
While
a
simple
indicator
of
poverty
sta-
tus,
this
designation
has
been
shown
to
work
with
adolescent
respondents.42·43
Because
of
the
complex
patterns
of
in-
tercorrelation
between
variables
from
each
of
the
3
contexts,
the
total
variance
in
each
dependent
variable
explained
by
a
combination
of
family,
school,
and
indi¬
vidual
context
measures
is
typically
less
than
the
sum
of
the
variances
explained
by
each
context
analyzed
independently.
To
ensure
adequate
control
for
demo¬
graphic
effects,
in
the
first
step
of
analy¬
ses
demographic
variables
were
forced
into
regression
equations
and
retained
regardless
of
their
statistical
significance.
In
the
second
step
of
analyses,
the
set
of
generic
independent
variables
was
in¬
troduced;
significant
generic
measures
along
with
demographic
variables
were
retained
in
subsequent
regression
mod¬
els.
In
the
third
step
of
analyses,
a
set
of
domain-specific
independent
variables
was
introduced
into
regression
models,
and
significant
domain-specific
measures
were
retained.
In
a
fourth
and
final
step,
the
models
developed
on
the
exploratory
sample
were
cross-validated
by
recom¬
puting
parameter
estimates
on
the
vali-
Downloaded From: https://jamanetwork.com/ by a Columbia University Libraries User on 03/22/2019
Table
4.—Distribution
(Percentage)
of
Risk
Behaviors
by
Demographic
Variables
Risk
Behavior
Demographic
Variables
Emotional
Distress*
Suicide
Attempt
(2:1)
Violence
Perpetration*
Smoke
-6
Cigarettes/d
Alcohol
(Beer
or
Wine)
Use
=:2
d/mo
Marijuana
Use
at
Least
Once
in
Past
Month
Had
Sex
Pregnancy
Historyt
Grade
7th-8th
17.7
3.7 9.2
3.2 7.3
6.9
17.0
9th-12th
18.4
3.6
7.8
12.8
23.1
15.7
49.3
19.4
Sex
Male
15.7
2.1
11.0
10.0
39.9
Female
5.1
5.7
9.2
15.6
11.9
37.3
Geography
Urban
18.6
3.5
6.9
14.9
37.7
22.4
Suburban
17.5
3.7
8.2
10.2
19.2
13.8
38.2
Rural
19.9
3.6 8.4
11.9
17.7
10.5
41.9
15.7
Region
West
20.7
4.3
8.5
5.4
15.3
14.9
31.7
18.9
Midwest
20.4
12.8
18.4
15.0
38.0
18.9
Northeast
18.2
3.6
11.3
19.5
15.0
35.6
13.9
South
17.4 3.0
8.2 17.7
9.0
42.8
19.7
Poverty
Parents
receive
welfare
10.2
24.0
Parents
do
not
receive
welfare
7.6
9.5
12.4
37.9
"Continuous
variable
reported
as
a
mean
score;
higher
score
indicative
of
higher
risk.
tPercentage
of
those
who
are
sexually
active.
Table
5.—Percent
Variance
in
Dependent
Variables
Explained
by
Each
Context
Independently,
After
Controlling
for
Demographic
Factors*
Demographic
Factorst
Family
Context
School
Context
Individual
Characteristics
3
Models
and
Demographics
Combined:):
Dependent
Variables
Grades
7-8
Grades
9-12
Grades
7-8
Grades
9-12
Grades
7-8
Grades
9-12
Grades
7-8
Grades
9-12
Grades
7-8
Emotional
distress
4.2
5.9 14.6 13.5
17.6
13.1
21.8
21.0 30.0
27.1
Suicidality
1.0
1.2
7.0
3.1
3.0
2.5
5.9
9.9
Violence
6.6 8.0 6.5
4.6
5.8
43.9
49.6 50.6
Substance
use
Cigarette
use
2.2
6.2
6.4
3.7 5.7
10.0
14.5
14.4
Alcohol
use
1.0
2.9 8.5
6.1
5.6 4.3
7.1
7.3
13.7
12.5
Marijuana
use
1.6
2.0 5.6
8.6
5.6
4.8
7.4
10.2
13.7
*For
history
of
pregnancy
and
age
of
sexual
debut,
no
f?2
available
using
logistic
regression
or
Cox
regression.
tThe
factors
include
poverty
status,
family
structure,
race,
ethnicity,
and
sex.
^Explanatory
variables
significant
In
the
3
context-specific
analyses
were
retained
in
the
combined
analysis
regardless
of
changes
in
significance
due
to
¡ntercorrelations
among
them.
dation
sample,
with
all
retained
variables
from
the
estimation
analysis
forced
into
the
validation
analysis.
Thus,
indepen¬
dent
variables
found
in
final
models
in¬
cluded
the
full
set
of
demographic
vari¬
ables
as
well
as
generic
and
domain-spe¬
cific
measures
that
remained
significant
on
cross-validation.
For
linear
regression
analysis,
potential
design
effects
result¬
ing
from
the
use
of
a
cluster
sampling
de¬
sign
were
adjusted
with
the
use
of
a
mixed-models
linear
regression
proce¬
dure
(SAS
PROC
MIXED)44
with
speci¬
fied
use
of
a
block
diagonal
covariance
structure.
RESULTS
Prevalence
of
Behaviors
by
Demographic
Variables
The
distribution
of
key
risk
behaviors
in
the
national
sample
of
adolescents
is
presented
in
Table
4.
Prevalence
data
are
presented
by
grade
group,
place
of
resi-
dence,
region,
self-reported
poverty
sta¬
tus,
and
sex.
Emotional
Distress
and
Suicidal-
ity.—Two
indicators
of
risk
to
adoles¬
cents'
emotional
well-being
were
as¬
sessed:
emotional
distress
(a
recent
history
of
physical
and
emotional
symp¬
toms
of
distress)
and
suicidality
(a
history
of
suicidal
ideation
and
attempts
in
the
past
year).
Overall,
87.4%
(10
010/11453)
of
adolescents
indicated
that
they
had
nei¬
ther
suicidal
thoughts
nor
attempts
over
the
past
year.
A
total
of
10.2%
of
girls
(599/5745)
and
7.5%
of
boys
(428/5708)
re¬
ported
having
considered
suicide
without
having
attempted
it
over
the
past
year,
while
3.6%
of
all
adolescents
(415/11453)
(5.1%
of
girls
[295/5745]
and
2.1%
[120/
5708]
of
boys)
reported
suicide
attempts.
Of
adolescents,
3.6%
(412/11438)
re¬
ported
a
parental
suicide
attempt
during
the
previous
year,
while
0.9%
of
the
young
people
surveyed
(103/11438)
reported
suicide
completions
among
their
parents.
Family
Context.—Family
context
vari¬
ables
explained
14%
to
15%
of
the
vari¬
ability
in
emotional
distress
(9th-12th
graders
and
7th-8th
graders,
respec¬
tively)
and
5%
to
7%
of
the
variability
in
suicidality
for
all
adolescents
(Table
5).
The
key
aspect
of
family
context
that
ac¬
counted
for
these
relationships,
after
con¬
trolling
for
the
influence
of
demographic
factors,
was
parent-family
connected¬
ness
(Table
6).
The
presence
of
parents
at
key
times
during
the
day
(at
waking,
af¬
ter
school,
at
dinner,
and
at
bedtime),
shared
activities
with
parents,
and
high
pa¬
rental
expectations
for
their
child's
school
achievement
were
also
moderately
pro¬
tective
against
emotional
distress
for
both
younger
and
older
adolescents.
A
recent
family
history
of
suicidality
was
associ¬
ated
with
higher
distress
as
well
as
ado¬
lescent
suicidality.
Except
for
parent-family
connected¬
ness,
no
family
context
variables
sig¬
nificantly
protected
against
adolescent
Downloaded From: https://jamanetwork.com/ by a Columbia University Libraries User on 03/22/2019