Effects of a School-Based Human
Immunodeficiency Virus Infection and Sexual Risk
HIV/AIDS HIV/AIDS Resource Center, The Journal of American
pp. 961-970, Oct. 1998
M. Siegel, MD, MPH; Marilyn J. Aten, PhD, RN; Klaus J.
Roghmann, PhD; Maisha
To determine the short-term effect of a middle and high
school-based human immunodeficiency virus and sexuality
intervention (Rochester AIDS Prevention Project for Youth
[RAPP]) on knowledge, self-efficacy, and behavior intention.
Nonrandomized intervention study with 2 intervention groups
and 1 control group. Setting: Middle and high school health
classes in an urban, predominantly minority school district.
Middle and high school students (N=3635) enrolled in health
classes in 9 schools; 50% African American, 16% Hispanic, 20%
white, and 14% other. Less than 10% of students refused
There were 3 study conditions: (1) Control, usual health
education curriculum taught by classroom teacher; (2) RAPP
adult health educator, intervention curriculum implemented by
ethnically diverse male-female pairs of highly trained health
educators; and (3) RAPP peer educator, intervention
implemented by male-female pairs of extensively trained high
school students. Health classes within schools were assigned
to 1 of the 3 conditions each semester, and simultaneous
implementation of the control program with health educators or
peer educators in the same school and during the same semester
was not permitted.
A confidential questionnaire administered to all study
subjects before and immediately after the intervention,
containing scales to measure knowledge, sexual self-efficacy,
and safe behavior intention.
Preintervention data indicated that the study population was
involved in sexual activity and other risk behaviors at rates
comparable to those of other urban adolescent populations.
Examination of 3 outcome constructs as dependent variables
(knowledge, sexual self-efficacy, and safe behavior intention)
revealed that the health educators and peer educators
increased students' knowledge significantly more than did the
control condition for both middle (females, P<.01; males,
P<.01) and high (females, P<.001; males, P<.001) school.
Comparisons of self-efficacy changes across intervention
groups did not reach statistical significance, and safe
behavior intention changes differed significantly by
intervention group for high school but not for middle school
students. For all analyses, the preintervention scores for
each outcome variable were the most powerful predictors of
postintervention scores, and analysis of variance models
predicted substantial overall variance.
At short-term follow-up, the RAPP intervention had a powerful
effect on knowledge for all students and a moderate effect on
sexual self-efficacy and safe behavior intention, particularly
for high school students. The peer educators were found to be
equally and, for some variables, more effective than the
highly trained adult educators. The substantial effect of the
baseline scores and the high prevalence of risk behavior
already evident by seventh grade indicate the importance of
early implementation of school-based sexuality programs. Arch
Pediatr Adolesc Med. 1998;152:961-970
Editor's Note: The 2 interventions seem to be effective in
changing short-term knowledge. I hope that the authors plan a
follow-up on student-reported behavior . . . and then wouldn't
it be great to determine actual practice. I can dream, can't
I? Catherine D. DeAngelis, MD
Adolescent sexual risk behaviors continue to represent one of
the most serious public health problems in the United
States.[1-4] Consequences of these activities include
pregnancy,[2,5,6] sexually transmitted diseases (STDs)[7-12]
and, most recently, human immunodeficiency virus infection and
the acquired immunodeficiency syndrome (HIV/AIDS).[13-18]
While adolescents still represent less than 1% of the nation's
identified HIV/AIDS population,[16,17] the disease incubation
period extends well beyond 10 years and it is currently
estimated that 1 in 5 Americans with AIDS was infected during
adolescence. In response to this increasing and profound
HIV/AIDS risk, as well as those of STDs and pregnancy, a
multitude of strategies have been developed to address sexual
risk reduction among adolescents.
preventive and risk reduction efforts include school-based
curricula reflecting a wide variety of informational content
and methods. Program goals can be categorized as abstinence
only, sex education, or HIV/STD education. Key distinguishing
features exist among these program categories. Abstinence
programs do not include discussion of birth control aside from
contraceptive failure and/or disease prevention.[19,20] In
contrast, sex education and HIV/STD education programs include
information about abstinence, sexuality, contraception, and
disease prevention. A range of methods have been used by
school-based interventions to disseminate information and
impart behavioral skills. These include teaching by peers,
classroom teachers, and/or adults from outside agencies;
incorporation of highly interactive exercises and skill-based
methods with or without didactic presentations; and the direct
or indirect involvement of parents and guardians.[19-21,22-25]
Curricula also vary widely in duration, consisting of anywhere
from 1 to 30 classroom sessions.
effectiveness, as measured by changes in knowledge, attitudes,
self-efficacy, behavioral intention, and behavior, has varied.
The importance of examining self-efficacy (the adolescent’s
belief in his or her ability to engage in a specific behavior)
and behavior intention (the adolescent's belief that he or she
will engage in a particular behavior within the next year) is
derived from theories of behavior change. Social learning
theory, the theory of reasoned action, and the theory
of planned behavior all hold that in addition to knowledge
about the ramifications of chosen behaviors, one's
self-efficacy regarding the behavior is an important predictor
of one's intention to behave in a certain way. Further,
behavior intention is proposed to be closely linked to
behavior. The ultimate effectiveness of risk reduction
programs can only be meaningfully assessed by measuring the
maintenance of safe behavior or adherence to safer sex
practices over a significant duration (eg, >6 months). As a
potential first step to long-term change it is also important
to address early program effects (1-3 months after
knowledge alone has not been found to be sufficient to change
behavior, it is certainly a necessary prerequisite.[29,30]
Several studies have reported success in improving students'
information base around sexuality and HIV/AIDS. Project SNAPP,
a randomized study based in 6 urban middle schools, used an
8-session, peer-taught, skills-based, highly interactive HIV
and pregnancy prevention intervention, which was compared with
the existing school curriculum. While a positive effect on
knowledge was noted, the 17-month follow-up revealed an
improvement in only 2 of 21 relevant attitudes or beliefs, and
there was no significant change in sexual or contraceptive
behaviors. Other investigators have similarly described
knowledge increases, but with mixed results in other measured
constructs.[29,31-33] Main et al, reporting on a
15-session, skills-based HIV prevention curriculum implemented
in Colorado, noted significant HIV knowledge increases among
students in 10 intervention schools as opposed to students in
7 comparison schools. The experimental students also expressed
greater intentions to engage in safer sex practices within the
next 2-month period.
review of the effectiveness of 40 interventions designed to
reduce AIDS risk in adolescents, Kim et al reported that
of the 12 studies that assessed changes in attitudes toward
personal preventive behavior, 7 (58%) found significant
improvement, but that most of these were nonrandomized
designs. Other articles describing knowledge and attitude
changes tended to find improvement in the former but not
consistently in the latter.[23,31-33] Weeks et al reported
significant increases in contraceptive self-efficacy among
middle school students in Chicago, Ill, after a 15-session
classroom-based intervention. Walter and Vaughan also
observed significant, albeit modest, changes in self-efficacy
related to HIV preventive actions among high school students
participating in a 6-session AIDS prevention curriculum.
However, Newman et al reported a decrease in middle school
students' self-efficacy related to AIDS prevention behaviors
as well as their level of communication with peers and family
members about AIDS following a 1-hour HIV education program
developed and taught by the Red Cross. In this study knowledge
scores also failed to increase as a result of the brief
intervention. Thus, school-based programs aimed at HIV risk
among adolescents seem to have some successes, consistently in
the area of knowledge change and somewhat in self-efficacy and
attitudes, but only in the context of substantial content and
duration. The reasons for intervention success or failure have
yet to be fully explained.
important factor both for implementation and evaluation of
school-based studies is student attendance. That is, students
may not be present for an entire intervention and yet they
participate in pretesting and posttesting and become part of
the outcome database. While this, of course, is consistent
with all clinical trials, the methodologic consideration is
whether to eliminate students who have not attended all
sessions (a severely compromising choice that ignores the
realities of generalizability) or make an attempt to measure
the "dose," or degree of exposure to the program.
Surprisingly, intervention dose and its relation to
intervention effect has been rarely considered in school-based
work. Additionally, studies often fail to include any measure
of the learning adequacy of the existing classroom
environment. Relevant variables include the physical
environment, support of the learning process, and control of
students in the classroom.
Rochester AIDS Prevention Project for Youth (RAPP) is a middle
and high school-based intervention trial. We report below on
preintervention to immediate postintervention changes in
knowledge concerning HIV/AIDS and sexuality, self-efficacy,
and behavior intention. The effects on these dependent
variables of dose as well as the adequacy of the learning
environment are included in the analyses.
PARTICIPANTS AND METHODS
subjects (N=3696, Table 1) were drawn from 9 urban schools in
Rochester, NY (population, 250,000). The criteria for study
inclusion were that students be (1) enrolled in required
health education classes and (2) fluent in either English or
Spanish. Ethnicity of the sample was diverse: 50% African
American, 16% Hispanic, 20% white non-Hispanic, and 14% other
ethnic backgrounds, including Asians, Native Americans, and
those who indicated that they were biracial. The socioeconomic
status (SES) of the sample was assessed by subject-reported
ZIP code and street address (socioeconomic area, or SEA
[described later]) and the mean SEA rating was 5.2 (SD=2.7),
slightly lower for middle than high school students.
Approximately 70% of the families with children in this school
district have incomes placing them below the federal poverty
Students were recruited within their regular school health
education classes to participate in RAPP, a
quasi-experimental, classroom-based intervention designed to
increase knowledge and skills aimed at safe behavior regarding
sexuality and HIV/AIDS. Classes were assigned within semesters
to 1 of 3 conditions: (1) control, the usual health education
curriculum taught by the regular health education teacher; (2)
RAPP adult health educator, the RAPP intervention implemented
by an ethnically diverse male-female pair of highly trained
adult educators; or (3) RAPP peer educator, volunteer high
school students who completed approximately 50 hours of
preparation by RAPP staff and taught the RAPP curriculum as
pairs of educators. Health education in middle school was
taught in seventh grade only, while in high school students
had the option to take health class in 10th, 11th, or 12th
grade; most students chose 10th or 11th grade.
semester assignments of classes to intervention condition was
based on feasibility issues and availability of peer
educators. The primary goals were that (1) all conditions were
to occur in all classes and schools by the conclusion of the
study; and (2) control and experimental conditions could not
coexist in the same school during a given semester. These
design features enhanced generalizability by ensuring that the
intervention was spread across a variety of different schools,
and helped to avoid contamination between intervention and
control classrooms. The RAPP intervention consisted of 10
(high school) or 12 (middle school) consecutive health class
sessions (usually 2 or 3 sessions per week) delivered for a
period of 2 to 7 weeks. The intervention was integrated into
the regular school health education schedule to avoid
disruption within schools and to build an intervention that
might generalize to other schools in the future. With one
exception during the intervention period of 2.5 years, all
study conditions took place at both middle and high schools.
content of the intervention was based on current literature
concerning school-based interventions, expertise of the RAPP
health educators, and principles from the theory of reasoned
action and normal adolescent development. Early sessions
emphasized self esteem and decision-making strategies, while
later classes progressed through in-depth discussion and
skill-based activities concerning sexuality, STDs, pregnancy,
and finally HIV/AIDS. This last topic received particular
emphasis, and all sessions included small and large group
activities such as games, role playing, and take-home
exercises, often requiring parental input. Priority was placed
on maximum engagement of the students in a highly interactive
and dynamic learning experience in both intervention
conditions. In this article we focus on the preintervention to
immediate postintervention measurement of knowledge, sex
self-efficacy, and behavior intention and compare observed
changes in intervention groups with each other and with the
Students were asked to complete a confidential survey before
intervention and immediately after intervention, as well as 6
and 12 months after intervention, after verbal and written
study explanation. Passive parental consent for student
participation was obtained. Parent(s) of all students
scheduled to take health education in the upcoming school year
are routinely sent a letter from the district Director of
Health and Physical Education informing them that family life
education, including sexuality, will be taught and they can
request their son or daughter not participate in that unit.
During the time of the study, a description of the RAPP
program was a part of this letter and parents were given the
opportunity to inquire further about RAPP and/or refuse
participation. Questions were directed to the study's
principal investigator (D.M.S.), who met with parents
individually to address their concerns. Very few (<10)
families withdrew their children from the program. The study
was reviewed and approved by both the administration of the
local school district and the university institutional
research review board. Students were assured that no names
would be used on any surveys, that their answers would be seen
only by research staff, and that they could participate in the
health classes in which the education and skills project
occurred without completing the research instrument. Few
eligible students refused to participate in the study; more
than 90% completed the survey before intervention. Subjects
were tracked over time by using (1) a school district-assigned
identification number; and (2) a RAPP study identification
number. This procedure ensured that, despite student mobility,
duplicate subject enrollment did not occur. The survey
instrument, available in both English and Spanish, was read to
students during class by the project health educators and
required approximately 40 minutes for completion.
survey questionnaire, pilot tested on 450 students preceding
the main study, was composed of sections measuring constructs
determined to be important in assessing the effects of the
RAPP curriculum. Those reported here include demographics,
knowledge, self-efficacy regarding sexual matters, behavior
intention within the next year, history of risk behaviors, and
history of sexual experiences. In addition to the
student-completed questionnaires, the RAPP health educators
measured the adequacy of the existing health education
learning environment in each class, resulting in a "class
years, gender, ethnicity, and a proxy for SES were measured.
Although the student population of the school district is
generally of low SES, there was concern that some differences
might exist across study subjects and potentially confound our
findings. For confidentiality reasons, and recognizing that
younger teenagers often do not know about household income or
employment and education of family members, we used an SES
proxy as follows. Street name and ZIP code for the student's
residence (as given on the questionnaire) were used to code
census tracts, and this allowed a 1 to 10 SEA ranking for each
student. The 10-point ranking was based on median house value,
rent, and family income, as well as educational level of the
adult population and proportion of professionals and
executives among the employed population within each census
tract. The median house value in the city in 1990 was $60,700,
the average monthly rent was $360, the mean annual family
income was $25,000, and 16% of the adults had a college
degree. While a family's SES might rarely be inconsistent with
that of the census tract in which they resided, we decided
that SEA was more reliable and valid than household-specific
income and educational data provided by the students. The
large study sample also minimized the influence of potential
remaining measurement error.
26-item knowledge scale tested information concerning human
reproduction, decision making, communication with others
concerning sexual matters, HIV/AIDS and other STDs, high-risk
behaviors and their sequelae, and other adolescent sexuality
items. Students responded to statements with yes if they
believed the statement was true, no if they believed the
statement was false, and "not sure" (a choice scored as
incorrect and included to minimize guessing and the possible
inflation of correct response scores). To avoid a ceiling
effect, individual items were included only if they were shown
to have less than 80% correct responses by middle and high
school students during the pilot phase. The scale score range
was from 0 to 26, and alpha reliability was .79.
items, each with a 7-point response scale, measured sex
self-efficacy. This was adapted from similar work developed by
Misovich et al and tested how hard (score of 1) or easy
(score of 7) it would be to carry out each of 8 behaviors in
relation to sexuality (eg, How hard or easy would it be for
you to "convince your partner that a condom must be used
before you have intercourse," "remain abstinent and avoid
having sex," and others). Efficacy was scored as the sum of
the 8 items and ranged from 8 to 56, alpha reliability was
.74, and test-retest reliability, based on 450 control
subjects during a 4-week period, was 0.66. Principal component
factor analysis supported a 1-factor solution (eigenvalue=2.9),
accounting for 36% of variance.
Behavior Intention Scale...
index of intention to behave in safe ways (Figure) was
developed using 9 items asking students to indicate their
agreement or disagreement (on 7-point response scales) with
statements such as "I will be abstinent (not do it) this year"
or "If someone wanted to have sexual intercourse (do it) with
me, I would probably do it." Items measured intention to
engage in the following risk behaviors: sexual behavior
(intention to be abstinent or have intercourse during the next
year, intention to have multiple partners), becoming a teenage
parent, disease risks (HIV/AIDS, STDs), and substance abuse.
Items were scored with anchors of risk (1) or safe intention
(7) and summed. The possible score range was 9 to 63; alpha
reliability was .74 (N=2385) and .74 (N=1526) for middle and
high school students, respectively. Test-retest correlations
across 2 to 4 weeks were 0.77 (N=381) and 0.81 (N=380) for
middle and high school students, and a principal component
factor analysis suggested a 1-factor solution (eigenvalue=3.2),
accounting for 35% of variance.
measure the risk history of each subject, 15 items from the
Youth Risk Behavior Survey were used, including questions
about school- and community-related behaviors (eg, skipping
school, getting into fights, carrying weapons, crime
conviction), substance abuse, and cigarette smoking. We asked
a panel of 25 experts in adolescent health (both clinicians
and behavioral scientists) to rank the behavior items from low
to high risk as follows: 0 for no or minimal risk (eg, missed
school without permission), 1 for some risk (eg, tried
marijuana), or 2 for substantial risk (eg, used marijuana
regularly). Students responded as to whether they had ever
participated in these behaviors. There was a possible score
range of 0 to 31, alpha reliability was .79, and test-retest
reliability during a 4-week period was 0.84. Factor analysis
again suggested a 1-factor solution (eigenvalue=4.3),
accounting for 25% of variance.
of Sexual Intercourse
intervention, students were asked about their history of
sexual intercourse as part of 7 different questionnaire items
addressing onset, frequency, and multiple partner experience.
We examined the degree to which students were consistent
across all 7 items in which there was an opportunity to answer
"I have never had sex," to be confident regarding the validity
of response, and, particularly for the younger students, to
assure that subjects understood the concept of sexual
intercourse prior to initiating the intervention. Students
were categorized as ever having had sexual intercourse (score
of 1) or never having had sexual intercourse (score of 0).
dose of intervention (number of classes attended) may
represent an important contribution to change in HIV
prevention studies. Thus, we asked students to indicate
the extent to which they attended RAPP classes from 1 (not at
all) to 5 (all classes).
for any differences across various learning settings that
might have influenced the effect of the intervention, the
learning adequacy of the existing health education class
environment was observed and scored by the adult RAPP
educators for all participating teachers and classrooms.
Working independently, each member of a pair of educators in a
classroom rated the physical environment and the regular
health teacher's facilitation of the RAPP curriculum. The 18
items were summed to form an overall "class climate" score
(scale score range, 0-36). Rater agreement was high (r>0.80)
and the 2 scores were averaged.
Comparison of Study Variables...
Recognizing that age and gender would likely significantly
affect baseline findings as well as intervention effect, we
stratifed all data into 4 groups: (1) middle school females,
(2) middle school males, (3) high school females, and (4) high
school males. Intervention effect was then tested within these
groups. Before intervention, all study variables were compared
within school level for the 3 intervention groups using the 2
statistic for categorical data and analyses of variance
(ANOVA) for continuous level variables To examine differences
between pretest and posttest scores, repeated-measure ANOVAs
were used with demographics (age, SEA), the existing life risk
score, the class climate score, and the relevant pretest score
for the scale in question (knowledge, self-efficacy, or
behavior intention) introduced first as covariates.
the factors of ethnicity and sex history were entered,
followed by the intervention level factor (1=control, 2=health
educator, 3=peer educator). Because the sample was large and
statistical significance may be easily reached with large
samples, a more rigorous significance threshold of P<.01
(rather than .05) was chosen.
for the dose effect, Pearson product moment correlations were
computed between the student's self-report of attendance and
the 3 outcome variables of interest. These analyses were
compared only for the students in the 2 RAPP intervention
groups (health educator and peer educator classes), because
controls were prevented from any RAPP class attendance. This
characteristic of control subjects (ie, by definition their
dose was 0) precluded entering dose in the ANOVA analyses.
PREINTERVENTION COMPARISONS BY SCHOOL LEVEL AND GENDER
total sample consisted of 1028 female and 971 male middle
school students and 877 female and 820 male high school
students. Within school level, comparable proportions of
students were assigned to each of the 3 intervention groups.
As compared with middle school, the high school students were
approximately 4 years older (F3,3631=7901.5, P<.001), and of
slightly higher SEA status (F=10.4, P<.001). There were ethnic
differences (29=44.4, P<.001), with somewhat greater
percentages of Hispanic and "other" ethnic backgrounds
represented among the younger students. The life risk history
mean scores by groups (in ascending order) were 5.7 (middle
school females), 6.8 (high school females), 7.2 (middle school
males), and 8.3 (high school males) (F=39.4, P<.001). There
were no significant differences across the 3 intervention
groups for middle school students. However, for high school
students, the peer educator group was slightly younger
(F=72.5, P<.000), of higher SEA (F=12.0, P<.000), included
fewer Hispanic students and more non-Hispanic white students
(2=35.4, P<.000), and were less likely to have reported a
history of sexual intercourse (2=21.1, P<.000) (Table 1).
Further, peer-taught high school students reported lower life
risk scores (F=5.6, P<.000) and greater safety intention
(F=13.3, P<.000) than controls or adult- taught students
(Table 2). In addition, there were several significant
gender-specific differences. While only 26.9% of the younger
females indicated that they had experienced intercourse, the
majority of the younger males (64.7%) indicated that they were
sexually experienced. For the older students, 67% of female
and 79% of male high school students reported that they had
been sexually active. In relation to the class climate score,
there were significant differences by school level (F=278.9,
P<.001), with the class environments of the older students
rated as being higher (that is more conducive to learning)
than those at middle school.
variables of interest for examination of intervention effects
(knowledge, self-efficacy, and behavior intention) were also
compared before intervention by school level and gender. As
would be expected, knowledge was greater at the high school
level (F=208.9, P<.001), while there were no gender
differences at either school level. For self-efficacy, there
were both school level and gender differences (F=94.8,
P<.001); self-efficacy was greater for females than for males
at both school levels, and mean scores were higher at high
school in comparison with middle school. Safe behavior
intention was greater for females than males overall, but
scores were lower for high school students in comparison with
middle school students (F=289.1, P<.001).
COMPARISON OF QUESTIONNAIRE SCORES FROM BEFORE INTERVENTION TO
(knowledge), Table 4 (self-efficacy), and Table 5 (behavior
intention) present preintervention to postintervention changes
in questionnaire responses, including the effect of the
interventions compared with each other and with controls using
ANOVA. Beginning with knowledge as the dependent variable
(Table 3), all covariates were significant except life risk,
and significant main effects were found for ethnicity and,
most important, for the intervention. There was no significant
difference for knowledge change based on sex history among any
of the 4 age and gender groups. In each of the 4 age and
gender groups, the pretest score for knowledge outstripped all
other covariates at striking F magnitudes (from 224-399). Age
was significant, even after controlling for differences
between middle and high school students, indicating that older
students did less well on knowledge. In relation to ethnicity,
white non-Hispanics had slightly higher mean knowledge scores
and Hispanics had somewhat lower mean scores than either the
African American or "other" groups.
intervention effect, there were significant differences
between the control and the 2 intervention groups among all 4
of the age and gender groups. Means for the intervention
students (both health educator and peer educator) were
significantly higher after intervention, while the control
group maintained their preintervention mean scores for the
middle school students and rose only about 1 to 1.5 points in
mean score at the high school level. There were notable (high
school females only) 2-way interactions for ethnic group x sex
history (F=3.7, P<.01) and sex history x intervention (F=4.3,
P<.01). Thus, there was a substantial effect of the
intervention beyond the covariates and independent of the
other factors. For the 4 age and gender groups the model
explained substantial variance, ranging from 41% to 55% (R2).
self-efficacy regarding sexual matters, there was statistical
significance for both the covariates and main effects across
the 4 groups of students (Table 4). Similar to the knowledge
scores, the covariates of age, SEA, class climate score, and
the self-efficacy pretest score were significant. In each of
the comparisons, the F for the pretest score (ranging from
182-554) was of much greater magnitude than for the other
covariates. While there were no mean differences in
self-efficacy by sex history, gender proved to be important,
with females reporting higher posttest self-efficacy scores at
both age levels. There were also significant differences by
ethnicity for middle and high school females (but not males).
Hispanic students tended to have mean scores that were
somewhat lower for middle school students (36-36.8) in
comparison with white non-Hispanic middle school students
(37.6-42.6), and for high school females. Hispanic and "other"
students had lower scores in comparison with African American
and white non-Hispanic students. There were no mean
differences for the ethnic groups among high school males.
Statistically significant differences were not found between
intervention and control but trends suggested an intervention
effect; that is, the means for the control subjects were lower
than for the health educator or peer educator intervention
groups. The 4 models predicted from 24% to 46% of variance in
self-efficacy, with most of the variance attributed to the
Finally, safe behavior intention was tested for the same set
of covariates, as well as the ethnicity, sex history, and
intervention factors (Table 5). Again, the covariates and main
effects were significant, but there was a different pattern to
the relationship with behavior intention than for knowledge or
self-efficacy. While the pretest score for intention was the
covariate with the greatest significance (F range, 277-447
across the 4 age and gender groups), the general life risk (F
range, 6.3-54) emerged as being inversely related to safe
behavior intention. In this analysis, there were no ethnic
differences in safe behavior intention but sex history status
was statistically significantly different in 3 of the 4 groups
(F range, 10.6-25.1). Thus, students who indicated that they
had already experienced sexual intercourse also reported less
intention to behave in safe ways. While not statistically
significant for high school males, the mean scores suggested
the same relationship (51.2 vs 41.8). Overall, middle school
students were more likely to intend to engage in safe
behaviors than were high school students. Intervention
students demonstrated greater safe behavior intention at
posttest than controls for high school males (F=4.5, P<.01)
and high school females (F=4.0, P<.05). The models explained
variance in behavior intention ranging from 0.45 to 0.55 (R2).
OF ATTENDANCE AT RAPP SESSIONS (DOSE OF INTERVENTION)
regarding the correlations between the student's self-report
of RAPP participation and knowledge, self-efficacy, and safe
behavior intention scores are presented in Table 6. The
magnitude of knowledge score increases from pretest to
posttest correlated positively with reports of RAPP
participation; that is, as self-report of attendance
increased, total knowledge scores increased with correlations
ranging from modest (0.14) to strong (0.50), and were most
significant at high school level. For sex self-efficacy, there
was less of a relationship with attendance report
(r=0.00-0.28) with only 1 of the correlations (health
educator, high school females) reaching significance. Overall,
correlations for females (range, 0.08-0.28) were greater than
for males (range, 0.00-0.06). There was no correlation between
safe behavior intention and participation reports with the
exception of a modest correlation for high school males
early examination of the effects of RAPP reveals first that
the population was comparable to other urban settings,
particularly with regard to the high risk attributable to male
gender[38,39,40] and age. Against this generalizable
sociodemographic backdrop we found that a large-scale,
school-based, explicit sexual risk reduction intervention can
be implemented and have a successful effect on important
outcomes. Limitations of this research must, however, be
considered when interpreting the results. To begin, all
longitudinal school-based studies are biased by inherent
subject attrition resulting from both graduation and school
dropout. The higher SEA score found among the high school
subjects is consistent with previous reports that urban
students who stay in school are more likely to be members of
families with greater income.[41,42] While the SEA ranking we
used may not precisely measure each subject's true SES, we
believe it is more valid than other self-reported SES data
among adolescents, which usually rely on youth to report
family income and parental education or occupation (as
discussed earlier in the "Participants and Methods" section).
finding that the high school classes were more conducive to
learning than were the middle school classes is probably
rooted in certain classroom characteristics related to the age
groups. High school classroom enrollments tended to be smaller
than in middle school and there may again be some contribution
of a dropout-induced bias toward more motivated students at
the higher grade levels. Older students were, perhaps, more
able to pay attention and participate in sexuality-focused
sessions than were younger students. The learning environment
clearly warrants measurement in school-based research and must
be factored into interpretation of intervention effectiveness.
higher levels of self-efficacy we found among females is
consistent with the recognition that many of our cultural and
educational messages around sexual safety are often directed
toward girls and young women as opposed to boys and young
men. Intention to behave in safer ways concerning sex was
also a female attribute in this study, a theoretically
consistent extension of the self-efficacy findings. The
inability of the older students to translate their greater
knowledge and self-efficacy into safer behavioral intention
points out the urgent need to focus prevention interventions
on the younger population. It may, however, also suggest that
for adolescents the link between self-efficacy and behavior
intention is not as tight as theory might otherwise propose.
examined differences between intervention and control groups,
the ANOVA models included important covariates that might
explain findings that would have been incorrectly attributed
solely to intervention effect in a less sophisticated
analysis. Knowledge gains observed in RAPP (which were greater
than those reported in other school-based programs) were
likely due to interactive teaching techniques, the use of
gender and ethnically diverse educator pairs, the careful
inclusion of this program within the regular school
environment, and the length of the intervention (10-12
sessions). It is notable that the peer educator condition
produced results comparable to the health educator condition
(Table 3). The RAPP study confirms that, at least in certain
content areas and over short follow-up, extensively prepared
high school students can be effective teachers for their
modest effect of RAPP on self-efficacy may reflect the
possibility that assessment immediately following the
intervention is too early to detect a difference in this
construct. If a knowledge, self-efficacy, and behavior
intention link does exist (as proposed by the theory of
reasoned action), knowledge change will temporally precede
observable efficacy change. Intervention effect on safe
behavior intention was positive among the high school
subjects, especially the females, but not for the middle
school students. In the case of middle school females, this
lack of intervention effect could be an artifact of
measurement. That is, these students scored quite high at
baseline in all 3 study conditions (mean score, 55; maximum,
63) and this "ceiling effect" limited the ability of our
analyses to detect a difference. These results might evidence
a pressure felt by 13-year-old girls to provide (at pretest)
what they perceive to be socially acceptable responses to
questions about safe sex behavior intention. The high school
students, on the other hand, did show greater increases in
safe behavior intention after the test in the intervention
groups than in control groups. Perhaps their developmental
attainment was better suited to the effect of the
intervention. Our future analyses will document the
longer-term status of these variables as well as the most
important outcome, that of behavior and its relationship to
behavior intention. Our findings regarding intervention dose
and its positive correlation with outcome measures (especially
knowledge) not only reinforces the conclusion that it was RAPP
curriculum exposure that affected posttest scores, but also
points out the importance of factoring attendance into
analyses of school-based interventions.
should not be forgotten that for the 3 constructs and for all
age and gender groups our models explained significant
variance, with R2 ranging from 0.41 to 0.58 for knowledge and
behavior intention and somewhat less for self-efficacy
(0.24-0.46) (Tables 3 through 5). As stated earlier, it is the
burden of the past (pretest scores) that casts a long shadow
over predictions of intervention-induced change in knowledge,
self-efficacy, and behavior intention. This finding not only
mandates the testing of interventions among subjects younger
than middle school age, but also illustrates the need for
researchers and clinicians to be methodologically sensitive to
removing the variance attributable to pretest scores when
interpreting intervention study data. Finally, despite
substantial predictive power of our model, the influences on
pretest scores go beyond age and personal experience to
include parental, family, cultural, and community forces. More
comprehensive and multidimensional interventions that
reinforce school-based activities with other sites and
contexts for prevention strategies must be considered.
the Department of Pediatrics (Drs Siegel and Roghmann and Ms
Enaharo) and the School of Nursing (Dr Aten), University of
Rochester, Rochester General Hospital, Rochester, NY.
Accepted for publication May 14, 1998.
research was supported by grant R01-MH 49037 from the National
Institutes of Mental Health, Rockville, Md.
thank Barbara Thompson for her tireless preparation of the
manuscript. We also thank the staff of the Rochester AIDS
Prevention Project for Youth; the health educators, Margaret
Cain, BA; Raul Corujo-Molina; Desiree Voorhies, RN, MSEd; and
Lennard Wedderburn, CSW; and research assistant Terri Vaughn,
CSW, for their dedication, commitment, and hard work on behalf
of the project. Special thanks to the staff and students of
the participating schools.
Corresponding author: David M. Siegel, MD, MPH, Department of
Pediatrics, Rochester General Hospital, 1425 Portland Ave,
Rochester, NY 14621 (e-mail: email@example.com).
Guttmacher Institute. Sex and America's Teenagers. New York,
NY: Alan Guttmacher Institute; 1994.
Harvey SM, Spigner C. Factors associated with sexual behavior
among adolescents: a multivariate analysis. Adolescence.
Rotheram-Borus MJ, Koopman C, Haignere C. Reducing HIV sexual
risk behaviors among runaway adolescents. JAMA.
Epner JEG, ed. Policy Compendium on Reproductive Health Issues
Affecting Adolescents. Chicago, Ill: American Medical
Forrest JD, Singh S. The sexual reproductive behavior of
American women, 1982-1988. Fam Plann Perspect.
Bayne Smith MA. Teen-incentives program: evaluation of a
health promotion model for adolescent pregnancy prevention. J
Health Educ. 1994;25:24-29.
T, Hein K. The adolescent and sexually transmitted diseases.
In: Holmes K, ed. Sexually Transmitted Diseases. New York, NY:
McGraw-Hill International Book Co; 1984:73-84.
Cates W. The epidemiology and control of sexually transmitted
diseases in adolescents. In: Schydlower M, Shafer M, eds.
Adolescent Medicine: State of the Art Reviews. Philadelphia,
Pa; Hanley & Belfus Inc; 1990:409-428.
Centers for Disease Control and Prevention, National Center
for HIV, STD and TB Prevention. Sexually Transmitted Disease
Surveillance, 1997. Atlanta, Ga: Centers for Disease Control
and Prevention; 1997.
Schacter J. Why we need a program for the control of Chlamydia
trachomatis. N Engl J Med. 1989;320:802-804.
Moscicki A, Paletsky J, Gonzales J, Schoolnik GK. Human
papilloma virus infection in sexually active adolescent
females: prevalence and risk factors. Pediatr Res.
Centers for Disease Control. Annual Report. Atlanta, Ga:
Centers for Disease Control; 1991.
Sonnenstein FL, Pleck JH, Ku LC. Sexual activity, condom use
and AIDS awareness among adolescent males. Fam Plann Perspect.
Romer D, Black M. Ricardo I, et al. Social influences on the
sexual behavior of youth at risk for HIV exposure. Am J Public
Levy SR, Handler AS, Weeks K, et al. Correlates of HIV risk
among young adolescents in a large metropolitan midwestern
epicenter. J Sch Health. 1995;65:28-32.
Centers for Disease Control and Prevention. HIV/AIDS
Surveillance Report. Vol 9. Atlanta, Ga: Centers for Disease
Control and Prevention; 1997.
Hein K. "Getting real" about HIV in adolescents. Am J Public
Boyer CB, Kegeles SM. AIDS risk and prevention among
adolescents. Soc Sci Med. 1991;33:11-23.
Kirby D. No Easy Answers: Research Findings on Programs to
Reduce Teen Pregnancy. Washington, DC: The National Campaign
to Prevent Teen Pregnancy; 1997.
Klein NA, Goodson P, Serrins DS, Edmundson E, Evans A.
Evaulation of sex education curricula: measuring up to the
SIECUS guidelines. J Sch Health. 1994;64:328-333.
Kirby D, Short L, Collins J, et al. School-based programs to
reduce sexual risk behaviors: a review of effectiveness.
Public Health Rep. 1994;109:339-360.
Kirby D. School-based programs to reduce sexual risk-taking
behaviors. J Sch Health. 1992;62:280-287.
Sunwoo J. Brennan A, Escobedo J, et al. School-based AIDS
education for adolescents. J Adolesc Health. 1995;16:309-315.
Kirby D, Korpi M. Adivi C, Weissman J. An impact evaluation of
Project SNAPP: an AIDS and pregnancy prevention middle school
program. AIDS Educ Prev. 1997;9(suppl A):44-61.
Walter H, Vaughan R. AIDS risk reduction among a multiethnic
sample of urban high school students. JAMA. 1993;270:725-730.
Bandura A. Social Foundations of Thought and Action: A Social
Cognitive Theory. Englewood, NJ: Prentice-Hall; 1986.
Ajzen I, Fishbein M. Understanding Attitudes and Predicting
Social Behavior. Englewood, NJ: Prentice-Hall International
Terry DJ, O'Leory JE. The theory of planned behavior: the
effects of perceived behavioural control and self-efficacy. Br
J Soc Psychol. 1995;34:199-220.
Kirby D, Barth R, Leland N, Fetro JV. Reducing the risk:
impact of a new curriculum on sexual risk taking. Fam Plann
Whitley BE, Schofield JW. A meta-analysis of research on
adolescent contraceptive use. Popul Environ. 1986;8:173-203.
Main DS, Iverson DC, McGloin J, et al. Preventing HIV
infection among adolescents: evaluation of a school-based
education program. Prev Med. 1994;23:409-417.
Brown LK, Barone VJ, Fritz GK, et al. AIDS education: the
Rhode Island experience. Health Educ Q. 1991;18:195-206.
N, Stanton B, Li X, et al. Effectiveness of the 40 adolescent
AIDS-risk reduction interventions: a quantitative review. J
Adolesc Health. 1997;20:204-215.
Weeks K, Levy SR, Zhu C, et al. Impact of a school-based AIDS
prevention program on young adolescents' self-efficacy skills.
Health Educ Res. 1995;10:329-344.
Newman C, DuRant RH, Ashworth CS, Gaillard G. An evaluation of
a school-based AIDS/HIV education program for young
adolescents. AIDS Educ Prev. 1993;5:327-339.
Miller BC, Paikoff RL. Comparing adolescent pregnancy
programs: methods and results. In: Miller BC, Card JJ, Paikoff
RL, Peterson JL, eds. Preventing Adolescent Pregnancy. Newbury
Park, NJ: Sage Publications; 1992:265-284.
Misovich SJ, Fisher WA, Fisher JD. Understanding and promoting
AIDS preventive behaviors: measures of AIDS risk reduction
information, motivation, behavioral skills, and behavior. In:
Davis CM, Yarbor WH, Bauserman R, Scheer G, Davis SL, eds.
Sexuality Related Measures: A Compendium. Newbury Park, NJ:
Sage Publications; 1998.
Warren CW, Kann L, Small ML, Santelli JS, Collins JL, Kolbe LJ.
Age of initiating selected health-risk behaviors among high
school students in the United States. J Adolesc Health.
Siegel DM, Aten MJ, Roghmann KJ. Self-reported honesty among
middle and high school students responding to a sexual
behavior questionnaire. J Adolesc Health. In press.
Aten MJ, Siegel DM, Roghmann KJ. Use of health services by
urban youth: a school-based survey to assess differences by
grade level, gender, and risk behavior. J Adolesc Health.
National Research Council. Losing Generations: Adolescents in
High Risk Settings. Washington, DC: National Academy Press;
Department of Education, National Center for Education
Statistics. Dropout Rates in the United States., Washington,
DC: US Dept of Education; 1997. NCES publication 97-473.
Hayes CD, ed. Risking the Future: Adolescent Sexuality,
Pregnancy, and Childbearing. Washington, DC: National Academy