Ajph Sos 2004

 RESEARCH AND PRACTICE 

An Outcome Evaluation of the SOS Suicide Prevention Program
| Robert H. Aseltine Jr, PhD, and Robert DeMartino, MD

Suicide among young people is one of the most
serious public health problems in the United
States. According to the National Center for
Health Statistics, the suicide rate for youths and
young adults aged 15 to 24 years has tripled
since 1950, and suicide is now the third leading cause of death in this age group.1,2 Recent
studies indicate that the incidence of suicide attempts among adolescents may exceed 10%
annually,3,4 although it is difficult to obtain reliable estimates because of the accompanying
stigma associated with attempting suicide.
A number of diverse approaches to suicide
prevention have been incorporated into high
school curricula in the past 15 years.5–7 Few,
however, have been subjected to rigorous evaluation, and those that have been scientifically
evaluated have produced mixed results. On the
positive side, a suicide awareness curriculum
developed by Spirito et al. yielded a significant
increase in knowledge about suicide and small
but statistically significant reductions in the use
of maladaptive coping strategies among ninthgrade students.8 Similarly, increases in personal
control, problem-solving coping, self-esteem,
and family support and decreases in depression
were observed among at-risk high school students who were exposed to brief supportive
counseling interventions developed by Randell
et al.9 However, these modest successes are
overshadowed by several other studies that
have failed to observe any effects of such interventions on students’ attitudes or behaviors.10,11
A relatively new approach to reducing the
incidence of suicide among adolescents is
found in Signs of Suicide (SOS), a school-based
prevention program. It incorporates 2 prominent suicide prevention strategies into a single
program by combining curricula to raise awareness of suicide and its related issues with a
brief screening for depression and other risk
factors associated with suicidal behavior.12 In
the didactic component of the program, SOS
promotes the concept that suicide is directly related to mental illness, typically depression, and
that suicide is not a normal reaction to stress or
emotional upset.13–17 Youths are taught to rec-

Objectives. We examined the effectiveness of the Signs of Suicide (SOS) prevention
program in reducing suicidal behavior.
Methods. Twenty-one hundred students in 5 high schools in Columbus, Ga, and Hartford, Conn, were randomly assigned to intervention and control groups. Self-administered
questionnaires were completed by students in both groups approximately 3 months
after program implementation.
Results. Significantly lower rates of suicide attempts and greater knowledge and more
adaptive attitudes about depression and suicide were observed among students in the
intervention group. The modest changes in knowledge and attitudes partially explained
the beneficial effects of the program.
Conclusions. SOS is the first school-based suicide prevention program to demonstrate significant reductions in self-reported suicide attempts. (Am J Public Health. 2004;94:446–451)

ognize the signs of suicide and depression in
themselves and in others, and they are taught
the specific action steps necessary for responding to those signs. The objective is to make the
action steps—ACT—as instinctual a response as
the Heimlich maneuver and as familiar an
acronym as CPR. ACT stands for acknowledge,
care, and tell: First, acknowledge the signs of suicide that others display and take those signs seriously. Next, let that person know that you care
and that you want to help. Then, tell a responsible adult.
The program’s teaching materials consist of
a video and a discussion guide. The video features dramatizations that depict the signs of suicidality and depression and the recommended
ways to react to someone who is depressed and
suicidal. It also includes interviews with real
people whose lives have been touched by suicide. Students also are asked to complete the
Columbia Depression Scale (CDS), a brief
screening instrument for depression, derived
from the Diagnostic Interview Schedule for
Children.10 The screening form is scored by the
students themselves; a score of 16 or higher on
the CDS is considered a strong indicator of clinical depression, and the scoring and interpretation sheet that accompanies the screening form
encourages students with such scores to seek
help immediately. Each school provides a description of the resources available to students
who wish to seek assistance.
The goal of the SOS program is to reduce
suicidal behavior among adolescents through

446 | Research and Practice | Peer Reviewed | Aseltine and DeMartino

2 mechanisms. First, the educational component of the program is expected to reduce suicidality by increasing students’ understanding
and recognition of depressive symptoms in
themselves and in others and by promoting
more adaptive attitudes toward depression and
suicidal behavior. Second, the self-screening
component of the SOS program helps students
assess and evaluate the depressive symptoms
and the suicidal thoughts they might be experiencing and prompts them to seek assistance
when dealing with these problems. Seeking
help need not be limited to referral for treatment by a mental health professional, which is
likely to be constrained by such factors as
availability and accessibility of providers,
health insurance coverage, and social stigma,
but should also be directed at the “indigenous
trained caregivers” in the school environment
(e.g., teachers and guidance counselors) as well
as from loved ones.18
In addition to its use of multiple suicide prevention strategies, the SOS program offers
other potential advantages. First, the focus on
peer intervention is developmentally appropriate for the targeted age group.7,19,20 During
adolescence, the peer group becomes the primary sphere of social involvement and emotional investment for most youths.21,22 The SOS
program capitalizes on a key feature of this developmental period by teaching youths to recognize the signs of depression and by empowering them to intervene when confronted with
a friend who is exhibiting these symptoms. Sec-

American Journal of Public Health | March 2004, Vol 94, No. 3

 RESEARCH AND PRACTICE 

ond, the program can be implemented on a
schoolwide basis by health educators with relative ease. Data from schools that offered the
SOS program during the 2001–2002 school
year indicate that the program can be implemented with minimal staff training and that the
program does not unduly burden teachers,
counselors, or administrative staff.23 Implementation of other suicide prevention programs that
include mental health screening can be costly,
difficult, and time-consuming.13
This article presents data from an outcome
evaluation of the SOS program conducted during the 2001–2002 school year in 5 high
schools in Hartford, Conn, and Columbus, Ga.
The primary goal of our research was to assess
the short-term impact of the program on suicidal behavior, seeking help, and knowledge of
and attitudes toward depression and suicide in
a diverse student population.

METHODS
Our study included 2100 public school students in 3 high schools in Hartford and 2 high
schools in Columbus. As indicated by the demographic profile of the sample (Table 1), these
schools provided a racially mixed and economically diverse sample of youths. The students in
the 3 Hartford schools (n=1435) were primarily economically disadvantaged youths from
diverse racial and ethnic backgrounds: approximately 59% of the Hartford sample was Hispanic and 20% was non-Hispanic Black.
Twenty percent of Hartford students had been
placed in a remedial English or bilingual program during high school. In contrast, the students in the Columbus schools (n=665) were
predominately from working- or middle-class
families, with approximately equal proportions
of White (39%) and Black (37%) youths.
The experimental design consisted of randomized treatment and control groups and
posttest-only data collection. In 4 of the 5 participating schools, students were randomly assigned to health classes (Hartford) and social
studies classes (Columbus) by a computerized
scheduling program. (Only ninth-grade classes
were eligible to participate in the Columbus
sites, because all other grades had received the
program during the previous year.) Because the
semester in which students were assigned to
these half-year classes was determined ran-

TABLE 1—Demographic Characteristics
by City, 2001–2002
Hartford, Conn Columbus, Ga
Race/ethnicity, %
Non-Hispanic White
Non-Hispanic Black
Hispanic
Multiethnic
Other
Gender, %
Male
Female
Grade, %
9th
10th
11th
12th
ESL classes during
high school, %
No
Yes

6
20
59
9
6

39
37
8
12
5

47
53

52
48

35
30
18
18

100
0
0
0

Measures and Instruments
80
20

85
15

Note. ESL = English as a Second Language. In Hartford
and Columbus, respectively, n = 1363 and 655 for
race/ethnicity; n = 1382 and 659 for gender;
n = 1352 and 655 for grade; and n = 1367 and 655
for ESL classes during high school. The race/ethnicity
numbers for Columbus add up to 101 because of a
rounding error.

domly, all students who took these classes during the first half of the school year were assigned to the treatment group and participated
in the program over a 2-day period from October through November 2001. Students who
took these classes during the second half of the
school year were assigned to the control group
and did not participate in the program until
after the evaluation was completed. The single
exception was a technical–vocational high
school in Hartford, where students were clustered in health classes according to their major
area of study and where class composition did
not change at midyear. For this school, random
assignment of classes to both the intervention
and the control conditions was achieved by flipping a coin. A number of potential concerns associated with the assignment of classrooms to
experimental conditions were minimized,24 because the same teachers and the same classrooms were used for both intervention and
control conditions in all 5 schools.
Students in both the treatment and the control groups were asked to complete a short

March 2004, Vol 94, No. 3 | American Journal of Public Health

questionnaire in a group setting during class
time approximately 3 months after implementation of the program. Trained interviewers
from the University of Connecticut’s Center
for Survey Research and Analysis and Columbus State University read aloud the questions
to each class, and students recorded their confidential written responses on anonymous
questionnaires. Parents were notified in writing about the objectives of the study and were
invited to contact their respective schools to
ask questions or to withdraw their child from
the study. Questionnaires were completed by
2100 of the 2258 students eligible for the
study (n = 1073 for the control group, n =
1027 for the treatment group), which resulted
in an overall response rate of 93%.

The questionnaire included items relevant
to 3 specific categories of outcome: (1) selfreported suicide attempts and suicidal ideation,
(2) knowledge and attitudes about depression
and suicide, and (3) help-seeking behavior.
The primary endpoint for our study was a
single-item measure of self-reported suicide attempts taken from the Centers for Disease
Control and Prevention’s (CDC) Youth Risk Behavior Survey: “During the past 3 months, did
you actually attempt suicide (yes or no)?” 4 Suicidal ideation also was assessed with a question taken from the survey: “During the past 3
months, did you ever seriously consider attempting suicide (yes or no)?”
The measures of knowledge and attitudes
about depression and suicide were adapted
from instruments previously used to evaluate
school-based suicide prevention programs.8,10
Knowledge of depression and suicide was measured with 10 true/false items that reflect the
central themes of the SOS program (e.g., “People who talk about suicide don’t really kill
themselves”; “Depression is an illness that doctors can treat”). Scores on this variable reflected
the number of correct answers. The measure
of attitudes toward depression and suicide was
an 8-item summary scale that assessed attitudes toward suicidal people and suicidal behaviors (e.g., “If someone really wants to kill
him/herself, there is not much I can do about
it”; “If a friend told me he/she is thinking about
committing suicide, I would keep it to myself”).
Responses to these questions ranged from

Aseltine and DeMartino | Peer Reviewed | Research and Practice | 447

 RESEARCH AND PRACTICE 

TABLE 2—Descriptive Characteristics of Measures of Suicidal Behavior, Knowledge,
and Attitudes

Treated for depression/suicidal ideation, %
Talked with adult about depression/suicidal ideation, %
Talked with adult about friends’ emotional problems, %
Suicidal ideation during past 3 months, %
Suicide attempt during past 3 months, %
Knowledge of depression/suicide, mean (SD)
Attitudes toward depression/suicide, mean (SD)

“strongly disagree” to “strongly agree” on a
5-point scale, with higher values indicating
more adaptive attitudes about depression and
suicide (Cronbach α=.74). Three questions
were used to assess help-seeking behavior. Students were asked whether in the past 3
months, “ . . . you received treatment from a
psychiatrist, psychologist, or social worker because you were feeling depressed or suicidal
(yes or no)”; whether “. . . you talked to some
other adult (like a parent, teacher or guidance
counselor) because you were feeling depressed
or suicidal (yes or no)”; and whether “. . . you
talked to an adult about a friend you thought
was feeling depressed or suicidal (yes or no).”
Subjects who had missing values on any variable in a particular analysis were excluded
from that analysis. Although 84 youths assigned to the treatment group did not actually
participate in either of the central elements
of the program—the video and depression
screening—mainly because of absences from
school, they were retained in the analysis so
that we could estimate “intention to treat” effects. After exclusions for missing data, the
effective sample size for these analyses ranged
from 1894 to 1912. Descriptive statistics for all
dependent variables used in this analysis are
shown separately by treatment status in Table 2.

RESULTS

Control
(n = 1073)

Treatment
(n = 1027)

Total Sample
(N = 2100)

Valid N

9.9
18.7
13.0
12.2
5.4
6.49 (1.68)
3.80 (0.658)

8.5
15.9
11.9
10.1
3.6
7.18 (1.68)
4.05 (0.644)

9.2
17.3
12.4
11.2
4.5
6.67 (1.97)
3.93 (0.662)

2039
2041
2042
2034
2042
2090
2041

ences in the composition of treatment and control groups by race/ethnicity or gender. However, significant differences were observed for
grade (χ2 =23.6, df=3) and for ESL status
(χ2 =7.8, df=1): 10th-grade students were
slightly more likely than students in other
grades to be assigned to the treatment group
(e.g., 58% of 10th-grade students were in the
treatment group vs an expectation of 50%),
while ninth-grade students were slightly less
likely than students in other grades to be assigned to the treatment group (44% in treatment), and only 40% of those who had taken
ESL or bilingual classes during high school
were assigned to the treatment group.

Assessing Effects of the SOS Program
To account for the assignment of classrooms
to experimental conditions, we used HLM 5
software25 to perform multivariate analyses of
program effects. HLM was developed to address generic problems in the analysis of hierarchical data structures—that is, data in which
characteristics of 1 unit of analysis (e.g., individuals) are nested within and vary among larger
units (e.g., social groups or contexts). In our
analysis, the effect of exposure to the SOS program on each outcome variable was estimated
in a 2-level HLM model, where students (the
level-1 unit of analysis) were nested with classrooms (the level-2 unit of analysis). The basic
level-1 model for these outcomes was

Comparability of Treatment
and Control Groups

(1)

Preliminary analyses were conducted to assess the comparability of treatment groups and
control groups in terms of race/ethnicity, gender, grade, and English as a Second Language
(ESL) status. Chi-square tests revealed no differ-

where Y represents the predicted value of each
outcome variable for each individual (i) in the
classroom ( j ); Female, Race, and ESL represent

Yij =B0j + B1j Femaleij + B2–5j Raceij +
B6j ESLij + B7–9jGradeij + eij

448 | Research and Practice | Peer Reviewed | Aseltine and DeMartino

a series of dummy variables for the demographic control variables included in the analysis; and e represents random error. To reduce
the error variance in the outcome measures
and to control for differences in the composition of the treatment and control groups,26 all
level-1 models included dummy variables for
race/ethnicity (non-Hispanic Black, Hispanic,
multiethnic, and other race vs non-Hispanic
White), gender (female vs male), grade (10, 11,
and 12 vs 9), and ESL status (ESL vs no ESL).
Because exposure to the SOS program was
determined at the classroom level, treatment effects were assessed for each outcome by inserting a dummy variable for exposure to the program into the level-2 equation for the level-1
intercept term:
(2)

B0j =G00 + G01Treatmentj + U0j

The random error in this equation (U0j )
represents residual variability in treatment effects across classrooms. All demographic control variables were modeled as fixed effects
(i.e., B1j = G10).
The effects of the SOS program on students’
knowledge of and attitudes toward depression
and suicide, help-seeking behavior, and suicidal
ideation and self-reported suicide attempts are
shown in Table 3. For the analysis of attitudes
and knowledge, this table shows coefficients
from a standard 2-level HLM analysis; for helpseeking behavior, suicidal ideation, and suicide
attempts, coefficients are derived from nonlinear 2-level HLM models that used the logit link
function. The top row in Table 3 shows the effects of exposure to the SOS program on the
various outcome measures included in our
study. First and most important, the coefficients
shown in column 1 of Table 3 indicate that exposure to the SOS program was associated with
significantly fewer self-reported suicide attempts. The coefficient for the effect of the
SOS program on attempts is –.467, which
when converted to an odds ratio (OR) indicates that the youths in the treatment group
were approximately 40% less likely to report
a suicide attempt in the past 3 months compared with youths in the control group (OR=
e –.467 =0.628). The magnitude of the difference between the treatment group and the control group also is indicated in the descriptive
statistics shown in Table 2; the rate of selfreported suicide attempts among students in the

American Journal of Public Health | March 2004, Vol 94, No. 3

 RESEARCH AND PRACTICE 

TABLE 3—Effects of Signs of Suicide Program on Students’ Knowledge of and Attitudes Toward Depression and Suicide, Seeking Help, and
Suicidal Ideation and Suicide Attempts

(SE)

Intercept
SOS program
Female
Hispanic
Non-Hispanic Black
Multiracial
Other race
ESL status
10th grade
11th grade
12th grade
ICC

Attempts

Ideation

Knowledge

Attitudes

Treatment

Adult

Adult/Friend

–3.447* (.133)
–.467* (.207)
1.022* (.313)
–.193 (.218)
–1.478* (.378)
–.025 (.392)
–1.307* (.659)
.753* (.273)
–.434 (.337)
–.540 (.438)
–.281 (.426)
.000

–2.196* (.078)
–.272 (.147)
.764* (.183)
–.245 (.144)
–1.027* (.202)
–.095 (.232)
–.510 (.342)
–.113 (.198)
.117 (.191)
–.387 (.306)
–.016 (.226)
.002

6.803* (.054)
.689* (.109)
.349* (.077)
–.626* (.108)
–.589* (.104)
–.432* (.145)
–.495* (.194)
–.569* (.103)
.176 (.137)
.228 (.151)
.336 (.142)
.088

3.914* (.019)
.255* (.038)
.136* (.031)
.097* (.038)
.039 (.032)
–.038 (.054)
–.050 (.070)
–.029 (.086)
–.040 (.039)
.057 (.059)
.050 (.054)
.071

–2.459* (.094)
–.217 (.181)
.719* (.189)
–.147 (.299)
–.999* (.288)
–.147 (.299)
–.692 (.388)
.495* (.177)
–.217 (.288)
–.071 (.268)
.105 (.251)
.011

–1.759* (.081)
–.233 (.146)
1.266* (.193)
.091 (.158)
–.415* (.187)
.344 (.214)
.032 (.292)
.332* (.155)
–.595* (.201)
–.132 (.193)
–.166 (.208)
.011

–2.114* (.074)
–.147 (.138)
1.152* (.165)
.132 (.186)
–.388 (.199)
–.138 (.274)
–.520 (.392)
.314 (.170)
–.057 (.206)
–.141 (.233)
–.115 (.221)
.000

Note. ESL = English as a Second Language; ICC = intraclass correlation coefficient for each outcome.
*P < .05.

control group was 5.4%, compared with only
3.6% among students in the treatment group.
Similarly, exposure to the SOS program resulted in greater knowledge of depression and
suicide and more adaptive attitudes toward
these problems (Table 3, columns 3 and 4).
The effects of the SOS program on knowledge
and attitudes were modest in magnitude and
resulted in effect sizes of slightly more than one
third of a standard deviation (e.g., knowledge:
.689/1.98=.35). The effects of the SOS program on both attitudes and knowledge remained statistically significant at the .0071 and
.0083 levels, respectively, when Holm adjustments were applied to correct for multiple tests
that involved these secondary endpoints.27,28 In
contrast, the effects of the SOS program on
help-seeking behavior did not achieve statistical
significance. The negative coefficients for treatment effects in columns 3, 4, and 5 of Table 3
indicate that the treatment group was slightly
less likely than the control group to seek help
for emotional problems, but these effects did
not achieve statistical significance at either a
nominal or a corrected .05 α level. Finally, although the descriptive statistics in Table 2 indicate lower levels of suicidal ideation among the
treatment group, this difference fell short of statistical significance at the .05 level in the full
multilevel model (Table 3, column 2).
With regard to the impact of the demographic control variables on these outcomes,

the patterns observed in Table 3 are consistent
with those observed in national data from the
1999 Youth Risk Behavior Surveys.4 The female coefficients used in these models indicate
that girls, compared with boys, had significantly greater knowledge and more constructive attitudes about depression and suicide,
were more likely to seek help when depressed
and to intervene on behalf of friends, and were
significantly more likely to report suicidal
ideation and suicide attempts in the past 3
months.29 Students in high school ESL programs had less accurate knowledge about
depression and suicide and had a higher prevalence of self-reported suicide attempts. However, ESL status was positively related to seeking help, as students in these programs were
more likely to seek treatment or to talk with an
adult when feeling depressed.
Significant effects of race/ethnicity on
knowledge of depression and suicide, 2 of the
help-seeking outcomes, and suicidal ideation
and self-reported suicide attempts also were
observed. White students were more knowledgeable about depression and suicide compared with those in other race and ethnic categories. However, Black students reported
lower rates of suicidal ideation and suicide attempts than did White students and were less
likely to seek professional help for these problems, consistent with previous epidemiological
research that showed lower rates of suicidal

March 2004, Vol 94, No. 3 | American Journal of Public Health

ideation and depression among Blacks.1,4
A reparameterization of the models shown in
Table 3 (adding a dummy variable for White
race and removing the Black term) indicated
that Black students also had significantly
lower rates of suicidal ideation, self-reported
suicide attempts, and seeking professional
help than did Hispanic students. Differences
in these outcomes by grade did not exceed
what would be expected by chance (only
1 significant effect out of 21 contrasts).
Finally, the intraclass correlation coefficient
for each outcome variable is shown in the bottom row of Table 3. The coefficients range
from nearly 0 (for self-reported suicide attempts, suicidal ideation, and talking with an
adult about a troubled friend) to a high of .07
to .09 (for the measures of knowledge and attitudes). These coefficients indicate that there
is a high degree of independence among observations within classrooms for each outcome
variable; at the most, only 7% to 9% of the
variance in these outcomes occurred at the
classroom level.

Effects of the SOS Program
on Suicide Attempts
The impact of the SOS program on suicidal
behavior may in part be due to its role in fostering greater knowledge about and more constructive attitudes toward depression and suicide. These 2 measures were included as

Aseltine and DeMartino | Peer Reviewed | Research and Practice | 449

 RESEARCH AND PRACTICE 

TABLE 4—Role of Knowledge and
Attitudes in Mediating the Effects of
the Signs of Suicide Program on
Suicide Attempts
Suicide Attempts

Intercept
SOS program
Knowledge
Attitudes

Model 1,
β (SE)

Model 2,
β (SE)

–3.447 (.133)
–.467* (.267)
...
...

–3.615 (.146)
–.264 (.207)
–.195* (.055)
–.605* (.165)

Note. All models controlled for gender, race/ethnicity,
grade, and English as a Second Language status.
*P < .05.

predictor variables in the level-1 model for selfreported suicide attempts (Table 4) so that we
could examine the role of knowledge and attitudes in explaining the effects of the SOS program on suicidality. More adaptive attitudes toward depression and suicide and greater
knowledge of depression and suicide were both
significantly associated with a lower probability
of self-reported suicide attempts. When we controlled for these variables, the effect of the SOS
program on self-reported attempts was substantially reduced, as demonstrated by the finding that the coefficient that captured the effect
of the program on this outcome was reduced
by approximately 40% ([(–.467)–(–.264)]/
–.467) and was no longer statistically significant. Although there is some causal ambiguity
regarding the associations between these concurrent measures of attitudes and behavior, our
analysis suggests that a substantial portion of
the effect of the SOS program on self-reported
suicide attempts may be explained by the subjects’ improved understanding of and attitudes
about depression and suicide.

DISCUSSION
It is clear from these data that the SOS suicide prevention program had a substantively
important short-term impact on the attitudes
and behaviors of high school–aged youths in
high-risk settings. By significantly reducing
rates of self-reported suicide attempts in the 3
months following exposure to the program,
SOS appears to have had a substantial impact
on the ultimate target of suicide prevention

programs. Efficacy in increasing students’
knowledge of and promoting more adaptive
attitudes toward depression and suicide also
was demonstrated, and further analysis highlighted the importance of these variables in
potentially accounting for the beneficial effects
of the SOS program on self-reported suicide
attempts. Although further research is necessary to determine whether the effects of the
SOS program are enduring, the short-term impact of this program on students’ attitudes and
behaviors was noteworthy. This is the first
school-based suicide prevention program for
which a reduction in self-reported suicide attempts has been documented with a randomized experimental design.
In contrast, significant effects of the SOS program on suicidal ideation and help-seeking behaviors were not observed. The fact that selfreported suicide attempts were reduced by a
much greater extent than were thoughts of suicide is most likely a result of the SOS program’s
relatively greater emphasis on action and behavior. Reductions in levels of suicidal ideation
are expected to be an ancillary benefit of the
SOS program, particularly if the program’s efforts to encourage active engagement and communication with peers about these issues fosters a general mobilization of peer support.22
However, suicide prevention programs that
place a greater emphasis on personal growth
and positive youth development will likely have
a greater relative impact on outcomes such as
depressed mood and suicidal ideation.
Although significant effects of the intervention on help-seeking behaviors were expected,
further investigation revealed several likely explanations for the absence of program effects
on help-seeking behaviors for this particular
sample. First, a process evaluation that included
site coordinators at schools that implemented
the SOS program during the 2000–2001
school year found evidence that the number of
youths who sought help from school personnel,
either because of their own emotional problems or because of those of their friends, was
generally lower in urban communities. Second,
there were several barriers to seeking help that
were specific to schools involved in our study,
particularly in Hartford. Administrators in the
Hartford schools reported a serious shortage of
available staff for helping students with mental
health concerns. Moreover, a series of informal

450 | Research and Practice | Peer Reviewed | Aseltine and DeMartino

discussions conducted in 12 classes from 3
Hartford schools several months after exposure
to the program revealed that students were unlikely to seek out school personnel to discuss
emotional problems, primarily because of confidentiality concerns. Instead, students reported
that friends were the first people they would
turn to when feeling depressed, a finding that is
corroborated in previous research.7
Some may question the rates of self-reported
suicide attempts in our sample (4.5% over a
3-month period), which appear to be somewhat higher on an annualized basis than recent 1-year national prevalence estimates
from the CDC’s Youth Risk Behavior Surveys
(8.5%–10.5%).4 Although there is ample reason to expect higher rates of suicidal behavior
in our sample because of the predominance
of seriously disadvantaged youths at high risk
for depression and substance abuse, research
has shown that data collected during shorter
recall periods cannot be “annualized” through
simple multiplication (i.e., multiplying the
3-month prevalence by 4). For example, epidemiological data from the National Comorbidity Survey on the course of major depression among adolescents indicate that the
1-month prevalence rate for major depression
is approximately one half that observed for
the past year, because of chronicity and the
lengthy duration of depressive episodes.30 Applying this logic to the 3-month prevalence
rates obtained in our study yields annual
prevalence rates that are not inconsistent with
the national data published by the CDC. No
suicides were reported in any of the participating schools during the study period.
Finally, our study has a number of limitations that must be acknowledged. First, our
evaluation should be replicated in more socially
and geographically diverse locations. The significant positive impact of the SOS program on
high-risk youths in urban settings is certainly an
important finding, but replication in rural and
suburban settings that contain fewer disadvantaged youths is necessary to determine whether
these findings are generalizable to a broader
population. Second, the effects of the SOS program were observed over a very short postintervention period. A longer-term follow-up of
youths exposed to the SOS program is necessary to determine whether the observed effects
are enduring. Third, pretest measures of the

American Journal of Public Health | March 2004, Vol 94, No. 3

 RESEARCH AND PRACTICE 

outcomes assessed in our study would add confidence that the assignment of classes to experimental conditions resulted in equivalent
groups. Fourth, our study has revealed some of
the challenges facing school-based programs
designed to foster help-seeking behaviors
among students. Uncertainty about confidentiality may be acting to suppress interaction between students and school personnel regarding
serious mental health concerns, which may
lead to acute problems among youths in highrisk settings who possess limited parental and
financial resources. Relatedly, future research
should assess the degree to which help-seeking
behaviors among emotionally troubled adolescents are directed toward friends and siblings.
Future research also should assess the impact
of the support received from these relationships
on suicidal behavior. Finally, some may question whether our results are tainted by the desire of the students exposed to the program to
provide what they perceive to be the “right answers” when responding to survey questions
about attitudes and behavior; however, suicide
prevention programs have historically demonstrated very little efficacy. Adolescents have not
felt compelled to select what they thought were
the “right” answers in previous research, and
there does not appear to be anything unique
about this sample that would lead students to
do so in our study. Additionally, if students
were endorsing the “right” answers rather than
their true feelings and experiences, it is reasonable to expect that treatment effects would be
observed universally. The selective impact of
the SOS program on the various outcomes assessed in our study provides fairly strong evidence to the contrary.

About the Authors
Robert H. Aseltine Jr is with the Department of Behavioral
Sciences and Community Health, University of Connecticut
Health Center, Farmington, Conn. Robert DeMartino is with
the Center for Mental Health Services, Substance Use and
Mental Health Services Administration, Rockville, Md.
Requests for reprints should be sent to Robert H. Aseltine
Jr, PhD, Dept of Behavioral Sciences and Community Health,
MC 3910, University of Connecticut Health Center 263
Farmington Ave, Farmington, CT 06030-3910 (e-mail:
aseltine@uchc.edu).
This article was accepted August 27, 2003.

Contributors
R. Aseltine conceived of the study and took primary responsibility for data analysis and writing of the article. R.

DeMartino contributed to the study design and data interpretation and reviewed drafts of the article.

12. Shaffer D, Craft L. Methods of adolescent suicide
prevention. J Clin Psychiatry. 1999;60(suppl 2):70–74.

Acknowledgments

13. Jacobs DG, Brewer M, Klein-Benheim M. Suicide assessment: an overview and recommended protocol. In:
Jacobs DG, ed. Guide to Suicide Assessment and Intervention. San Francisco, Calif: Jossey-Bass; 1999:3–39.

Support for this project was provided by the Center for
Mental Health Services, Substance Abuse and Mental
Health Services Administration, and by a grant from the
Robert Leet and Clara Guthrie Patterson Trust.
We would like to thank Barbara Kopans, Amy Bloom,
and Gene Wallenstein for assisting us with this study and
Douglas Jacobs and Ross Baldessarini for helpful comments on earlier drafts of our article. We also would like
to acknowledge the hard work and cooperation of teachers, counselors, and administrators in Hartford, Conn, and
Columbus, Ga.

Human Participant Protection
The procedures used to collect these data were approved
by the institutional review board of the University of Connecticut Health Center.

References
1. Moscicki EK. Epidemiology of suicide. In: Jacobs
DG, ed. Guide to Suicide Assessment and Intervention. San
Francisco, Calif: Jossey-Bass; 1999:40–51.

14. Brent DA, Kolko DJ. The assessment and treatment
of children and adolescents at risk for suicide. In: Blumenthal SJ, Kupfer DJ, eds. Suicide Over the Life Cycle:
Risk Factors, Assessment, and Treatment of Suicidal Patients. Washington, DC: American Psychiatric Press;
1990:253–302.
15. Lewinsohn PM, Rohde P, Seeley JR. Psychosocial
risk factors for future adolescent suicide attempts. J Consult Clin Psychol. 1994;62:297–305.
16. Andrews JA, Lewinsohn PM. Suicidal attempts
among older adolescents: prevalence and co-occurrence
with psychiatric disorders. J Am Acad Child Adolesc Psychiatry. 1992;31:655–662.
17. Velez CN, Cohen P. Suicidal behavior and ideation
in a community sample of children: maternal and youth
reports. J Am Acad Child Adolesc Psychiatry. 1988;27:
349–356.
18. Gullotta TP. Prevention’s technology. J Primary Prev.
1987;7:176–196.

2. Murphy SL. Deaths: final data for 1998. In: National Vital Statistics Report 48(11). Hyattsville, Md: National Center for Health Statistics; 2000.

19. Davis JM, Sandoval J. Involving peers in suicide prevention. In: Davis JM, Sandoval J, eds. Suicidal Youth:
School-Based Intervention and Prevention. San Francisco,
Calif: Jossey-Bass; 1991:140–149

3. Joffe RT, Offord DR, Boyle MH. Ontario Child
Health Study: suicidal behavior in youth age 12–16
years. Am J Psychiatry. 1998;145:1420–1423.

20. Kellam SG, Koretz D, Moscicki EK. Core elements
of developmental epidemiologically based prevention research. Am J Community Psychol. 1999;27:463–482.

4. Kann L, Kinchen SA, Williams BI, et al; Centers for
Disease Control and Prevention. Youth risk behavioral
surveillance—United States. MMRW CDC Surveill Summ.
2000;49(5):1–32.

21. Coleman JS. The Adolescent Society. New York, NY:
Free Press of Glencoe; 1961.

5. Ploeg J, Ciliska D, Dobbins M, Hayward S, Thomas
H, Underwood J. A systematic overview of adolescent
suicide prevention programs. Can J Public Health. 1996;
87:319–324.

22. Aseltine RH Jr, Gore S, Colten ME. Depression and
the social developmental context of adolescence. J Pers
Soc Psychol. 1994;67:252–263.
23. Aseltine RH Jr. An evaluation of a school-based suicide prevention program. Adolesc Fam Health. 203;2:
81–88.

6. Garland A, Whittle B, Shaffer D. A survey of youth
suicide prevention programs. J Am Acad Child Adolesc
Psychiatry. 1989;28:931–934.

24. Cook T, Campbell D. Quasi-Experimentation: Design
and Analysis Issues for Field Settings. Boston, Mass:
Houghton Mifflin; 1979.

7. Kalafat J, Elias MJ. Suicide prevention in an educational context: broad and narrow foci. In: Silverman MM,
Maris RW, eds. Suicide Prevention: Toward the Year 2000.
New York, NY: Guilford Press; 1995:123–133.

25. Raudenbush S, Bryk A, Cheong YF, Congdon R.
HLM 5: Hierarchical Linear and Nonlinear Modeling. Lincolnwood, Ill: Scientific Software International; 2000.

8. Spirito A, Overholser J, Ashworth S, Morgan J,
Benedict-Drew C. Evaluation of a suicide awareness curriculum for high school students. J Am Acad Child Adolesc Psychiatry. 1988;27:705–711.
9. Randell BP, Eggert LL, Pike KC. The immediate
post-intervention effects of two brief youth suicide prevention interventions. Suicide Life Threat Behav. 2001;
31:41–61.

26. Rossi PH, Freeman HE. Evaluation: A Systematic
Approach 5. Newbury Park, Calif: Sage Publications;
1993.
27. Sankoh AJ, Huque MF, Dubey SD. Some comments
on frequently used multiple endpoint adjustment methods in clinical trials. Stat Med. 1997;16:2529–2542.
28. Aicken M, Gensler H. Adjusting for multiple testing
when reporting research results: the Bonferroni vs Holm
methods. Am J Public Health. 1996;86:726–728.

10. Shaffer D, Garland A, Vieland V, Underwood M,
Busner C. The impact of curriculum-based suicide prevention programs for teenagers. J Am Acad Child Adolesc
Psychiatry. 1991;30:588–596.

29. Overholser JC, Hemstreet AH, Spirito A, Vyse S.
Suicide awareness programs in the schools: effects of
gender and personal experience. J Am Acad Child Adolesc
Psychiatry. 1989;28:925–930.

11. Vieland V, Whittle B, Garland A, Hicks R, Shaffer
D. The impact of curriculum-based suicide prevention
programs for teenagers: an 18-month follow-up. J Am
Acad Child Adolesc Psychiatry. 1991;30:811–815.

30. Kessler RC, Walters EE. Epidemiology of DSM-III-R
major depression and minor depression among adolescents and young adults in the National Comorbidity Survey. Depress Anxiety. 1998;7:3–14.

March 2004, Vol 94, No. 3 | American Journal of Public Health

Aseltine and DeMartino | Peer Reviewed | Research and Practice | 451