Vous êtes sur la page 1sur 12

KenIN CoyLE, PHD r KensN BasrN-ENGeursr, PUD r Doucras Ktnnv, PUD

Guv PaRCEL, PHD r STEpHEN BaNspAcH, PHD r JANET CotLtNs, PHD

Ertza.BETH BAUnnLEn, PsD r Scorr CaRVAJAL, PHD r RoNnLo HAnntsr, PsD

Safer Choices: Reducing Teen


Pregnancy,HIV and STDs

Drs. Coyle, Kirby, and Carvalal are with


ETR Associates, Scotts Valley, California. SYNOPSlS
Dr. Coyle is Director of Research, Dr.
K i r b y i s S e n i o r R e s e a r c hS c i e n t i s t ,a n d Objectives.This study evaluatedthe long-term effectivenessof Sofer
Dr. Carvajal is Senior Research Associate.
d ,u l t i - c o m p o n eendt u c a t i o n aplr o g r a md e s i g n etdo
C h o l c e sa,t h e o r y - b a s e m
Dr. Basen-Engquistis an Assistant
Professor, University of Texas MD reducesexualriskbehaviorsand increaseprotectivebehaviorsin preventing
Anderson Cancer Center. At the time of , n d p r e g n a n caym o n gh i g hs c h o o sl t u d e n t s .
H l V ,o t h e rS T D s a
this study, she was an Assistant Professor,
Center for Health Promotion Research . h e s t u d yu s e da r a n d o m i z e cd o n t r o l l e dt r i a l i n v o l v i n g2 0 h i g h
MethodsT
and Development, University of Texas, schoolsin Californiaand Texas.A cohort of 3869 ninth-grade
studentswas
Houston. Dr. Parcel is Director and Dr.
r 9 9 3( b a s e l i n et o) s p r i n gs e m e s -
t r a c k e df o r 3 | m o n t h sf r o m f a l l s e m e s t e 1
Harrist is an Associate Professor, UT
Houston Center for Health Promotion . a t aw e r e c o l l e c t e du s i n gs e l f - r e p o rst u r -
t e r 1 9 9 6( 3 l - m o n t hf o l l o w - u p )D
Research and Development. Dr. Banspach v e y sa d m i n i s t e r ebdy t r a i n e dd a t a c o l l e c t o r sR. e s p o n s rea t e a t 3 l - m o n t h
and Dr. Collins are with the Centers for
follow-upwas79%.
Disease Control and Prevention. Atlanta,
Georgia. Dr. Banspachis Chief, Evaluation involvingcondom
had its greatesteffecton measures
Results.SoferChorces
R e s e a r c hS e c t i o n ,S u r v e i l l a n c ea n d
Evaluation Research Branch, Division of
use.The programreducedthe frequencyof intercoursewithout a condom
Adolescentand School Health. Dr. d u r i n gt h e t h r e e m o n t h sp r i o rt o t h e s u r v e yr,e d u c e dt h e n u m b e ro f s e x u a l
Collins is Deputy Director, National partnerswith whom studentshad intercoursewithout a condom, and
C e n t e r f o r C h r o n i c D i s e a s eP r e v e n t i o n
increaseduse of condomsand other protectionagainstpregnancyat last
and Health Promotion. At the time of this
study, she was Chief, Surveillance and intercourse.Sofer Choicesalso improved7 of l3 psychosocial
variables,
Evaluation Branch, Division of Adolescent effectupon rates
manyrelatedto condomuse,but did not havea significant
and School Health. Dr. Baumleris a
o f s e x u ailn i t i a t i o n .
strategic modeler with Eucid Inc.,
C h i c a g o ,l l l i n o i s .A t t h e t i m e o f t h i s s t u d y ,
Conclusions. programwas effectivein reducingimportant
The SaferCholces
she was an Assistant Professor, UT
Houston Center for Health Promotion
r i s k b e h a v i o rfso r H l V , o t h e r S T D s ,a n d p r e g n a n cayn d i n e n h a n c i nm
g ost
R e s e a r c ha n d D e v e l o o m e n t . p s y c h o s o c idael t e r m i n a n ot sf s u c hb e h a v i o r .

PUBLIC HEALTH REPORTS . 2OOI SUPPLEMENT 1 . VOLUME I I6


chool-agedadolescentsare an important target cents in the comparisongroup, after adjustmentfor indi-
population for health promotion programs, vidual-level baseline values. Further, we hypothesized
particularly programs addressingsexual behav- that studentsin the interventiongroup who reported sex-
iors. The number of AIDS casesamongadoles- ual intercourse would report fewer sexual partners,
cents and young adults ages 13 to 24 remains greatercondom use at first intercourse(amongthosewho
relatively small (26,518 reported cases through June, initiated intercourseduring the study),greatercontracep-
2000);r however, it has been estimated that approxi- tive use at last intercourse,lessalcohol and drug use prior
mately one-half of all new HIV infections occur among to sexualintercourse,and more frequent testing for HIV
young people between the ages of 13 and 24.2 In addi- and other STDs. Finally, we hlpothesized that interven-
tion, roughly one in four adolescentsages 13-19 who tion students would score more favorably on the psy-
have had sexualintercourseacquire an STD each year.3 chosocialscalesthan would comparisonstudents.
Finally, approximately10% of 15- to l9-year-old females The evaluation results after the first year of program
become pregnant each year;aan estimated B5Voof these implementation showed a positive effect on condom use
pregnanciesare unintended.5 and most psychosocialmediating variables among stu-
Numerous HIV, other STD, and pregnancypreven- dents in the cohort.li In this paper, we discuss the 31-
tion programs have been implemented in secondary month follow-up results, conducted approximatelyone
schoolsacrossthe nation, but relatively few have had a year following the intervention.
significant effect on sexualrisk behaviors,6or the effects
have been modest or short term.7-r0Given the significant MsrHoDS
consequencesof unprotected sexualintercoursefor ado-
lescentsand the limited results of past HIV, other STD, Procedures and assignment. The Safer C'hoicesinter-
and pregnancy prevention studies, we developed Safer vention was implemented during the 1993-1994 and
Ch.oices, an innovativemulti-componentprogram. 1994-1995 school years. The evaluation used a random-
SaferChoicesis a2-year,school-basedHIV, other STD, ized controlled trial involving 10 public schoolsin an urban
and pregnancy prevention program for high school stu- area in northern California and l0 public schools in an
dents. It has recently been identified as a "program that urban area in southeastTexas.The schools ranged in size
works" by the Centers for DiseaseControl and Prevention from 961 to2733 students(mean= 1767\.Five schoolsin
(CDC). The primary aim of SaferChoicesis to reduce the each state were randomly assigned to the Safer Ckoices
number of students engagingin unprotected sexualinter- program;the remaining schoolswere assignedto a compar-
course by reducing the number of students who begin or ison program (a standard, knowledge-basedHIV preven-
have sexualintercourse during their high school years and tion curriculum). We used a restricted randomization
by increasingcondom use among those students who have process in which schoolswere ranked based on an index
sex. In addition, the program seeksto modifr severalfac- score of possible confounding variables (such as percent
tors related to sexualrisk-taking behavior:knowledgeabout collegebound and number of studentsin school);adjacent
HIV and other STDs; attitudes about sexualbehavior and schools in the ranking were then paired and randomly
condom use; normative beliefs regardingsexualintercourse assignedto either the intervention or control condition.r2
and condom use; students'beliefs in their ability (self- The school districts were chosen because they served
efficacy) to refuse sexualintercourseor unprotected sexual diverse populations in terms of ethnicity and socioeco-
intercourse,use a condom, and communicateabout safer nomic status, were in areasof the country with high HIV
sexualpractices;perceivedbarriers to condom use; per- prevalencerates, and were in close proximity to the inves-
ceived risk of becoming infected with HIV or other STDs; tigative team. We chose districts in two different areasof
and communication with parents. the country to improve the generalizabilityof study results.
This randomizedcontrolled trial tested the effects of We collected self-reportdata from a cohort of ninth-
Safer Ckoices. We hypothesizedthat adolescentswho gradestudentsat all 20 schoolsusing trained data collec-
received Safer Choice.swould initiate sexual intercourse tors. The baselinedata were collectedin fall 1993,imme-
at a slower rate than adolescentsin the comparison diately before the intervention. Follow-up data were
group. We also hypothesized that adolescentsin the collected in spring 1994 (7 months after baseline and
intervention group who had ever had sexualintercourse immediately following the first year of the intervention),
would have unprotectedintercoursefewer times and with spring 1995 (19 months after baseline and immediately
fewer partners in a 3-month period than would adoles- following the secondyear of the intervention),and spring

PUBLIC HEALTH REPORTS.2OOl SUPPLEMENT I. VoLUME II6


ET AL.

1996 (31 months after baselineand l2 months followinq brief overview of each component. A more detailed
the secondyear of the intervention). descriptionof the interventionis providedelsewhere.le
Active parental consent was required for surveypar- During eachyear of the program,interventionschools
ticipation; B}Vo ol students returned parental consent implemented activities acrossall five components. Stu-
forms (5184 of 6488): a total of 4733 studentshad con- dents receivedtheir most intensive exposureto the pro-
sent to take the survey.Baselinesurveyswere completed gram from the 2O-lessoncurriculum and school-wide,
by 9\Vo of these students.Becauseschoolswere the unit peer-sponsored events.
of randomization,we used multi-level statisticalanalyses
to accountfor the clusteringof studentswithin schools.A Measures. The surveyconsistedof items assessing demo-
detailed discussion of the evaluation methods is pub- graphic characteristics,sexuality-relatedpsychosocialfac-
lished elsewhere.r2We also collectedcross-sectionaldata tors, sexual behaviors, and program exposure. The psy-
at three time points to assessthe school-wideeffects of chosocial scales were: HIV knowledge; other STD
the intervention. The cross-sectionalresults are Dre- knowledge; attitudes about sexual intercourse; attitudes
sentedin a separatepaper.13 about condoms;normativebeliefsabout sexualintercourse;
normative beliefs about condoms; self-efficacyin refusing
Participants. The cohort consistedof 3869 ninth-grade sex;self-efficacyin using condoms;self-efficacyin commu-
studentswho completedthe baselinesurveyin the fall of nicating with partners; barriers to condom use; HIV risk
1993 and who were officially enrolled in the secondyear perceptions;other STD risk perceptions;and communica-
of the intervention (fall 1994). We excludedstudents (at tion with parents.The scalesand their psychometricprop-
both intervention and comparisonschools)who were in ertiesare discussedin more detailelsewhere.IrJ}2o
llth or l2th grade;we also excludedstudentswho left
schoolduring the 1993-1994schoolyearand did not re- Prirnary outcolnes. The survey measured three primary
enroll in the fall for the 1994*1995 school year. These behavioraloutcomes:(a) whether studentsdelayedinitia-
criteria were adopted becausethe intervention program tion of sexualintercourse:(b) the number of times stu-
was multi-year. Baselinedemographiccharacteristicsfor dents had intercoursewithout a condom in the last three
the cohort are summarizedin Table 1. months (amongthose reporting intercourse);and (c) the
A total of 3058 (79Vo)of the 3869 studentsin our number of sexualpartnerswith whom studentshad inter-
final cohort were surveyedat the 3l-month follow-up. course without a condom in the last three months
Responseratesfor the 7-month and I9-month follow-ups (amongthose reportinglntercourse).
w ere 9 5% and B3Vo,respectively.
Secondaryoutcow.es. The survey also assessednumerous
Intervention.The SaferChoicesinterventionis basedon secondarybehavioraloutcomes:use of a condom at first
social cognitive theoryla social influence theory15-17
and intercourse among students who initiated sexual inter-
models of school change.lBThe program consistsof five course following baseline;use of protection at last inter-
primary components: school organization (a School course;number of times had sexualintercoursein the last
Health Promotion Council involving teachers,students, three months; number of sexualpartnersin the last three
parents,administrators,and community representatives); months; use of alcohol or drugs before sexualintercourse
curricuium and staff development (a sequential 20- in the last three months; and being testedfor HIV and for
session classroom curriculum for 9th- and l0th-grade other STDs. The surveyitem assessinguse of protection
students that includes 10 sessionsat each grade level at last intercourse was analyzedas two dichotomous vari-
taught by trained teachers);peer resourcesand school ables: use of a method that effectively protects against
environment(a SaferC'hoicespeer team or club that hosts HIV and other STDs (condom) and use of a method that
school-wide activities); parent education (activities for effectively protects against pregnancy (condom alone,
parents including parent newsletters, student-parent condom and birth control pills, birth control pills alone).
homework activities, and other parent events); and
school-community linkages (activities to enhance stu- Statistical analysis. Multilevel models were used to
dents'familiarity with and accessto support servicesout- adjust for the correlation between students within
side school, such as homework to gather information schools, and correlation within students because of
about local services,resource guides, presentationsby repeatedmeasurementsover time. The three-levelmod-
speakerswho are HlV-positive). The Figure includes a els included surveymeasurementoccasionas level 1, stu-

PUBLIC HEALTH REPORTS . 2OO1 SUPPLEMENT' 1 . VOLUME 116


&lii:r{i::;r::r:i::!:i::r:"'.:

'., ', '


.; uu"a91 ",

.. ..:':;.i--::::- . - 7 O 6 - . - ' - ; - . ; . i " ' : : - : r f i n ' ' a : : . : : : - - : 4 " ' ; ' : ' Z ( l / .{ : : ' : : : : : : ' : ' - ' - ' : ' ) ' f' ' l ' , : " " ; ; ; .
"'-''..'f9'.I
':'.,

dents as Ievel 2, and schoolas level 3. We used linear and negativebinomial regressionmodel is a generalizationof
logistic multilevel models to analyze continuous and the Poissonregressionmodel and is often used for model-
dichotomous data, respectively,and Poissonor negative ing over-dispersedcount data in which the variable of
binomial multilevel models to analyzecount data. The interestis the number of occurrencesduring a given time

PUBLIC HEALTH REPORTS . 2OO1 SUPPLEMENT I . V O L U N , T EI I 6


Figure. Features ofthe SoferChoicesprogram

Progromcomponent Feotures

School organization A SchoolHeolth PromotionCouncilinvolvingteochers,students,parents,odministrotors,and


The Council plansand conducts Programactivities.
communityrepresentotives.

Curriculum and staff A sequentiol,20-sessionc,ossroomcurrkulum for ninth-ond tenth-grodestudents(10 lessons


develooment ot eochgrodelevel).The curriculum includesin-classpeer leadersto facilitateselected
activities(e.g.,leadingsmall-grouprole playing)and is implementedby classroom
teachers trained by prolect staff.

Peer resourcesand school A Sofer Choicespeer teom or club ot everyschoo,.The club members meet with an adult
environment peer coordinator to plan and host six types of school-wideactivitiesdesignedto alter
the normative culture of school. Peer teams also run a resource area on camPus.
Additionally,proiect staff developedrole model stories in which teens tell their per-
sonal stories modelingpositivebehaviors;the stories are presentedin a poster format
(and were presentedin a monthly calendarduring the study).

Parenteducation Activitiesfor porents.Parents receive three project newsletters ayear that provide
information about the program;functionalinformation regardingHIV/AIDS,other
STDs, and pregnancy;and tips on talkingwith teenagersabout these issues.The cur-
riculum includesstudent-parenthomework activitiesto facilitatecommunication
regardingHlV, other STDs and pretnancy.Parentsalso serve on the health Promotion
councilsand help plan other parent-relatedevents.

School-communitylinkages Activitiesdesrgnedto enhoncestudents' fomiliority with ond occessto support servicesoutside


school.The curriculum includeshomework assignmentsrequiringstudentsto gather
informationabout local resourcesand services.lt also includesa lessonin the lOth
grade involvingH|V-positivespeakers.Studentsand teachersreceiveresourceguides
that provided a list of HlV, other STDs,and pregnancy-related servicesfor adolescents.

NOTE: The publishedprogram materialsare availablefor purchasethrough ETR Associates(800) 321-4407(customer service).

period. We carried out computations for the multilevel variabledenoting whether the surveydata were collected
models using MLn Softwarefor Multilevel Analysis,Ver- via mail or in school. Variableswere included as covari-
sion1.0a.21 ates for a particular outcome if they were significantly
We examined the effect of the intervention from related to the outcome and intervention condition, and
baselineto the final follow-up measurement,a period of remained significantin the final stageof multilevel mod-
approximately 31 months. The analyses included the eling. The multilevel models provided a flexible frame-
baselineand three follow-up measuresof each outcome work for handling missing data. Students with incom-
to provide a test of the overall intervention effects. The plete data (missing one time point, fbr example) were
follow-up measureswere modeledas dependentvariables included in the analysisand contributed to the estimation
in random effectsmodels.We modeledthe following vari- of the overall intervention effects acrosstime. Students
ables as predictors for each outcome: participants'base- with missing data on the covariateswere excluded from
line responseson the outcome; intervention group; geo- the analyses.
graphic location (Texas or California); an intervention We used two-tailed tests and made no adjustments
group-by-geographic location interaction term; measure- for multiple tests of significance.Our primary and sec-
ment occasion; intervention group-by-measurement ondary hypotheseswere stated a priori and were limited
occasioninteraction terms; and a set of outcome specific in number. All other analysesbeyond those for the pri-
covariates.We also created a variable representingthe mary and secondary hypotheseswere considered
number of weeks between baselineand follow-up, and a exploratory.Statisticalsignificancewas set at P <0.05.

PUBLIC HEALTH REPORTS .2001 SUPPLEMENT 1 . VOLUNIE ll6


Estimates of the magnitude of effects were calculated A m o n g t h e s e c o n d a r yo u t c o m e s , s t u d e n t s w h o
for all outcomes. We used odds ratios to estimate overall reported having sexual intercourse during the three
effects for dichotomousbehavioralvariables.The effects for months prior to the survey in the intervention schools
the Poissonor negativebinomial models representthe ratio were 1.68 times more likely to have used condoms (P =
of the adjusted mean for the intervention goup to the 0.04), and 1.76 times more likely to use an effectivepreg-
adjusted mean for the comparison group. Thus, an effect nancypreventionmethod (birth control pills, birth control
sizeof 1.00indicatesno difference,<1.00 indicatesa lesser pills plus condoms,or condoms alone) (P = 0.05) at last
mean for the intervention group relative to the comparison intercoursethan were studentsin the comparisonschools.
group, and >1.00 indicatesa greatermean for the interven- We found no significant differences at final follow-up
tion group relative to the comparison goup. For the psy- between intervention conditions on any other secondary
chosocial variables, we used group coefficients flom the behavioraloutcomesafter we adjustedfor baselinevalues
multilevel models as an er?ression of effect size. This (P = 016 to 0.51), althoughthe mean differences
approach to estimating the magnitude of program effects between the two groups are in the desired direction for
was used becausefurther computations of effect size (for five of the six remainingoutcomes.
example,conversionto Cohen's d or r family effect sizes)
would not fully use all the elementsof the multilevel ana\tic Psychosocial factors. Thble 3 shows the results of the
model. These group coefficients represent the difference multi-level analysesfor the 13 psychosocialscales.Signifi-
between the two adjustedmeans (interventionvs. compari- cant differences-all favoring the intervention schools-
son group) on the scaleof the outcome variablebeing mea- were found for 7 of the 13 psychosocialscales.Intervention
sured (such as HIV knowledgeor self-efficacyfor refuslng students scored significantly higher than comparison stu-
sex).Becausethe coefficientsare unstandardized,onlythose dents on the HIV and other STD knowledgescales(by an
that are measuredon the samescalecan be compared. adjusted mean difference of I I and 9 percentagepoints);
expressedsignificantly more positive attitudes about con-
RssuLrs doms (P = 0.01); and reportedgreatercondom-useself effi-
cacy(P = 0.00),fewerbarriersto condomuse (P = 0.01),and
Estimates of the averageoverall effects of the interven- higher levelsof perceivedrisk for HIV (P = 0.02) and other
tion are presented fbr the behavioral and psychosocial STDs (P = 0.04). Interventionstudentsalsoreportedgreater
measuresover the 3l-month follow-up period. We also normativebeliefs about condom use and more communica-
discusstrends over the three follow-up assessments. tion with parentsthan did comparisonstudents;thesediffer-
encesnearedsignificance(P = 0.06 for eachvariable).
Behavioral factors. Among the primary outcomes, all There were no significant differences between stu-
three were in the desireddirection and two were statisti- dents in the two programgroupsin their attitudesregard-
cally significant (Table 2). Sexuallyexperiencedstudents ing sexual intercourse (P = 0.95), normative beliefs
in the intervention schools reported having intercourse regardingsexual intercourse (P = 0.79), self-efficacyto
without a condom fewer times during the three months refuse sex (P = 0. l0), or self-efficacyto communicate
preceding the follow-up survey than did sexuallyexperi- with a partner about sexuallimits (P = 0.60).
enced studentsin the comparisonschools(P = 0.05) by a There were two significantgroup-by-locationinterac-
ratio of 0.63. The group-by-locationinteraction was sig- tions-knowledge about other STDs and condom use
nificant for this variable (P = 0.05); it indicated the self-efficacy.The data suggestthe effects on other STD
effects were greater in Texasthan in California. Similarly, knowledge were greater in Texas than in California, and
Safer Choice.sstudents reported having fewer partners the effects on condom-useself-efficacywere greater in
with whom they had sexual intercourse without a con- California than in Texas.
dom during the previous three months than did sexually
experienced students in the comparison schools (P = Trends over time. As discussedin the analysissection,
0.02); the ratio of the adjusted means was 0.73. We we included group-by-time interaction terms in the
found no statistically significant difference in the inci- multilevel models to examine changesin the magnitude
dence of sexualinitiation between students in the Safer of effects over time. These interactionterms demonsuare
Choieesand comparison schools at the final follow-up; that both the behavioraland psychosocialeffects endured
the effect was in the desireddirecrion (oddsratio = 0.83), over the 3l-month follow-up period. Only one of the
but not closeto significant (P = 0.39). interaction terms (number of partners unprotected) was

PUBLIC HEALTH REPORTS . 2OOI SUPPLEMENT I . VOLUME 1I6


Covrr ET AL.

Table 2. Parameter estimates for multilevel models of behavioralvariables:cohort sample, 3l-month follow-up
Rotio of group
ktimated estimote'to
efbct Stondord group stondard
Behovior vorioble r'f" siz€ effor et?or' 95"/.Cl P.

Primoryoutcomes
Sexualinitiation 2029 0.83 0.22 0.s4.t.27 o.3e
l
Frequency of intercourse
without
a condomin last3 months 137| 0.63 0.23 -1.97' 0'05
Nymber 9f :sexual'rPartnerS..with9!1t
,::r
a condom in last 3 months. 137| 0.73 o.l4 -2.37 0;0-1

Secondoryoutcomes
Use of condoms at first intercourse '
(amon8Inrtratorsonly) . . 733 t.44 0.27 0.85,?.44 o 1.'
Use of condoms at last
intercourse s49 r.68 0.25
Use of protection againstpregnancy
at last intercourse. .. 549 1.76 0.29
Number of times had sexual
r37t
'i' ,,,
,.,:,intercou!:se
in,last 3 months, . . . . ., 0.81

*rf
Number of sex partners in last
3 months 13,71 ''u,,',
Use of alcohol and other drugs
before sex in last 3 months r37l 0.04 oe7
Testedfor HlV. . . 36t6 t.20
Tested for other STDs 7627 r.52 :.1
"n representsrhe number of individualsincludedin the analyses.The rcsts of significance are basedon the number of observationsrather chan
the number of individuals.Eachindividualhad from two to four observations.
bOdds ratios were used to estimate overall intervention effects for dichotomous outcomes; ratio of adjusted means was used to estimat€ ov€r-
all effecs for Poissonor negativebinomialmodels.
= 2.57
"Group-by-locarion interaction was significantat P = 0.05; ratio of interaction estimate to inrcraction standard error

statistically significant (P = 0.0a); it indicated that the in the previous three months; number of partners with
interwentionhad a smaller effect at the final follow-up whom studentshad sex without a condom; use o[ a con-
relative to the first follow-up. The other group-by-time dom at last intercourse; and use of protection against
interaction terms also indicated that the interventron pregnancy at last intercourse.Three of these four out-
effects diminished somewhatover time [data not shown], comeswere statisticallysignificant and favoredthe inter-
althoughnot to a significantdegree. vention at the 7-month follow-up; one of the four was sta-
We also analyzedthe data following each time period tistically significant at the 19-month follow-up; and all
to examine the averageeffects up to a given time point. four were statisticallysignificantby the final follow-up.
(For example,at the 19-month foilow-up the effects rep- Among the 13 psychosocialvariables(Table 5), 9 were
resentthe averageof the 7-monthand 19-monthresults). significant and favoredthe intervention condition among
The majority of the findings were relatively consistent cohort studentsat the 7-month and 19-monthfollow-ups.
over time. Among the behavioralfactors (Table 4), four They were HIV knowledge,other STD knowledge,atti-
were consistentlysignificantat one of the three follow-up tudes about condoms,normativebeliefs about condoms,
assessments: frequencv of intercoursewithout a condom self-efficacyin using condoms, barriers to condom use,

PUBLIC HEALTH REPORTS.200l SUPPLEMENT 1 ' VOLUN'{E 116


l

Vanable,

t4 3736 oOr 0,26 ,


Condoms . .. .. t4 3584 0.06 l.g8 ,
Yt:"*o
RefusingS*.t......, r-3 3633 l:65
.OO4
C o n d o mu s e . . . . . . . . . . . . . . . . . . . . r-3 .3592 o . tI 4 : 1 5 "r
C o * r u n i . " t i o n. . . . . . . . . . . . . . . . . l-3. 3s76 0.01 o:53
Barriers to condom use . . ., t4 "3682 - 0 . tI .2.51
Risk perceptions

. . . . .
r-5 3584 0J I ',os,
.o0li'
-3'.1";111
Communicationwith parents
r-s
r-3
3685
3632
0.09
0.05 ili 0:06:r
NOTES:.All models were adiustedfor baselinevaluesand other relevant covariates(suchas gender, parens' education,nurnber parents
of
or guardians in household, grade point average,-andethnicky)- hH modek also includld indepindem variables:group (intervention =
|,
control = 0); location (California= l, Texas =.0), and groui-by-l6citibn interaction terms (Calrforniainterven'tjbn= i, re-"ining = 0).

(3-5, depending on the scale) r.ufl".. ror. of the desired anribuce, excep c for variables barriersto
IiSh:: yrllu: ' use and risk perceptlons'
r.rirrr,i.r' nigh"l: r-."Jr"'n-".a s.i"ou.. b1..1"r. and risk perceptions. "condom

3"m;r:,:5Jffiff..;Til:flfiHHf,*J,*:::*:';J5#ff""1''*'.u..*e onthbnumber
arebased ofobservations
rather
than
"Escimateof magniude of effe

[I]"iH,f'tHrff::ii,,::[Tl,J"ffiJ#]rutro;+"*ru;"'*r"i"[""?'i:'ffi.tt
dGroup-by-location
interacdoRsignificantat P <O05; ratio of interacdonestimareto interacdonstandarderror = -2. lB
€roup-by-focation interacdon significant at P <0.05: ratio of interaction estimare to inreracdon standard error = 2.97

HIV risk perceptions,other STD risk perceptions,and the psychosocialvariables, particularly HIV and other
communication with parents. Seven of these variables STD knowledgeand variablesrelatedto condom use.
remained significant at the 3l-month follow-up (all but Among students who reported having sexual inter-
normativebeliefs about condomsand parent communica- course, the Safer Choicesstudents had decreasedinter-
t i o n .n h i c h w e r eb o r hs i s n i f i c a nat t P = O . O O
.f coursewithout a condom by slightly more than one-third.
Although muiti-level analysesdo not enable direct calcu-
DrscussroN lation of the adjustedmeans,we can extrapolatefrom the
raw means.Over the study period, students in the com,
Our findings suggestthat the Sat'erChcticesprogram pro- parisongroup reported having intercoursewithout a con-
duced numerouspositiveand programmaticallyimportant dom an averageof 3.82 times during the previous three
effects among cohort students.In general,SaferChoices months.An effect size of 0.63 indicatesthat, after proper
had a consistenteffect on condom use and use of other adjustments,students in the intervention group had
protectivemethods.It alsohad a positiveeffect on most of intercoursewithout a condom an averageof 2.40 times.

PUBLIC HEALTH REPORTS . 2OOI SUPPLEMENT I . VOLUME I I6


j
" -:'^*,.

r nffi,1[ffi"*,,.,..r,,,ii; -'.,],f,-t.n.,fir"-filtt
.$ffi',,,,.
" ,. ,,.ffiffi
..'ifi'"il*',
ffi-#,;*ffiffi:ffi;';'"Tffi**
;i:r:i:::l

.qga':i
ngsqlve

The results also suggest that Safer Choices students that the social norms supportingsexualactivity were too
reduced the number of sexualpartners with whom they strong for such an intervention to reverse.Indeed, from
had intercoursewithout a condom by nearly one-quarter: one-fourth to one-third of the students in the study had
comparison students reported having unprotected sex reported having sex at baseline, and many others were
with an averageof .69 partners during the three months probably considering it before the intervention began.
prior to the follow-up survey, and Safer Choices students Given the potential influence of such norms, it is possible
reported .50 partners. Tke Safer Ckoices students also that high schoolis too late to have a substantialeffect on
significantlyincreasedthe use of condomsand pregnancy the initiation of sexualintercourse,and that condom use
prevention methods (condoms, birth control pills, con- behaviormay be more salient for this population. Several
dom plus birth control pills) at last intercourse. studiesof HIV preventionProgramshave found that pro-
Counter to our expectations,the intervention did not grams are often more effective at changing condom use
significantly delay the onset of sexual intercourse, than at changingthe incidence of sexualintercourse.s'e'22
although the effect was in the desired direction for this Finally, it is also possible that the lack of a significant
variable.Notably, the intervention clearly did not hasten effect on delaying sexual initiation was statistical in
the onset of intercourse.The programemphasizedchoos- nature. Becausea limited number of students initiated
ing not to have sex as the safest choice for preventing sexbetweenbaselineand the 3l-month follow-up(14Vo),
HIV, other STD, and unplanned pregnancy; nearly all we may have lacked statisticalpower to detect the differ-
lessonsin the curriculum reinforced this message.Many ential rate of initiation between the two grouPs.
of the school-wideactivities also emphasizedthe impor- The intervention produced its most pronounced effects
tance and value of choosingnot to have sex.It is possible on HIV knowledge.The program also had a positive effect

PUBLIC HEALTFI REPORTS . 2OO1 SUPPLEMENT I . VOLUME 116


on other STD knowledgeand most other psychosocialvari- have restricted potential gains for California students.
ables,particularlythoserelatedto condom use. Similarly,Texasstudentshad higher baselinecondom use
The program did not significantly affect students'per- self-efficacyscores.By the final follow-up, the mean val-
ceived ability to refuse intercourse or unprotected inter- ues for thesevariableswere similar in both states,suggest-
course or their perceived abillty to communicate with a ing the intervention was equally effective in both sites,
partner about not har.ingsexor usingprotection, despitethe and that there may be a ceiling effect for thesemeasures.
fact that theseskills were amongthe core skills addressedin The results are strengthenedby their relative consis-
the cur"riculum.Howeveqstudents'baseline scoreson both tency over time. The majority of effects remained signifi-
the refusal self-efficacy and communicarion self-efficacy cant over time, although the magnitude of many of the
scaleswere high initially (2.4 and2.7 out of 3.0), thereby effects dissipated somewhat. This pattern is typical of
limiting room for positive changefor thesevariables. other health promotion intervention trials.23'24 Despite
We noted statistically significant group-by-location the general consistencyof effects over time, there were
interactions for two of the psychosocialeffects in the fewer significant behavioral effects at the l9-month
cohort sample: oth.erSTD knrnutedgeand comd.om useself- follow-up than there were at the 7-month and 31-month
,ff Studentsin both statesdemonstratedsignificant follow-ups. This may be due to the difficulty we experi-
"o"y.
gains on both scales.The gains in other STD knowledge enced in tracking the cohort in the Texasschoolsat the
were strongerin Texas,whereasthe gainsin condom use 19-month fbllow-up, resulting in lower statisticalpower
self-efficacywere strongerin Califomia. Studentsin the at that time point. This hlpothesis is supported by the
California schools had higher baseline scores on other fact that the trends for the behavioralvariablesare in the
STD knowledge than did students in Texas, which may desired direction at the 19-month assessment,and the

PUBLIC HEALTH REPOR.I.S.2OOI SUPPLEMENT I . VoLUME 116 9l


magnitudeof the effects are fairly sizeable.The presence ual intercourse,decreasethe frequencyof intercourse,or
of significant effects at both the 7-month and 3l-month reduce the number of sexualpartners;however,it did not
assessments iend further support to this hypothesis. increasethesebehaviors,either.
When the results of this cohort study are combined
Study limitations. The results of this study are encour- with the results of the cross-sectionaistudy that exam-
aging;howeveqseveralmethodologicallimitations should ined school-wideeffects ol SaferCh'oices,t3the data sup-
be noted. The outcome data were collectedby using self- port the value of a comprehensive,multi-component
report questionnairesand there are few, if any,acceptable interventionthat includes an intensive2O-lessonsequen-
approachesfor examlning the criterion validity of stu- tial curriculum for 9th- and i0th-grade students sup-
dents'responses.However,it is reasonableto assumethat ported by broad-reaching, school-wide activities. The
privacy and confidentiality affect the veracity of self- intensivecomponent providesan opportunity for individ-
report.25'26Thus, severalapproacheswere used to createa ual skill practice and mastery that is not easily gained
safe and comfortable environment for completing the through school-wide activities. The school-wide activi-
questionnaire.These included using trained data collec- ties, on the other hand, reinforce the classroomcompo-
tors, providing students with paper to cover their nent by serving as environmental cues; they also con-
answers,and providing a formal assuranceof confiden- tribute to a more supportive environment for practicing
tiality. Although it is impossible to rule out potential healthful behaviors.
biases due to self-report, some evidence supports the These results representstrongerand more consistent
general reliability and validity of adolescents'reportsof behavioral findings than reported in other randomized
sexualand contraceptivebehaviors.25'27 school-basedevaluationsof HIV other STD, and preg-
Studentsdroppedfrom the cohort or lost to follow-up nancy preventionprograms.28'3o Severalrandomizedstud-
differed from studentsretainedin the cohort. Dropped or ies conducted in communitl' 5"111tt*thave also found
lost studentswere older, less likely to live with both par- behavioralchanges,but they were conductedwith adoles-
ents, more likely to be males,reportedmore risk behavior cents who volunteered to participate in the project.B'lo
at baseline,and had less favorablescoreson many of the Such participants are likely to be more motivated and
psychosocialscales. In light of these differences, our interestedin making behavioralchangesthan were the par-
study results may not generaiizeto all adolescentswho ticipants in our study.The population-basedapproachwe
are absent frequently from school or who have left used is a more conservativetest of the intervention effects.
school.Yet, the studentsin our study representan impor- A high level of rigor was employed in evaluatingthe
tant heterogeneouspopulation of young people. program. The study included random assignmentof 20
Finally, we did not correct the statisticalprocedures schools, a large cohort sample, long-term follow-up
to adjust for multiple tests of significance.However,we assessments,assessmentof multiple sexual and contra-
did limit our analysesto a small number of primary and ceptive behaviors,and the use of multilevel analysesto
secondaryhypotheses,and further testing was regarded adjust for the clusteringof studentswithin schools.Con-
as exploratory.We also consideredthe overall pattern of fidence in these results is also strengthenedby the con-
the results: if selected results had occurred by chance, sistencyof results acrosstime and acrossmultiple mea-
one might expectresultsin both the positiveand negative suresinvolvingcondom use.
direction. The significant results of this study all were This study indicatesthat theory-driven,school-based,
consistentwith the program'stheoreticalfoundation and multi-component programs with a clear messagecan
consistentover time, suggestingthat the observedresults enhance psychosocialvariables and reduce sexual risk
are not attributableto chance. behaviorsrelatedto HIV, other STDs, and pregnancypre-
vention among high school students.It also suggeststhat
CONCLUSION additional research is needed to identify successful
approachesto delaying sexual initiation and enhancing
SaferChoiceswas successfulin changingfour of five out- potentially important determinants(such as self-efficacy
comes addressingcondom use and other protective to refuse intercourse and communication with partners
behaviors.The program also enhancednumerous behav- about sexuallimits), and that even larger samplesmay be
ioral determinants,particularly those related to condom neededto measurethe effect on sexualinitiation.
use.These positiveeffects lastedover a 31-month period. SaferCkoi.ceshas some of the strongestdata suPPort-
SaferChoicesdid not significantlydelay the onset of sex- ing its effectivenessin reducingimportant sexualrisk-tak-

PUBLIC HEALTH REPORTS . 2OO1 SUPPLEMENT 1 . VOLUME 1I6


ing behaviors. Several other programs also have been Choices,the published program materials include
shown to be effective.6To the extent feasible,school dis- detailed implementation manuals and other materials
tricts interested in addressingthese issuesshould adopt that provide extensiveguidelinesregardinghow to imple-
programsthat have been shown to be effective through ment each programcomponent.3l
research and implement them with fidelity. Our inter-
views with schoolpersonnelinvolvedin the programsug- This study was supported by Contract No. 200-91-0938from the Centers
gestedthat SaferChoicesis well receivedbv studentsand for DiseaseControl and Prevention-
staff and can be implementedwith appropriateplanning.
Indeed, the year following the study,most schoolsin the The authors gratefullyacknowledgethe contributionsto this research
project by Nancy Calvin,BarbaraCollins, Chris Harvey, Deborah lvie,
study continued offering parts of the program, although
Chris Markham,JesseNodora, MargoParr,MarshaWeil, and Duane
we do not have extensivedata regarding the extent of
Wilkerson.
implementation. For school districts interested in Safer

Ref e r e n c e s

l. Centers for DiseaseControl and prevention(US). HIV/AIDSsurveil_ 18. Marsh D, Brown E, Crocker P, Lewis H. Buildingeffectivemiddle
lancereport 2000;| 2( | ):22. schools:a studyof middleschoolimplementation in Californiaschools.
2. HIV preventionstrategicplanthrough 2005.Atlanta(GA): Centersfor Los Angeles(CA): Schoolof Education,Universityof SouthernCalifor-
DiseaseControl and Prevention.In press. nia;1988.
3. DonovanP. Testingpositive:sexuallytransmitteddiseaseand the oub_ 19. Coyle K, Kirby D, ParcelG, et al. Saferchoices:a multicomponenr
lic healthresponse.New York Alan GuttmacherInstitute;| 993. school-based HIV/STDand pregnancypreventionprogramfor adoles-
4. Henshaw SK. US teenage pregnancy statistics. New york: Alan cents.J SchoolHealth 1996;66:89-94.
GuttmacherInstitute;| 998. 20. Basen-Engquist K, MasseLC, Coyle K, et al. Sexualrisk behaviorbeliefs
5. Alan GuttmacherInstitute.Sex and America'steenagers.New york: and selfefficacyscales.In: DavisCM, YarberWL, et al.,editors.Hand-
Alan GuttmacherInstitute:1994. book of sexuality-related measures.
ThousandOaks (CA): Sage;1998.
6. Kirby D. Emerginganswers:researchfindingson programsto reduce p. 54 | -4.
teen pregnancy. Washington(DC): The NationalCampaignro prevent 21. Rasbash J, WoodhouseG. MLn commandreference.London:Institute
Teen Pregnancy; 200| . of Education,Universityof London;| 995.
7. Kirby D, Barth R, LelandN, Fetro Reducingthe risk: a new curncu- 22. Kim N, StantonB, Li X, DickersonK, GalbraithJ. Effectiveness
J. of the
t u mt o p r e v e n ts e x u arl i s k - t a k i n gF.a mp l a n np e r s p e c tl 9 9 l ; 2 3 : 2 5 3 _ 6 3 . 40 adolescentAIDS-riskreductionintervention:a ouantitativereview.
8. JemmottJB lll, Jemmott LS, Fong GT. Reductionsin HIV risk_associ_
J of Adol HealthCare 1997;20:204-15.
ated sexualbehaviorsamongblackmaleadolescents: effectsof an AIDS 23. JemmottJB,JemmottLS,FongGT. Abstinenceand safersex HIV risk-
preventionintervention.Am J publicHealth 1992;92:372-7. reduction interventionsfor African Americanadolescents: a random-
9. MainDS, lversonDC, McGloinJ,et al. preventingHIV infectronamong izedcontrolledtrial. JAMA 1998;2791529-36.
adolescents:evaluationof a school-basededucationprogram. prev 24. ResnicowK, Botvin G. School-based substanceuse preventionpro-
tlled 1994:23:409-17. grams:why do effectsdecay?Prev Med 1993;22:484-90.
10. St LawrenceJS,JeffersonKW, AlleyneE, BrasfieldTL. Comparisonof 2 5 . B r e n e rN D , C o l l i n s J LK, a n nL , W a r r e n C W , W i l l i a m sB l : R e l i a b i l i toyf
educationversus behavioralskills training interventionsin lowering t h e y o u t h r i s k b e h a v i o r s u r v e y q u e s t i o n n a i r eA
. m J Epidemiol
sexualHIV risk behaviorof substancedependentadolescents. Con_ 1995;l4l:575-80.
J
s u l ta n dC l i n P s y c h o|l 9 9 5 ; 6 3 : 2|2- 3 7 . 26. SudmanS, BradburnNM. Responseeffectsin surveys.Chicago:Aldine
| |. Coyle K, Basen-Engquist K, Kirby D, parcelG, Banspach S, Harrist R, et P u b l i s h i n g19; 7 4 .
al. Shorcterm impacrof a multi-componentschool_based HlV, other 27. SonensteinF, Ku L, PleckJ:Measuringsexualbehavioramongteenage
STD and pregnancy preventionprogram.J SchoolHealth I 999;69:I g | _g. males in the US. Paper presentedat ResearchingSexualBehavior:
12. Basen-Engquist K, ParcelG, Harrist R, et al. Methodolosicatissuesin Methodologicallssues,KinseyInstitutefor Researchin Sex, Gender,
s c h o o l - b a s e dh e a l t h p r o m o t i o n i n t e r v e n t i o n r . r u r r J , r h e s a f e r and Reproduction;| 996 Apr 26-28; Universityof Indiana,Indianopolis.
choicesproject.J SchoolHealth | 997;67:365-7l. 28. LevySR,PerhatsC, Weeks K, HandlerAS, Zhu C, FlayBR.lmpactof a
13. Basen-Engquist K, Coyle K, parcel G, et al. School-wideeffectsof a school-based AIDS preventionprogramon risk and protectivebehav-
multicomponentHlV, STD,and pregnancypreventionprogramfor high ior for newlysexuallyactivestudents.J SchoolHealth | 995;65;| 45-5| .
schoolstudents.HealthEducBehav200| ;2g(2):| 66-95. 29. Mitchell-DiCensoA, ThomasBH, Devlin MC. et al. Evaluationof an
14. BanduraA. Socialfoundationsof thoughtand action.EnglewoodCliffs educationalprogram to prevent adolescentpregnancy.Health Educ
( N J ) :P r e n t i c eH a l t ;t 9 8 6 . B e h a v1 9 9 7 ; 2 4 : 3 0 0 - 1 2 .
| 5. FisherJD. Possibleeffectsof referencegroup-based socialinfluenceon 30. Kirby D, Korpi M, Adivi C, WeissmeanJ. An impactevaluationof pro-
A I D S - r i s kb e h a v i o rasn dA I D S .A m p s y c h o ll g g g ; 4 3 : 9 1 4 - 2 0 .
lect SNAPP:an AIDS and pregnancypreventionmiddle school pro-
16. McGuireW. Socialpsychology. ln: Dodwell pC , editor. New horizons g r a m .A I D SE d u cP r e v 1 9 9 7 ; 9 ( 5 A ) : 4 4 - 6 1 .
in psychology. Middlesex(England): penguinBooks; 1972.p.219_42. 31. ETRAssociates,Center for HealthPromotionResearchand Develoo-
17. McGuire W, PapageorisD. The relative efficacyof various types of ment, University of Texas-Houston, Health ScienceCenter. Safer
prior belief-defense in producingimmunityto persuasion. J Abnorm choices:preventingHlV, other STD and pregnancy.SantaCruz (CA):
Soc Psychol 196l;62:237-331. ETRAssociates;
1998. I

PUBLIC HEALTH REPORTS . 2OO1 SUPPLEMEN.f 1 . VOLUME 116